Classifications

• • G—PHYSICS
  • G09—EDUCATION; CRYPTOGRAPHY; DISPLAY; ADVERTISING; SEALS
  • G09G—ARRANGEMENTS OR CIRCUITS FOR CONTROL OF INDICATING DEVICES USING STATIC MEANS TO PRESENT VARIABLE INFORMATION
  • G09G5/00—Control arrangements or circuits for visual indicators common to cathode-ray tube indicators and other visual indicators
  • G09G5/003—Details of a display terminal, the details relating to the control arrangement of the display terminal and to the interfaces thereto
  • G09G5/006—Details of the interface to the display terminal
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L1/00—Arrangements for detecting or preventing errors in the information received
  • H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
  • H04L1/0041—Arrangements at the transmitter end
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L1/00—Arrangements for detecting or preventing errors in the information received
  • H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
  • H04L1/0056—Systems characterized by the type of code used
  • H04L1/0057—Block codes
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L1/00—Arrangements for detecting or preventing errors in the information received
  • H04L1/20—Arrangements for detecting or preventing errors in the information received using signal quality detector
  • H04L1/203—Details of error rate determination, e.g. BER, FER or WER
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L1/00—Arrangements for detecting or preventing errors in the information received
  • H04L1/20—Arrangements for detecting or preventing errors in the information received using signal quality detector
  • H04L1/206—Arrangements for detecting or preventing errors in the information received using signal quality detector for modulated signals
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
  • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
  • H04N19/85—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using pre-processing or post-processing specially adapted for video compression
  • H04N19/89—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using pre-processing or post-processing specially adapted for video compression involving methods or arrangements for detection of transmission errors at the decoder
• • G—PHYSICS
  • G09—EDUCATION; CRYPTOGRAPHY; DISPLAY; ADVERTISING; SEALS
  • G09G—ARRANGEMENTS OR CIRCUITS FOR CONTROL OF INDICATING DEVICES USING STATIC MEANS TO PRESENT VARIABLE INFORMATION
  • G09G2370/00—Aspects of data communication
  • G09G2370/04—Exchange of auxiliary data, i.e. other than image data, between monitor and graphics controller
  • G09G2370/045—Exchange of auxiliary data, i.e. other than image data, between monitor and graphics controller using multiple communication channels, e.g. parallel and serial
  • G09G2370/047—Exchange of auxiliary data, i.e. other than image data, between monitor and graphics controller using multiple communication channels, e.g. parallel and serial using display data channel standard [DDC] communication

Definitions

• the invention pertains to methods and systems for transmitting encoded video data and at least one other stream of encoded data (e.g., encoded video data and packets of encoded audio and/or other auxiliary data) over a serial link, and to transmitters and receivers for use in such systems.
• the serial link is a transition minimized differential signaling ("TMDS") link, or a link having some but not all of the characteristics of a TMDS link.
• TMDS transition minimized differential signaling
• base 10 decimal (base 10) numbers are represented using no additional prefixes or suffixes, and the following notation is sometimes used:
• bit 0 denotes the least-significant bit of a byte or word
• prefix “Ox” denotes that the following symbol is a hexadecimal representation of a number (for example “OxC” denotes a binary number 1100); and the prefix “0b” denotes that the following symbol is a binary (base-2) representation of a number (for example "0b 1000” denotes a decimal number 8).
• Elements of this invention are based upon properties of a serial link.
• Various serial links for transmitting data and clock signals are well known.
• TMDS transition minimized differential signaling interface
• video data are encoded and then transmitted as encoded words (each 8-bit word of digital video data is converted to an encoded 10-bit word before transmission); a. the encoding determines a set of "in-band” words and a set of "out- of-band” words (the encoder can generate only "in-band” words in response to video data, although it can generate "out-of-band” words in response to control or sync signals. Each in-band word is an encoded word resulting from encoding of one input video data word. All words transmitted over the link that are not in- band words are "out-of-band” words); b.
• the encoding of video data is performed such that the in-band words are transition minimized (a sequence of in-band words has a reduced or minimized number of transitions); c. the encoding of video data is performed such that the in-band words are DC balanced (the encoding prevents each transmitted voltage waveform that is employed to transmit a sequence of in-band words from deviating by more than a predetermined threshold value from a reference potential. Specifically, the tenth bit of each "in-band" word indicates whether eight of the other nine bits thereof have been inverted during the encoding process to correct for an imbalance between running counts of ones and zeroes in the stream of previously encoded data bits);
• the encoded video data and a video clock signal are transmitted as differential signals (the video clock and encoded video data are transmitted as differential signals over conductor pairs); 3. three conductor pairs are employed to transmit the encoded video, and a fourth conductor pair is employed to transmit the video clock signal; and
• TMDS serial link is the "Digital Visual Interface" interface ("DVI" link) adopted by the Digital Display Working Group. It will be described with reference to Fig. 1.
• DVI link can be implemented to include two TMDS links (which share a common conductor pair for transmitting a video clock signal) or one TMDS link, as well as additional control lines between the transmitter and receiver.
• DDC Display Data Channel
• HPD Hot Plug Detect
• the Display Data Channel standard specifies a protocol for bidirectional communication between a transmitter and a monitor associated with a receiver, including transmission by the monitor of an Extended Display Identification (“EDID") message that specifies various characteristics of the monitor, and transmission by the transmitter of control signals for the monitor.
• Transmitter 1 includes three identical encoder/serializer units (units 2, 4, and 6) and additional circuitry (not shown).
• Receiver 3 includes three identical recovery/decoder units (units 8, 10, and 12) and inter-channel alignment circuitry 14 connected as shown, and additional circuitry (not shown).
• circuit 2 encodes the data to be transmitted over Channel 0, and serializes the encoded bits.
• circuit 4 encodes the data to be transmitted over Channel 1 (and serializes the encoded bits), and circuit 6 encodes the data to be transmitted over Channel 2 (and serializes the encoded bits).
• Each of circuits 2, 4, and 6 responds to a control signal (an active high binary control signal referred to as a "data enable" or "DE" signal) by selectively encoding either digital video words (in response to DE having a high value) or a control or synchronization signal pair (in response to DE having a low value).
• DE data enable
• encoder 2 receives horizontal and vertical synchronization signals (HSYNC and VSYNC); encoder 4 receives control bits CTLO and CTLl; and encoder 6 receives control bits CTL2 and CTL3.
• each of encoders 2, 4, and 6 generates in-band words indicative of video data (in response to DE having a high value), encoder 2 generates out-of-band words indicative of the values of HSYNC and VSYNC (in response to DE having a low value), encoder 4 generates out-of-band words indicative of the values of CTLO and CTLl (in response to DE having a low value), and encoder 6 generates out-of-band words indicative of the values of CTL2 and CTL3 (in response to DE having a low value).
• each of encoders 4 and 6 In response to DE having a low value, each of encoders 4 and 6 generates one of four specific out-of-band words indicative of the values 00, 01, 10, or 11, respectively, of control bits CTLO and CTLl (or CTL2 and CTL3).
• HDCP High- bandwidth Digital Content Protection
• a DVI transmitter implementing HDCP outputs a 24-bit bus, known as cowt[23:0], during the video active period (i.e. when DE is high).
• This 24-bit cout data is "Exclusive Ored” (in logic circuitry in the transmitter) with the 24-bit RGB video data input to the transmitter in order to encrypt the video data.
• the encrypted data is then encoded (according to the TMDS standard) for transmission. The same cout data is also generated in the receiver.
• the cout data is processed together with the decoded video in logic circuitry in order to decrypt the decoded data and recover the original input video data.
• the transmitter and receiver communicate bidirectionally with each other to execute an authentication protocol (to verify that the receiver is authorized to receive protected content, and to establish shared secret values for use in encryption of input data and decryption of transmitted encrypted data).
• the transmitter calculates the initial set of encryption keys (for encrypting the first line of input video data) in response to a control signal and sends the control signal to the receiver (during each vertical blanking period, when DE is low) to cause the receiver to calculate an initial set of decryption keys (for decrypting the first received and decoded line of transmitted video data).
• each of the transmitter and receiver performs a re-keying operation during each blanking (vertical or horizontal) interval to generate a new set of keys for encrypting (or decrypting) the next line of video data, and actual encryption of input video data (or decryption of received, decoded video data) is performed using the latest set of keys only when DE is high (not during the blanking intervals).
• Each of the transmitter and receiver includes an HDCP cipher circuit (sometimes referred to herein as an "HDCP cipher") including a linear feedback shift register (LFSR) module, a block module coupled to the output of the LFSR module, and an output module coupled to an output of the block module.
• LFSR linear feedback shift register
• the LFSR module is employed to re-key the block module in response to each assertion of an enable signal, using a session key (Ks) and frame key (Ki).
• Ks session key
• Ki frame key
• the block module generates (and provides to the LFSR module) the key Ks at the start of a session and generates (and applies to the LFMS module) a new value of key Ki at the start of each frame of video data (in response to a rising edge of a control signal which occurs in the first vertical blanking interval of a frame).
• the block module comprises two halves, known as "Round Function K” and "Round Function B.”
• Round Function K includes 28-bit registers Kx, Ky, and Kz, seven S-Boxes (each a 4 input bit by 4 output bit S-Box including a look-up table), and a linear transformation unit K.
• Round Function B includes 28-bit registers Bx, By, and Bz, seven S-Boxes (each a 4 input bit by 4 output bit S-Box including a look-up table), and a linear transformation unit B.
• Round Function K and Round Function B are similar in design, but Round Function K performs one round of a block cipher per clock cycle to assert (to the output module) a different pair of 28-bit round keys (Ky and Kz) each clock cycle in response to the output of the LFSR module, and Round Function B performs one round of a block cipher per clock cycle, in response to each 28-bit round key Ky from Round Function K and the output of the LFSR module, to assert (to the output module) a different pair of 28-bit round keys (By and Bz) each clock cycle.
• the transmitter generates value An at the start of the authentication protocol and the receiver responds to it during the authentication procedure.
• the value An is used to randomize the session key.
• the block module operates in response to the authentication value (An) and an initialization value (Mi, also referred to as an integrity verification key) which is updated by the output module at the start of each frame.
• Each of linear transformation units K and B outputs 56 bits per clock cycle.
• Each diffusion network of linear transformation unit K produces seven output bits in response to seven of the current output bits of registers Ky and Kz.
• Each of four of the diffusion networks of linear transformation unit B produces seven output bits in response to seven of the current output bits of registers By, Bz, and Ky, and each of the four other diffusion networks of linear transformation unit B produces seven output bits in response to seven of the current output bits of registers By and Bz.
• the output module performs a compression operation on the 28-bit keys (By, Bz, Ky and Kz) asserted to it (a total of 112 bits) by the block module during each clock cycle, to generate one 24-bit block of pseudo-random bits cout [23:0] per clock cycle.
• Each of the 24 output bits of the output module consists of the exclusive OR ("XOR") of nine terms.
• logic circuitry receives each 24-bit block of cout data and each input 24-bit RGB video data word, and performs a bitwise XOR operation thereon in order to encrypt the video data, thereby generating a word of encrypted RGB video data.
• the encrypted data subsequently undergoes TMDS encoding before it is transmitted to a receiver.
• logic circuitry receives each 24-bit block of cout data and each recovered 24-bit RGB video data word (after the recovered data has undergone TMDS decoding), and performs a bitwise XOR operation thereon in order to decrypt the recovered video data.
• TMDS-like link will be used to denote a serial link capable of transmitting encoded data (e.g., encoded digital video data) and a clock for the encoded data, from a transmitter to a receiver, and also capable of transmitting (bidirectionally or unidirectionally) one or more additional signals (e.g., encoded digital audio data or other encoded data) between the transmitter and receiver, that is or includes either a TMDS link or a link having some but not all of the characteristics of a TMDS link.
• encoded data e.g., encoded digital video data
• additional signals e.g., encoded digital audio data or other encoded data
• TMDS-like links include links that differ from TMDS links only by encoding data as N-bit code words (e.g., with N 10) that are not 10-bit TMDS code words, and links that differ from TMDS links only by transmitting encoded video over more than three or less than three conductor pairs.
• N-bit code words e.g., with N 10
• TMDS-like links There are several conventional TMDS-like links.
• transmitter is used herein in a broad sense to denote any device capable of encoding data and transmitting the encoded data over a serial link (and optionally also performing additional functions, which can include encrypting the data to be transmitted and other operations related to encoding, transmission, or encryption of the data).
• receiver is used herein in a broad sense to denote any device capable of receiving and decoding data that has been transmitted over a serial link (and optionally also performing additional functions, which can include decrypting the received data and other operations related to decoding, reception, or decryption of the received data).
• transmitter can denote a transceiver that performs the functions of a receiver as well as the functions of a transmitter.
• the term transmitter (with reference to a device that transmits non- audio auxiliary data over a TMDS-like link or other serial link) can denote a transceiver that is configured to receive video data and audio data over the link and to transmit the non-audio auxiliary data over the link.
• Some TMDS-like links encode input video data (and other data) to be transmitted into encoded words comprising more bits than the incoming data using a coding algorithm other than the specific algorithm used in a TMDS link, and transmit the encoded video data as in-band characters and the other encoded data as out-of-band characters.
• the characters need not be classified as in-band or out-of-band characters based according to whether they satisfy transition minimization and DC balance criteria.
• classification criteria could be used.
• the classification (between in-band and out-of-band characters) need not be based on just a high or low number of transitions.
• the number of transitions of each of the in-band and out-of-band characters could (in some embodiments) be in a single range (e.g., a middle range defined by a minimum and a maximum number of transitions).
• the data transmitted between the transmitter and receiver of a TMDS-like link can, but need not, be transmitted differentially (over a pair of conductors). Also, although a TMDS link has four differential pairs (in the single pixel version), three for video data and the other for a video clock, a TMDS-like link could have a different number of conductors or conductor pairs.
• the primary data transmitted by a TMDS link are video data. What is often significant about this is that the video data are not continuous, and instead have blanking intervals. These blanking intervals provide an opportunity (exploited in some embodiments of the present invention) for auxiliary data to be transported, and they represent unused bandwidth.
• many serial links do not transmit data having blanking intervals, and thus do not encode input data (for transmission) in response to a data enable signal. For example, audio serial links would typically transmit continuous data.
• auxiliary data is used in a broad sense herein to denote digital audio data or any other type of data other than video data and timing information for video data (e.g., a video clock).
• timing information for audio data e.g., a clock for recovering transmitted audio data
• auxiliary data transmitted in accordance with the invention include computer keyboard signals, still image data (generated by a camera, for example), text data, control signals for a power supply, picture in picture data, monitor control information (audio volume, brightness, power state), control signals for indicator lights on a monitor or keyboard, non-audio or video control information, etc.
• the term "stream” of data denotes that all the data are of the same type and is transmitted with the same clock frequency.
• channel refers to that portion of a serial link that is employed to transmit data (e.g., a particular conductor or conductor pair between the transmitter and receiver over which the data are transmitted, and specific circuitry within the transmitter and/or receiver used for transmitting and/or recovery of the data) and to the technique employed to transmit the data over the link. Because it is desirable to transmit many different streams of auxiliary data in important applications of the invention, preferred embodiments of the invention provide multiple channels for transmission of auxiliary data, including channels for transmission of auxiliary data in both directions over the link (that is, with and against the direction of the video data).
• a channel is employed to transmit one stream of auxiliary data. In other implementations, a channel is employed to transmit more than one stream of auxiliary data. In some embodiments of the invention, two (or more than two) streams of serial video data are transmitted (over one, two, or more than two channels), and either one, two, or more than two streams of serial auxiliary data are also transmitted.
• synchronization words can be transmitted over the link during "preamble" periods in which encoded video data are not transmitted.
• the synchronization words can be transition maximized words that are members of a second subset (disjoint from the first subset) of the set of code words.
• 5,999,571 teaches that several (e.g., three) repetitions of a synchronization word should be transmitted consecutively, to allow the decoder (in the receiver) rapidly and accurately to identify a specific transition (e.g., the leading edge) of one of the synchronization words and thus to accomplish synchronization with the encoder (in the transmitter.
• U.S. Patent 6,151,334, issued November 21, 2000 teaches transmission (over a TMDS link) of several different types of encoded control words, each distinguishable from transition minimized code words indicative of data. At least some of the control words can be transition maximized words.
• One of the control words is a "data stream separation" word that is transmitted before or after a burst of data and is indicative of the start or end of a burst and the type of data transmitted during the burst.
• Another one of the control words is an "isochronous data transfer" word that is a synchronization character typically transmitted at the beginning or end of a blanking interval and indicates the type of the blanking interval (e.g., horizontal or vertical) and distinguishes between the beginning and the end of the blanking interval.
• a first isochronous data transfer word indicates the start of a vertical blanking interval
• a first data stream separation word indicates the start of a burst of data in the vertical blanking interval
• a second data stream separation word indicates the end of such data burst
• a second isochronous data transfer word indicates the end of the vertical blanking interval
• Each of the first isochronous data transfer word, the first data stream separation word, the second data stream separation word, and the second isochronous data transfer word is a transition maximized code word
• a transition minimized code word can indicate each word of data of the data burst (transmitted in the vertical blanking interval)
• the vertical blanking interval can be followed by an active video period comprising a third data stream separation word (indicative of the start of a stream of video data) followed by a stream of transition minimized code words indicative of the video data itself.
• the invention is a communication system including a transmitter, a receiver, and a serial link (having at least one channel), for transmitting a stream of encoded video data and at least one other stream of encoded data (e.g., encoded audio data) over each of one or more channels of the link from the transmitter to the receiver.
• the serial link is a TMDS or TMDS-like link.
• the transmitter sends encoded video to the receiver over each of one or more video channels of the link during active video periods, and sends packets including encoded auxiliary data (e.g., audio data) to the receiver over each of at least one of the video channels during data islands, wherein each of the data islands is a time interval that neither coincides with nor overlaps any of the active video periods.
• the transmitter preferably also transmits control data to the receiver over each of at least one of the video channels during control data periods, wherein the control data periods neither coincide with nor overlap any of the data islands and neither coincide with nor overlap any of the active video periods.
• transmitters for use in formatting and encoding multiple streams of data for transmission over a serial link
• receivers for receiving and decoding multiple streams of encoded data transmitted over a serial link
• methods for sending multiple streams of encoded data over a serial link are transmitters for use in formatting and encoding multiple streams of data for transmission over a serial link, receivers for receiving and decoding multiple streams of encoded data transmitted over a serial link, and methods for sending multiple streams of encoded data over a serial link.
• One or more packets can be transmitted in each data island.
• control data periods sometimes referred to as a "control data interval” or “control period”
• encoded control words preferably including sync words and preamble words
• 8 -bit video data words are transmitted over the video channels of a TMDS link (or other TMDS-like link having multiple channels for transmitting serial video) during active video periods.
• packets containing 4-bit words typically including 4-bit audio data words
• each encoded according to the TMDS encoding algorithm as a 10-bit, transition-minimized code word, and preferably as a 10- bit golden word are transmitted over each of at least some of the video channels.
• the transmitter sends control words (each encoded as a 10-bit, transition-maximized code word indicative of two control bits: CTLO and CTLl, or CTL2 and CTL3) and sync words (each encoded as a 10-bit, transition-maximized code word indicative of two sync bits: HSYNC and VSYNC) over the video channels.
• control words each encoded as a 10-bit, transition-maximized code word indicative of two control bits: CTLO and CTLl, or CTL2 and CTL3
• sync words each encoded as a 10-bit, transition-maximized code word indicative of two sync bits: HSYNC and VSYNC
• transition-minimized code words indicative of HSYNC and VSYNC bits are sent (e.g., one code word per pixel clock cycle, each word indicative of an HSYNC bit, a VSYNC bit, a packet header bit, and another bits) over one channel (e.g., CHO) during each data island.
• the invention is a transmitter configured to be coupled to a serial link having at least one video channel.
• the transmitter has inputs for receiving video data and audio data (including I 2 S format audio data and/or DSD format audio data), at least one output configured to be coupled to the link, and circuitry configured to generate encoded video data by encoding the video data, to generate code words indicative of packets of auxiliary data, to assert the encoded video data to at least one said output during active video periods, and to assert the code words indicative of packets of auxiliary data to at least one said output during data islands.
• the auxiliary data include the audio data.
• inventions of the inventive transmitter receive multiple streams of input audio data (which can but need not be I 2 S format audio data and/or DSD format audio data).
• the transmitter includes multiplexer circuitry and data packetizing circuitry.
• the multiplexer circuitry is coupled and configured to receive the streams of input audio data and assert predetermined ones of the streams with predetermined timing to the data packetizing circuitry.
• the invention is a transmitter configured to be coupled to a serial link having at least one video channel.
• the transmitter has inputs for receiving at least one stream of audio data having an embedded clock and video data, at least one output configured to be coupled to the link, and circuitry configured to generate encoded video data by encoding the video data, to generate resampled audio data by sampling the stream of audio data with a second clock having no known phase relationship with the embedded clock, to generate code words indicative of packets of auxiliary data, to assert the encoded video data to at least one said output in response to a pixel clock during active video periods, and to assert the code words indicative of packets of auxiliary data to at least one said output in response to the pixel clock during data islands.
• the auxiliary data include time code data and the resampled audio data, and the time code data with the pixel clock are indicative of the second clock.
• the invention is a transmitter configured to be coupled to a serial link having at least one video channel.
• the transmitter has inputs for receiving at least one stream of audio data having an embedded clock and video data, at least one output configured to be coupled to the link, and circuitry configured to generate encoded video data by encoding the video data, to generate resampled audio data by sampling the stream of audio data with a pixel clock having no known phase relationship with the embedded clock, to generate code words indicative of packets of auxiliary data, to assert the encoded video data to at least one said output in response to the pixel clock during active video periods, and to assert the code words indicative of packets of auxiliary data to at least one said output in response to the pixel clock during data islands.
• the auxiliary data include the resampled audio data.
• the inventive transmitter generates code words indicative of packets of auxiliary data, and at least some of the packets are audio sample packets.
• Each audio sample packet has a header and can include at least one sample of the audio data.
• the header includes header data in at least one predetermined time slot indicative of whether the audio sample packet includes an audio data sample and whether each said audio data sample is a flatline sample.
• the invention is a receiver configured to be coupled to a serial link having a video clock channel and at least one video channel.
• the receiver has at least two inputs configured to be coupled to the link (including a clock input to be coupled to the video clock channel), outputs for asserting video data and audio data recovered from the link, and circuitry configured to receive a pixel clock from the clock input, to generate recovered video data including by decoding encoded video data received at at least one of the inputs during active video periods in response to the pixel clock, to assert the recovered video data to at least one of the outputs, to generate decoded data including by decoding code words of packets of encoded auxiliary data received at at least one of the inputs during data islands in response to the pixel clock, to generate at least one stream of I 2 S format audio data (and/or DSD format audio data) from the decoded data, and to assert each such stream of audio data to at least one of the outputs.
• At least some of the decoded data are time code data indicative of time stamps.
• the inventive receiver recovers multiple streams of audio data and includes multiplexer circuitry for asserting each stream of recovered audio data with predetermined timing to at least one predetermined output.
• the inventive transmitter is configured to generate encoded video data that are indicative of video having a second format in response to input video having a first format.
• the inventive receiver is configured to generate recovered video data having a first format in response to encoded video data indicative of video having a second format.
• Fig. 1 is a block diagram of a conventional system including a Digital Visual Interface (“DVI”) link.
• DVI Digital Visual Interface
• Fig. 2 is a block diagram of an embodiment of the inventive system.
• Fig. 3 is a table showing data patterns transmitted in "auxiliary preamble” and “video preamble” portions of a control data period, and guard band code words transmitted after such auxiliary preamble and video preamble portions, in some embodiments of the invention.
• Figs. 4A and 4B are first and second parts, respectively, of a table showing a set of seventeen of the inventive code words (including two guard band words) that are employed in some embodiments of the invention.
• the table also shows other code words that are mapped to each of these seventeen code words by receivers that are designed in accordance with these embodiments.
• Figs. 4A and 4B collectively as "Fig. 4.”
• Fig. 5 is a timing diagram of signals input to the transmitter during a control data period and data island of an embodiment of the inventive system, and encoded signals transmitted over a TMDS link of such system in response thereto.
• Fig. 6 is a timing diagram of signals input to the transmitter during the video preamble portion of a control data period (and during a subsequent active video period) of an embodiment of the inventive system, and encoded signals transmitted over a TMDS link of such system in response thereto.
• Fig. 7 is a diagram of a mapping of clusters (e.g., clusters S a and S b ) of received code words to individual transmitted code words (e.g., code words "a” and "b") in accordance with some embodiments of the invention.
• clusters e.g., clusters S a and S b
• individual transmitted code words e.g., code words "a" and "b
• Fig. 8 is a diagram of the format in which video data (and guard bands) are transmitted in active video periods, packets of data (and guard bands) are transmitted during data islands, and preamble words and control words are transmitted in control data periods, in preferred embodiments of the invention.
• Fig. 9 is diagram of the format in which data are transmitted in packets (during data islands) in preferred embodiments of the invention.
• Fig. 9A is a schematic diagram of a circuit which is included in a class of preferred embodiments of the invention, for generating BCH parity bits to be included with packetized data and then transmitted.
• Fig. 9B is a schematic diagram of a circuit which is included in a class of preferred embodiments of the invention, for generating BCH parity bits to be included with packetized data and then transmitted.
• Fig. 9C is a schematic diagram of a circuit which is included in a class of preferred embodiments of the invention, for generating a syndrome from packetized data.
• Fig. 9D is a schematic diagram of a circuit which is included in a class of preferred embodiments of the invention, for generating a syndrome from packetized data.
• Fig. 10 is a timing diagram showing the format in which RGB video data are transmitted over a TMDS link in some embodiments of the invention.
• Fig. 11 is a timing diagram showing the format in which YCbCr 4:4:4 video data are transmitted over a TMDS link in some embodiments of the invention.
• Fig. 12 is a timing diagram showing the format in which YCbCr 4:2:2 video data are transmitted over a TMDS link in some embodiments of the invention.
• Fig. 13 is a block diagram of a preferred embodiment of the inventive transmitter.
• Fig. 13 A is a block diagram of circuitry employed in some embodiments of the inventive transmitter for determining whether to insert a data island between active video periods.
• Fig. 13B is a timing diagram of some of the signals received and generated by the Fig. 13 A circuitry during operation.
• Fig. 13C is a block diagram of circuitry employed in some embodiments of the inventive transmitter for inserting a data island between active video periods.
• Fig. 14 is a block diagram of a preferred embodiment of the inventive receiver.
• Fig. 15 is a block diagram of the auxiliary data clock transmission and regeneration circuitry employed in typical embodiments of the inventive system.
• Fig. 16 is a block diagram of the auxiliary data clock regeneration circuitry employed in typical embodiments of the inventive receiver.
• FIG. 17, 18, 19, 20, and 21 is a timing diagram of a link integrity check operation that can be implemented by some embodiments of the inventive system (or by other systems in which data are transmitted over a serial link).
• Fig. 22 is a timing diagram of audio and pixel clocks that can be asserted to the inventive transmitter.
• the audio and pixel clocks shown are not quite synchronous.
• Fig. 23 is a timing diagram of signals generated by a preferred embodiment of the inventive system, showing how CTS counts by the transmitter and receiver counting can be resynchronized.
• the term "stream” of data denotes that all the data are of the same type and are transmitted with the same clock frequency
• channel refers to that portion of a serial link that is employed to transmit data (e.g., a particular conductor or conductor pair between the transmitter and receiver over which the data are transmitted, and specific circuitry within the transmitter and/or receiver used for transmitting and/or recovery of the data) and to the technique employed to transmit the data over the link.
• auxiliary data When transmitting audio (or other auxiliary) data via a serial link, is it often desired to transmit multiple streams of the auxiliary data, and it is often valuable for multiple channels of the link to be available for transmission of the auxiliary data. For example, there can be two audio streams (left and right streams of stereo audio), six streams (e.g., those of "5.1" surround sound), or up to eight streams (e.g., those of "7.1" surround sound). Alternatively, it may be desired to transmit even more streams of audio data with video, or to transmit streams of non-audio auxiliary data (for providing non-audio effects that are synchronized to the video) with audio and video.
• auxiliary data are typically on the same time base, but alternatively there can be a need for some of the audio (or other auxiliary) data to be based upon another time base, or to have a different sampling rate.
• transmission of six streams of pulse code modulated (PCM) audio data over the link can be based upon one clock.
• Another two streams of compressed audio data, possibly a down-mix (for playback on a reduced number of speakers), might be transmitted with the video and PCM data as well.
• PCM pulse code modulated
• TMDS encoded video data are transmitted over each of three channels of the TMDS link, and encoded auxiliary data (e.g., audio data) can be transmitted over one or more of these three channels during blanking intervals between the active video periods.
• auxiliary data e.g., audio data
• auxiliary data are desirably encoded using a subset (comprising "golden words" of the transition- minimized TMDS code words that are used to encode the video data for transmission.
• a class of embodiments of the invention can be implemented by a system of a type shown in Fig. 2.
• the TMDS link between transmitters 1 ' and 2' of Fig. 2 is identical to the TMDS link between transmitters 1 and 3 in Fig. 1, although some of the conductors thereof are shown in Fig. 1 but not in Fig. 2 (for simplicity).
• Fig. 2 system performs the functions of the Fig. 1 system, and is also configured to encode audio data (or other auxiliary data) as well as to encode video data in the same conventional manner as in the Fig. 1 system, to transmit the encoded auxiliary data over one or more of Channel 0, Channel 1, and Channel 2 of the TMDS link (and also transmit encoded video data over each such channel), and to decode the encoded auxiliary data (as well as the encoded video data).
• Transmitter 1 ' and receiver 2' of Fig. 2 conespond respectively, to transmitter
• Transmitter 1' of Fig. 2 is an element of a source device that also includes MPEG2 decoder 13 and microcontroller 15, coupled as shown. Decoder 13 asserts input video ("DigVideo") to a video data processing subsystem of transmitter 1 ', and input audio data ("SPDIF") and audio reference clock (“MCLK”) to an audio data processing subsystem of transmitter 1 '. Input audio SPDIF can be indicative of two or more streams of audio data (e.g., left and right stereo signals).
• EEPROM 14 stores key values and identification bits for use in HDCP encryption of content to be transmitted to receiver 2'.
• Receiver 2' of Fig. 2 is an element of a sink device that also includes EDID
• EDID ROM 23 is coupled to the DDC channel of the TMDS link, and stores status and configuration bits which can be read by microcontroller 15 over the DDC channel.
• Receiver 2' also includes an interface (e.g., interface 201 of Fig. 14) for communication via the DDC channel with microcontroller 15.
• Microcontroller 25 is coupled for I2C communication with receiver 2'.
• EEPROM 24 stores key values and identification bits for use in HDCP decryption of content received from transmitter 1 '.
• the sink device also includes display circuitry 26 which receives analog and/or digital video recovered by receiver 2', and an audio digital-to-analog converter 27 (DAC 27) which receives digital audio recovered by receiver 2'.
• display circuitry 26 which receives analog and/or digital video recovered by receiver 2'
• DAC 27 audio digital-to-analog converter 27
• the Fig. 2 system preferably transmits a video clock over a conductor pair (labeled "Channel C in Fig. 2) of the TMDS link, and also transmits a clock for the auxiliary data over at least one channel of the link.
• transmitter 1 ' transmits video data to receiver 2' over Channels 0, 1, and 2 (which are identical to the identically numbered channels of the Fig.
• Audio data e.g., left and right stereo signals
• a video clock e.g., determined by the rising edges of a binary waveform
• time stamp data over one or more of Channels 0, 1, and 2 with each burst of the audio data.
• the time stamp data determine a clock for the audio data, as described in above-cited U.S. Patent Application No. 09/954,663, filed on September 12, 2001.
• Receiver 2' is configured to process the time stamp data to regenerating the audio clock employed to transmit the audio data.
• the clock for a stream of audio data has a much lower frequency than the pixel clock for a stream of video.
• the audio clock needs to be more accurate than the pixel clock, to reduce jitter. This is true since distortion in analog audio (that has been generated from digital audio data having jitter) is more easily discernible (to one experiencing the analog audio) than is the distortion in a displayed video program generated from digital video having the same amount of jitter.
• each 10-bit code word is decoded back to the original 8-bit word if no errors are present.
• Each code word comprises a 9-bit base pattern (a transition- minimized member of a set of 2 9 nine-bit patterns, whose most significant bit indicates that the base pattern is transition-minimized, concatenated with a tenth bit indicating whether the eight least-significant bits of the base pattern have or have not been inverted in accordance with a DC balancing algorithm).
• each 8-bit source word is first encoded to one of the 9-bit base patterns, and a stream of the 9-bit base patterns are then encoded as a stream of the 10-bit code words (in a manner that achieves improved DC balancing of the transmitted stream of 10-bit code words).
• the decoded video data can include errors (especially when the relevant channel has significant ISI), depending on the specific channel media and the specific data patterns of the transmitted serial bit stream.
• transmitter 1' and receiver 2' were operated to encode and decode the auxiliary data in the same way that they encode and decode the video data, and to send both types of encoded data over the same channel of the serial link, the decoded auxiliary data would be subject to error at the same error rate.
• This error rate can be unacceptably high for auxiliary data (especially when the auxiliary data are audio data), even if it is acceptable for video data.
• transmitter 1 ' can be configured to encode the auxiliary data using "golden words” as explained below.
• transmitter 1 ' is configured also to encode the video data using "golden words" (or to be operable in a mode in which it encodes both the video data and auxiliary data using "golden words”).
• the inventive transmitter and receiver distinguish between at least three portions of each blanking interval (between active video periods): an initial portion (in which a "data island preamble” can be transmitted) followed by an auxiliary data portion (sometimes refened to as a "data island”) followed by a final portion (in which a "video preamble" can be transmitted).
• a blanking interval comprising at least one auxiliary guard band word followed by a burst of a different stream of encoded auxiliary data
• an initial portion between the falling edge of DE (at the start of the blanking interval) and the start of the first data island, an additional portion (including another data island preamble) before each subsequent data island in the blanking interval, and a final portion (including a video preamble) between the last data island and the next active video period.
• repetitions of code words indicative of specific patterns of control bits CTL3, CTL2, CTLl, and CTLO, and optionally also initial bit patterns are transmitted.
• repetitions of code words indicative of other specific patterns of control bits CTL3, CTL2, CTLl, and CTLO, and optionally also initial bit patterns are transmitted.
• packets of code words indicative of encoded auxiliary data and guard band words are transmitted.
• two or more data islands can be transmitted consecutively (without an active video period between them).
• two or more active video periods can be transmitted consecutively (without a data island between them), or data islands can alternate with active video periods.
• the following signals are transmitted during the initial data island preamble (as indicated in Figs.
• ISI inter-symbol interference
• data e.g., audio data transmitted during data islands between active video periods
• bit patterns that are less susceptible to ISI during fransmission over the link than are the patterns determined by conventionally encoded versions of the same data.
• data are encoded using a subset (a "robust" subset) of a full set of code words.
• the code words in the full set have equal length (e.g., each consists of N bits).
• the robust subset will sometimes be referred to herein as a "selected” or “inventive” set of code words, and the code words in the robust subset will sometimes referred to as the “inventive” code words (or “golden words").
• the robust subset is selected such that each transmitted stream of encoded data (coded using only members of the inventive code word set) has patterns that are less susceptible to ISI during transmission over the serial link than are patterns determined by a transmitted, conventionally encoded version of the same data (that has been coded using code words of the full set other than members of the inventive code word set, as well as members of the inventive code word set). Since there are more code words in the full set than there are inventive code words, fewer words of data can be fransmitted over the link per unit time if the transmitted data are encoded using only the inventive code words than if the transmitted data are encoded conventionally using the full set of code words.
• Encoding of data in accordance with the invention is particularly beneficial in applications in which the encoded data are fransmitted over very long conductors or under other conditions in which there would otherwise be a high risk of error due to ISI during transmission.
• transmitter 1 ' is configured to encode auxiliary data transmitted between active video periods in accordance with the invention as follows.
• a subset of the full set of 10-bit TMDS code words is selected as the "inventive" code word set such that each transmitted stream of 10-bit words of encoded auxiliary data (consisting only of the inventive code words, sometimes referred to as “golden words”) has a pattern that is less susceptible to inter-symbol interference than is the pattern determined by a transmitted stream of a TMDS-encoded version of the same data (including not only inventive code words but also members of the full set that are not inventive code words).
• a 2 M — bit subset (where M ⁇ 8) of the full set of 10-bit TMDS code words is selected to be the inventive code word set.
• the inventive code word set also includes one or more code words of the full set that are used as guard band words.
• Receiver 2' is implemented to decode each received one of the inventive 10-bit code words as an auxiliary data word of length M bits.
• Receiver 2' performs the same decoding operations on the encoded auxiliary words received during blanking intervals that it performs on the conventionally encoded video words received during the active video periods.
• transmitter 1 ' does not perform the conventional DC balancing steps that it performs during its conventional encoding of source video data (in which the eight least significant bits of the "N+l"th encoded video word are inverted, and the resulting nine bits are concatenated with a distinctive tenth, most significant bit when the cumulative DC drift of the N previous encoded video words reaches a predetermined threshold, and otherwise does not invert the eight least significant bits of the "N+F'th encoded video word and instead concatenates the word with another distinctive, tenth, most significant bit).
• transmitter 1 ' is configured simply to replace each 4-bit source word of auxiliary data with the corresponding one of the inventive code words, regardless of the cumulative DC drift of the resulting stream of inventive code words (and regardless of whether the MSB of the inventive code word is a one or zero).
• inventive code words are preferably chosen so that when the bits of a stream of the inventive code words are transmitted over a serial link as sequence of rising and falling voltage transitions, the bit pattern of such stream of the inventive code words is DC balanced (or is likely to be DC balanced) in the sense that the voltage drift that it determines over time is limited to an acceptable amount.
• transmitter 1 ' does perform the same DC balancing steps during its encoding of source auxiliary data (using the inventive code words) and during its conventional encoding of source video data. This is taken into consideration in the selection of the inventive code word set.
• each code word of the inventive code word set has a 9-bit base pattern that is a member of a selected subset of the 9-bit base pattern space of the full set of 10-bit TMDS code words, and during encoding of 4-bit words of source auxiliary data (to replace them with the inventive 10- bit code words), the eight least-significant bits of this 9-bit base pattern are either inverted and the resulting pattern concatenated with a tenth (and most significant) bit having a first value, or the base pattern is not inverted and is instead concatenated with a tenth (and most significant) bit having a second value, depending on whether the cumulative DC drift of the stream of previously encoded auxiliary words has reached a predetermined threshold.
• receiver 2' is implemented to perform the same decoding operations on the encoded auxiliary data words received during blanking intervals that it performs on the conventionally encoded video data words received during the active video periods, and then to map each 8-bit word (generated as a result of conventional decoding of one of the 10-bit encoded auxiliary data words) to one of the 2 M auxiliary data words each having M-bit length.
• M i.e., selecting a smaller inventive set of code words from the full set
• BER bit-error rate
• increasing the parameter M results in an increased data rate but at the cost of increased BER.
• This code word set is a subset of the full set of conventional TMDS 10-bit code words, and is useful for encoding 4-bit words of auxiliary data for transmission over a TMDS (or TMDS-like) link over which 8-bit video words (conventionally encoded using the full set of TMDS 10-bit code words) are also transmitted, in cases when it is adequate to transmit the auxiliary data at half the data rate as the video data.
• 8-bit input words of binary auxiliary data are buffered, the four least-significant bits of each are encoded (e.g., in transmitter 1' of Fig. 2) as one of the sixteen 8-bit words "ADO-AD 15" in the left column (labeled
• each code word is the LSB and (in the case of each 10-bit code word) is the first bit to be fransmitted over the serial link. Also, the right bit of each code word is the MSB and (in the case of each 10-bit code word) is the last bit to be transmitted over the serial link.
• an input auxiliary data word 10000000 (whose LSB is 1) would be split into two halves (1000 and 0000) and the two halves then encoded as AD1 and ADO, respectively. Then, the 8-bit word ADO is encoded as the 10-bit inventive word "0011100101" and the 8-bit word AD1 is encoded as the 10-bit inventive word
• each 10-bit inventive word is decoded into one of the 8-bit words AD0-AD15, and the original 8-bit input auxiliary data words can be reconstructed from the recovered words AD0-AD15 since there is a one-to-one mapping between each word AD0-AD15 and one half (four bits) of each 8-bit input auxiliary data word.
• the input auxiliary data asserted to the transmitter can be 4-bit words, in which case the transmitter would not need to split (or otherwise pack) received input auxiliary data words into 4-bit format before encoding them as a sequence of the words AD0-AD15.
• the input auxiliary data can be pre-encoded as a sequence of 8-bit words AD0-AD15, and the pre-encoded auxiliary data then provided to the transmitter in the form of a sequence of the 8-bit words AD0-AD15.
• the encoded auxiliary data are transmitted in the same channels (CHO, CHI, and CH2) of a TMDS link in which video data are transmitted, but the auxiliary data are transmitted during the blanking intervals between the active video periods of video data transmission.
• Figs. 5 and 6 are timing diagrams of signals transmitted during such an embodiment of the invention.
• the upper nine signals of Fig. 5 represent signals input to the transmitter during a control data period and data island (of a blanking interval), and the lower three signals of Fig. 5 represent the auxiliary data (encoded using the 10-bit words of Fig. 4) and encoded control and sync signals (to be discussed below) that are transmitted over channels CHO, CHI, and CH2 during the control data period and data island in response to the upper nine signals.
• the upper nine signals of Fig. 6 represent signals input to the transmitter in a control data period at the end of a blanking interval (the blanking interval of Fig. 5) and during the active video period that follows such control data period, and the lower three signals of Fig. 6 represent the auxiliary data (encoded using the 10-bit words of Fig. 4), video data (conventionally encoded), and encoded control and sync signals (to be discussed below) that are transmitted over channels CHO, CHI, and CH2 in response to the upper nine signals.
• 24-bit words of input data are provided to the encoding circuitry of the transmitter for encoding.
• Fig. 5 pertains to those of such words (each identified as D[23:0] in Fig.
• Fig. 6 pertains to those of such words (each identified as D[23:0] in Fig. 6) that are words of video data.
• Eight bits of each input word (D[23:16]) are encoded, serialized, and transmitted on channel CH2 (as 10-bit encoded words CH2[0:9]), another eight bits of each such word (D[15:8]) are encoded, serialized, and transmitted on channel CHI (as 10-bit encoded words
• CHI [0:9]) and another eight bits of each such word (D[7:0]) are encoded, serialized, and transmitted on channel CHO (as 10-bit encoded words CH0[0:9]).
• the video data are in RGB format (and the red, green, and blue pixels are fransmitted on channels CH2, CHI, and CHO, respectively).
• channels CH2, CHI, and CHO are sometimes referred to herein (such as in Fig. 3) as the red (or "R”) channel, the green (or "G” channel), and the blue (or "B”) channel, respectively.
• the video data that are encoded (and then transmitted) are in luminance-chrominance format; the waveform "DCK" indicates the data clock.
• phase shifting circuitry is used to generate multiple, phase-shifted versions of the clock DCK which are then used (with the clock DCK itself) to clock the encoding, transmission, and decoding operations.
• a clock having ten times the frequency of DCK could be used to clock the encoding, fransmission, and decoding operations, and one code bit would be transmitted during each cycle of this faster clock; the waveform "DE" (of Fig.
• auxiliary data (identified as D[23:16], D[15:8], and D[7:0] in Fig. 5) are encoded, and serialized 10-bit words of the encoded video are transmitted over channels CHO, CHI, and CH2.
• auxiliary data (identified as D[23:16], D[15:8], and D[7:0] in Fig. 5) are encoded, and serialized 10- bit words of the encoded auxiliary data are transmitted over channels CHO, CHI, and CH2.
• a 10-bit transition-minimized code word indicative of HSYNC and VSYNC is sent each pixel clock cycle (e.g., one code word indicative of an HSYNC bit, a VSYNC bit, and two other bits is sent per pixel clock cycle) over channel CHO.
• each packet sent during a data island includes a 32-bit packet header, and each code word indicative of HSYNC and VSYNC that is transmitted during a data island is also indicative of one bit of the packet header.
• 32 pixel clock cycles are required to transmit an entire 32-bit packet header.
• Figs. 5 and 6 have been described with reference to two data enable signals, "DE” and “AUX DE,” it is contemplated that the inventive transmitter can be implemented with a portion (e.g., "core” processor 114 of transmitter 100 of Fig. 13) configured to perform all the described encoding, serialization, and transmission in response to a single data enable signal (e.g., a combined enable signal indicative of the result of performing a logical "OR” operation on the signals DE and AUX DE), and a single set of data inputs (D[23:0]) indicative of either video or auxiliary data.
• a single data enable signal e.g., a combined enable signal indicative of the result of performing a logical "OR” operation on the signals DE and AUX DE
• D[23:0] a single set of data inputs
• Additional circuitry of the transmitter outside the core is configured to receive separate sets of auxiliary data (e.g., 24-bit auxiliary data words) and video data (e.g., 24-bit video data words), and both a video data enable signal DE, and an auxiliary data enable signal "AUX DE.”
• the data enable signals can also occur with other sequences of values, including non-repeating sequences. For example, in some circumstances, auxiliary data are transmitted in some but not all video blanking intervals.
• the additional circuitry of the fransmitter can include logic circuitry that "ORs" together the signals DE and AUX DE to produce a combined data enable signal.
• the additional circuitry can also pack the auxiliary data into 4-bit format, encode each 4-bit portion of the auxiliary data as one of the words AD0-AD15 shown in Fig. 4, add guard band words with appropriate timing into the sfream of ADO-AD 15 auxiliary data words, and add video guard band words into the stream of video data (or alternatively replace, with appropriate timing, words of the video data with video guard band words).
• the additional circuitry can assert a sequence of bursts of the video data (with video guard band words) and auxiliary data (with guard band words) to the core (e.g., alternating bursts of the video data with video guard band words, and auxiliary data with guard band words), and also assert the combined data enable signal to the core.
• the core performs all the encoding, serialization, and transmission operations described with reference to Figs. 5 and 6 in response to the combined data enable signal (rather than separate DE and AUX DE signals) and the bursts of video and auxiliary data.
• the "additional circuitry" of the fransmitter is coupled and configured to receive and encode two or more sets of auxiliary data (each set comprising a different type of auxiliary data).
• the additional circuitry is also coupled and configured to receive a set of video data, an auxiliary data enable signal for each set of auxiliary data (e.g., first and second auxiliary data enable signals "AUXl DE” and “AUX2 DE”) and a video data enable signal (“DE”), and to assert a sequence of bursts of the video data and bursts of the encoded auxiliary data to the transmitter's core.
• the additional circuitry can include logic circuitry that "ORs" together the signals DE, AUXl DE, and AUX2 DE to produce a combined data enable signal, and can assert the combined data enable signal (rather than the individual video data enable and auxiliary data enable signals) to the core.
• the combined data enable signal (rather than the individual video data enable and auxiliary data enable signals) to the core.
• each of at least one channel of a serial link e.g., in each of channels CH2 and
• an appropriate one of the inventive code words is (or two or more appropriate ones of the inventive guard band words are) preferably transmitted (as a guard band word or set of guard band words) at the start of each burst of encoded auxiliary data (i.e., immediately after each "auxiliary preamble" of each blanking interval), at the end of each burst of encoded auxiliary data, and at the start of each burst of encoded video data (i.e., immediately after the "video preamble" of each blanking interval).
• the source data to be transmitted during data islands are encoded using a "robust" subset of a full set of code words.
• Each "robust" subset consists of code word sets (sometimes referred to herein as
• golden sets consisting of one or more code words (sometimes referred to herein as “golden words”).
• Each "golden word” of a golden set is indicative of a single source data value (e.g., a source data word).
• a golden set consists of two or more golden words, each of these golden words is indicative of the same source data value. Clusters of code words in the full set are determined.
• Each cluster includes a "golden set” and optionally also one or more additional code words of the full set, where each of the additional code words is "similar" to a golden word of the cluster's golden set in the sense that each additional code word is likely to be generated as a result of probable bit errors in fransmission, or fransmission and decoding, of such golden word.
• Each received code word in one of the clusters is mapped to the source data value determined by the cluster's golden set.
• Each mapping of a cluster of received code words to a single source data value can provide error correction by mapping an error-containing word in the cluster back to the source data value most likely to correspond to the error-containing word.
• the full set of code words can be used to encode one type of data (e.g., video data) for transmission over a channel of a serial link, and the robust subset can be used to encode another type of data (e.g., audio data or other "auxiliary" data related to or useful with video data) for transmission over the same channel.
• one type of data e.g., video data
• the robust subset can be used to encode another type of data (e.g., audio data or other "auxiliary" data related to or useful with video data) for transmission over the same channel.
• each code word in each golden set (and each code word in the full set) is an N-bit word that is an encoded version of an M-bit word, where M is an integer less than N.
• each received N-bit code word can differ from one of the golden words (if a transmission error has occurred) or it can be identical to one of the transmitted golden words.
• Each received N-bit code word in one of the clusters is decoded to generate a decoded M-bit word, and each such decoded M-bit word is mapped to the source data value determined by the cluster's golden set.
• the full set of code words is the set of 10-bit, transition-minimized TMDS-encoded words that are indicative of 256 eight-bit source words.
• the robust subset ofthe full set consists of eight-bit "golden words" indicative of a subset ofthe full set of 256 eight-bit source words.
• the robust subset consists of sixteen golden sets, each golden set consists ofthe 10-bit TMDS code words indicative of one eight-bit source word, and each cluster ofthe 10-bit TMDS code words includes one ofthe golden sets and at least one 10-bit TMDS code words similar to the code words in such golden set.
• each received 10-bit code word in one ofthe clusters is decoded in accordance with the TMDS decoding algorithm (or a modified version thereof) to recover an eight-bit word, and each recovered eight-bit word is mapped to the eight-bit source word determined by the cluster.
• the full set of code words (which can be used to encode primary data, for example when auxiliary data are encoded in accordance with the invention using only “golden words” ofthe full set) are those code words (the "2 N space") that can be used to encode 2 N different source data words, each source data word being an ordered set of N bits.
• the golden words (the "2 n space”) are a subset of the code words ofthe full set that can be used to encode 2 n different source data words, each such source data word being an ordered set of "n” bits (where "n” is an integer less than N).
• raw source data (which can consist of words of any length) can be buffered and packed into an n-bit format (i.e., into n-bit members of a set of 2" source data words).
• Each different n-bit source data word can then be encoded as one ofthe golden words (in the "2 n space") and fransmitted over a serial link (typically over a single channel ofthe link).
• the transmission can result in enor or it can be error free.
• Clusters of the full set of code words are predetermined such that each cluster includes a "golden set" (of one or more ofthe golden words) and optionally also one or more additional code words ofthe full set, where each ofthe additional code words is similar to a golden word ofthe cluster's golden set.
• cluster "S a” includes the golden set consisting of each ofthe golden words that encodes source word "a”
• cluster “S b” includes the golden set consisting of each ofthe golden words that encodes source word "b”).
• Each received code word in one ofthe clusters is mapped to the source data value determined by the cluster's golden set.
• each code word ofthe 2 N space is a 10-bit TMDS-encoded word
• the 2 n space is a subset ofthe full set of 10-bit TMDS-encoded words.
• Each transmitted 10-bit TMDS code word is decoded in accordance with the TMDS decoding algorithm (or a modified version thereof) to generate an 8-bit code word.
• the transmission of golden words (and any decoding thereof) can result in enor or can be error free.
• the cluster containing such golden word preferably includes all the N-bit members ofthe 2" space likely to result from occunence of such a transmission enor during transmission ofthe golden word.
• the clusters are disjoint, an N-bit word included in one cluster is omitted from all the other clusters.
• Each embodiment ofthe invention employs at least one (and typically more than one) cluster that contains at least one code in addition to each golden word of a golden set. Some embodiments ofthe invention employ at least one cluster that contains at least one code in addition to each golden word of a golden set, and also at least one other cluster that contains no code other than each golden word of a golden set. Since all the clusters (e.g., S a , S b , etc. of Fig. 7) are mutually disjoint, then regardless of whether or not an error occurs during transmission (or transmission and decoding), if a cluster contains a received N-bit code, the received code is mapped back to the conect source word. In a class of implementations of Fig.
• each cluster includes a golden set ofthe TMDS code words including golden words indicative of one ofthe 16 different source data words.
• 4-bit words of raw source data are preliminarily encoded as the 16 different 8-bit source data words, and each resulting 8-bit source data word is then encoded as one ofthe 10-bit members ofthe 2 n space for transmission.
• the robust subset (of the full set of TMDS code words) consists of those 10-bit TMDS code words that (when decoded in accordance with the TMDS decoding algorithm or a modified version thereof) determine the 16 predetermined 8-bit source data words (ofthe 256 eight-bit source data words determined by the full set of TMDS code words).
• Each cluster preferably includes not only the golden words of a golden set, but also at least one 10-bit TMDS code word "similar" to one ofthe golden words in the sense that it is more likely to result from bit enors in transmission ofthe golden word than are other TMDS code words (e.g., the code word "similar" to the golden word may differ from the golden word only in the value of one of its bits).
• the process of selecting the golden sets from the full set of code words is very important.
• the best choice for the specific golden sets selected from a full set of binary code words depends on the particular coding implemented by the full set (i.e., the details of which bits of each code word in the full set are zeroes and which are ones).
• the code words ofthe golden sets are selected to be those whose serial patterns (during transmission) have fewer contiguous zeros and ones (e.g., on the average), and thus are less susceptible to ISI during transmission, than do those code words in the full set that are not selected (e.g., the average number of contiguous zeros and ones, per code word, ofthe golden words is less than the average number of contiguous zeros and ones, per code word, ofthe code words in the full set that are not selected as golden words).
• the golden words are selected to be those satisfying the criterion that the Hamming distance between any golden word in one cluster and any golden word in any other cluster exceeds a threshold, or the criterion that the Hamming distance between golden words in different clusters is maximized (in some sense) to the extent practical (e.g., the criterion that an average Hamming distance between golden words in different clusters is maximized) subject to the constraint that the clusters are mutually disjoint.
• This helps to increase the number of "errored codes" (codes other than golden codes of one golden set) that can be included in each cluster, while keeping the constraint that the clusters are mutually disjoint.
• the receiver e.g., receiver 200 of Fig. 14
• the receiver is preferably configured to include a predetermined table that outputs one value (e.g., a source data word) determined by each cluster in response to each input indicative of an element ofthe cluster (e.g., in response to each of four received code words, where the cluster consists of one golden word and three code words similar to the golden word).
• one value e.g., a source data word
• the table implements a mapping of each received code word in each cluster to the source word determined by the cluster's golden set, or of each received code word in each cluster to a golden word ofthe cluster's golden set (which is then mapped by conventional hardware or software in (or external to) the receiver to a source word determined by the cluster's golden set), or of a decoded version of each received code word in each cluster to a source word determined by the cluster's golden set, or of a decoded version of each received code word in each cluster to a single decoded code word (which is then mapped by conventional hardware or software in, or external to, the receiver to a source word determined by the cluster's golden set).
• receiver 200 of Fig. 14 includes core processor 214 (which includes code word recovery circuitry), and mapping and decoding circuitry 20 (of subsystem 208) which implements such a table.
• Circuitry 214 is configured to recover 10-bit TMDS code words from the data received on each of TMDS channels CHO, CHI, and CH2.
• Circuitry 20 is configured to map each recovered word in one ofthe clusters to a golden word ofthe cluster (using the table), to decode each golden word to generate an 8-bit decoded value (one ofthe seventeen 8-bit words in the left column of Fig. 4).
• the 8-bit decoded values are indicative of source words.
• mapping and decoding circuitry would generate a 4-bit raw source data word in response to each decoded 8-bit word (indicative of one ofthe words ADO-AD 15 in Fig. 4), circuitry 20 of Fig. 14 asserts 8-bit decoded words to decryption circuitry 21.
• the clusters (and thus the inputs to the above-mentioned table in the receiver) can be a partition ofthe full set of code words (e.g., the 2 N space of Fig. 7), so that the union ofthe clusters covers the whole space of code words (e.g., the entire 2 space of Fig. 7) and the clusters are mutually disjoint.
• the mapping of words in each cluster to a golden code can be performed in a manner that is in some sense "on the fly," for example using clocked logic to substitute a golden code for each received word that belongs to a predetermined cluster that contains the golden code.
• the receiver includes logic (e.g., in circuitry 20 of Fig. 14) that applies rules to determine which one ofthe bits of a received code word is likely to be enoneous, and then "map" the received code word to a golden code by flipping each bit that it determines to be likely to be enored.
• the mapping of words in each cluster to a golden code is performed using a look-up table which outputs a golden code in response to each word of a cluster that contains the golden code.
• an algorithm can be implemented (e.g., by software or logic circuitry) to determine which bits (to be refened to as "vulnerable" bits) in a golden word are most vulnerable to enor, a process is then implemented wherein vulnerable bits of golden words are replaced by enored versions thereof to determine "likely enor- containing code words," and a process is then implemented to compare all (or a subset of all) ofthe "likely enor-containing code words" to the actual received code words and thereby to map each received code word that is a "likely enor-containing code word" to the conesponding golden word.
• the algorithm for identifying vulnerable bits preferably identifies bits that are most susceptible to inter-symbol interference. Single '0' bits within a stream a ' 1 ' bits would qualify.
• the vulnerable bits are flipped in the golden words, thereby producing "likely enor-containing code words".
• An alternative approach is to determine manually which bit patterns include vulnerable bits and store a series of masks with each bit pattern that the designer wishes to address, along with the golden words. Logic would then only need to flip the relevant bits in the golden words to generate the "likely enor-containing code words.”
• the process of choosing which code words to include in each cluster can be automated using logic circuitry (or software).
• two or more clusters can conditionally include the same code word (to be denoted as a "multi-value” code word for convenience), where a multi-value code word is "conditionally included” in a cluster in the sense that it is mapped to the cluster's golden word only provided that a predetermined condition is met.
• the conditions associated with the multi-value code word in the different clusters should differ from each other, and should be chosen such that only one ofthe conditions can be met in any particular case, so that a receiver will be able to map the multi-value code word to one specific cluster's golden word in each particular case. In such embodiments, the receiver would apply predetermined rules to pre-existing information to determine which of the conditions is met in each particular case.
• a received enor-containing word "A” was most likely transmitted as word “B” under condition “X,” and was most likely transmitted as word “C” under a different condition “Y.”
• a first cluster includes word “B” and conditionally includes word “A” (subject to condition “X”).
• a second cluster includes word “C” and conditionally includes word “A” (subject to condition “Y”).
• the receiver would apply an algorithm to pre-existing information to determine which of conditions X and Y is met during transmission ofthe code words.
• the receiver could determine whether the bit pattern of a sfream ofthe "Z" most recently inventive code words (where Z is a number) is DC balanced, and determine that condition X (but not condition Y) is met if the bit pattern is DC-unbalanced in the positive direction, and that otherwise condition Y (but not condition X) is met.
• the pre-existing information is the DC-balance level ofthe specified bit pattern.
• the pre-existing information could be a history of types of errors, or the pixel clock frequency, or a type of jitter, or other information.
• the received code that is less likely to occur is preferably excluded from the relevant cluster.
• a first cluster includes a first golden word
• a second cluster includes a second golden word
• a received code word that is not a golden word
• P2 probability of transmission
• the first cluster should include the received code word, but the received code word should not be included in the second cluster.
• two disjoint clusters containing code words A, B, and C are determined as follows: a first cluster including golden word B and word A (and typically other words similar to B); and a second cluster including golden word C and one or more other words similar to C that are not included in the first cluster.
• the receiver includes enor conection circuitry (ECC), and the receiver is configured to assert to the ECC the code words that result from mapping of received words (belonging to the clusters) to golden words.
• ECC enor conection circuitry
• the ECC determines that such a word B has a bad bit, and the word B would be transformed to word C by flipping this bit, then if the bit is the only bad bit for the entire ECC block, the ECC could conect the enor by flipping the bad bit and thereby transform the word B to the word C.
• some implementations ofthe inventive receiver are configured to perform a two-stage mapping of received versions ofthe golden words to source data values: a first stage in which each received code word in a cluster is mapped to a golden word ofthe cluster's golden set; and a second stage in which the golden words determined during the first stage is mapped to the source word determined by the cluster's golden set.
• an additional block of enor conection code is transmitted with each set of golden words, and the receiver is configured to perform the first stage of mapping before performing enor conection using the enor conection code (to conect enor-containing received code words that are not members of any ofthe clusters, to replace them with golden words to the extent possible).
• the inventive golden words and clusters are preferably chosen so as to implement mappings that exploit the degree of freedom provided by the performance of enor conection.
• the golden words can be selected to satisfy the criteria that the Hamming distance between any two golden words in different clusters is minimized to the extent practical (or otherwise not maximized), and that the clusters are mutually disjoint.
• clusters including such golden words the number of enoneous bits detected by the enor conection circuitry in the receiver can be minimized and hence, the overhead ofthe enor conection code can be minimized.
• receiver 200 of Fig. 14 includes decryption circuit 21 which receives and decrypts 8-bit decoded words from "mapping and decoding" circuitry 20, and enor conection circuitry 22 coupled to receive the decrypted, decoded words output from circuit 21.
• Circuitry 20 is configured to perform a first stage of mapping to generate (and assert to circuit 21) a golden word in response to each recovered code word that is a member of one ofthe inventive clusters, but enors in some recovered words may prevent circuitry 20 from mapping the enor-containing recovered words to golden words.
• Decrypted versions of the golden words determined by circuitry 20 the enor-containing recovered words, and enor conection code recovered with the code words are asserted to enor conection circuitry 22.
• Circuitry 22 can pass through the decrypted golden words that it receives, and perform enor conection using the enor conection code to conect enor-containing code words that are not members of any of the inventive clusters, thereby replacing such enor-containing code words with golden words to the extent possible.
• enor conection is not performed and circuitry 22 is omitted.
• a set of seventeen golden words and a set of code word clusters (each including one ofthe golden words) employed in a class of prefened embodiments ofthe invention.
• the third column shows the seventeen golden words, which have been described above.
• Sixteen ofthe golden words are used to encode sixteen different 8-bit source data words (each 8-bit source data word being indicative of four bits of raw source data), and the other golden word (the word "1100110010" in the first row) is used only as a guard band word.
• Each ofthe golden words is a 10-bit TMDS encoded word.
• the fourth column shows some possible enor cases for each 10- bit golden word
• the fifth column shows the 8-bit word resulting from decoding ofthe conesponding 10-bit word in the fourth column in accordance with the conventional TMDS decoding algorithm
• the sixth column simply shows the hexadecimal representations ofthe conesponding elements in the fifth column.
• the seventh column includes an indication as to whether each conesponding word in the fourth column is or is not a member ofthe cluster containing the conesponding golden word in the third column. Specifically, the term "IGNORE" in the seventh column indicates that the conesponding word in the fourth column is not a member ofthe cluster that contains the conesponding golden word in the third column.
• clusters there are seventeen clusters (separated by the horizontal bars in Fig. 4): a first cluster including the golden word “1100110010” and the code words “1110110010” and “1100010010” (all mapped to the "pre-data” auxiliary guard band word “01010101”); a second cluster (for encoding the source word ADO) including the golden word “0011100101” and the code words "1011100101,” "0001100101,”
• a third cluster for encoding the source word AD1 including the golden word "0110001101” and the code words "0111001101” and "0110000101” (all mapped to the source word AD1); and the fourteen other indicated clusters (each including a different one ofthe fourteen other golden words, and consisting of words mapped to a different one ofthe source words AD2-AD 15).
• the receiver When the receiver recovers a code word (in the fourth column of Fig. 4) in one ofthe indicated clusters (there will be no "IGNORE" symbol in the seventh column conesponding to such recovered code word), the receiver will map the recovered code word to the source word (in the first column of Fig. 4) determined by the cluster's golden word (which is equivalent to mapping the recovered code word to the cluster's golden word in the third column).
• code words in the fourth column marked with the term "IGNORE” in the seventh column are not members ofthe cluster that contains the conesponding golden word.
• the code word "1100111010" in the third row of Fig. 4 is not a member of the first cluster (containing golden word "1100110010") because this code word is a member ofthe cluster that contains golden word "1000111010” and the clusters should be disjoint.
• the receiver would recover code word "1100111010” as a result of a single bit enor (having relatively high probability) in transmission of golden word "1100110010" (in the first row of Fig.
• the receiver would map the received word to 8-bit source word "AD 10" (which is equivalent to mapping the received word to the golden word "1000111010”) rather than mapping to the source word ("01010101") determined by golden word "1100110010.”
• AD 10 which is equivalent to mapping the received word to the golden word "1000111010”
• the source word 01010101
• the receiver would recover the code word "1100110010” if the fransmission is enor- free (which has very high probability).
• the receiver would decode this recovered word in accordance with the conventional TMDS-decoding algorithm to determine decoded 8-bit word "01010101” and map the decoded 8-bit word to the source word "01010101” (which is the pre-data auxiliary guard band word).
• the receiver would recover the word "0011001111” (in the 52 nd row of Fig.
• each guard band word can be a golden word (as in the Fig. 4 example) in which case there can be a cluster for each guard band word, each such cluster containing two or more code words (including a guard band word).
• the guard words are not golden words but they are reliably distinguishable from the golden words.
• each guard band word should have a bit pattern which allows the receiver to more reliably identify the relevant transition (indicated by the guard band word or words) between encoded control (or sync) word fransmission and encoded data fransmission.
• the golden set should includes appropriate guard band words (i.e., the guard band words are golden words), or each golden word ofthe golden set should be reliably distinguishable from each guard band word to be employed.
• the set of 17 golden words shown in Fig. 4 includes a special auxiliary guard band word (having bit pattern "1100110010,” and shown in the third column of the first row of Fig.
• auxiliary guard band word is preferably transmitted at the start of each burst of encoded auxiliary data (i.e., just after each auxiliary preamble) in each of channels CH2 and CHI. Since the last bit of each specific encoded control word transmitted in channels CH2 and CHI (during the auxiliary preamble) is "0" as explained above, the first transmitted bit ofthe code word chosen as the pre-data auxiliary guard band word is "1" to increase the reliability with which the receiver can identify the start of a fransmitted burst of auxiliary data.
• the set of 17 golden words shown in Fig. 4 also includes a word (the golden word "0011001101" that conesponds to input word ADl 1) that is used to identify the end of an auxiliary data burst, and is also used as a video guard band word.
• a word the golden word "0011001101” that conesponds to input word ADl 1
• two repetitions of this "post-data" auxiliary guard band word are preferably transmitted at the end of each burst of encoded auxiliary data (i.e., just before each video preamble) in each of channels CH2 and CHI.
• the pre-data auxiliary guard band word need not be repeated (fransmitted twice) at the start of each auxiliary data burst, and the post-data auxiliary guard band word need not be repeated at the end of each auxiliary data burst.
• each is repeated in order to allow the receiver more easily to recognize and conect for data shift enors between channels that can occur during transmission and recovery ofthe data (e.g., enor in the trace length ofthe received data on channel CHI relative to that ofthe data received on channel CH2).
• an auxiliary guard band word is repeated more than twice (or is transmitted only once) at the start of each auxiliary data burst and/or more than twice (or is transmitted only once) at the end of each auxiliary data burst.
• the golden word "0011001101" (that conesponds to input word ADl 1) is used as a video guard band word to identify the start of a video data burst, in addition to being used as a code word for encoding the four-bit quantity of auxiliary data indicated by input word ADl 1, and as a post-data auxiliary guard band word.
• two repetitions of this video guard band word are preferably transmitted at the start of each burst of encoded video data (i.e., just after each video preamble).
• the first two transmitted bits ofthe video guard band word are chosen to be "00" to increase the reliability with which the receiver can identify the start of a transmitted burst of video data.
• the video guard band word need not be repeated (transmitted twice) at the start of each video data burst. In the embodiment shown in Fig. 6, it is repeated in order to ensure transmission (on each of channels CHO, CHI, and CH2) of code words indicative of an even number of pixels during the burst. In other embodiments, a video guard band word is repeated more than twice (or is transmitted only once) at the start of each video data burst.
• two (or more than two) streams of video data are transmitted (over one, two, or more than two channels).
• two or more streams of video data can be transmitted in time-multiplexed fashion over each of one or more of Channels 0, 1, and 2 of Fig. 2.
• different video guard band words can be transmitted at the start (and/or the end) of each burst, with each different sfream being identified by a different video guard band word.
• two (or more than two) streams of auxiliary data can be fransmitted over one, two, or more than two channels). If bursts of different streams of auxiliary data are sequentially transmitted over one channel, different auxiliary guard band words can be transmitted at the start (and/or the end) of each burst, with each different stream being identified by a different guard band word.
• DE shifts in individual channels can be conected independently (in accordance with the invention) by using guard band words in each channel. Since there can be misalignment between the DE transitions indicated by the bits transmitted over multiple channels of a TMDS link (or TMDS-like link or other serial link) by one pixel clock cycle (or more than one pixel clock cycle) in either direction (due to ISI or other noise sources on the link), a set of identical guard band words (each a member ofthe set of inventive code words) is preferably transmitted in accordance with the invention at the start and/or end of each burst of data encoded using the inventive code words that is transmitted over each channel (e.g., at the end of each auxiliary preamble of each channel, and/or at the start ofthe video preamble of each channel, and/or at the end of the video preamble of each channel). This can improve the channel-to-channel alignment and data integrity.
• the need to have available the appropriate number of guard band words is a
• the purpose of repeating the transmission of a guard band word is to prevent two types of misidentification of transitions: identifying the transition too early and identifying the transition too late.
• the invention prevents such pixel shift enors up to N pixels in either direction.
• the invention ensures that when there is an N pixel shift to the left, the last data value is not lost (only the post-data guard band word is lost).
• a sequence of only N post-data guard band words is needed for use with a sequence of N pre-data guard band words.
• the auxiliary guard band words transmitted at the start and end of each auxiliary data burst on channels CH2 and CHI are not transmitted at the start and end of each auxiliary data burst on channel CHO. Rather, special encoding is used to determine the first two and last two 10-bit inventive code words transmitted in each auxiliary data burst on channel CHO. Specifically, two bits of input auxiliary data are encoded and fransmitted during each of the first two and last two pixel clock cycles of each auxiliary data burst (whereas four bits of input auxiliary data are encoded and transmitted during all the other pixel clock cycles of each auxiliary data burst as described above).
• the two bits of input auxiliary data to be transmitted during the first clock cycle are encoded as one ofthe words ADO, ADl, AD2, and AD3 in Fig. 4 and the two bits of input auxiliary data to be transmitted during the second clock cycle are encoded as another one ofthe words ADO, ADl, AD2, and AD3.
• the first two 10-bit words transmitted in the burst are versions of the inventive code word indicative of these two words ADO, ADl, AD2, and AD3 (and are thus indicative ofthe first four bits of input auxiliary data).
• the two bits of input auxiliary data to be transmitted during the next to last clock cycle are encoded as one ofthe words ADO, ADl, AD2, and AD3 in Fig.
• control or synchronization bits can produce different enors on video (or auxiliary) data bits that are fransmitted just after the control characters, when ISI is present on the serial data transmission channel. This is preferably recognized and used as a factor in selecting the inventive code word set for use in transmitting the video (or auxiliary) data.
• the control codes sent just before the data are controlled to reduce the ISI effect.
• bursts of encoded auxiliary data and bursts of encoded video data are transmitted over a serial link (which need not be a TMDS link), and the auxiliary data are encoded in accordance with the invention using a set of inventive code words.
• the set of inventive code words includes a "video" guard band word that is transmitted at the start of each encoded video data burst, and an "auxiliary" guard band word that is transmitted at the start of each encoded auxiliary data burst.
• the video guard band word is also used for a second purpose: to encode auxiliary data.
• a video guard band word is transmitted at the start of each active video period.
• Each blanking interval comprises an "auxiliary" preamble period (between the falling edge ofthe video data enable signal and the start of a burst of auxiliary data) in which control (or sync) signals of a specific type are fransmitted, at least one auxiliary data period after the auxiliary preamble period (each auxiliary data period comprising an auxiliary guard band word followed by a burst of encoded auxiliary data), and a "video" preamble period between the last auxiliary data period and the next active video period.
• guard band words in accordance with the invention is to guarantee that the receiver can recognize the transition between the first guard band word transmitted at the start of an encoded data burst and the last bit transmitted before such guard band word, and between the last guard band word transmitted at the end of an encoded data burst and the first bit transmitted after such guard band word.
• a conventional encoding algorithm is used to encode primary data (which can but need not be video data) for transmission in bursts over a serial link, and auxiliary data (e.g., audio data or data of another type that can be transmitted with a lower data rate than the primary data) are encoded in accordance with the invention for transmission in bursts (between bursts ofthe encoded primary data) over a serial link.
• auxiliary data e.g., audio data or data of another type that can be transmitted with a lower data rate than the primary data
• the full set of code words used for encoding the primary data has at least one code word for each of 2 N different words ofthe primary data (sometimes refened to as source data words).
• the inventive subset of such full set has at least one code word for each of not more than 2 M different words (where M ⁇ N) ofthe auxiliary data (also refened to sometimes as source data words).
• the auxiliary data are buffered and packed into M-bit format (i.e., into words each consisting of M bits).
• Each possible value ofthe M-bit source data has a preselected code in the 2 M word space provided by the inventive code words.
• the M-bit words of auxiliary data are mapped to inventive code words in the 2 M word space which are then transmitted over the link.
• the inventive golden words to employ to transmit encoded data e.g. auxiliary data distinct from video data
• the DC balancing bits and transition control bits e.g., bits Q[9] and Q[8]
• Any bit enor occurring during processing ofthe DC balancing and transition control bits can affect other bits ofthe multi-bit encoded words.
• a one-bit enor in one ofthe critical bits is translated into a burst enor. This effect is preferably considered in selecting the inventive code words from a full set of TMDS-encoded words.
• a transmitter e.g., transmitter 1' of Fig. 2 sends 8-bit video data words (each encoded using the TMDS encoding algorithm as a 10-bit, transition-minimized code word) over the video channels (CHO, CHI, and CH2) of a TMDS link during active video periods in which a control signal (DE) is high.
• the transmitter sends packets containing 4-bit words (which typically include 4-bit audio data words), each encoded according to the TMDS encoding algorithm as a 10-bit, transition-minimized code word, and preferably as a 10-bit golden word, over the video channels.
• 4-bit words which typically include 4-bit audio data words
• control words each encoded as a 10-bit, transition-maximized code word indicative of two confrol bits: CTLO and CTLl, or CTL2 and CTL3
• sync words each encoded as a 10- bit, transition-maximized code word indicative of two sync bits: HSYNC and VSYNC
• HSYNC, VSYNC, CTLO, CTLl, CTL2, and CTL3 are assumed by the receiver (e.g., receiver 2' of Fig. 2) to maintain the values that they had when the active video period started.
• transition-minimized code words indicative of HSYNC and VSYNC bits are sent over channel CHO (e.g., one code word, indicative of a HSYNC bit, a VSYNC bit, and two other bits, is sent per pixel clock cycle).
• each packet sent during a data island includes a 32-bit packet header, and each code word indicative of HSYNC and VSYNC that is transmitted during a data island is also indicative of one bit ofthe packet header.
• each control data period is an entire horizontal (or vertical) blanking interval between two consecutive active video periods.
• the transmitter inserts one or more data islands in each horizontal blanking interval and vertical blanking interval (or in each of a subset ofthe blanking intervals).
• a "data island” preamble word is transmitted repeatedly (in a confrol data period) just before each data island, and a video preamble word is fransmitted repeatedly (in a control data period) just before each active video period.
• a burst of eight identical data island preamble words is preferably transmitted just before each data island, and a burst of eight identical video preamble words is preferably fransmitted just before each active video period.
• guard band words and the preamble words described in the preceding paragraph are chosen to have patterns such that a receiver can reliably identify each leading and trailing edge of each data island and each leading edge of an active video period.
• Each data island of channel CHO begins and ends with a burst of two identical guard band words (a fransition minimized TMDS 10-bit code word indicative of any ofthe values OxC, OxD, OxE and OxF, depending upon the values of HSYNC and VSYNC such that the receiver can reliably identify each leading and trailing edge ofthe data island).
• TMDS code words are "golden words”.
• the least significant bit of each code word (q_out[0]) is the first bit to be transmitted, the most sigmficant bit (q_out[9]) is the last to be transmitted, the word "Video GB(CHl)” is a video guard band word transmitted in channel CHI, the word “Video GB(CH0,CH2)” is a video guard band word fransmitted in channels CHO and CH2, and the word “Data Island GB (CH1,CH2)” is a data island guard band word fransmitted in channels CHI and CH2.
• the video words fransmitted over channels CHO, CHI, and CH2 during active video periods can encoded using the full set of transition-minimized TMDS code words.
• Such combination of guard band words and preambles enables the receiver to reliably identify the beginning of each active video period and the beginning and end of each data island.
• special encoding is preferably used to determine the first two and last two 10-bit inventive code words transmitted in each data island on channel CHO.
• two bits of input auxiliary data are preferably encoded and fransmitted on channel CHO during each ofthe first two and last two pixel clock cycles of each data island (whereas four bits of input auxiliary data are encoded and transmitted on channel CHO during all the other pixel clock cycles of each data island as described above).
• each data island contains at least one packet (limiting its minimum duration to 36 pixel clock cycles, including two pixel clock cycles for leading guard bands and two pixel clock cycles for trailing guard bands).
• Each data island must contain an integer number of packets, and preferably (in order to assure the reliability ofthe data within the data island) the maximum number of packets in each data island is fourteen.
• Enor detection and conection preferably BCH enor detection and conection, a well-known enor detection and conection technique named after the developers Bose, Chauduri, and Hocquenghem is applied each packet sent during each data island.
• the subpackets are labeled Subpacket 0 (or "SPO"), Subpacket 1, Subpacket 2, and Subpacket 3.
• Each subpacket includes 56 data bits (i.e., 7 bytes, each byte conesponding to one 10-bit TMDS code word) followed by 8 BCH parity bits.
• Nine data bits of a packet are transmitted per pixel clock cycle.
• one data bit (a header bit) is transmitted (in the form of one 10-bit TMDS code word, which also encodes HSYNC and VSYNC bits) on channel CHO per pixel clock cycle, and four data bits are fransmitted (in the form of one TMDS code word) on each of channels CHI and CH2 per pixel clock cycle.
• the TMDS decoding circuitry in the receiver recovers four bits of data from channel CHO (i.e., from one TMDS code word), four more bits ofthe data from channel CHI, and four more bits ofthe data from channel CH2.
• the data and parity bits of Subpacket 0 are identified as "BCH block 0," the data and parity bits of Subpacket 1 are identified as “BCH block 1,” the data and parity bits of Subpacket 2 are identified as “BCH block 2,” and the data and parity bits of Subpacket 3 are identified as "BCH block 3.”
• the packet header includes 24 data bits (3 bytes of data) followed by 8 BCH parity bits (these data and parity bits are identified in Fig. 9 as "BCH block 4").
• Prefened embodiments ofthe inventive transmitter include the Fig. 9B circuit for generating BCH parity bits for the 56 data bits of each subpacket, and also the Fig. 9A circuit for generating BCH parity bits for the 24 data bits of each packet header.
• Prefened embodiments ofthe inventive receiver for receiving packets whose BCH parity bits have been generated using the Fig. 9A and 9B circuits include the Fig. 9D circuit (for generating a BCH syndrome from the 64 bits of each subpacket) and also the Fig. 9C circuit (for generating a BCH syndrome from the 32 bits of each packet header).
• circuits of Figs. 9B and 9D embody an aspect ofthe present invention that is useful in a variety of contexts for generating enor conection code bits (e.g., parity bits or syndrome bits) from parallel streams of input data, for reducing the number of cycles needed to generate enor conection bits from cyclic polynomials, such as BCH polynomials, and for reducing the number of cycles needed to calculate the syndromes of codes created by cyclic polynomials.
• enor conection code bits e.g., parity bits or syndrome bits
• This aspect ofthe invention is useful in a variety of systems in which enor conection code (e.g., BCH parity bits, and syndromes generated from BCH parity bits and data) must be created for partially parallelized data streams, and in which a small number (e.g., two) ofthe input bits of the next block of input data are available during generation ofthe enor conection code for the cunent block of input data.
• enor conection code e.g., BCH parity bits, and syndromes generated from BCH parity bits and data
• a small number e.g., two
• the even bits of each block of input data can be processed simultaneously (in parallel) with the odd bits from each input data block in accordance with the invention, to generate the enor conection code for each block in a reduced number of clock cycles.
• a clock whose frequency is a multiple (e.g., 2x, 3x, or 4x) ofthe frequency ofthe system clock (the clock used by the system for purposes other than enor conection code generation) in order to generate the same enor conection code in the same time that the code could be generated (using the system clock) in accordance with the invention.
• Utilizing a multiple of a system clock is disadvantageous since it requires the clocked logic to run faster and requires the existence ofthe frequency-multiplied version ofthe system clock (generation of such a frequency-multiplied clock is usually expensive since it requires a clock multiplier circuit, which typically includes a PLL).
• the enor conection code generation circuitry of each of Figs. 9A and 9B employs a control bit Cl which controls whether the circuitry passes through to its output(s) the input data received at its input(s), or whether the circuitry asserts enor conection code at its output(s).
• the control bit Cl is high, allowing a z-bit block ofthe input data (INPUT) to feed into the shift register, to generate the enor conection code, while the block also passes through multiplexer 300 to the output.
• the Fig. 9B circuit operates similarly to that of Fig. 9A, but includes a shift register having two input nodes and two output nodes, each input node receiving a different subset ofthe input bits (rather than a shift register having one input as in Fig. 9A), and requires only (N/2) + 1 clock cycles to generate eight BCH parity bits (whereas the Fig. 9A circuit requires N + 1 clock cycles to do so) and four additional clock cycles to shift out the eight BCH parity bits (by shifting four BCH parity bits from register Q6 while shifting four BCH parity bits from register Q7). Given the register values (in each of registers Q0-Q7 of Fig.
• the BCH parity bits asserted from the two shift registers to multiplexers 301 and 302 for the cunent block of input data are always determined.
• the Fig. 9B circuit is implemented in a prefened embodiment ofthe inventive transmitter (e.g., a prefened implementation of transmitter 100 of Fig. 13)
• the 56 data bits of a subpacket are received at the inputs (and passed through to the outputs) during 28 clock cycles while Cl is high, with a different pair ofthe subpacket bits, IN[0] and IN[1], being received per clock cycle.
• variations on the Fig. 9B design that employ a shift register with "m” input nodes and receive "m” bits ofthe next block of input data (one ofthe “m” bits at each shift register input) while generating enor conection code for the current block, can generate enor conection code (e.g., BCH parity bits) for a cunent block of input data in "k/m +p" clock cycles and can shift out the enor conection code in "n/m” cycles (e.g., when Cl is low), where "n,” "p,” and
• k are parameters determined by the size of each input block and the number of bits of enor conection code to be generated for each input block.
• the first even bit (“IN[0]”) and first odd bit (“IN[1]”) ofthe next block of input data are asserted at the inputs ofthe Fig. 9B circuit at the time when Cl undergoes a high-to-low transition, and these values of
• Q7 Q5 ⁇ ((Q6 ⁇ IN[1]) & C1), where " A " denotes an exclusive OR operation, and "&” denotes an AND operation.
• the bits Q0, Ql, Q2, Q3, Q4, Q5, Q6, and Q7 are generated when Cl is high.
• the Fig. 9B circuit functions as a simple shift register to shift out Q0, Q2, Q4, and Q6 through multiplexer 301 , and Ql , Q3, Q5, and Q7 through multiplexer
• a syndrome is an indicator of whether a transmitted sequence with its enor conection code has been corrupted. If there is not corruption, the syndrome should be zero. If any ofthe bits has been corrupted, the syndrome will likely become nonzero. Depending on the number of bits that have been corrupted and the Hamming distance ofthe encoded sequence, the syndrome can be used to conect the corrupted bits.
• the syndrome generation circuitry of each of Figs. 9C and 9D employs a control bit C2 which controls whether the circuitry passes through to its output(s) the input data received at its input(s), or whether the circuitry asserts syndrome bits at its output(s).
• the control bit C2 is high, allowing a z-bit block ofthe input data (INPUT) to feed into the shift register and also to pass through multiplexer 303 to the output.
• the Fig. 9D circuit operates similarly to that of Fig. 9C, but includes a shift register having two input nodes and two output nodes, each input node receiving a different subset ofthe input bits (rather than a shift register having one input node as in Fig. 9C), and requires only requires only (N/2) + 1 clock cycles to generate eight syndrome bits (whereas the Fig. 9C circuit requires N + 1 clock cycles to do so) and four additional clock cycles to shift out the eight syndrome bits (by shifting four syndrome bits from register Q6 while shifting four syndrome bits from register Q7). Given the register values (in each of registers Q0-Q7 of Fig.
• the syndrome bits asserted from registers Q6 and Q7 to multiplexers 304 and 305) for the cunent block of input data are always determined.
• the Fig. 9D circuit is implemented in a prefened embodiment ofthe inventive receiver (e.g., a prefened implementation of receiver 200 of Fig.
• the 64 bits of a subpacket are received at the inputs (and passed through to the outputs) during 32 clock cycles while C2 is high, with a different pair ofthe subpacket bits, D [0] and IN[1], being received per clock cycle, and a different pair ofthe subpacket bits,
• Fig. 9D design employ a shift register having "m" input nodes and receive "m” bits ofthe next block of input data (one ofthe “m” bits at each shift register input node) while generating a syndrome for the cunent block, and can generate syndrome bits for a current block of input data in "k/m + p" clock cycles and can shift out the syndrome bits in "n/m” cycles (e.g., when C2 is low), where "n,” “p,” and “k” are parameters determined by the size of each input block and the number of syndrome bits to be generated for each input block.
• the first even bit (“IN[0]”) and first odd bit (“IN[1]”) ofthe next block of input data are asserted at the inputs ofthe Fig. 9B circuit at the time when C2 undergoes a high-to-low transition, and these values of IN[0] and IN[1] continue to be asserted at the inputs while C2 remains low.
• Q7 Q5 ⁇ Q6, where " A " denotes an exclusive OR operation, and "&" denotes an AND operation.
• the bits Q0, Ql, Q2, Q3, Q4, Q5, Q6, and Q7 are generated when C2 is high.
• the Fig. 9D circuit When C2 is low, the Fig. 9D circuit functions as a simple shift register to shift out Q0, Q2, Q4, and Q6 through multiplexer 305, and Ql, Q3, Q5, and Q7 through multiplexer
• the transmitter is configured to automatically insert zero bits in the input data stream (when grouping the input data stream into subpackets) so that the inserted zero bits occur at known positions among the 56 slots of each subpacket that are available for data bits.
• the enor detection circuitry in the receiver should be configured to check for zeros in the expected slots of each recovered subpacket of data, and to use the resulting information to identify enors in each recovered subpacket of data.
• the third bit (“bit 2" in Fig. 9), of each four-bit word determined by the 10-bit code word transmitted on CHO during one ofthe 32 pixel clock cycles of a packet transmission, is one ofthe bits ofthe packet header.
• the four-bit words determined by the 10-bit code words fransmitted on CHI during the 32 pixel clock cycles of a packet transmission, are the even bits ofthe subpackets.
• the four-bit words determined by the 10-bit code words transmitted on CH2 during the 32 pixel clock cycles of a packet fransmission, are the odd bits ofthe subpackets.
• each code word sent on CHI during the "N"th pixel clock cycle determines an even bit of each of subpackets 0, 1, 2, and 3, and each code word sent on CH2 during the "N"th pixel clock cycle determines an odd bit of each of subpackets 0, 1, 2, and 3.
• Each code word sent on CHO during the "N"th pixel clock cycle determines one bit ofthe packet header (as well as one HSYNC bit and one VSYNC bit).
• Each code word sent on CHI during the "N+l"th pixel clock cycle determines the next even bit of each of subpackets 0, 1, 2, and 3, and each code word sent on CH2 during the "N+l”th pixel clock cycle determines the next odd bit of each of subpackets 0, 1, 2, and 3.
• Each code word sent on CHO during the "N+l"th pixel clock cycle determines the next bit ofthe packet header (as well as the next HSYNC bit and the next VSYNC bit).
• the BCH parity bits for the packet header are calculated over the packet header's 24 data bits, and are determined by the code words fransmitted over CHO during the last 8 pixel clock cycles ofthe packet.
• the BCH parity bits for each subpacket are calculated over the subpacket's 56 data bits, and are determined by the code words transmitted over CHI and CH2 during the last four pixel clock cycles ofthe packet (the parity bits for each subpacket include bits transmitted over two channels: CHI and CH2).
• the transmitter not only transmits each packet (during a data island) such that the packet is spread over three channels of a TMDS link (CHO, CHI, and CH2), and each subpacket of a packet is spread over two channels of the link (CHI and CH2), but the transmitter separately generates BCH parity bits for each subpacket of a packet and transmits the BCH parity bits for each subpacket such that these BCH parity bits are spread over two channels ofthe link (CHI and CH2).
• the first byte of each packet header is a "Packet Type" code
• each ofthe second and third bytes of each packet header indicates packet-specific data.
• ACR audio clock regeneration
• a transmitter can transmit a null packet anytime, but the receiver does not recover data from a null packet.
• the receiver identifies a null packet (by identifying a 0x00 value in the first byte of a packet header), the receiver ignores (treats as "undefined") all values in the null packet after the first byte ofthe packet header.
• An audio clock regeneration (ACR) packet contains both the CTS (Cycle Time Stamp) value and N values (described below) that are used for audio clock regeneration.
• the four subpackets of an ACR packet are identical. Each subpacket contains a 20-bit "N" value and a 20-bit "CTS" value. A CTS value equal to zero indicates that there is no new value of CTS.
• An audio sample packet can include one, two, three, or four pairs of audio samples (each sample pair determined by a different subpacket ofthe packet). These can be different samples, different partial samples (e.g., 2 of 6 channels), or repeated samples. Bits in the second and third bytes of an audio sample packet's header indicate the configuration of each subpacket ofthe packet, indicate whether each subpacket does or does not contain a sample, and indicate whether each subpacket contains a "flatline" sample (i.e., one whose magnitude is zero).
• each ofthe four LSBs ofthe second byte ofthe packet header indicates whether the conesponding subpacket includes an audio sample
• each ofthe four LSBs ofthe third byte ofthe packet header indicates whether the conesponding subpacket includes a "flatline” sample, but a "1" value (indicating a flatline sample) of any ofthe LSBs ofthe third byte is valid only if the conesponding "present" bit ofthe second byte (one ofthe four LSBs ofthe second byte) indicates that the subpacket contains a sample.
• Each subpacket of an audio sample packet can contain channel status bits and parity bits as well as bits indicative of audio samples.
• the audio data of each subpacket are formatted as a structure that closely resembles an IEC_60958 or IEC_61937 frame, but with a "start" bit (in the packet header) replacing the multi-bit preamble of each sub- frame of an 1EC_60958 or IEC_61937 frame.
• Each subpacket of an audio sample packet can include two 28-bit words (e.g., a 28-bit word of a left stereo stream and a 28-bit word of a right stereo stream), each 28-bit word conesponding to one ofthe two sub-frames of an IEC_60958 frame, and including 24 bits of audio data as well as a valid bit, a parity bit, a channel status bit, and a user data bit. All fields within each subpacket of an audio sample packet preferably follow conesponding rules specified in the IEC_60958 or IEC_61937 specification.
• each ofthe four MSBs ofthe third byte ofthe packet header of an audio sample packet is a "start" bit indicating whether the conesponding subpacket contains the first frame of a block of 1EC 60958 (or 1EC_61937) frames of audio samples.
• the transmitter sets the start bit to "1" when the conesponding subpacket contains a first "B, W” frame of a block and clears the start bit to "0" when the conesponding subpacket contains a first "M, W" frame of a block.
• the inventive system can transmit an IEC_60958 or 1EC 61937 audio stream having sample rate of 32kHz, 44.1kHz or 48kHz (which can determine a stereo or compressed surround-sound sfream), and can transmit as many as four streams of IEC 60958 or 1EC_61937 data (e.g., 8 stereo channels) at sample rates of up to 96KHz, or at least one stream of IEC 60958 or IEC 61937 data at a sample rate of 192KHZ.
• IEC_60958 or 1EC 61937 audio stream having sample rate of 32kHz, 44.1kHz or 48kHz (which can determine a stereo or compressed surround-sound sfream)
• IEC 60958 or 1EC_61937 data e.g., 8 stereo channels
• An InfoFrame packet can include format infonnation and related information regarding audio and/or video data being transmitted.
• the EIA CEA-861B standard defines an "InfoPacket" data structure which consists of one or more "InfoFrames” containing format information and related information regarding transmitted digital audio and/or digital video data.
• each InfoFrame packet transmitted by the inventive system includes the information of an EIA/CEA-861B InfoPacket, with the following restrictions: the inventive InfoFrame packet can contain only the information of a single EIA/CEA-861B InfoFrame having a maximum size of 30 bytes (including the seven LSBs ofthe first byte ofthe packet header which indicate an InfoFrame type code, the second byte ofthe packet header which indicates a version number field, and the third byte ofthe packet header which indicates an InfoFrame length field); and type code must be within the range of 0 to 127 (since it is indicated by the seven LSBs ofthe first byte ofthe packet header).
• InfoFrame type codes are 0x02 (indicating that the InfoFrame packet includes the data of an Auxiliary Video information ("AVI") InfoFrame), 0x03 (indicating that the InfoFrame packet includes the data of an Audio InfoFrame), and 0x04 (indicating that the InfoFrame packet includes the data of a DVD confrol InfoFrame).
• the InfoFrame length field indicates InfoFrame packet's length in bytes as per the EIA/CEA-861B standard, including the three bytes ofthe packet header and any valid bytes in the packet including the checksum (so that the maximum value ofthe length is "31").
• An auxiliary video information (“AVI") InfoFrame packet indicates various aspects ofthe cunent video stream being fransmitted (or to be transmitted) to the receiver, such as whether the video data are in RGB, YCbCr 4:2:2, YCbCr 4:4:4 format, whether the video data are overscanned or underscanned, the picture aspect ratio (and whether the picture has been horizontally or vertically scaled), the video component bit depth, the number of pixels per line and lines per frame, and the number of pixel repetitions (i.e., each pixel is not repeated, or is sent N times, where N is an integer greater than one).
• An Audio InfoFrame packet indicates various aspects ofthe audio data being transmitted (or to be transmitted), such as the number of streams of audio data, the sampling frequency for each sfream, and the sample size (number of bits).
• the inventive system can employ video pixel replication as needed to provide adequate audio bandwidth.
• video pixel replication as needed to provide adequate audio bandwidth.
• a frame of video consists of L lines (whose pixels are transmitted in non-repeated fashion at a first pixel clock frequency), and 2L data islands (each inserted in a blanking interval following a line of video) would be needed to transmit the audio bits that conespond to one frame ofthe video
• the system is preferably operable in a mode in which transmission of each pixel is repeated.
• the pixel clock frequency is not changed, repeated transmission of each pixel effectively halves the video transmission rate (since two lines of repeated pixels are needed to transmit all the pixels of each non-repeated line) but allows the audio bits for a full frame ofthe video non-repeated video to be transmitted in 2L data islands, each following one ofthe 2L active video periods in which pixels ofthe frame are transmitted repeatedly.
• the pixel clock frequency is increased (relative to the pixel clock frequency employed for non-repeated pixel transmission) during each mode in which pixels are transmitted repeatedly.
• audio data can be transmitted at a doubled data rate (with time stamps based on the doubled pixel clock) between active video periods but the effective video pixel transmission rate is not changed.
• transmission of each pixel is repeated more than once (e.g., each is transmitted three times, using a frequency-tripled version ofthe pixel clock).
• the receiver When receiving video data transmitted in any repeated pixel transmission mode, the receiver drops all but one sample of each set of identical video samples received during each active video period (e.g., it drops all odd video samples, or all even video samples where each pixel is transmitted twice) but the receiver processes all samples of packetized data received during data islands between the active video periods.
• the transmitter's use of pixel-repetition is preferably indicated by a Pixel Repetition Count field in a fransmitted AVI InfoFrame packet.
• a Pixel Repetition Count field indicates to the receiver how many repetitions of each unique pixel are transmitted.
• the value ofthe field is zero.
• the value ofthe field indicates the number of pixels (of each set of consecutively received pixels) that should be discarded by the receiver.
• the video samples transmitted during the first pixel clock cycle of each line are unique (and are not discarded) but each repetition of these samples is dropped in modes with pixel repetition.
• the inventive system is configured to transmit video data in any of several different formats over a TMDS link, and the receiver is configured to recover the data in any of such formats.
• the system can transmit video over the CHO, CHI, and CH2 channels of such a link in any of YCbCr 4:4:4 format, YCbCr 4:2:2, or RGB format, with or without pixel repetition. In all cases, up to 24 bits can be transmitted over these channels per pixel clock cycle. As shown in Fig.
• an 8-bit red color component can be transmitted over channel CH2
• an 8-bit green color component can be fransmitted over channel CHI
• an 8-bit blue color component can be transmitted over channel CHO, during each pixel clock cycle.
• the R,G, and B components ofthe first pixel of each line are transfened during the first pixel clock cycle after a video guard band (e.g., after the second of two consecutively transmitted video guard band codes).
• an 8-bit “Cr” sample can be transmitted over channel CH2, an 8-bit luminance ("Y”) sample fransmitted over channel CHI, and an 8-bit “Cb” sample fransmitted over channel CHO during each pixel clock cycle.
• Y and Cb per pixel clock cycle, more bits are allocated per component.
• the available 24 bits are split into 12 bits for the Y component and 12 bits for the C components.
• a 12-bit luminance ("Y") sample can be transmitted over channels CHO and CHI and a 12-bit Cb sample can be transmitted over channels CHO and CH2.
• the four least significant bits ofthe Y sample and the four least significant bits ofthe Cb sample are determined by the TMDS code word transmitted over CHO.
• the next 12-bit Y sample can be transmitted over channels CHO and CHI, and a 12-bit Cr sample can be transmitted over channels CHO and CH2.
• the four least significant bits of this Y sample and the four least significant bits ofthe Cr sample are determined by the TMDS code word transmitted over CHO.
• the bits of the Cb sample transmitted during the Nth clock cycle are for the pixel that includes the
• the bits ofthe Cr sample transmitted during the "N+l"th clock cycle are for the pixel that includes the Y sample transmitted during the same ("N+l"th) clock cycle. If each Y, Cr, and Cb sample consists of fewer than 12 bits, the valid bits should be left-justified (the most significant valid bits should be aligned) and zeroes should pad the bits below the least significant bit.
• all ofthe data sent during the Nth pixel clock cycle are sent again during the next ("N+l"th) pixel clock cycle, the next set of data are sent during the next ("N+2"th) pixel clock cycle, and so on.
• the inventive transmitter includes a processor (e.g., microcontroller 15 of Fig. 2) programmed to support HDCP encryption of data and to communicate with the receiver (e.g., over the DDC channel of a TMDS link) to initiate an HDCP authentication exchange with the receiver and to query registers in the receiver as appropriate in connection with implementation of HDCP content protection.
• the processor is preferably programmed with system software for setting up the transmitter (e.g., via interface 101 of Fig. 13) and the receiver (e.g., via the DDC channel) for HDCP encryption during Data Island Mode operation (explained below).
• the system software causes the transmitter to enter the Data Island Mode and to trigger the receiver's entry into the Data Island Mode by sending a packet to the receiver (as explained below), and then (via the DDC interface) to query the appropriate registers in the receiver to verify that the receiver has indeed fransitioned to Data Island mode.
• the processor in the fransmitter executes HDCP authentication software and (upon successful completion of an authentication exchange) causes the fransmitter to encrypt data to be fransmitted during active video periods and data islands.
• all video data transmitted over channels CHO, CHI, and CH2 of a TMDS link during active video periods are HDCP encrypted (except that video guard band words are not encrypted), and during each data island, only the code words transmitted over channels CHI and CH2 are HDCP encrypted (but neither leading data island guard band words nor trailing data island guard band words are encrypted).
• a high value of DE indicates that video data are being transmitted (or optionally, that auxiliary data are being transfened between active video periods), and all data transfened while DE is high are encrypted.
• HDCP rekeying is triggered by each high-to-low transition of DE.
• Ml is high only when transition-minimized code words indicative of video data (but not video guard bands) are transmitted over channels CHO, CHI , and CH2 of a TMDS link.
• M2 is high only when transition-minimized code words (but not leading data island guard band words or trailing data island guard band words) are transmitted over channels CHI and CH2 during data islands.
• DE is high but neither Ml nor M2 is high.
• the HDCP cipher engine in the transmitter performs a bitwise XOR operation thereon using a stream of 24-bit pseudo-random values that are generated in a well known manner (as described above).
• the HDCP cipher engine in the transmitter performs a bitwise XOR operation on the eight audio data bits using a selected 8-bit subset of the bits of each 24-bit pseudo-random value ofthe conventionally generated pseudo-random value sfream.
• HDCP rekeying is performed in response to each high-to-low transition of Ml (i.e., at the end of each active video period).
• the rekeying must be complete within Y pixel clock cycles (e.g., within 58 pixel clock cycles ofthe high-to-low transition of Ml), and the fransmitter is configured to place each data island only in time slots between active video periods in which its HDCP encryption circuitry has completed rekeying and is ready to encrypt the packetized data to be transmitted in the data island.
• the HDCP encryption indicator is indicated by two transition-maximized TMDS control words (indicative ofthe predetermined values of control bits CTLO, CTLl, CTL2, and CTL3) that are transmitted over channels CHI and CH2 of a TMDS link during a blanking interval (other than during a data island or active video period), and preferably during a vertical blanking interval.
• TMDS control words indicative ofthe predetermined values of control bits CTLO, CTLl, CTL2, and CTL3
• key control data or “decryption control” data increases the reliability with which the receiver determines when to perform the frame key calculation (after successful completion of an HDCP authentication exchange by the transmitter and receiver). Since the encryption indicator depends on all four confrol bits (rather than just control bit CTL3), reliability is improved and future expandability is enabled.
• WOO having predetermined duration (preferably 72 pixel clock cycles) following an active edge of VSYNC (i.e., following detection of a code word indicative of a rising edge of VSYNC).
• Such constraints on the HDCP encryption indicator are added to improve the reliability with which the receiver detects the indicator when the indicator is transmitted by the transmitter and expected by the receiver.
• confrol characters CTLO, CTLl, CTL2, and CTL3 set to specific values
• the first two code words (transmitted during the first two pixel clock cycles of every data island and active video period are guard band codes designed to create a very robust DE rising edge.
• the fransmitter should be configured never to encrypt guard band code words, and the receiver should be configured never to decrypt guard band code words.
• the transmitter and receiver ofthe inventive system are coupled by at least one TMDS link, and are operable in a digital visual interface ("DVI") mode or a "Data Island” mode.
• DVI digital visual interface
• the receiver In the DVI mode, the receiver expects conventional transition maximized (out-of-band) code words indicative of HSYNC/VSYNC on TMDS channel CHO from the fransmitter, distinguishes blanking intervals from active video periods (and optionally also from bursts of auxiliary data in blanking intervals) by recognizing whether the incoming code words on channels CHO, CHI, and CH2 are out-of-band code words (in blanking intervals) or in-band code words (in active video periods and bursts of auxiliary data), and monitors the incoming data for an indication that it should enter the Data Island mode.
• the receiver In the Data Island mode, the receiver expects packets of audio data (or other auxiliary data) in data island between active video periods, expects transition-minimized code words indicative of HSYNC, VSYNC, and packet header bits on channel CHO during data islands, and identifies active video periods by detecting leading video guard bands.
• the fransmitter and receiver undergo synchronized (and nearly simultaneous) transitions from the DVI mode to the Data Island mode.
• each DVI-to-Data Island mode fransition occurs when the fransmitter enters the Data Island mode and sends a data island which is detected by the receiver.
• a preamble sent over both CHI and CH2
• at least one distinctive guard band code preferably two consecutive identical guard band codes, each having the value 0x55, on each of CHI and CH2 following the preamble.
• the receiver In response to detecting a data island (e.g., by detecting the data island preamble, and then (in the first pixel clock cycle after the data island preamble) detecting the guard band word 0x55, and then (in the second pixel clock cycle after the data island preamble) detecting the guard band word 0x55) the receiver enters the Data Island mode.
• the receiver sets a bit (indicative of Data Island mode operation) in a register that is readable by the fransmitter.
• the receiver sets the bit in a register in the HDCP register space (e.g., registers 203 of Fig. 14) that is readable by the fransmitter via the DDC channel.
• this bit can be cleared only upon hardware reset ofthe receiver.
• circuitry in the receiver monitors the incoming data for each preamble/guard band combination indicative of an active video period, and each preamble/guard band combination indicative of a data island.
• the transmitter can determine the receiver mode status (i.e., whether the receiver is in the DVI or Data Island mode, and what type of HDCP decryption the receiver can perform) via the DDC channel. If the transmitter reads register bits (via the DDC channel) indicating that the receiver is in the Data Island mode and supports a particular type of HDCP, the transmitter will operate in the Data Island mode (including by sending VSYNC/HSYNC during data islands as bits determined by transition-minimized TMDS code words rather than by transition- maximized code words, and inserting video guard band codes in the video stream) and will employ the appropriate HDCP state machine to encrypt data to be transmitted (but will not encrypt any video guard band).
• the receiver mode status i.e., whether the receiver is in the DVI or Data Island mode, and what type of HDCP decryption the receiver can perform
• the fransmitter In the Data Island mode, the fransmitter preferably uses the DDC channel to determine the capabilities and characteristics ofthe receiver by reading (e.g., from ROM 23 of Fig. 2) an EDID data structure (preferably including the EIA/CEA-861 B- defined extensions to EDID).
• the transmitter can use the DDC channel to determine its own physical device address, for use in various cluster-wide control operations.
• the fransmitter preferably determines what type of encryption to use in response to determining the decryption capabilities (if any) ofthe receiver, and informs the receiver of the encryption mode chosen. In prefened embodiments, the following steps must be taken by the fransmitter to initiate Data Island mode operation with HDCP encryption of fransmitted data:
• the transmitter's state is reset to an initial state
• the transmitter reads the EDID data ofthe receiver (e.g., EDJD data stored in ROM 23 of Fig. 2 over the DDC channel) to determine whether the receiver is capable of Data Island mode operation;
• the EDID data ofthe receiver e.g., EDJD data stored in ROM 23 of Fig. 2 over the DDC channel
• the transmitter reads a distinctive mode bit in a register ofthe receiver (e.g., a Data Island Mode bit in a predetermined location in registers 203 of Fig. 14) via the DDC channel;
• a distinctive mode bit in a register ofthe receiver e.g., a Data Island Mode bit in a predetermined location in registers 203 of Fig. 14
• both the transmitter and receiver operate in the Data Island Mode, and the fransmitter initiates an HDCP authentication procedure;
• the transmitter sets a bit (e.g., a bit "HDCP Authenticated") in the transmitter's configuration register, in response to which all the transmitter's HDCP engines are enabled.
• a bit e.g., a bit "HDCP Authenticated”
• the receiver disables its video and audio outputs, because there will be times during startup when the transmitter is in the Data Island mode, but the receiver is still in the DVI mode, which could cause undesirable video output from the receiver.
• Fig. 13 is a block diagram of fransmitter 100 which is an embodiment ofthe inventive transmitter.
• Transmitter 100 includes video subsystem 106 which performs pipelined encryption and other pipelined processing (e.g., reformatting, upsampling, and/or color space conversion) on video data D[23:0] received from an external source.
• a video data clock IDCK, refened to herein as a "pixel clock”
• a video data enable signal (DE)
• HSYNC and VSYNC horizontal and vertical sync control signals
• Transmitter 100 is typically configured to operate in response to a pixel clock having frequency in the range from 25MHz to 112 MHz.
• Subsystem 106 encrypts the input video data (or a reformatted or otherwise processed version ofthe input video) using pseudo-random values from cipher engine 104 and asserts the encrypted video to a first input of multiplexer 118.
• cipher engine 104 uses bits in registers 103. Registers 103 can be loaded with values received via interface 101 and/or interface 102.
• interface 101 is coupled for I2C communication with a microcontroller (e.g., microcontroller 15 of Fig. 2).
• the microcontroller is also coupled to the DDC channel of a TMDS link, and to an input data source (e.g., a video source).
• Interface 101 can implement an I2C slave protocol to receive information and configuration bits (e.g., InfoFrame bits) and other bits (e.g., key values received during an HDCP authentication procedure), and load such bits into registers 105 and 103.
• information and configuration bits e.g., InfoFrame bits
• other bits e.g., key values received during an HDCP authentication procedure
• transmitter 100 is implemented as an integrated circuit
• an EEPROM preloaded with key values and identification bits can be implemented as a separate chip in a secure multi-chip module ("MCM") that includes both fransmitter 100 and the EEPROM.
• MCM secure multi-chip module
• An example of such an EEPROM is EEPROM 14 of Fig. 2.
• Interface 102 provides the interface between transmitter 100 and the EEPROM. interface 102 can retrieve values from the EEPROM at appropriate times, such as during an HDCP authentication exchange with a receiver.
• Interface 102 preferably uses the output of ring oscillator 113 (typically having frequency 64 MHz , or frequency in the range 51.2 MHz to 76.8 MHz) to generate a clock (e.g., a 100 KHz clock) for use for I2C communication with the EEPROM.
• ring oscillator 113 typically having frequency 64 MHz , or frequency in the range 51.2 MHz to 76.8 MHz
• a clock e.g., a 100 KHz clock
• Transmitter 100 also includes audio subsystem 108 which performs pipelined formatting, packetizing, encryption, and other pipelined processing on audio data AUD received from an external source (although AUD can be auxiliary data other than audio data, we will refer to it as audio data for simplicity).
• transmitter 100 can accept audio data AUD in S/PDIF format with sample rate Fs in the range from 32kHz to 48 kHz, or in any of a number of other formats (e.g., 2-channel uncompressed PCM data or a compressed bitsfream indicative of multi-channel data).
• An audio reference clock (“MCLK”) is received with the input data AUD.
• the clock MCLK (also refened to as a master clock) has a frequency of at least 256*Fs (or 384*Fs) in prefened embodiments.
• Main phase lock loop (“PLL") 116 generates a stabilized version ofthe pixel clock IDCK.
• Reset circuitry 112 is coupled to a reset pin for receiving a reset bit from an external processor.
• Transmitter 100 is configured to reset itself to an initial state in response to a predetermined value ofthe reset bit.
• Test circuitry 110 is coupled to a test pin for receiving a test mode bit from an external source.
• Transmitter 100 is configured to operate in either a test mode or a normal operating mode depending on the value of the test mode bit.
• Subsystem 108 can sample the audio data using the stabilized pixel clock (provided that the pixel clock frequency is greater than 2*128Fs), the clock MCLK, or a frequency-multiplied version of MCLK).
• Subsystem 108 generates packets that contain the sampled audio data, encrypts the data in the packets, and encodes the encrypted data using TERC2 or TERC4 encoding, and asserts the packets containing encoded, encrypted data to a second input of multiplexer 118.
• subsystem 108 encodes the four least-significant bits of each encrypted sample as one of sixteen "golden words" (e.g., the 8-bit words "AD0-AD15" in the left column of Fig. 4) and the four most significant bits of each encrypted sample as another golden word.
• Subsystem 108 also determines the timing (relative to DE) with which data islands (each data island including one or more ofthe packets) are asserted to multiplexer 118. Subsystem 108 also time-division-multiplexes control words with the data islands, including control words indicative of: a data island preamble (e.g.
• subsystem 108 inserts eight pixel clock cycles of auxiliary data preamble words immediately before each data island), HSYNC and VSYNC (e.g., subsystem 108 inserts at least twelve pixel clock cycles of synchronization words before each burst of auxiliary data preamble words), leading and trailing data island guard bands (e.g., subsystem 108 inserts two pixel clock cycles of leading guard band words as the first two words of each data island and two pixel clock cycles of trailing guard band words as the last two words of each data island), a video preamble (e.g. subsystem 108 inserts video preamble words after each data island), and video guard bands (e.g., subsystem 108 inserts two pixel clock cycles of video guard band words after each video preamble).
• leading and trailing data island guard bands e.g., subsystem 108 inserts two pixel clock cycles of leading guard band words as the first two words of each data island and two pixel clock cycles of trailing guard band words as the last two words of each data
• multiplexer 118 In response to a confrol signal indicative of DE (the DE signal received at interface 107), multiplexer 118 passes either video data from subsystem 106 (when DE is high) or the output of subsystem 108 (when DE is low) to TMDS core processor 114.
• Core processor 114 operates in response to the stabilized pixel clock (generated by PLL 116) and performs the above-described operations of encoding 8-bit data words as 10-bit TMDS code words, serializing the data, and transmitting the serialized encoded data (and the stabilized pixel clock) over a TMDS link to receiver 200 of Fig. 14 (an embodiment ofthe inventive receiver).
• receiver 200 includes core processor 214 which in operation is coupled to the TMDS link.
• Processor 214 recovers the pixel clock from the link's clock channel, and main PLL 216 generates a stabilized pixel clock in response to the recovered pixel clock.
• processor 214 performs the above-described operations of de-serializing the data received over the link, decoding the de-serialized 10-bit TMDS code words to recover 8-bit code words, and asserting the 8-bit code words to splitting unit 218.
• Unit 218 also receives a signal indicative of DE, and the stabilized recovered pixel clock, from processor 214.
• Unit 218 detects the beginning and end of each data island and each active video period in the code word stream from processor 214 (including by identifying guard bands and preamble code words ofthe above- mentioned types), routes each audio data packet (of each data island) to pipelined audio subsystem 208, and routes the remaining data (including all bursts of video data) to pipelined video subsystem 206.
• the data asserted by unit 218 to subsystem 206 includes HSYNC and VSYNC code words.
• Video subsystem 206 performs decryption and other processing (e.g., reformatting, upsampling or subsampling, and/or color space conversion) on the video data received from unit 218.
• Subsystem 206 decrypts the video data from unit 218 (to generate 8-bit decrypted words) using pseudo-random values from cipher engine 204 and asserts the decrypted video to the pipelined circuitry for performing other processing (e.g., reformatting, subsampling, and color space conversion) thereon.
• the latter circuitry outputs the decrypted, processed video bits Q[23:0], typically after final reformatting, and with conesponding DE, HSYNC, and VSYNC signals and a pixel clock.
• subsystem 206 also includes a digital-to-analog converter that generates and outputs analog video signals (AnRPr, AnGY, and AnBPb), which can be red, green, and blue color component signals or luminance and chrominance signals, in response to 8-bit decrypted and processed video words.
• cipher engine 204 uses bits in registers 203. Registers 203 can be loaded with values received via interface 201 and/or interface 202.
• an EEPROM preloaded with key values and identification bits can be implemented as a separate chip in a secure multi-chip module (“MCM”) that includes both receiver 200 and the EEPROM.
• MCM secure multi-chip module
• Interface 202 provides the interface between receiver 200 and the EEPROM. Interface 202 can retrieve values from the EEPROM at appropriate times, such as during an HDCP authentication exchange with the transmitter, interface 202 preferably uses the output of ring oscillator 213 (typically having frequency 64 MHz , or frequency in the range 51.2 MHz to 76.8 MHz) to generate a clock for use for I2C communication with the EEPROM.
• interface 201 can be coupled to the DDC channel ofthe TMDS link, and can implement an I2C slave protocol to communicate with the transmitter over the DDC channel (for example, to perform HDCP authentication including by loading key values received from the transmitter over the DDC channel into registers 203.
• receiver 200 when receiver 200 is a repeater (configured to operate as a transmitter that sends data to another receiver, in addition to being configured to receive data from transmitter 100), receiver 200 includes interface 207.
• Interface 207 can be coupled to a host device and can implement an I2C slave protocol to receive information and configuration bits and load such bits into registers 205 and 203.
• interrupts When appropriate (i.e., in response to predetermined status, information, or enor conditions), interrupts (“INT") are asserted from registers 205 to a host device to request software attention.
• Receiver 200 also includes pipelined audio subsystem 208 which receives and processes packets of audio data from unit 218.
• Subsystem 208 decodes the golden words in the packets to determine the 8-bit word indicated by each pair of golden words" (when the transmitter employed TERC4 encoding to encode the audio data), or the 4-bit word indicated by each pair of golden words (when the transmitter employed TERC2 encoding to encode the audio data).
• Subsystem 208 also decrypts the decoded audio samples (using pseudo-random bits from cipher engine 204), performs enor conection on the decrypted, decoded samples, unpacks the enor-conected, decrypted, decoded samples from the packets (and routes configuration and status bits that are unpacked from the packets to appropriate registers), and performs other processing on the unpacked audio samples (for example, by organizing the audio data for S/PDIF and I S output engines and then processing the organized data in one or both of such engines to generate output data in one or both of S/PDIF or I S format).
• subsystem 208 can output audio data in any of a number of different formats (e.g., as 2- channel uncompressed PCM data or as a compressed bitsfream indicative of multichannel data).
• subsystem 208 asserts (with the output audio data) one or more of a bit clock (“SCK”) whose frequency is the audio bit rate, an audio reference clock (“MCLK”) which is a recovered version ofthe MCLK clock employed by transmitter 100, a serial data output clock (SDO) for use in demultiplexing a time-division multiplexed output audio data stream indicative of two audio channels, and a word select clock ("WS").
• SCK bit clock
• MCLK audio reference clock
• SDO serial data output clock
• WS word select clock
• Main PLL 216 generates a stabilized version ofthe pixel clock recovered from the clock channel ofthe TMDS link, for use by interface 214 and other elements of receiver 200.
• Reset circuitry 212 is coupled to a reset pin for receiving a reset bit from an external processor.
• Receiver 200 is configured to reset itself to an initial state in response to a predetermined value ofthe reset bit.
• Test circuitry 210 is coupled to a test pin for receiving a test mode bit from an external source.
• Receiver 200 is configured to operate in either a test mode or a normal operating mode depending on the value ofthe test mode bit.
• each fransmitter and receiver for use in an embodiment ofthe inventive system is manufactured to meet a detailed specification.
• each transmitter and receiver is manufactured in such a manner that it can be tested efficiently for compliance with the specification.
• the inventive transmitter includes a test pattern generator (e.g., test pattern generator 119 of fransmitter 100 of Fig. 13) and the inventive receiver includes an identical test pattern generator (e.g., in test circuitry 210 of receiver 200 of Fig. 14).
• the test pattern generator in the receiver is preferably controllable by the transmitter in a test mode, or the test pattern generators in the fransmitter and receiver are separately controllable (e.g., by a host device), to allow determination of enor rates on the serial link between the transmitter and receiver.
• transmitter 100's test pattern generator 119 asserts test data via multiplexer 118 to core processor 114, and processor 114 encodes the test data and transmits the encoded test data to receiver 200 over a TMDS link.
• receiver 200 receives and decodes the data and asserts the recovered data to test circuitry 210.
• Enor detection and accumulation logic in test circuitry 210 compares the recovered data with the test data produced by the test pattern generator in circuitry 210.
• the test results, including the measured enor rate of the recovered test data can be sent back to fransmitter 100 over the DDC channel ofthe link (e.g., via interface 201) and/or asserted to a host device (e.g., via interface 207).
• the test pattern generator in the fransmitter e.g., pattern generator 119
• the identical pattern generator in the receiver includes a linear feedback shift register (LFSR) that generates a pseudo-random test pattern, and both pattern generators start from the same initial state (during test mode operation) so that they generate the same pseudo-random test pattern during test mode operation.
• LFSR linear feedback shift register
• the fransmitter e.g., microcontroller 15 of Fig. 2
• the transmitter preferably also can control test mode operation ofthe receiver via the DDC channel or another channel.
• Use ofthe DDC channel (or another channel of a serial link) as an enor-detection back-channel during test mode operation allows the transmitter (e.g., microcontroller 15 of Fig. 2) to change one or more parameters ofthe transmitter, and to determine the enor rate (at the receiver) that results from each set of fransmitter parameters.
• the transmitter can determine the reliability ofthe link as a function of various sets of operating parameters, and the fransmitter can optimize its operating parameters to reduce or minimize the enor rate at the receiver.
• each ofthe inventive receiver and the inventive fransmitter is implemented to support a robust set of test features, enabled through the assertion of a test pin (the pin to which test circuit 110 is coupled in Fig. 13, or to which test circuit 210 is coupled in Fig. 14).
• Individual test modes can be selected through input-capable pins that are redesignated as test modifier pins during test mode operation, or through register bits.
• test mode enabling is accomplished by loading bits into registers (e.g., registers 205 of receiver 200 of Fig. 14, or registers 105 of transmitter 100 of Fig. 13), for example via an I2C interface (such as interface 207 of Fig. 14 or interface 101 of Fig. 13).
• I2C interface such as interface 207 of Fig. 14 or interface 101 of Fig. 13.
• test mode enabling via the loading of register bits requires a larger set of vectors to enable a test mode, it can allow the transmitter or receiver to be designed with fewer input pins.
• An 8-bit test mode register would allow 64 different
• the test modes include a logic test (e.g., a full scan), EEPROM test (e.g., a checksum BIST, or built-in-self-test), a TMDS Core test (e.g., in which core 114 or 214 is isolated and then tested), a DAC test (e.g., in which the receiver DAC circuitry is isolated and then tested using test patterns), a PLL test, a ring oscillator test, and one or more debug tests (e.g., in which internal signals are multiplexed to input/output pins).
• a logic test e.g., a full scan
• EEPROM test e.g., a checksum BIST, or built-in-self-test
• TMDS Core test e.g., in which core 114 or 214 is isolated and then tested
• DAC test e.g., in which the receiver DAC circuitry is isolated and then tested using test patterns
• PLL test e.g., in which
• a full scan logic test would require dedicated input and output pins, which are configured as test pins during logic test mode operation.
• a checksum-based BIST method is implemented to test an external key EEPROM (e.g., EEPROM 14 or EEPROM 24 of Fig. 2).
• test circuitry Upon being enabled, test circuitry would read the contents ofthe EEPROM, perform a checksum calculation, and then compare with a checksum read from the EEPROM.
• EEPROM e.g., EEPROM 14 or EEPROM 24 of Fig. 2
• test circuitry Upon being enabled, test circuitry would read the contents ofthe EEPROM, perform a checksum calculation, and then compare with a checksum read from the EEPROM.
• EEPROM e.g., EEPROM 14 or EEPROM 24 of Fig. 2
• test circuitry Upon being enabled, test circuitry would read the contents ofthe EEPROM, perform a checksum calculation, and then compare with a checksum read from the EEPROM
• the regenerated pixel clock can also be multiplexed to an output pin to allow eye diagram testing.
• an input clock pin provides a test clock directly to the DAC (e.g., in circuitry 206 or 208) to be tested, so that an external device can directly confrol clocking ofthe DAC. Then, via counters and state machines, specific test patterns are applied to the DAC to provide well-controlled tests.
• a LOCK indicator (from each PLL to be tested) is multiplexed to an output pin.
• the clock signal ofthe ring oscillator to be tested e.g., ring oscillator 113 or 213 is multiplexed to an output pin.
• the inventive receiver responds to control signals (including preamble code words) that are transmitted over a serial link (by the transmitter) during control data periods (i.e., other than during data islands and active video periods).
• control signals including preamble code words
• Each such control signal is determined by a bit pattern (e.g., a 10- bit TMDS code word indicative of bits CTLO and CTLl, or bits CTL2 and CTL3).
• the receiver is configured to respond to a control signal only upon detecting a minimum number of repetitions ofthe bit pattern, or upon detecting the bit pattern (or a minimum number of repetitions thereof) in a predetermined region (time window) ofthe confrol data period.
• the inventive receiver implements equalization by filtering the incoming signals on channels CHO, CHI, and CH2 of a TMDS link in a manner designed to compensate for the degradation that the signals suffer during propagation (typically over a long cable) from the transmitter to the receiver.
• a prefened implementation of core processor 214 of receiver 200 of Fig. 14 includes such a filter.
• the equalization filter the filtering effects ofthe cable are analyzed to determine what "cable filter" is effectively applied to the signals by the cable over which they propagate, and an inverted version ofthe cable filter is chosen as the equalization filter.
• the equalization filter compensates for (and preferably cancels) the cable filter.
• Some prefened embodiments ofthe inventive system not only employ an equalization filter in a receiver during recovery of transmitted auxiliary and video data, but they use only "golden words" (the above-described subsets of full sets of code words) to encode the auxiliary data for fransmission, Both techniques are effective to reduce enor rates in the recovered data.
• the inventive receiver includes enor conection circuitry (e.g., circuitry 22 of receiver 200 of Fig. 14), for conecting detected enors in at least the audio data (or other auxiliary data) transmitted during data islands and optionally also for conecting detected enors in video data transmitted during active video periods.
• enor conection circuitry e.g., circuitry 22 of receiver 200 of Fig. 14
• each code word of a packet that is transmitted in a data island can be repeatedly transmitted.
• triple repetition coding in which each code word to be transmitted is fransmitted three times during three consecutive clock cycles
• the packetized data can be recovered reliably, e.g. using the same techniques employed conventionally in the video channels of TMDS links to identify the special out-of-band words that define the blanking intervals.
• Enors occurring during transmission of triple repetition encoded data can typically be conected by implementing a "bubble conection" enor conection scheme in the receiver. Bubble conection is the removal and replacement of a suspect data value (having a single clock width) with reference to its neighboring data values.
• enor conection circuitry in the receiver replaces the center bit (the "N+l” bit) with one ofthe adjacent bits (i.e., the Nth bit) to accomplish bubble conection.
• a sampling point for each sequence of single recovered auxiliary data bits can be determined in essentially the same manner that is performed conventionally by a digital PLL of a TMDS receiver to identify blanking intervals of a TMDS-encoded sfream of video data.
• the circuitry for determining the sampling point for the auxiliary data typically will not need to operate as fast as the digital PLL of a TMDS receiver during identification of blanking intervals of a TMDS-encoded video data stream.
• the inventive receiver includes enor conection circuitry that employs interpolation to conceal detected enors in recovered data. Any of a variety of interpolation techniques can be implemented, including linear interpolation techniques and those which employ higher-order curve-fitting.
• the inventive transmitter determines the temporal placement and duration of each data island transmitted between two active video periods such that the data island does not collide with transmission of other signals (e.g., signals employed for re-keying cipher engines) and such that encryption protocol (e.g., HDCP) requirements are taken into account.
• the transmitter does so according to the following constraints on data island format: a data island must contain at least one packet (thus its minimum duration is 36 pixel clock cycles); each island must contain an integer number of packets; and (in order to assure the reliability of data within a data island) no data island can contain more than 14 packets.
• the fransmitter is configured to determine automatically, from previous lines or frames, where the active video periods and blanking periods of a sequence of video frames (to be encoded for transmission) occur, so that the transmitter will know when these periods will occur in subsequent frames and lines to be transmitted and when other events (such as HDCP re-keying events) will need to occur within the subsequent frames and lines.
• Such automatic detection of blanking periods allows the transmitter to decide whether to initiate a data island (which will require a minimum of 44 pixel clock cycles before the next active video period or next HDCP rekeying or other required event: 32 clock cycles for the packet plus 4 clock cycles for data island guard band words plus 8 pixel clock cycles for video preamble code words, assuming that any required synchronization signals have been transmitted) or to continue a data island with another packet (which would require at least 42 pixel clock cycles: 32 clock cycles for the packet plus 2 pixel clock cycles for trailing guard band words plus 8 pixel clock cycles for video preamble code words).
• the transmitter would typically postpone transmission of a data island (or fransmission of another packet within a data island) if it expects another active video period (or other required event) to begin before transmission ofthe data island (or additional packet) can be completed. If the fransmitter were transmitting a data island when a new line of video data were asserted to the transmitter from a video source, then it would need to ignore the incoming video pixels from the source until it could complete the data island, transmit the mandatory synchronization signals (which requires 12 pixel clock cycles in a control data period, in prefened embodiments), then transmit the video leading guard band, and only then start to transmit encoded video pixels. Of course, it would be undesirable for the transmitter to drop pixels that were intended to be displayed.
• the transmitter is configured to initiate data island transmission based on register indication of blanking interval duration.
• the transmitter includes registers (e.g., registers 105 of Fig. 13) that can store bits indicating when the active video and blanking periods of a sequence of video frames will occur (in subsequent frames and lines to be transmitted) and when other required events (such as HDCP re-keying events) must occur during transmission ofthe subsequent frames and lines.
• registers e.g., registers 105 of Fig. 13
• a host device e.g., the video source
• the transmitter eliminates the need for the transmitter to include circuitry for determining automatically, from previous lines or frames, when the active video and blanking periods and critical events of a sequence of video frames will occur (as in the embodiments described in the previous paragraph).
• the critical values indicated by the bits in the registers need to be tied to some event whose timing is known to the transmitter, for example, to the occunence of HSYNC or assertion (to the transmitter) of a rising edge of Video DE (the start of an active video period) from a video source.
• the transmitter could include registers that store bits indicating that a rising edge of Video DE will occur X pixels after a rising edge of HSYNC (or a rising or falling edge of Video DE).
• a counter in the fransmitter starts to count.
• the fransmitter decides that it wants to send a data island, it checks the counter and determines from the cunent count whether or not there are enough pixel clock cycles between the present time and the next expected rising edge of video DE (or the next other required event) for transmission of a whole data island.
• the circuitry of Figs. 13 A and 13C is implemented in prefened embodiments of the inventive fransmitter to determine whether to insert a data island between two active video periods, determine the temporal placement and duration of each data island to be inserted, and insert each data island in the appropriate time slot.
• the circuitry of Figs. 13A and 13C is coupled to multiplexer 118, core 114, and subsystems 106 and 108 in prefened implementations of fransmitter 100 of Fig. 13.
• the Fig. 13 A circuitry generates a signal ("ok to xmit") indicative of whether enough time remains in the cunent blanking interval for the transmitter to insert a data island (or to add another packet to a data island already inserted) in the blanking interval.
• Each of counters 140 and 142 and logic units 141 and 143 of Fig. 13A receives the pixel clock and the data enable signal "DE" asserted to the transmitter, and a reset signal.
• Fig. 13B is a timing diagram of signals v_blk_inc, ok_to_xmit_h, ok_to_xmit_v, and ok_to_xmit generated by, and signal DE received by, the Fig. 13 A circuitry during operation.
• unit 141 During each vertical blanking interval, unit 141 asserts timing signal "v_blank_inc" (a sequence of pulses whose rising edges occur once per line ofthe input video asserted to the transmitter, in phase with the rising edges of DE) to counters 140 and 142.
• v_blank_inc a sequence of pulses whose rising edges occur once per line ofthe input video asserted to the transmitter, in phase with the rising edges of DE
• counter 142 During each vertical blanking interval, counter 142 generates a count (“v_blk_count[5:0]) indicative ofthe number of pixel clock cycles elapsed since the start ofthe vertical blanking interval, in response to signals "v_blank_inc," DE, and the pixel clock ("elk”).
• Counter 140 generates a count (“h_count[ll:0]) indicative ofthe number of pixel clock cycles elapsed since the last rising edge of DE or the last rising edge of "v_blank_inc,” whichever edge occurs later.
• counter 140 continues to generate the h_count[l 1 :0]) count even during vertical blanking intervals.
• Logic unit 143 receives the "v_blk_count[5:0]” value from counter 142, and a value “v_blank_lth[5:0]” (typically from a configuration register in the transmitter) that indicates the predetermined length of each vertical blanking interval ofthe input video asserted to the fransmitter. In response, unit 143 generates the output signal "ok_to_xmit_v.” The rising edge of output signal "ok_to_xmit_v” coincides with an edge ofthe first pulse of "v blank inc” received in a vertical blanking interval, and the falling edge of "ok_to_xmit_v” occurs when there is no longer enough time remaining in the vertical blanking interval to insert a data island therein.
• Logic unit 141 receives the "h_count” value from counter 140, and a value “h_total[l 1 :0]" (typically from a configuration register in the fransmitter) that indicates the predetermined length of each horizontal line ofthe input video asserted to the transmitter. In response, unit 141 generates the output signal "ok o xmit h.”
• the output signal "ok_to_xmit_h” has a rising edge when a predetermined number of pixel clock cycles have elapsed since the end ofthe last horizontal blanking interval (or the last pulse of "v_blank_inc”) and a falling edge when there is no longer enough time remaining in a horizontal blanking interval to insert a data island therein.
• OR gate 144 receives the signal "ok_to_xmit_h,” the other input of OR gate 144 receives the signal “ok_to_xmit_v,” and OR gate 144 asserts the signal "ok_to_xmit” at its output.
• Logic unit 151 ofthe Fig. 13C circuitry receives the signal “ok_to_xmit” from OR gate 144. Logic unit 151 also receives the output of VSWOO ("VSYNC window of opportunity") generator 150, and signals "packet_req” (which indicates whether the transmitter is ready to send a packet in a data island), “HDMI_mode” (which indicates whether the transmitter is operating in a mode in which it asserts packets in data islands to a serial link), and “hdcp_enc_en” (which indicates whether the transmitter's HDCP cipher engine is enabled). Logic unit 151 is configured to generate the timing and control signals required for inserting a data island between active video periods and transmitting each guard band and packet to be transmitter in the data island.
• VSWOO VSYNC window of opportunity
• unit 151 outputs the following signals: vidDEnoGB and audDEnoGB (which are asserted to the HDCP cipher engine to indicate, respectively, a portion of an active video period containing no video guard band and a portion of a data island containing no data island guard band), vidDE and audDE(which indicate respectively an active video period and a data island), DE (vidDE logically "Ored” with audDE), hdcp_vs (indicating the time interval in which the transmitter can transmit code words indicative ofthe above-described HDCP encryption indicator on channels CHI and Ch2), data_sel[2:0] (which indicates whether video data, a video guard band, a leading or trailing data island guard band, or packetized data island data are to be transmitted at times other than during control data periods), confrol [3:0] (which indicates values of CTLO, CTLl, CTL2, and CTL3 to be encoded for transmission during control data periods), and load_pkt (which indicates that data
• Unit 151 operates in response to a count (cnt) from general purpose counter 152, and asserts control bits (ld_cnt) to counter 152 that determine the value to which counter 152 should count.
• Counter 152 is coupled to receive these confrol bits, the pixel clock, and a reset signal.
• VSWOO generator 150 receives the data enable (DE), pixel clock ("elk”), and VSYNC signals input to the transmitter, and a reset signal, and generates signals indicative ofthe above-described window of opportunity (“WOO”) having predetermined duration following each active edge of VSYNC, in which the fransmitter should transmit the above-described HDCP encryption indicator.
• generator 150 is configured to set the duration ofthe WOO to be 72 pixel clock cycles, and generator 150 outputs the following signals: vswoo (which is high during the WOO), vswoo_start_p and vswoo_end_p (which indicate the beginning and end ofthe WOO), vswoo_48_p (which indicates the 48 th pixel clock cycle after the end of assertion of a VSYNC pulse to the fransmitter), vsync_end_p (which indicates the end of assertion of a VSYNC pulse to the transmitter), and de_start_p and de_end_p (which indicate the beginning and end of each blanking interval).
• vswoo which is high during the WOO
• vswoo_start_p and vswoo_end_p which indicate the beginning and end ofthe WOO
• vswoo_48_p which indicates the 48 th pixel clock cycle after the end of assertion of a VSYNC pulse to the fransmitter
• the timing signals asserted by unit 151 cause transmission of each data island preamble as of (and not earlier than) 48 pixel clock cycles after each falling edge of vidDEnoGB.
• prefened embodiments ofthe inventive transmitter send a burst of eight identical data island preamble code words just before each data island and a burst of eight identical video preamble code words just before each active video period, over each of channels CHI and CH2 of a TMDS link.
• the transmitter sends default (“idle") control words over each of channels CHI and CH2 during control data periods prior to assertion of each burst of data island preamble code words.
• a pixel clock rate change will frequently cause the inventive system to lose synchronization between the transmitter and receiver.
• the encryption and decryption circuitry in the transmitter and receiver can respond to a pixel clock rate change without losing synchronization. This is preferably accomplished by sending a warning (typically indicated by a control bit) to the fransmitter.
• the transmitter enters a mute state, includes a receiver warning (typically indicated by one or more mute state control bits) in a packet, and transmits the packet to the receiver during a data island.
• the receiver warning the receiver enters a mute state.
• the fransmitter In response to the warning, the fransmitter typically sets a flag (an "AVMUTE” flag). In typical implementations of transmitter 100 of Fig. 13, the warning would be received at interface 101 and would cause the fransmitter to set an "AVMUTE” flag in registers 103 and/or registers 105. In response to the receiver warning, the receiver typically also sets a flag (an "AVMUTE” flag). In typical implementations of receiver 200 of Fig. 14, the receiver warning causes the receiver to set an AVMUTE flag in registers 203 and/or registers 205. In response to the AVMUTE flag in the transmitter, the encryption circuitry (e.g., cipher engine 104 in transmitter 100 of Fig.
• the encryption circuitry e.g., cipher engine 104 in transmitter 100 of Fig.
• the decryption circuitry (e.g., cipher engine 204 in receiver 200 of Fig. 14) transitions to an idle state after completing processing ofthe cunent frame of video.
• the fransmitter enters the mute state when the encryption circuitry transitions to its idle state, and the receiver enters the mute state when the decryption circuitry transitions to its idle state.
• operation of the encryption circuitry freezes (and enters an idle state) at the end of a window of opportunity (the above-described "WOO") following the active edge of VSYNC in the next vertical blanking interval that occurs after setting ofthe AVMUTE flag.
• the transmitter enters the mute state and stops transmitting useful auxiliary data (e.g., audio data) in packets and stops sending useful video information (e.g., the transmitter's video and audio output does not change in response to the input audio and video asserted thereto).
• the transmitter in the mute state then sends the receiver warning to the receiver during a data island (e.g., as one or more confrol bits in a packet sent during a data island when VSYNC is inactive, and preferably within one HSYNC period of a previous active video period).
• a data island e.g., as one or more confrol bits in a packet sent during a data island when VSYNC is inactive, and preferably within one HSYNC period of a previous active video period.
• operation ofthe decryption circuitry freezes (and enters an idle state) at the end of a window of opportunity (the above-described "WOO") following the active edge of VSYNC in the next vertical blanking interval that occurs after setting ofthe AVMUTE flag in the receiver.
• the receiver enters the mute state and stops asserting useful auxiliary data (e.g., audio data extracted from packets) at its outputs and stops asserting useful video information at its outputs.
• the receiver in its mute state will prevent any useful information from being seen, heard, or re-transmitted from any of its outputs.
• useful information e.g., video and/or audio
• the encryption circuitry in the transmitter and the decryption circuitry in the receiver have entered their idle states, a pixel clock rate change can occur. After the pixel clock settles and any PLLs in the receiver have settled, a message is preferably sent permitting the transmitter and receiver to leave the mute state (in response to a "mute off command), thus resuming encryption and decryption and "unmuting" the audio and video.
• the pixel clock rate change preferably occurs while the receiver is in the mute state, and the output ofthe receiver is muted during the mute state. Exit from the mute state is preferably accomplished by sending a "mute off signal (typically indicated by a control bit) to the transmitter.
• the transmitter leaves the mute state, includes "receiver mute off data (typically indicated by one or more mute state control bits) in a packet, and transmits the packet to the receiver during a data island.
• the receiver exits the mute state.
• the mute off signal causes the transmitter to clear the above- mentioned AVMUTE flag and the "receiver mute off data causes the receiver to clear its AVMUTE flag.
• the encryption circuitry e.g., cipher engine 104 in the transmitter is free to leave the idle state in response to an encryption indicator that may be asserted thereto following the active edge of VSYNC in the next vertical blanking interval occurring after the AVMUTE flag is cleared (the transmitter leaves the mute state upon receiving the encryption indicator).
• the decryption circuitry e.g., cipher engine 204 in the receiver is free to leave the idle state in response to an encryption indicator (e.g., the above-mentioned HDCP encryption indicator) that may be received in a window of opportunity following the active edge of VSYNC in the next vertical blanking interval that occurs after the AVMUTE flag is cleared.
• the receiver leaves the mute state upon receiving the encryption indicator.
• the invention is a multi-chip module (“MCM") that includes at least a first chip that includes most or all ofthe required circuitry for recovering and processing transmitted data (e.g., receiver 2' of Fig. 2, or receiver 2' of Fig.
• the inventive transmitter includes a PROM
• the EIA/CEA-86 IB standard defines an InfoPacket data structure which consists of one or more InfoFrames containing format infonnation and related information regarding digital audio and/or digital video data.
• the InfoFrame data stored in the PROM would preferably be indicative of such format infonnation and related information.
• the PROM is implemented as a separate chip in an MCM, and the MCM also includes at least one other chip that implements most or all ofthe required circuitry for encoding and transmitting data (e.g., circuitry 1' of Fig. 2).
• Such a separate PROM chip could be programmed at the factory with the InfoFrame data as well as HDCP or other encryption key data, and then securely embedded in the MCOM.
• the inventive transmitter or receiver is an MCM that includes a chip including a ROM that has selectable host micro-accessible regions (e.g., regions accessible by a host device during test mode operation, for example to read test results or configure the fransmitter or receiver for testing), and also non-host accessible regions (that can be used, for example, to securely store keys and identification data used for encryption or decryption).
• EEPROM 14 of Fig. 2 can be implemented as such a chip
• transmitter 1 ' of Fig. 2 can be implemented as another chip, and both chips can be included in an MCM.
• the non-host accessible regions of such an integrated circuit implementation of EEPROM 14 can store HDCP keys and/or other encryption key data (and optionally also other data) that are accessible only by the integrated circuit implementation of transmitter 1 ' and not by any device other than transmitter 1'.
• the inventive transmitter or receiver is implemented as an integrated circuit that includes at least one PROM for storing keys and identification data used for encryption or decryption (e.g., HDCP keys) and/or Enhanced Display Identification Data.
• a PROM of such an integrated circuit implementation ofthe inventive fransmitter could be programmed at the factory with HDCP or other encryption keys.
• a PROM of such an integrated circuit implementation ofthe inventive receiver could be programmed at the factory with HDCP or other decryption keys and/or with EDID data indicative ofthe receiver's configuration and/or capabilities.
• the inventive fransmitter is configured to send packets (in data islands) in any of a variety of formats, preferably including the above- mentioned InfoFrame packet format and information packets having other formats.
• a receiver is configured to receive packets in a special format, it is preferably configured to store data indicative of such format (e.g., in an EDID ROM) and the fransmitter is preferably configured to access this data (e.g., via a DDC channel of a TMDS link) and respond thereto by sending packets to the receiver in such format.
• the receiver can otherwise signal the transmitter to define the format to be used by the transmitter to send packets to the receiver. Implementing the transmitter in this way allows for future definitions of packet formats (e.g., information packet formats), either for one-time fransmission to a receiver or for repeated transmission to a receiver.
• the transmitter and receiver are configured to use a 2-bit semaphore (or alternatively, some other signaling scheme) to confrol the automatic transmission of packets (e.g., modifiable packets).
• the transmitter and receiver are configured to send two-bit codes to each other to confrol the transmission of InfoFrame packets or information packets having other format.
• One bit is a "send” bit.
• the receiver sends the send bit to the transmitter to cause the fransmitter to transmit or stop transmission of a packet.
• the second bit is a "repeat" bit.
• the transmitter could be configured to respond to a send bit from a receiver by overwriting the "repeat” bit with a value indicating that the fransmitter is sending repeated packets, and then to send a two-bit code including such a "repeat” bit to the receiver.
• the receiver could then respond by asserting another send bit to the transmitter.
• the 2-bit codes are sent over a DDC link (or other bidirectional channel between the transmitter and receiver) to cause packet transmission over one or more other channels (e.g., channels CHO, CHI , and CH2 of a TMDS link).
• the inventive fransmitter and/or the inventive receiver includes register space accessible by an external device (e.g., via a I C link) for storing data extracted from (or to be included in) InfoFrame packets (or information packets having other format), such as audio and/or video format information and related information regarding audio and/or video data fransmitted or to be transmitted.
• an external source can load such data into such registers in the fransmitter (e.g., registers 105 of Fig. 13), and the transmitter can then read the registers to obtain data to be included in packets for fransmission during data islands to the receiver.
• the transmitter is configured with a mapping ofthe register space to specific time slots of an InfoFrame packet (or other information packet), for example by including circuifry configured to read bits (received from an external source and loaded into the registers) from the registers and to insert the bits in specific predetermined time slots of a data sfream being assembled for transmission as an InfoFrame packet.
• the receiver can extract data from packets transmitted to the receiver (e.g., over CHO, CHI, and CH2 of a TMDS link during data islands) and load the exfracted data into accessible registers (e.g., registers 205 of Fig. 14). An external device can then access the registers in the receiver, preferably over an I 2 C link.
• the fransmitter could access the registers in the receiver via the DDC channel of a TMDS link, or a host device could access such registers in the receiver via an I 2 C link other than a DDC channel of a TMDS link.
• the receiver is configured to store unrecognized information packets (received over a TMDS or other serial link), so that a host (e.g., a CPU or other processor coupled to the receiver) can access the stored packets.
• a host e.g., a CPU or other processor coupled to the receiver
• the inventive receiver is configured with hardware that parses received Auxiliary Video information (AVI) InfoFrame packets (or other information packets) received over a TMDS link (or other serial link), and to auto-configure color space conversion (CSC) circuitry, chroma up/down sampling circuitry, and optionally also other video processing circuitry in the receiver in the manner determined by the packets.
• AVI Auxiliary Video information
• CSC color space conversion
• the inventive receiver is configured with hardware that parses received Audio InfoFrame (Al) packets (or other information packets) received over a TMDS link (or other serial link), and to auto-configure audio data processing circuitry in the receiver in the manner determined by the packets.
• Al Audio InfoFrame
• the inventive receiver can respond to the data in the Al packet by appropriately configuring its audio data processing circuitry to process the audio.
• An Al packet can indicate various information such as whether or not the audio is compressed.
• the inventive fransmitter has an input interface (e.g., a prefened implementation of interface 109 of Fig.
• I 2 S is used in a broad sense herein to denote a three-conductor link including a first conductor for data (typically audio data), a second conductor for a bit clock, and a third conductor for a word clock (e.g., a Left/Right word select clock identifying the channel ofthe data being sent over the first conductor).
• the receiver has at least one I 2 S output port and is configured to extract the audio data from the transmitted packets (e.g., Audio Sample Packets), to reformat the extracted data as necessary for assertion over at least one I 2 S link, and to assert the exfracted data in I 2 S format to the at least one I 2 S output port.
• the transmitted packets e.g., Audio Sample Packets
• the inventive fransmitter has an input interface configured to accept PCM input data from an I 2 S link (i.e., non-compressed I 2 S data) and in response thereto to reformat the data for transmission in Audio Sample packets having the same format as used in the above-described prefened embodiments ofthe inventive system to transmit S/PDIF audio data.
• the transmitter employs register programmed Channel Status bits indicative ofthe input data format (from the input data source) to perform the necessary reformatting.
• the transmitter can sample the input data on the selected edge ofthe I 2 S data bit clock and insert the samples (via a FIFO) into the time slots of a subpacket for a "Left" channel S/PDIF sample, and then in the next cycle ofthe I 2 S data word clock, continue to sample the input data on the selected edge ofthe I 2 S data bit clock but insert the samples (via the FIFO) into the time slots ofthe subpacket for a "Right" channel S/PDIF sample.
• appropriate channel status, valid, user data, and parity bits would preferably be inserted (via the FIFO) into the appropriate time slots ofthe subpacket, and a start bit would be inserted (via the FIFO) to the packet header with appropriate timing.
• the inventive transmitter has an input interface capable of accepting input audio data from one to four I 2 S links.
• the transmitter is capable of accepting input audio data from a multichannel system on one, two, three, or four I 2 S links, in which the I 2 S links all use the same bit and word clocks (so that all the information in the three, six, nine, or twelve input I 2 S conductors is included in three, four, five, or six ofthe input I 2 S conductors: one bit clock conductor; one data clock conductor, and one, two, three, or four data conductors) and formatting the data for transmission as packets (preferably as Audio Sample Packets having the above-described format).
• the fransmitter has an input interface capable of accepting input audio data from one to four I 2 S inputs, even when each I 2 S input has an independent clock (so that the bits received on all ofthe input I S conductors must be packetized), and formatting the data for transmission as packets (preferably as Audio Sample Packets having the above- described format).
• the inventive transmitter includes multiplexer circuitry (e.g., in the data capture block of interface 109 of Fig. 13) that can reconfigure the stream/channel mapping of input audio data to fit within constraints imposed by available Channel/Speaker allocations for packetizing the data for transmission.
• multiplexer circuitry e.g., in the data capture block of interface 109 of Fig. 13
• the inventive transmitter can accept input audio data (e.g., input audio from one or more I S links) in any of a set of formats (each with a unique stream channel mapping) and reformat the data into one of different set of allowed formats for fransmission as packets (preferably as Audio Sample Packets having the above-described format).
• a stream of S/PDIF data has its own clock (embedded among the data samples).
• the inventive fransmitter is configured to sample an input stream of S/PDIF data (typically, S/PDIF audio data) with a pixel clock having no known phase relationship with the clock embedded among the input data, and then to transmit the pixel-clock-sampled data.
• the transmitter is configured to sample an input sfream of S/PDIF data with an audio reference clock (sometimes refened to herein as "MCLK" or master clock) having frequency of at least 256*Fs (or 384*Fs in prefened embodiments), where Fs is the sample rate ofthe S/PDIF data, where there is no known phase relationship with the clock embedded among the input data.
• the inventive transmitter is configured to accept and synchronize multiple independent, synchronizable S/PDIF streams.
• the inventive fransmitter is configured to translate a flatline S/PDIF input (consisting entirely of zeroes) to a format for efficient transmission by packets.
• the fransmitter is configured to set four bits of a packet header to indicate whether a conesponding subpacket includes a "flatline" sample (i.e., a sample consisting entirely of zeroes).
• the transmitter is configured to set the four LSBs ofthe third byte ofthe packet header (as described above) to indicate whether a conesponding subpacket includes a "flatline” sample, in which case the receiver is preferably configured to treat a "1" value of any of these LSBs as valid only if a conesponding "present” bit ofthe second byte (one ofthe four LSBs ofthe second byte) ofthe header indicates that the conesponding subpacket contains a sample.
• the inventive transmitter is configured to accept DSD (Direct Stream Digital) input data and format this input data for transmission in IEC60958-type packets during data islands.
• Audio data in the conventional DSD (Direct Sfream Digital) format e.g., SACD or "Super Audio CD” data
• SACD Direct Sfream Digital
• SCCD Super Audio CD
• the receiver has at least one DSD output port and is configured to extract the audio data from the transmitted packets (e.g., Audio Sample Packets), to reformat the extracted data as necessary for assertion as a DSD output stream, and to assert the exfracted data in DSD format to the at least one DSD output port.
• the transmitted packets e.g., Audio Sample Packets
• the exfracted data in DSD format to the at least one DSD output port.
• the inventive transmitter has an input interface configured to accept DSD input data and in response thereto to reformat the data for transmission in Audio Sample packets having the same format as used in the above-described prefened embodiments ofthe inventive system to transmit S/PDIF audio data.
• the transmitter employs register programmed Channel Status bits indicative ofthe input data format (from the input data source) to perform the necessary reformatting.
• the inventive fransmitter has an input interface capable of accepting multiple streams of DSD format input audio.
• the fransmitter has an input interface capable of accepting DSD format input audio data from multiple DSD links, even when each DSD stream has an independent clock (so that the bits received on all ofthe input DSD conductors must be packetized), and formatting the data for transmission as packets (preferably as Audio Sample Packets having the above-described format).
• the fransmitter is configured to accept input audio data in I S format (on three-conductors: one conductor for a bit clock; one for a L/R clock; and another for data) as described above, it preferably also is configured to accept and transmit audio data in DSD format at the same 3-conductor input port.
• I S format on three-conductors: one conductor for a bit clock; one for a L/R clock; and another for data
• Such a fransmitter would preferably include multiplexers that map the three input streams appropriately to the circuitry for formatting the input data for transmission in Audio Sample Packets having the above-described format.
• the inventive receiver has at least one I 2 S output interface, and is configured to recover PCM data from Audio Sample packets and translate the recovered data to I 2 S format (e.g., using register programmed channel status bits that are transmitted in an Audio InfoFrame (Al) packet sent by the transmitter to the receiver with the Audio Sample Packets).
• the receiver is configured to output audio data
• I S recovered from Audio Sample packets
• I S format on from one to four I S outputs (each output configured for connection to a 3-conductor I 2 S link);
• the inventive receiver is configured to output audio data (recovered from Audio Sample packets) in I S format on from one to four I S outputs (including by asserting different I 2 S data streams, having different and independent clocks, to different outputs).
• the inventive receiver includes multiplexer circuitry that can reconfigure the stream/channel mapping of audio data extracted from packets in a format within constraints imposed by available Channel/Speaker allocations for packetizing the data.
• a receiver can output audio data (e.g., on one or more I S links) in any of a set of formats (each with a unique stream/channel mapping) after having exfracted the data from packets (preferably Audio Sample Packets having the above-described format) and reformatted the exfracted data from one of different set of allowed formats in which it was packetized.
• the inventive receiver is configured to output S/PDIF format data.
• the receiver is configured to translate packetized codes indicative of flatline S/PDIF data (consisting entirely of zeroes) into flatline S/PDIF format output data.
• the receiver is configured to recognize four bits of a packet header as being indicative of whether a conesponding subpacket includes a "flatline" sample (i.e., a sample consisting entirely of zeroes).
• the receiver is configured to recognize the four LSBs ofthe third byte ofthe packet header (as described above) as being indicative of whether a conesponding subpacket includes a "flatline” sample, and to treat a "1" value of any of these LSBs as valid only if a conesponding "present” bit ofthe second byte (one ofthe four LSBs ofthe second byte) ofthe header indicates that the conesponding subpacket contains a sample.
• the inventive receiver is configured to output data in DSD format in response to audio sample packets including an encoded version ofthe DSD data.
• the receiver is configured to output multiple DSD data streams with independent clocks, in response to Audio Sample Packets that include an encoded version ofthe DSD data.
• the receiver is configured to output audio data in I 2 S format at an output port (for connection to a three-conductor link comprising one conductor for a bit clock; one for a L/R clock; and another for data) in response to Audio Sample packets, it preferably also is configured to output audio data in DSD format at the same output port (e.g., the receiver includes multiplexers that map the three input streams recovered from the Audio Sample Packets to appropriate conductors of the output port).
• ACR Audio Clock Regeneration
• prefened embodiments ofthe inventive system transmit time stamps ("CTS" values) indicative ofthe audio clock (for audio data being transmitted in data islands over a TMDS or other serial link) and preferably transmit "audio clock regeneration” packets that contain both the "CTS” values and "N” values used for audio clock regeneration.
• the "N” value is a "numerator” value, indicative ofthe numerator of a fraction by which the receiver is to multiply the frequency ofthe pixel clock to regenerate the audio clock.
• Fig. 15 is a block diagram ofthe overall system architecture model used by prefened embodiments ofthe inventive system for audio clock regeneration.
• the transmitter determines the fractional relationship between the pixel clock (the video clock) and an audio reference clock (known as "MCLK" or master clock, where MCLK has frequency equal to Z*Fs, where Fs is the audio sample rate, and Z is typically equal to Y*128, wherein Y is a small integer such as 1, 2, or 3).
• the transmitter passes the numerator (“N”) and denominator (“CTS”) for this fraction, and the pixel clock, to the receiver across the serial link (via the transmitter's formatting, encoding, and transmitting circuitry 138).
• N numerator
• CTS denominator
• Receiving and decoding circuitry 139 in the receiver then recovers the pixel clock and data indicative ofthe N and CTS values, and the receiver recreates the MCLK clock from the pixel clock (whose frequency is denoted
• Clock multiplier 133 is typically implemented as a PLL including a phase detector 134, low-pass filter 135, voltage-controlled oscillator 136, and frequency divider 137, connected as shown in Fig. 16.
• the fransmitter determines the value ofthe numerator "N" (which is typically input thereto by the input data source) and stores data indicative of "N" in register 129, as shown in Fig. 15.
• the value N is used in clock divider 130, which also receives the MCLK (having frequency Z*Fs) to generate an intermediate clock that is slower than the MCLK clock by the factor N.
• the pixel clock and intermediate clock are asserted to counter 131.
• Counter 131 generates a CTS (Cycle Time Stamp) count by counting the number of pixel clock cycles that have elapsed (since assertion ofthe last CTS value) during each cycle ofthe intermediate clock.
• Data indicative ofthe values of N and CTS are transmitted to the receiver (preferably in each Audio Clock Regeneration Packet transmitted by formatting, encoding, and transmitting circuitry 138 over a TMDS link during a data island), and the pixel clock is asserted to the receiver.
• the pixel clock is preferably transmitted (e.g., by circuitry 138 of Fig. 15) to the receiver on the clock channel of a TMDS link (e.g., channel CHC of Fig. 2).
• the successive CTS values will quickly converge to a constant value. If the MCLK and pixel clocks are asynchronous, or there is some amount of jitter between them, then the successive CTS values will typically alternate between two or possibly three different values.
• the inventive transmitter is configured to accept an MCLK having frequency equal to any of several multiples of 128*Fs (not just an
• the inventive transmitter is configured to accept an audio clock that is based on a "divided down" version ofthe audio clock (whose frequency is the greatest common factor of all the possible audio sample frequencies).
• the frequency of such audio clock is Z*Fs/M, where Z and M are integers, Fs is the audio sample frequency, and Fs/M is the greatest common factor of all the possible audio sample frequencies.
• the receiver is configured to regenerate the "divided down" version ofthe audio clock from fransmitted time stamps, and preferably also to generate the original audio clock from the regenerated clock by frequency multiplying the regenerated clock by an appropriate factor. This factor is communicated to the receiver, preferably in an "audio clock regeneration" packet.
• the data rates are multiples of 32kHz or 48kHz.
• a PLL designed to support both of these frequencies requires a reference frequency range from 32kHz to 48kHz. If the data are transmitted using a divided-down audio clock (whose frequency is the greatest common factor, 16kHz), the clock regeneration PLL can be implemented so that the reference frequency is 16kHz (although the receiver would also need to be configured to include frequency multiplication circuitry, with multiplication factors of 2 and 3, to regenerate any ofthe original audio clocks.
• the receiver is implemented with a frequency divider (an implementation of circuit 132 of Fig.
• the transmitter is configured to accept an audio clock that is based on a "divided down" version of an input audio clock (whose frequency, B, is the greatest common factor of all the possible input audio clock frequencies), and to assert time stamps (CTS values), numerator values "N” and “L,” denominator values "D,” and a pixel clock to the receiver in response to such audio clock.
• the receiver is implemented with a frequency divider (an implementation of circuit 132 of Fig. 15) that regenerates an intermediate clock (having the frequency B/N) from the time stamps and pixel clock, and regeneration circuit (an implementation of circuit 133 of Fig.
• the fransmitter does not assert the "L” values and the regeneration circuit in the receiver regenerates only the "divided down” version ofthe input audio clock (having frequency equal to B) from the intermediate clock and the "N" values.
• the transmitter is configured to implement auto-phase adjustment for synchronizing the audio clock (e.g., the MCLK asserted to the transmitter of Fig. 15) and the pixel clock, when these clocks are coherent in the sense that both are derived from a common clock. This ensures that the time stamp values do not vary over time.
• the pixel clock also refened to below as the "video clock”
• audio clock are synchronous or nearly synchronous
• the audio clock information can be sent across the link by counting the number of video clock periods per audio clock period. If the audio and video is synchronized, this number should be a constant.
• prefened embodiments ofthe invention use an oversampling technique at the clock domain "boundary" (the boundary between the audio and video clocks) to automatically select delays which result in maximized setup and hold timing.
• multiple phase-shifted versions ofthe audio clock are generated (e.g., by "oversampling” the audio clock using multiple versions ofthe video clock , such as the video clock itself and the inverse ofthe video clock), and the one ofthe resulting streams of samples that results in the least CTS jitter is selected.
• This can prevent any "CTS jitter” as long as the input audio and video clocks are synchronized and reasonably stable.
• An important benefit of this design is to simplify the system level design (since the invention eliminates the need to worry about the audio to video clock skew).
• the transmitter is implemented to generate two sets of
• CTS values using both the video clock (VCLK) and an inverted video clock (VCLK bar) to count the audio clock edges, and to compare the two sets of CTS values to determine which shows less variation over time.
• the VCLK-based output could be selected (as the CTS values to be transmitted to the receiver) initially, with the option to switch to the VCLK_bar-based output if it shows less CTS value variation over time.
• Other embodiments generate multiple sets of CTS values by employing several delay cells to generate multiphase versions ofthe video clock, and generating a set of CTS values using each different version ofthe video clock.
• the transmitter than monitors the sets of CTS values for transitions and selects the set having the fewest transitions. This method would use little area on a transmitter chip, but would require hand design and layout for the timing critical sections.
• the rate at which the "autoaligning" is done should be kept fairly low (on the time scale of seconds, or only at startup and when the audio or video frequency changes). This is because the audio clock and video clock timing is not expected to vary much in a particular design if coherent audio and video is being sent, unless there is some slow (i.e. temperature related) phase shift between the audio and video clocks. If relative timing ofthe video and audio clocks is changing fairly often (i.e. in the case of noncoherent audio and video), then the CTS value sent across the link should change since that is the purpose ofthe CTS counting method.
• the autoalignment circuit in the inventive transmitter should have selected the "rising edge VLCK" CTS value.
• the transmitter would count the difference between the accumulated CTS values and the "rising and falling edge VCLK" CTS values, and when the autoalignment circuitry switches over to a new VLCK edge, the difference count (between the accumulated count ofthe CTS values sent and the accumulated count based on the new VCLK edge selected) should be sent to the receiver to avoid this enor.
• Another way to reduce amplified jitter in the regenerated audio clock (regenerated using transmitted CTS values) due to jitter in the audio clock (resulting from the above-described phase alignment in the audio and pixel clocks), which is an alternative to above-described autoalignment technique (of selecting different CTS values based on different versions ofthe pixel clock employed to sample the audio clock), is to implement the fransmitter so that it samples the audio clock with a clock having higher frequency than the pixel clock.
• jitter in the original audio clock can cause jitter in the CTS values that can affect the quality ofthe regenerated audio clock, even to a degree such that a small amount of jitter in the original audio clock is amplified (in the sense that the regenerated audio clock has worse quality than the original audio clock).
• a worst-case situation would be if the audio clock edge was close to the video clock edge, but it moved back and forth over the edge.
• a small amount of audio or video clock jitter e.g., one nanosecond jitter
• This CTS jitter could result in a lower quality recovered audio clock.
• Another benefit of more precise measurement ofthe audio clock is to allow the actual jitter in the regenerated audio clock to be tracked more closely. More accurate measurement of the jitter could also reduce the frequency range that the MCLK regeneration circuit needs to cover, and under some circumstances can make that section ofthe receiver design easier to implement.
• a simple oversampling method for producing a finer resolution measurement of the audio clock period is to use both the rising and the falling edge ofthe video clock to measure the audio clock period. If the video clock duty cycle is close to 50%, this would result in measurements with twice the resolution attainable using only the rising (or falling) edges ofthe video clock.
• Other oversampling methods are to use oversampled pixel clocks (which are typically generated for other purposes in a TMDS core processor, such as processor 114 of transmitter 100 of Fig. 13), which can produce lOx or even 3 Ox improvements in the timing resolution, but would require more complex circuitry, and would require use of packets having format adequate to transport these finer resolution measurements across the link. Also, the receiver would need to be configured to use the additional information.
• the inventive fransmitter can be implemented to include a programmable CTS register (rather than circuitry, e.g., elements 130 and 131 ofthe Fig. 15 fransmitter, for producing CTS values).
• An external device loads an updated CTS value into the CTS register from time to time (preferably periodically), and the transmitter is configured to include the current CTS value (in the CTS register) in each Audio Clock Regeneration Packet transmitted to the receiver.
• the audio clock regeneration (“ACR") circuitry is configured to assert (at an output) a regenerated version audio clock (MCLK) having frequency equal to any of several multiples M*Z*Fs of a first frequency Z*Fs (where Z is typically equal to 128), in response to the pixel clock, a time stamp value CTS, and at least one other value from the transmitter (e.g., a single value ofthe above-described numerator value "N,” and a frequency multiplication value "M” asserted with any of several different values from the transmitter).
• the inventive receiver is configured to operate in two modes: a first mode in which it employs internal audio clock regeneration (ACR) circuifry (including a PLL) to regenerate an audio clock (MCLK) and optionally also to generate one or more clocks having frequency equal to multiples of this audio clock; and a second mode in which the internal ACR circuitry is bypassed and instead the CTS and N values and pixel clock received from the transmitter are asserted to an external device (e.g., external PLL unit 160 of Fig. 15) for MCLK clock regeneration.
• ACR internal audio clock regeneration
• MCLK audio clock
• the receiver in the second mode can use the regenerated MCLK from the external device (including by stabilizing it in an internal PLL circuitry having a VCO) and/or pass it through to another device.
• the receiver of Fig. 15 is modified slightly so that it is configured to output the intermediate clock generated by clock divider 132 in the receiver's second mode of operation (this intermediate clock has frequency equal to Z*Fs/N) to drive an external ACR device (typically including a PLL), and to receive the regenerated MCLK from the external ACR device.
• the Fig. 15 receiver can either output the regenerated MCLK that it receives from the external device, or can further process the regenerated MCLK (e.g., by stabilizing it in a PLL implementation of clock multiplier 133, or other internal PLL circuifry having a VCO).
• the external ACR device that provides the regenerated MCLK to the modified Fig. 15 receiver (in its second mode of operation) includes a PLL (e.g., one including a voltage controlled crystal oscillator or "VCXO") that provides a stable recovered MCLK but only within a relatively narrow frequency range.
• a PLL e.g., one including a voltage controlled crystal oscillator or "VCXO" that provides a stable recovered MCLK but only within a relatively narrow frequency range.
• the receiver can be configured to support such an external PLL by performing automatic clock stretching or shrinking (i.e., to generate a reduced or increased frequency version ofthe regenerated MCLK signal from the external PLL when needed), or by automatic sample repetition or dropping (i.e., by using the regenerated MCLK signal from the external PLL to sample audio data recovered from packets, but inserting repeated samples into or omitting some samples from, the recovered audio sample stream that it outputs, to produce an output audio sample stream having a sample rate that differs from the frequency ofthe recovered MCLK signal).
• automatic clock stretching or shrinking i.e., to generate a reduced or increased frequency version ofthe regenerated MCLK signal from the external PLL when needed
• automatic sample repetition or dropping i.e., by using the regenerated MCLK signal from the external PLL to sample audio data recovered from packets, but inserting repeated samples into or omitting some samples from, the recovered audio sample stream that it outputs, to produce an output audio sample stream having a sample rate that
• the inventive receiver employs a "fractional- N" frequency synthesizer PLL for audio clock regeneration.
• a PLL implementation of clock multiplier 133 ofthe Fig. 15 receiver is configured, effectively, to multiply the frequency ofthe clock signal from circuit 132 by a non- integer value of "N.”
• a frequency synthesizer PLL could be implemented by elements 134, 135, 136, and 137 of Fig. 16, where frequency divider 137 is a dual-modulus frequency divider.
• phase detector 134 receives the outputs of frequency dividers 132 and 137 and generates in response a signal indicative of their relative phase
• loop filter 135 receives and low-pass filters the output of phase detector 134 to generate an enor signal indicative ofthe relative phase ofthe outputs of dividers 134 and 137
• VCO 136 asserts an output clock (having frequency very nearly equal to Z*Fs) in response to the enor signal from filter 135.
• N is a non-integer value that satisfies I ⁇ N ⁇ 1+1, where I is an integer
• frequency divider 137 receives the output clock and outputs a clock having frequency Z*Fs/I or Z*Fs/I+l in response thereto.
• the inventive receiver employs circuitry including a nanow-range PLL for audio clock recovery (e.g., a PLL implementation of clock multiplier 133 ofthe Fig. 15 receiver that is stable and accurate only when the frequency ofthe clock input thereto, from frequency divider 132, is within a nanow VCO frequency range).
• a nanow-range PLL for audio clock recovery e.g., a PLL implementation of clock multiplier 133 ofthe Fig. 15 receiver that is stable and accurate only when the frequency ofthe clock input thereto, from frequency divider 132, is within a nanow VCO frequency range.
• the ACR circuitry in such a receiver would multiply the frequency ofthe regenerated clock signal output from the nanow-range PLL by any of several different numbers, where the PLL is capable only of outputting a regenerated clock having frequency in a nanow range, and regenerated clocks having frequency outside this range are or may be needed.
• the frequency multiplication will itself typically require a PLL, so that these embodiments ofthe invention are typically useful only if the combination ofthe nanow-range PLL (for clock regeneration) together with one or more additional PLLs (for frequency multiplication) is simpler or less expensive than a wider-range PLL for clock regeneration.
• the inventive receiver consists of at least two chips.
• a first chip includes most or all ofthe required circuitry except for a PLL employed for audio clock regeneration, and a second chip includes the PLL.
• MCM multi-chip module
• the second chip e.g., circuit 161 of Fig. 16
• the rest of the PLL and other clock recovery circuitry is implemented in the first chip.
• low-pass filter 135 could be omitted and its functions performed by external filter 161, or filter 135 could be included and the receiver configured to operate using either internal filter 135 or external filter 161.
• the inventive receiver includes audio clock regeneration circuitry that uses a stable clock (e.g., a crystal clock) to generate "stable" time stamps (CTSx) from the pixel clock at the same time that the transmitter generates time stamps (CTSv) from the same pixel clock.
• the receiver uses the stable time stamps (CTSx) in place ofthe normally transmitted time stamps (CTSv) from the receiver when appropriate, to reduce enors in the regenerated audio clock (the audio clock regenerated using either the CTSx or the CTSv time stamps) that would otherwise occur due to jitter in the pixel clock.
• Prefened embodiments ofthe inventive system use a time-stamp method to transmit the audio clock (e.g., MCLK) between fransmitter and receiver, in which digital cycle time stamp (CTS) values are determined by counting the number of pixel clock cycles (VCLK cycles) in each period of MCLK and the CTS values are transmitted to the receiver.
• CTS digital cycle time stamp
• VCLK cycles pixel clock cycles
• VCLK cycles pixel clock cycles
• MCLK and VCLK may have independent sources
• the CTS value will, in general, vary with time and therefore introduce jitter into the reconstructed MCLK. While these and other sources of jitter are important and can affect the final audio quality, they can all, with the exception ofthe VCLK jitter, be managed by other techniques (disclosed herein) so as not to degrade the recovered audio quality.
• the problem addressed by the embodiments to be described next is degradation of recovered audio quality due to VCLK jitter. Such degradation is clearly seen in the following situation.
• the MCLK frequency is 1kHz (1ms period) and VCLK has a 0.5kHz (2ms period) jitter component
• the CTS value is measured during the first half cycle ofthe VCLK jitter (when the VCLK period is longer than average, for instance, resulting in a CTS count that is smaller than if the average period were used) and then this smaller-than-average CTS value is transmitted across the link and used to regenerate MCLK during the second half of the VCLK jitter cycle (when the VCLK period is shorter than average).
• Using this smaller-than-average CTS value to count out VCLK periods that are also shorter-than- average will result in a regenerated MCLK period being shorter than the ideal MCLK.
• the MCLK jitter is an amplified (doubled) version ofthe VCLK jitter. This is a result of using VCLK on each side ofthe link. This is the worst case and could potentially degrade the audio quality significantly.
• the problem of transferring VCLK jitter to MCLK occurs when the jitter on VCLK is in one phase during the measurement of CTS and in the other phase during MCLK regeneration.
• the particular time interval between the CTS measurement and the MCLK regeneration is important only to the extent that it affects the maximum frequency of VLCK jitter that can be conected in accordance with the present invention.
• VCLK jitter if the VCLK jitter is changing quickly compared to the time needed to measure, transmit, and use a particular CTS value, then the VLCK jitter cannot be effectively filtered out. So, a longer delay in the CTS path means that instead of filtering out VCLK jitter up to 500 Hz (the best case for a 1kHz CTS clock rate), for example, the filtering can only conect jitter having frequencies much lower than 500 Hz.
• a stable (negligible jitter) time base is used at the receiver, and "CTS" VCLK cycles are counted off in the receiver using the stable time base at the same time that the "CTS" VCLK cycles are counted off in the transmitter to measure the MCLK period.
• the stable time base at the receiver is preferably implemented by using a crystal external to the receiver to provide a stable clock (XCLK).
• the procedure requires that the measurement ofthe number of VCLKs in a MCLK period in the transmitter side ("CTSv”) occurs at the same time that the number of XCLK cycles for the same MCLK period are counted off in the receiver (the latter count is "CTSx"). That is, the counting must start at the same (or nearly same) time on both sides ofthe link. This will require a method of synchronizing the start of counting in the transmitter and receiver.
• the second difficulty is that the receiver needs to count (in XCLK cycles) the same CTSv number of VCLK cycles that is simultaneously being measured in the transmitter, and since the CTSv values can vary from measurement to measurement the receiver will not know exactly what value was measured in the transmitter until the packet containing the CTSv value is sent across the link. This will require a method for the receiver to anticipate the CTSv value currently being measured in the transmitter, such that the counting also stops at the same (or nearly same) time on both sides ofthe link.
• the synchronization will require using a signal that is common to both the fransmitter and receiver (such as VSYNC) as an origin from which to define the interval (such as some number of VCLK cycles) at which time the counting should start.
• a signal that is common to both the fransmitter and receiver such as VSYNC
• the CTSx counting in the receiver which may have started earlier, can be delayed by the appropriate amount to synchronize the counting in the transmitter and receiver. Any resyncing process should be able to deal with the situation that the MCLK and video signal sources are independent.
• the method by which the receiver anticipates the CTSv value cunently being measured in the transmitter, such that the counting also stops at the same (or nearly same) time on both sides ofthe link can exploit the fact that the CTS values are typically not expected to vary widely over time (typically, the CTS values are expected to vary from a typical value by as little as plus or minus one). This then allows the receiver to count and store CTSx values for a small range of CTSv values, and then when the actual CTSv value is sent across the link, the receiver merely uses the actual CTSv value as a pointer to the conect value of CTSx.
• This method requires that the CTSx values be buffered to the extent ofthe worst case time delay ofthe CTSv packet being sent across the link before it can be used to regenerate a period of MCLK. This is typically feasible since the stable time base allows such latency without introducing jitter.
• the timing diagram of Fig. 23 shows one example of how the transmitter and receiver counting can be resynchronized.
• the value ofthe CTSv counter in the fransmitter is captured at the leading edge of VSYNC (this value is identified as "CTS_sync” in Fig. 23) and sent across the link with high priority.
• CTS_sync the value of CTSv for the conesponding MCLK period
• the delay between the transmitter counting and the receiver counting can be determined if the receiver counting initially started at the leading edge of VSYNC. Then the counting in the receiver is approximately synced to the transmitter counting, at which point the VCLK jitter will no longer affect the regenerated MCLK (because the receiver will use the appropriate CTSx values for MCLK regeneration).
• a time stamp for a reference clock (generated from MCLK) is sent across the link, instead of a timestamp for MCLK itself.
• the regenerated reference clock is then used to reproduce MCLK.
• latency ofthe synchronizing signal between the transmitter and receiver is handled carefully. This latency could introduce a fixed offset between the fransmitter and receiver counting since the receiver will actually sync to a delayed version ofthe VSYNC signal asserted to the fransmitter. This can be acceptable if this offset is small, but the maximum latency would need to be confrolled.
• Another option is to send the transmitter latency across the link in a packet so that the receiver can use it to determine how to make the appropriate adjustment.
• the CTS sync value is sent once per video frame (assuming it is based on VSYNC). This allows the receiver to make small conections on a frame basis that would prevent long term drift ofthe transmitter- receiver counting synchronization.
• the receiver asserts an interrupt to the host when appropriate to indicate to the host that it needs to have the transmitter send a CTS_sync value in order to synchronize counting.
• VCLK jitter is not excessively severe, it is possible in some cases to use the receiver without a crystal clock in a standard mode of operation in which VCLK jitter is not addressed (in accordance with the invention) for lower quality audio applications, while for higher audio quality applications the external crystal is provided and employed as described above to remove the effect of VCLK jitter.
• the described technique for preventing the transfer of VCLK jitter to the regenerated MCLK can allow the support of spread spectrum clocking on the link (for which the frequency ofthe VCLK is purposely varied in order to reduce EMI).
• the implementation in the receiver would need to support the worst case frequency variation.
• An alternative method to storing many values of CTSx and then using the value of CTSv to select the conect stored value is to implement the receiver to measure only the CTSx value conesponding to the typical CTSv value, then calculate the conection to CTSx when the CTSv value is available.
• the latter method is expected to be useful in situations where the CTSv variation is much larger than +/-1. This may be the case when using spread spectrum clocking on the link. Since there would be many clock cycles to do the calculation, a hardware implementation would likely not require a large area penalty. Synchronizing and adjusting the start of counting in the receiver would still be required.
• Another class of prefened embodiments implements digital filtering (low pass filtering) ofthe CTS values sent to the receiver.
• clock regeneration e.g., audio clock regeneration
• transmitted CTS values a time-stamp methodology
• the actual CTS values that are transmitted can vary from their average value (averaged over a short period) due to quantization enor as well as pixel clock jitter (jitter in the clock employed for sampling the clock to be regenerated). This variation produces jitter on the regenerated clock.
• the CTS values are digitally filtered to reduce the regenerated clock jitter caused by quantization enor and pixel clock jitter.
• the digital filter functions as a low-pass filter to suppress the higher frequency component ofthe time-stamp variation, and conespondingly suppress the higher frequency component ofthe regenerated clock jitter.
• the required word length for the filter output can be larger than the original time-stamped value.
• the digitally filtered time-stamped value will have a fractional part in comparison to the original integer time-stamp (i.e. 74.25 instead of 74). If the longer time-stamp words are simply truncated to an integer, the filter will still function to attenuate the higher frequency jitter. However, the average ofthe truncated time-stamp values may not equal the average ofthe original sequence of time-stamp values, such that the regenerated clock frequency would not be accurate and the clock would have long term drift relative to the original clock.
• a fractional-N type PLL is employed in the receiver (as described above) to preserve the longer words.
• One important application of this class of embodiments is the transmission of a digital audio clock using the time-stamp method.
• jitter produced by time-stamp variation may be modulated into the final analog audio causing a degradation in audio quality.
• a digital filter By using a digital filter to suppress jitter frequencies in the audio band, the effect ofthe jitter on the audio quality can be reduced or eliminated.
• PLL's e.g., PLLs with bandwidths of on the order of 100 Hz
• a digital method once properly designed, is expected to be very straightforward to manufacture and test, and easy to apply on a board level design.
• the inventive fransmitter is configured to accept input video in ITU- BT.656 or YCbCr 4:2:2 format, and to convert input video in ITU- BT.656 format to YCbCr 4:2:2 format for transmission.
• the transmitter can accept (on one channel) two streams (time-division-multiplexed together) of ITU- BT.656 or YCbCr 4:2:2 video (each stream comprising N-bit components), and convert the input video to YCbCr 4:2:2 format (comprising 2N-bit components, each 2N-bit component indicative of two components ofthe input video) for fransmission.
• the inventive transmitter is configured to accept input video in DDR YCbCr 4:4:4 format (DDR denotes "double data rate.”
• DDR format video can be written to a DDR RAM, a type of synchronous DRAM, on falling clock edges as well as rising clock edges).
• the fransmitter can put such input video into conventional (non-DDR) YCbCr 4:4:4 format for fransmission.
• the transmitter is configured to accept input video in DDR YCbCr 4:2:2 format and put the input video into conventional YCbCr 4:2:2 format for transmission.
• the fransmitter can accept (on one channel) two streams of input video (time-division-multiplexed together) in DDR ITU- BT.656 format and place the input video in conventional YCbCr 4:2:2 format for fransmission.
• the inventive fransmitter is configured to accept input video in any of a variety of YCbCr formats, with sync bits embedded with the data or with separate HSYNC and VSYNC (to allow support for a large number of source devices), and is configured to reformat the input video into YCbCr 4:4:4 or YCbCr 4:2:2 format video for fransmission.
• the inventive fransmitter is configured to accept input video in YCbCr 4:2:2 format and includes upconversion circuitry for converting the input data to data in YCbCr 4:4:4 format for fransmission.
• the inventive fransmitter is configured to accept input video in any of a variety of YCbCr formats and includes YCbCr to RGB color space conversion circuitry (with programmable or fixed coefficients) and universal mapping circuitry for converting the input video to RGB format for transmission.
• the inventive receiver is configured to output a stream of video in ITU- BT.656 format and includes multiplexer circuitry for placing the Y, Cr, and Cb components of recovered video in YCbCr 4:2:2 format into the ITU- BT.656 fonnat.
• the receiver can assert, with a stream of ITU- BT.656 output video, separate HSYNC and VSYNC signals (whose timing matches that ofthe sync bits embedded in the ITU- BT.656 sfream).
• the receiver is configured to output (on one channel) two streams (time-division-multiplexed together) of ITU- BT.656 video (each sfream comprising N-bit components) in response to recovered YCbCr 4:2:2 video (comprising 2N-bit components, each 2N-bit component indicative of two components ofthe output video).
• the receiver is configured to put recovered video in YCbCr 4:2:2 format into a desired output format that can include sync bits embedded in a stream of output Y, Cr, and Cb components (for example, the ITU-BT.656 format), and/or HSYNC and VSYNC output separately from a stream of output Y, Cr, and Cb components.
• a desired output format that can include sync bits embedded in a stream of output Y, Cr, and Cb components (for example, the ITU-BT.656 format), and/or HSYNC and VSYNC output separately from a stream of output Y, Cr, and Cb components.
• the inventive receiver is configured to output video in DDR YCbCr 4:4:4, DDR YCbCr 4:2:2, or DDR ITU- BT.656 format (DDR denotes "double data rate") and includes circuitry for reformatting recovered video in conventional (non-DDR) YCbCr 4:4:4 or YCbCr 4:2:2 format into the DDR YCbCr 4:4:4, DDR YCbCr 4:2:2, or DDR ITU- BT.656 output format.
• DDR denotes "double data rate
• the inventive receiver is configured to output video in any of a variety of YCbCr formats and includes universal mapping circuifry that converts recovered video in conventional YCbCr 4:4:4 or YCbCr 4:2:2 format into any desired one ofthe output formats.
• the inventive receiver includes circuifry for performing downconversion or upconversion on recovered video (e.g., to convert recovered video in YCbCr 4:4:4 format to output video in YCbCr 4:2:2 format or recovered video in YCbCr 4:2:2 format to output video in YCbCr 4:4:4 format).
• the inventive receiver is configured to output video in any of a variety of YCbCr formats and includes RGB to YCbCr color space conversion circuitry (with programmable or fixed coefficients) and universal mapping circuifry for converting recovered video in RGB format into any desired one ofthe output formats.
• One verification method is to periodically read a "result" vector from each side ofthe link. If the vector is changing, and doing so in the expected sequence (usually with the same value at each side ofthe link), then this is a strong indication that the system is working properly.
• the conventional HDCP protocol employs such a method, known as a "link integrity check.”
• a link integrity check is not a foolproof method in all cases, however. Such a method by itself cannot sample both sides ofthe link at exactly the same moment in time. Further, the sampling rate may be different than the rate at which the result vectors change (indeed, the two events may be entirely asynchronous). These factors lead to anomalies in which the link may appear to be broken, when it is not. Even worse, the link may appear fine, when it is in fact broken.
• Prefened embodiments ofthe inventive system implement an improved method for link integrity checking. While the following discussion assumes that the HDCP content protection protocol being used, some embodiments ofthe inventive system that embody a link integrity checking aspect ofthe invention implement a content protection protocol other than HDCP.
• the inventive link integrity checking method and apparatus is also useful in systems which transmit encrypted video over a serial link, whether or not the systems are also configured to transmit packets of encrypted audio data (or other auxiliary data) over the link in data islands or otherwise between active video periods.
• the transmitter ofthe inventive system encrypts video and packets of audio data using HDCP, and performs a link integrity check function once every 128 video frames to capture a 16-bit result vector (called R,) from both the fransmitter and receiver, and to compare these values every two seconds or so to make sure that they match.
• R 16-bit result vector
• This scheme has the following faults: there is no guarantee that the cipher engines in the fransmitter and receiver are exactly synchronized. If they are off only slightly, then the R, values will appear to match most ofthe time; the rate at which "snapshots" (samples) ofthe R, values are taken is not related to the rate at which they are compared.
• the rates might be completely asynchronous, in which case the fransmitter and receiver will occasionally show different values because the sample straddles a snapshot boundary; and conversely, the sample rate might be synchronous (or nearly so). This prevents the occasional straddling enor. However, if integrity checks ever do straddle a snapshot boundary then the situation becomes pathological, and may never be resolved.
• Fig. 17 As indicated in Fig. 17, the R, snapshots for the transmitter and receiver are exactly synchronized, as they should be.
• the link integrity checks occur at a similar, though slightly different rate. Most ofthe time both the fransmitter integrity check and the receiver integrity check occur in the same "window" between Ri snapshots. This means that both R, values will match, and the link integrity check will succeed.
• the fransmitter and receiver parts ofthe integrity check straddles an R, snapshot. This means that the R, values seen will differ.
• systems that implement HDCP are configured to perform the integrity check again a short time later if such an event occurs.
• the integrity check may straddle an R, snapshot (as in the fourth integrity check from the left in Fig. 20, but there is no guarantee that this will occur in a timely fashion, or at all.
• the length of time that elapses between such lucky events is inversely proportional to the beat frequency formed by the R, snapshot interval and the link integrity interval, and this beat frequency may be zero.
• the link integrity frequency is synchronous (or nearly so) to the R. snapshots, but out-of-phase far enough that it is blind to the sequence enors in the cipher state machines.
• This system is not working properly, but the link integrity checks cannot recognize that fact until some external agent (e.g., a frustrated user) intervenes.
• the transmitter and receiver are implemented so as to reduce or eliminate the enors that can cause the transmitter and receiver cipher engines to get out-of-phase; the link integrity check is done in a manner that ensures that R, snapshots occur at the same time in both the receiver and transmitter, regardless of phase; the link integrity checks occur with a known relationship to the R, snapshots. This does not mean they must be synchronous, only that they cannot be totally asynchronous.
• Link Integrity Check does not straddle an R, snapshot; and a method is implemented for directly checking the received, decrypted data for conditions indicative of improper system operation (e.g., "snow" conditions). For example, periodically (or from time to time), a known data pattern is input to the fransmitter, encrypted, fransmitted, received, decrypted, and then checked to see if the received decrypted pattern matches the known data pattern.
• the fransmitter and receiver cipher engines can get out of phase if certain confrol signals glitch. For example, if the receiver does not see a VSYNC signal it will fall behind. If the receiver gets an exfra VSYNC signal it will get ahead. If this happens then the receiver will not be able to decrypt properly, and the screen will show "snow”. If the point where the receiver takes a "snapshot" ofthe Ri value also moves in the same way (falls behind or gets ahead), then the receiver will continue to produce the conect R, values. This is because the R, sequence is a function only of constant values, and the frame sequence. It does not depend on the data in-between. This leads to a situation very like that in Fig. 20 or Fig.
• VSYNC could be used to control the advance ofthe cipher engine
• CTL3 could be used to confrol the advance ofthe snapshot counter. If the snapshot occurs at a different relative point in the cipher engine's cycle, then the R f sequence will be very different, and this will become immediately obvious.
• Another alternative is to provide either the transmitter or receiver with an "interrupt" capability that allows it to signal when a Link Integrity Check should occur.
• This function should generally be in the element ofthe system that initiates these checks (e.g., in the transmitter in a system that implements HDCP).
• This method works well if the interrupt latency time is always less than one video frame, because it guarantees that there will be no straddling, and that enors in which the receiver gets one or more frames ahead will be detected. It is expected that this will be the more common case, because it is more likely that a glitch will cause an additional VSYNC edge than that it will entirely eliminate a VSYNC.
• This method will not work well if the interrupt latency time spans multiple frames (unless it is unusually predictable), or if the receiver cipher engine gets one or more frames behind.
• the snapshots always occur at the same time on both sides ofthe link.
• the Link Integrity Check can be reliably completed before the next snapshot (due to the interlocks built into the protocol).
• the Link integrity Checks can be done as frequently or as rarely as necessary to suit the needs ofthe system. The rate can even vary from one moment to the next, as necessary. On this last point, it may be useful or desirable to include a "deadman timer" in the transmitter to verify that the Link integrity Checks are performed at least as often as some minimally acceptable interval.
• the receiver could be configured to include the deadman timer, although it will typically be preferable for the fransmitter to include it, since the transmitter is typically responsible for guaranteeing that protected content never leaves a system without suitable encryption, and so the transmitter is the logical candidate for ensuring the proper checks and balances are in place.
• the receiver of a class of embodiments ofthe inventive system determines directly whether there has been loss of synchronization between cipher engines in the fransmitter and receiver, using only data transmitted from the transmitter to the receiver.
• the transmitter and receiver implement the HDCP content protection protocol, the transmitter transmits encrypted video data to the receiver over video channels of a TMDS link (during active video periods), and packets containing encrypted audio (or other auxiliary data) are fransmitted over the video channels in intervals between active video periods.
• Each packet includes a header and at least two sub-packets, the header of each packet includes header data (which are typically not encrypted) and BCH parity bits for the header data, and each sub-packet includes data and BCH parity bits for the data ofthe sub-packet.
• the receiver is configured to check the types of enors that are occurring in the received, decrypted data and to determine from this infonnation whether most ofthe enors are transmission enors or HDCP decryption enors.
• the receiver determines that its TERC4 decoding operation results in few (or no) non-golden code receptions while its BCH enor conection circuitry detects and conects consistent enors in the decrypted subpacket data, the receiver concludes that HDCP decryption is failing.
• the receiver determines that there is a significant difference between the rate of BCH enor conection of subpacket data (which are fransmitted as encrypted data and then decrypted in the receiver) and packet header data (which are transmitted as non- encrypted data), the receiver concludes that HDCP decryption is failing.
• the receiver can be implemented to employ simple processing that allows it to make such conclusions reliably. Filtering ofthe detected enor rate information (generated by the receiver) across multiple packets (e.g., across a field or frame of video, or over a predetermined number of pixels or a predetermined time) can provide flawless or nearly flawless detection.
• the transmitter transmits packets (in data islands) frequently, to cany audio and other auxiliary data. Each such packet preferably comprises 24 bits of packet header data and four 56-bit blocks of subpacket data.
• the receiver performs BCH(32,24) enor detection and conection on the packet header data, and BCH(64,56) enor conection and detection on each ofthe subpackets.
• the transmitter performs HDCP encryption on each 4 bit quantity to be encoded as a TMDS code word for transmission on Channel 1 or Channel 2, but no HDCP encryption is performed on data encoded for fransmission on Channel 0.
• the transmitter performs TERC4 encoding ofthe data (packet header data) to be transmitted on Channel 0 and of the data (subpacket data) to be transmitted on Channels 1 and 2.
• the receiver performs the inverse of this processing: TERC4 decoding ofthe data received on Channels 0, 1, and 2; HDCP decryption ofthe decoded data received on Channels 1 and 2; BCH enor conection and detection on the decoded data from Channel 0 and the decoded and decrypted data from Channels 1 and 2; and de-packetization.
• the circuifry for implementing some of these operations in the receiver typically has (or can readily be modified to have) the ability to detect (and sometimes conect) certain types of enors. This allows the receiver to determine the likely point- of- failure for enors that occur.
• a transmission enor will typically be detected by the TERC4 decoding circuitry by determining whether the received value directly matches one ofthe possible transmitted values (i.e., one ofthe golden code words).
• An HDCP loss-of-synchronization enor which causes a decryption enor, can readily be detected by typical implementations of BCH enor detection and conection circuitry.
• BCH enor detection and conection circuitry indicates a high level of enors when the TERC4 decoding circuitry indicates a low level of enors, this strongly suggests that all or most ofthe enors occuned in the HDCP decryption process.
• the receiver can determine whether decryption is failing by comparing enor rates for these two types of data. If the BCH enor rate for subpacket data is significantly higher than that for packet header data, this also suggests that decryption is the culprit.
• Integrating two or more enor rate detection and comparison mechanisms in a receiver can provide an even greater degree of confidence that the receiver will detect decryption failures and will not reach "false positive" conclusions in which the decryption circuitry is functioning properly but the receiver concludes that it has failed.
• a known data pattern is provided to the transmitter, or existing data is modified in a predetermined way, and the result is then encrypted and transmitted. Then when the receiver decrypts the data it can then check for the expected results. If the receiver finds the expected results then the decryption can be assumed to be working properly. Otherwise it should be assumed that the encryption/decryption function is not working properly.
• This approach has the advantage of being very direct: it does not rely on indirect evidence from result vectors such as Rj.
• This approach is also backward compatible with existing protocols, and can be added to conventional HDCP systems with a small amount of external or internal logic, and no other changes to the link's behavior. For example, suppose that one defines the first two pixels in every frame as "A" and "B" respectively.
• the exact algorithm employed can vary among implementations, and can even be adaptive (such that it has the least possible impact on the overall picture). The only requirement is that the end result is modified data have predetermined (known) form.
• the receiver can then check for predetermined pattern in the decrypted data.
• the receiver should do this in a way that accounts for a small number of bit enors in the link. It might do this by setting a threshold for the perceived result. For example, satisfying two ofthe three above-specified equations might be enough. Or it might average the results over some number of frames. In either case, the receiver should be configured to find the proper result only if it the encryption decryption function is working properly. The receiver can then signal this result back to the host in some way.
• One method for doing so is the direct method of setting a bit in a register. Another method of signaling the result is to do so indirectly (e.g., by intentionally altering the R, value). The latter method has the advantage of working well in a repeater environment, in which a break in any link (the link between any transmitter-receiver pair) can readily be made to propagate upstream toward a host.
• the known data pattern can be implemented in other ways that do not affect the video picture in any way.
• special "test” packets can be inserted into the data stream, or "test” data can be inserted in other packets or places.
• Video lines could also be stretched to accommodate the "test” data, and then shrunk back to original size in the receiver.

Landscapes

• Engineering & Computer Science (AREA)
• Signal Processing (AREA)
• Computer Networks & Wireless Communication (AREA)
• Quality & Reliability (AREA)
• Multimedia (AREA)
• Physics & Mathematics (AREA)
• Computer Hardware Design (AREA)
• General Physics & Mathematics (AREA)
• Theoretical Computer Science (AREA)
• Television Systems (AREA)
• Data Exchanges In Wide-Area Networks (AREA)
• Compression Or Coding Systems Of Tv Signals (AREA)

Applications Claiming Priority (9)

Publications (1)

Family

ID=27488352

Family Applications (1)

Country Status (4)

Families Citing this family (33)

Family Cites Families (4)

• 2002
  • 2002-12-05 CA CA002471541A patent/CA2471541A1/en not_active Abandoned
  • 2002-12-05 JP JP2003559026A patent/JP2005514849A/ja active Pending
  • 2002-12-05 EP EP02795758A patent/EP1459532A2/en not_active Withdrawn
  • 2002-12-05 WO PCT/US2002/038948 patent/WO2003058826A2/en active Application Filing

Non-Patent Citations (1)

Also Published As

Similar Documents

Legal Events