GB2449428A - Redundantly transmitting message segments over a dual purpose transmission line - Google Patents

Abstract

A method of transmitting a message segment on a dual purpose transmission line, such as a power line, on which are liable to occur noise spikes, comprises coding the message segment as components of a burst of carriers of different frequencies, and transmitting on the line two successive bursts encoding the same message segment, preferably within an interval which is substantially less than twice the repetition period of the noise spikes. The message segment is therefore transmitted redundantly to in successive bursts to allow successful reception if one of the bursts is affected by noise spikes. The two bursts may be preceded by a start symbol at the commencement of said interval. The method may include monitoring the transmission line for the occurrence of a noise spike and initiating a transmission of a start symbol and the two bursts on detection of the occurrence of a noise spike.

Description

SIGNAL TRANSMISSION OVER NOISY MEDIA EMPLOYING FREQUENCY

DIWS ION MULTIPLEXING

This invention relates to the transmission of information employing a burst of multiple frequency signals such as for example in orthogonal frequency division multiplexing (OFDM), the carrier signals being modulated in amplitude or phase or both by message segments which may represent data packets. It particularly though not necessarily exclusively relates to the performance of OFDM by a modem intended for subsea communication, such as by means of an umbilical including a power transmission line, which may be of significant length.

Although a dedicated line may be employed for the transmission of the OFDM signals, it is feasible and in various circumstances convenient to couple the transmitted signals to a power line or existing line not especially adapted or dedicated for the OFDM signals. This is particularly so if the frequency range of the carriers in the burst are not particularly high, for example in a range from about 2.5 or 3kHz to 25k}Iz. However, such lines are liable to produce noise in the form of power switching spikes and even when the, noise spikes are comparatively short in relation to the signal burst' they may well be sufficient to corrupt the transmitted data. This is particularly serious when message segment is transmitted as a burst of modulated carriers, because multiple carriers may be affected at the same time by the same noise spike.

Summary of the Invention

The present invention principally overcomes the problem of switching spikes on a dual purpose by transmitting, for each message segment which is transformed into a multiple frequency burst, two successive identical bursts.

Where, as will frequently be so, the power switching spikes occur with a known repetition period, the duration of a burst may be selected to be substantially less than the known repetition period.

Preferably, in order to synchronise a receiver and preferably also to set a gain control for the bursts, the bursts are preceded by a start symbol which may comprise a few cycles of a sinusoidal signal modulated in amplitude by a Gaussian pulse waveform. It is preferable than the interval comprising the start symbol and the two bursts be less than about two-thirds of twice the aforementioned repetition period.

Alternatively or additionally the occurrence of switching spikes may be monitored and the transmission cycle comprising the two bursts and preferably also the start symbol may be commenced in response to the occurrence of a switching spike.

In a preferred system, a data stream is converted by means of an appropriate transform, such as an inverse fast Fourier transform (FFT) into, for each segment of the data signal, a burst of modulated carriers of which the spectra overlap but which are chosen so that they are orthogonal and therefore, despite the overlap, do not significantly mutually interfere. As will be described with reference to a particular example later, each transmission cycle preferably comprises a set of carriers each of which has an amplitude and phase determined by a complex number itself determined in accordance with a respective segment of the data stream. The inverse FFT converts the complex number into a time domain signal. In a preferred example, four bits of the digitized signal may determine the phase of the relevant carrier and two bits may determine the amplitude. On reception, the time domain signals in the burst may be sampled and reconverted by means of a fast Fourier transform to recover the original message segment.

Various techniques may be used to obtain a reconstituted message segment from the double transmission of the burst. A preferred technique comprises transforming the bursts into respective sets of digital samples, partitioning each set into a first half and a second half, and decoding a single set comprising a first half-set selected from either one of the sets and a second half-set selected from either one of.the sets. The technique may further comprise monitoring a parameter of power in each half-burst and selecting between corresponding half-bursts in the sets according to which half-burst exhibits the lesser power.

There follows a detailed description of an embodiment of the invention as constituted within a subsea communication system.

Brief Description of the Drawings

Figure 1 is a schematic diagram of a subsea communication system.

Figure 2 is a schematic diagram of a modem for use in the system shown in Figure 1.

Figure 3 is a schematic diagram of a transmit channel.

Figure 4 is a schematic diagram of a receive channel.

Figure 5 is a schematic diagram of a transmit and receive process.

Figure 6 is a waveform diagram illustrating an example of noise on a power line used as a data transmission medium.

Figure 7 is a graphical representation of a transmission cycle Figure 8 is a summary schematic of a transmission operation Figure 9 is a summary schematic of a different transmission operation Figure 10 is a summary schematic of a receiving operation

Detailed Description

Figure 1 of the drawings shows by way of example a subsea communication system 10. The reference 11 denotes a subsea control unit which may contain within it TCP/IP hosts 12 and 13, and a multi-port hub or switch 14. The subsea control unit, has for communicating with other units a modem, particularly an Ethernet modem 15, coupled, as hereinafter described, to an umbilical 16 which extends from the control unit to the locations of subsea electronics modules. Very typically the umbilical may be of very substantial length.

Reference 18 denotes schematically a subsea electronics module which may include S TCP/IP hosts and a switch as well as a modem 17 for communicating with other units by way of the umbilical. Other subsea electronics modules are denoted by the references 1 Ba, 1 8b and I 8c with their respective modems 1 7a, 1 7b and 1 7c.

Such a system provides the ability to connect TCP/IP enabled hosts together to allow them to communicate over long distances. The reference to the particular transmission protocol (TCP) and the addressing or network layer protocol (internet protocol) is given by way of example. The invention is not limited to such protocols.

Figure 2 illustrates one of the modems which may be used in Figure 1. In essence this modem will, for transmission, take a digital data stream and convert it into successive bursts of modulated carrier signals, preferably by means of OFDM bursts; on reception they will detect OFDM bursts employing a start signal as hereinafter described and apply transforms to reconvert the received signals back to a digital data stream.

The particular form of modem is not important. The modem shown in Figure 2 is expected to perform the routing of internet protocol packets between its network interfaces.

The modem in this example comprises a microcontroller 20 which can communicate by way of a data bus 21 with random access memory 22 as well as transmit and receive FIFOs 23 themselves coupled by way of a data bus 24 to a digital signal processor 25. This is coupled to a programmable logic device (PLD) 26 as well as analogue switches 26 that provide coupling to transmit channels amplifiers 28 and receive channel amplifiers 29 coupled to an umbilical 30. The microcontroller has an interface with the data bus 21. Coupled to the data bus 21 are a LAN interface 31, flash memory 32 and a UART 33. The LAN interface has coupling transformers 34.

The processor 20 also has a RS232 interface 35.

The primary interface is the LAN interface which is software configured in any desired manner to auto detect the speed of the connected network and whether it is full or half duplex. The microcontroller 20 is expected to implement a half-duplex logical bus topology and a data link protocol to carry the internet protocol packets on the umbilical.

When the microcontroller has a packet to transmit on an umbilical channel it writes data to its transmit FIFO, one word at a time. It then signals to the digital signal processor that there is data to transmit. The digital signal processor will receive the microcontroller start signal as an external interrupt indicating that a packet is in the transmit FIFO waiting to be transmitted. The digital signal processor reads the packet data a word at a time from the FIFO into an internal random access memory until it sees a flag denoting that the FIFO is empty. The DSP encodes the packet data into a digitised analogue signal to transmit onto the umbilical. It writes the signal by means of a serial interface to the relevant channel of a codec (Figure 3) to transmit the data.

In this example, the various elements on the system may be arranged in a token bus topology but such is not essential to the present invention.

The DSP constantly reads data from both channels of the codec looking for a start symbol. It then determines, as will be described later, where the start of the encoded packet will begin.

Raw samples are loaded into internal random access memory and decoding is performed to reproduce a data packet. When the complete data packet has been recreated, the DSP writes it one word at a time into the receive FIFO for that channel.

The DSP then clears the relevant receive FIFO bit.

Figure 3 illustrates a transmit channel. The digital signal processor 25 controls a codec 36 which drives two redundant transmit channels. The upper channel in the drawing includes an amplifier 37, a balanced line driver 38, switches 39 controlled by the digital signal processor, power amplifiers 40 and an output transformer 41. The same components are in the lower channel denoted (a).

Each receive channel comprises a transformer 42, a differential amplifier 43, a band-pass filter 44, an automatic gain control circuit 45, and a limiter 46 coupled to the codec 36. The lower (redundant) channel corresponds.

The transformers 41 and 41a couple the OFDM bursts to an AC power line extending along the umbilical. Likewise the transformers couple any received bursts from the power line to the modem.

Figure 5 illustrates the process of encoding and decoding a data stream. The transmitter and receiver in this example employ orthogonal frequency division multiplex to transmit data as a fixed length signal composed of multiple frequencies, each of which carry some data components.

This form of system is intended to provide reliable communication over possibly long distances, with high data rates and substantial immunity to noise. The system is intended to be either point-to-point or multi-drop but is not necessarily intended to provide any mechanism for media access control and assumes a shared bandwidth across all nodes.

As is shown in Figure 5, the modulation process starts with a data steam which is converted into a bit stream segments of which are encoded as complex numbers and converted by means of an inverse fast Fourier transform into a burst of sinusoids of orthogonal frequencies with amplitudes and phase determined by a respective complex signal. The OFDM burst is appended to a start signal, stage 54 and the resultant is transmitted, stage 55. The demodulation process commences with reception stage 56, the detection of a start symbol stage 57, the application of a fast Fourier transform stage 58, the frequency decoding stage 59 and conversion to bit stream 60 and finally conversion to the original data stream 61, For maximum power to transfer to occur when a modem is communicating, the modem line driver output impedance and the receiver input impedance should preferably by matched to that of the umbilical, which is approximately 100 ohms. The input impedance of the receiver is just that of the transmitter since they are shorted together. Therefore to ensure that the channel receiver has the correct impedance the -7.-transmit amplifier should be enabled at all times. Otherwise the receiver would be in a higher impedance node owing to the amplifiers' reflective impedance. In order to prevent unwanted circuit noise from being amplified and causing interference with the receiver, the ability to ground these amplifier inputs is provided by way of a switch.

The data stream is converted to a simple bit stream. The bit stream format depends upon the number of bits to be used in each of the amplitude and phase. For each of the available frequencies, the amplitude and phase is set based on the data in the bit stream. The inverse FFT is performed on the frequency information to provide a single in the time domain.

The DSP generates in memory a signal consisting of samples that the codec can output. Each sample is in the approximate range -32760 to +32760. The DSP first generates the start symbol and then generates a number of signal bursts' (to the length of the FF1) which contain the coded data. Once sufficient bursts have been generated to transmit the data packet to be sent, the codec is instructed to begin transmitting the signal. The codec may employ DMA (direct memory access) to read the DSP memory across the interconnecting serial link and outputs the signal to the analog output stages. The generated signal is then transmitted.

Preferably, as described in our co-pending application No. 0707334.9 filed 17 April 2007, each burst is preceded by a start symbol which comprises a few cycles of a sinusoidal signal modulated in amplitude by a Gaussian pulse waveform. As described in that co-pending application, this start symbol minimises the impulse response of the line to a start symbol, and facilitates the synchronisation of the receiver. Since each transmission burst is of a known, fixed length, the receiver can read in the correct number of samples after the start symbol. An FFT is performed on the samples to determine the amplitude and phase of each of the frequencies within the signal. The receiver knows how many bits per phase and amplitude to expect in each frequency and begins to decode the signal.

Figure 6 shows a typical waveform display of noise on a power line. The low level noise is of no great importance because it is readily possible, as described in the aforementioned co-pending application, to measure the power in successive cycles of the sinusoid constituting the start symbol to obtain a proper starting point for synchronisation of the receiver. However, The typical noise shown in Figure 6 includes repetitive noise spikes 64 which can corrupt the data signal.

S The present invention provides for a double transmission of the multiple frequency burs and in particular a transmission cycle comprising the start symbol and the double burst within an interval which is less than twice the expected repetition period of the noise spikes. Preferably the duration of each burst is less than two-thirds of the repetition period. There is then sufficient information transmitted between spikes to allow correct decoding of the signal.

Figure 7 illustrates one exemplary transmission cycle according to the invention. The cycle of transmission commences with a start symbol 70 comprising a few cycles of a sinusoid (as described in the aforementioned co-pending application) modulated in amplitude by a Gaussian waveform. It is followed by (e.g. a predetermined time after the peak of the Gaussian pulse) by a multiple frequency burst 71a obtained by transforming the respective message segment followed, preferably immediately, by a second, identical burst 71b. The duration of a burst is preferably short compared with the repetition period of the power spikes 72 (shown in an arbitrary phase relative to the bursts). The combined duration ti of the bursts and preferably also the interval t2 from the commencement of the start symbol to the end of the second burst 71b is less than two-thirds of the power spikes' repetition period.

The transmission cycle may be initiated without regard for the occurrence of a switching spike, as indicated in Figure 8. Alternatively the transmission line may be monitored for the occurrence of a switching spike and the cycle of start symbol and burst may be initiated immediately thereafter. This method minimises the likelihood of interference to the start symbol by a switching spike. This is of some importance if, as is preferred and as described in the co-pending application the power in each cycle of the start symbol is measured to obtain an indication of the peak of the symbol and to set the gain control for the subsequent bursts.

One result of transforming received bursts into raw samples is two sets of samples which might both be uncorrupted; the other likely result is that that at least some of the samples of a set corresponding to only one of the bursts will be corrupted by the occurrence of a switching spike on the line.

It may be feasible in some circumstances to employ check codes for all message segments so as to enable acceptance of a first (uncorrupted) set and discard of a second set of samples or rejection of a first (corrupted) set in favour of the second set.

However the use of such codes may be avoided employing the technique now described with reference to Figure 10.

In this technique the sample in a set are partitioned into two half-sets, so that the first set consists of half-sets I a and lb and the second set consists of the half sets 2a and 2b. The final' set of samples can be constituted by selecting according to an appropriate criterion one first half-set form the half-sets la and 2a and one half-set from the half-sets lb and 2b. Thus the selected half-sets may be selected form the same set or from different sets. The temporal spans of the sets in the analog domain are shown as Rxla, Rxlb, Rx2a and Rx2b in Figure 6.

One method of selection is to monitor a parameter of power in each half-burst and to select those samples which correspond to the respective half-burst having the lower power.

An example of a code sequence for this purpose is set out in Table 1.

Table I

void RxSignallnput() mt 1; float s = 1.0/32767.0 1/ II Get first signal into first half of m_DoubleData /1 for (i=O;i<Leadln;i++) ADCGetSample(); II Ignore Lead In for (i=O;i<FPrLength;i++) mDoubleData[iJ = s * (float)ADCGetSample(); I, Copy Data Section to Buffer for (m=0;i<LeadOut;i++) ADC_GetSample(); II Remove Lead Out From ADC Buffer /1 II Get second signal into second half of m_DoubleData // for (i0;i<Leadln;i++) ADCGetSampleO; II Ignore Lead In for (i=O;i<FFTLength;i++) m_DoubleData[FFTLength+i] S * (float)ADCGetSample(); II Copy Data Section to Buffer for (i=O;i<LeadOut;i++) ADCGetSample(); II Remove Lead Out From ADC Buffer

I-

1/ Divide each of the received signals in half and try to find a combination /1 of first half and last half that has the lowest power.

II Rxl:Rx2 becomes Rxla:Rxlb:Rx2a:Rx2b, combinations are therefore: // II Rxla:Rxlb II Rxla:Rx2b II P.x2a:Rx2b II Rx2a:Rxlb // mt half FFTLength = FFTLength / 2 II Measure the power in each half signal float Rxla SignalPower( &mDoubleData[0J, halfFFTLength); 1/ Riila float Rxlb = SignalPower( &mDoubleData[halfFFTLength], halfFFTLength); II Rxlb float R.x2a = SignalPower( &mDoubleData(FFTLengthJ, half FFTLength); II Rx2a float Rx2b SignalPower( &mDoubleData[FrrLength+halfFFTLength}, half FFTLength); // R2b /1 printf("\nLowestCombination is %d", LowestCombination( Rxla, Rxlb, Rx2a, Rx2b) ); switch( LowestCombinatiori( Rxla, Rxlb, Rx2a, Rx2b case 1: II Rxla+Rxlb for ( i=0; i<FrrLength; i++ Data(i = mDoubleData[i] break; case 2: II Rxla+Rx2b for ( 1=0; i<halfFFTLength; i++ II Rxla Data[i] = mDoubleData[iJ; for ( i0; i<halfFFTLength; i++ /1 Rx2b Data[i+halfFFTLerigth] = mDoubleData[FFTLength + half FFTLength + iJ break; case 3: II Rx2a+Rx2b for ( 1=0; i<FFTLength; i++ Data[iJ = m_DoubleData[F'FTLength+i) break; case 4: II Rx2a+R1b for ( i=0; i<halfFFTLength; i++ /1 Rx2a Data[i] = m_DoubleData[FFTLength+i] for ( i0; i<halfFFTLength; i++ II Rxlb Data [i+halfFFTLength) = mDoubleData [half FFTLength+i] break; default // Rla+Rxlb for ( i=0; i<FFTLength; i++ Datali] = m DoubleData[i]

-

mt LowestCon,bination( float A, float B, float C, float D // Possibilities: I/1A+B II 2 = A+D II 3 = c D //4=C+B // // Return which adds up to the lowest number.

mt result = 0; II A+B if (((A+B) < (A-1-D)) && ((A+B) < CC+D)) && ((A+B)<(C+B)) result = 1 II A+D if (((A+D) < (A-1-B)) && ((A+D) < (C+D)) && ((A+D)<(C+B)) result = 2 if (((C+D) < (Al-B)) && ((C+D) < (A+D)) && ((C+D)<(C+B)) ) result3; II C+B if (((C+B) < (A+B)) && ((C+B) < (A+D)) && ((C+B)<(Cl-D)) result 4 return result;

Claims (9)

1. -12 -Claims 1. A method of transmitting a message segment on a dual

   purpose transmission line on which are liable to occur noise spikes, comprising coding the message segment as components of a burst of carriers of different frequencies, and transmitting on the line two successive bursts encoding the same message segment.
2. 2. A method of transmitting a message segment on a dual purpose transmission line on which are liable to occur noise spikes having a predetermined repetition period, comprising coding the message segment as components of a burst of carriers of different frequencies, and transmitting on the line two successive bursts each encoding the same message segment and each having a duration which is less than the said repetition period.
3. 3. A method according to claim I or 2 in which the two bursts are preceded by a start symbol at the commencement of said interval.
4. 4. A method according to claim 2 and claim 3 in which the combined duration of the start symbol and the two bursts is less than twice the said repetition period.
5. 5. A method according to any foregoing claim and further comprising monitoring the transmission line for the occurrence of a noise spike and initiating a transmission of a start symbol and said two bursts on detection of the occurrence of a noise spike.
6. 6. A method according to any foregoing claim and further comprising transforming the bursts into respective sets of digital samples, partitioning each set into a first halfand a second half, and decoding a single set comprising a first half-set selected from either one of the sets and a second half-set selected from either one of the sets.
7. 7. A method according to claim 6 and further comprising monitoring a parameter of power in each half-burst and selecting between corresponding half-bursts in the sets according to which half-burst exhibits the lesser power.

   -13 -
8. 8. A method according to any forgoing claim in which each burst is a set of orthogonal frequency division multiplexed carriers.
9. 9. A method according to any foregoing claim in which the carriers are substantially within the frequency range 2.5 to 25 kHz.