GB2248008A - Data transmission system - Google Patents

Abstract

In a message transmission system, a transmitter 10 segments a message in store 20 into multi-word segments which are transmitted individually, each with a header (from 22) including a message identifier. A receiver 11 receives the segments and reassembles them into the message in store 30, and also returns acknowledgements. The lengths of the segments are selected in accordance with a predetermined pattern of different lengths. In the system shown, the segment length is alternated between 32 and 31 words by a segment length unit 23 in the transmitter and a segment length unit 35 in the receiver tracks this. Thus two successive segments can be sent, but a third segment (of length matching the first) cannot be sent until an acknowledgement (from 34) of the first has been received (at 24). For broadcasting, a receiver only accepts a message if the pattern of segment lengths matches the preset pattern. <IMAGE>

Description

Data TrSAD smi ssi OTE Sy stem The present system relates to data transmission systerrjs.

There are many types of data transmission system, having varying characteristics. The present invention is concerned with systems in which a plurality of stations may share 8 common transmission medium, the messages are usually relatively short (typically of sorne tens to hundr-eds of bits long) though their length may occasionally be very much longer, and the transmission medium is subject to oippreciable noise.

A typical example of such h system is a radio system including several mobile stations. Such a system may for example be in coniormity with the Department of Trade and Industry standard MPT 1327 and associated standards tcollectively known as Protocall 1327).

With such a system, the noisiness of the channel means that reception of messages is unreliable. Arrangements are therefore required for causing the selective repetition of messages which are not successfully received. (Forward error correction systems can te used, but it is found in practice that these alone are insufficiently reliable.) This implies that there is some arrangement for error detection, and also for acknowledgement of receipt of messages. In principle, a simple ACK (acknowledgement of correct receipt of message, can be used, but a PAK/ttAK type scheme (PAK = positive acknowledgement, NAK = negative acknowledgement, i.e. message received but with discovered errors in known positions) is often more convenient, as that allows the retransmission of only the erroneous parts of the message.

A further problem erises with lonE; messages, as the longer the message, the greater the chance of an error occurring in it. Thus if the length of a message is substantially greater than the mean tirne between errors, it will probably have to be sent many times before it is successfully received, so reducing the throughput of the system. It has therefore become well known to split long messages into distinct packets or sa nts, which are sent separately.It is the responsibility of the sending station to split tL;e message into segments, and of the receiving station to reassemble the segments into the oritrinal message. The segment length is chosen so that the chance of a segment being received successfully is good.

This segmentation technique introduces a problem of its own. If the message is long enough, it will consist of a large number of segments, and the chance of at least one of them not being detected by the receiving station will be high. The receiving station must therefore be able to detect when a received message has a segment missing. It must also be able to identify which segment is missing, so that the missing segment can be requested and sent again. Also, in the transmission systems being considered here there are usually several stations sharing the transmission medium, so there is the possibility of messages, or segments' of messages, between different pairs of stations becoming interleaved.

The receiving stations must then also be able to distinguish the different segments of a given message from other messages.

A technique has been proposed for solving these problems. This involves generating a unique identification for each message, and for segmented messages, generating a unique identifier for each segment. The message identifier may for example be, or include, the identifiers of the transmitting and receiving station, and the segments may for example be identified by number-ing them consecutively.

If interleaving can occur, each segment must also, of course, include the message identifier; the message and segment identifiers, possibly together with other control information, will normally form an initial header portion of each segment. The receiving station will then be able to recognize the segments of a given message, detect duplicated and missing segments, and request their retransmission if they are missing, and correctly insert the retransmitted segments into the message.

Looking at this in slightly more detail, there are two main ways in which the segment sequence may be disturbed. First, a segment header may not be received correctly by the receiving station; this results in a missing segment.

Second, a segment may be received correctly but its acknowledgement (sent back from the receiving station) may not be received correctly by the transmitting station; the transmitting station will then (after a time-out) retransmit the segment. If this retransmitted segment is received correctly by the receiving station, then that segment has been duplicated. (More complicated error situations can occur, but can generally be regarded as combinations of these two basic situations of segment header loss and acknowledgement loss.) For convenience, it is conventional for messages to consist of a number of "code words" of predetermined length, typically 64 bits.The length of a segment is limited to a predetermined number of code words: we will take this nurnber as 32 (although the current draft standard sets the maximum segment size as a header plus 62 code words). Obviously a short message will consist of a single segment of fewer than 32 code words, and the last segment of a long message will also usually be less than 32 code words long; the length of a segment is usually indicated in its header. The total length of a message can be indicated in various ways. One convenient way is to include its length in the header of the first (and possibly only) segment; another way is to use either an end-of-message or- a last segment indicator.

The technique of segment numbering is effective, but it involves substan tial overheads, since the identifiers have to be managed and the headers typically take up an appreciable portion of the segments. I he general object of the present invention is to provide a segmentation system in which segment identifiers are not required.

Accordingly the present invention provides, in one aspect, a message transmission system comprising transmitter means for segmenting a message into see- rr,ents which are transmitted individually, each with a message identifier, and receiving means for receiving and acknowledging the segments and reassembling them into the message. characterized in that the transmitter means selects the lengths of the segments in accordance with a predetermined pattern of different lengths, the receiving means assembles the segments into the message only insofar as the lengths of the received segments are in the predetermined pattern, and the sending of a segment of the same length as a previous segment is inhibited until the transmitter means has received an acknowledgement that that previous segment has been received by the receiver means.

The crux of this aspect of the present invention is thus the use of a plurality of different lengths for the segments, together with an acknowledgement system in which a segment of given length is acknowledged before another segment of the same length is allowed to be sent.

In its simplest form, the system distinguishes only between a repeated segment and a new segment. For this, the criterion is simply whether the current segment is the same length as the previous one. If it is, it is taken as e repetition of the previous one; if it Is not, it is taken as a new one. For this, the pattern of segment lengths need only ensure that successive segments are of different lengths; the lengths themselves need not be predetermined (though they can be, of course).

More elaborate forms are possible, in which the system is made capable of distinguishing between more than 2 segments. For this, the segment lengths in the pattern of segment lengths need to be predetermined.

The above discussion has been concerned essentially with message transmis- sion between a transmitter and a specific receiver. In many message transmission systems of the type being discussed, there is also a need for a transmitter to be able to broadcast to all receivers.

With broadcasting, a receiver may be inoperative (e.g. switched off) or receiving so badly that an- attempted transmission to it alone would be aborted as impossible. In eddition, there are many receivers, and it will generally not be feasible to employ an acknowledgement technique. For example, the acknowledgements from the different receivers would tend to interfere with each other; the burden on the transmitter of processing acknowledgements from multiple receivers may be excessive; and the total number and identities of the receivers may not be known to the transmitter (particularly if receivers are added to and removed from the system irregularly).

With broadcasting, therefore, a lower standard of reliability has to be accepted than for point-to-point transmission (i.e. sending a message from the transmitter to one specific receiver). (If high reliability is essential, then the same message can be sent individually to each receiver in turn.) Nevertheless, it is obviously desirable to achieve as high a levei of reliability as is technically feasible. In particular, it is desirable to minimize the chance of a receiver assembling a message incorrectly, i.e. the chance of e receiver which has not received all the segments of a massage from treating those which it has received as constituting a complete message.

It will be realized that the principles of the present system can also be applied to a broadcast system.

Accordingly the present invention provides, in another aspect, a message transmission system comprising transmitter means for segmenting a message into segments which are transmitted individually, each with a message identifier, and r-eceiving means for receiving the segments and reassembling them into the message, characterized in that the transmitter means selects the lengths of the segments in accordance with a predetermined pattern of different lengths, and the receiving means assembles the segments into the message only if the lengths of the received segments are in the predetermined pattern.

A broadcast system may be arranged to broadcast a message a plurality of times. Provided that the broadcast message is not so long that the pattern of segment lengths starts to repeat, then if a receiver in the present system has r-eceived only some of the segments of the message, it can obviously watch for the missing segments in repeat broadcasts.

A data transmission system embodying the invention will now be described, by way of example, with reference to the drawing, which is a block diagram of the system.

In the simplest form of the present systern, two segment lengths are used in the body of the message, one the maximum allowed segment length toe.8. 32 code words) and the other 1 code word shorter (31 code words). Of course, the final segment will normally be short, as it will end with the last code word of the message. The length of the segment (the number of code words) is included in the header of the segment, and in any acknowledgement of the segment.

The normal sequence of events in the sending of a multi-segment message is thus as follows: 1 The transmitting station sends the first segment, of say 32 code words.

2 The receiving station receives it correctly, and sends an acknowledgement.

3 The transmitting station receives the acknowledgement, and sends the next segment, of 31 code words length.

4 The receiving station receives it correctly, and sends an acknowledgement.

5 The transmitting station receives the acknowledgement, and sends the next segment, of 32 code words length.

and so on, with the segment length alternating between 32 and 31 code words.

Considering first the segment duplication type of error, in this a segment reaches the receiving station correctly and the acknowledgement is sent back but fails to reach the transmitting station. The transmitting station thereupon assumes that that segment has not been received, and will send it again after a time-out. The receiving station thus receives the segment same again.

It is 'essential for the receiving station to recognize this duplication, so that it does not take the second copy of the segment as being the next segment in the message. In the present system, the receiving station checks the length of the segment (as given in the header) and compares it with the length of the previously received segment. If the length is the same then this shows that the segment just received is a duplication of the preceding segment.

Since the segment just received is a duplicate of the previous segment, it is evident that the acknowledgement of that previous segment did not reach the transmitting station. An acknowledgement of the segment just received is therefore sent back to the transmitting station. However, the segment is otherwise rejected, instead of being added to the previously received segments which are being assembled together to form the complete message.

Considering now the segment loss type of error, this is handled in the present system in the same way as in the known system (segmentation with segment numbering). Typically, with this type of error a segment header is lost because of noise on the transmission medium; thus the receiving station cannot identify the segment and so cannot send an acknowledgement. The transmitting station awaits an acknowledgement of the segment, but the acknowledgement does not arrive. The transmitting station therefore retransmits the segment after a ti me-out.

This time the segment is received correctly by the receiving station, which - as it does for all segments - checks the length of the segment as given in the header and compares it with the length of the previously received segment. If the length is different, then this shows that the segment just received is not a duplication of the preceding segment.

Since the segment just received has not previously been received, it is a new segment, and its length is therefore different from that of the previously received segment. As far as the receiving station is concerned, there has been no error; it has no knowledge that the transmitting station had made a previous attempt to transmit the segment just received. The receiving station therefore sends back an acknowledgement of the segment, and adds the segment just received to the previously received segments which are being assembled together to form the complete message.

The drawing shows the logical organization of a simple or basic data transmission system embodying the present system. A transmitting station 10 and a receiving station 11 are coupled by means of a transmission medium 12, which may be a radio band. It will be realized that there may be further transmitting and/or receiving stations also coupled to the transmission medium 12, and that the transmission medium 12 may be logically divided into a number of channels e.g. by using separate frequencies). One station may be a fixed station exer-cising general system control, with the others being mobile stations.

The transmitting station 10 is generally controlled by a control unit 21, and the message to be transmitted is stored in a register 20, in the form of a series of code words (shown as a colurnn). A header generator 22. controlled by the control unit 21, generates the headers s for the successive segments, and the control unit 21 causes each segment to be transmitted as a header from generator 22 followed by a suitable number of code words from the stor-e 20.

A length control unit 23 in the control unit 21 controls the length of the segments being transmitted. This unit normally switches or toggles between 32 and 31, so that the segments are alternately 32 and 31 code words long. However, the first code word in each segment is the header, rather than a code word of the message itself, so the number of code words of the user message in the segments is is in fact alternately 31 and 30. Usually the last segment will be relatively short, as it is not normally padded out to the full 31 or 32 code wor-d length.

The transmitting station also includes an acknowledgement detector 24, which detects incoming acknowledgements from the receiving station 11. This unit matches the message identifier and length value in the acknowledgement with the message identifier and length value in the header generator 22 and segment length control unit 23. If there is a match, the unit 24 sends a signal to the control unit 21 to toggle the state of the length control unit 23 between the values 31 and 32; the control unit 21 thereupon causes the next segment to be sent.

The message code words of the segment being sent are also fed into a shift register 25, which is split into a main portion 25-1 and a last stage 25-2, so that its length is 30 or 31 code words depending on whether or not the last stage is utilized. The control unit 21 includes a timer 26, which is started when a segment is sent. If the timer 26 times out before the acknowledgement detector 24 detects an acknowledgement of the last segment sent, then the control unit 21 sends a segment in which the code words are taken frorn the shift register 25 instead of the message register 20. This results in the previously sent segment being retransmitted. The length control unit 23 is left unchanged, so the length of the retransmitted segment is unchanged.Also, the code words are recirculated back into the shift register 25, in case the segment gets lost again and has to be transmitted yet again.

The various signal paths for the code words, etc include gates (not shown) controlled by the control unit 21 so as to control the appropriate signal flows.

The receiving station 11 is controlled generally by a control unit 31, and includes a store 30 in which the code words of the message are accumulated.

An incoming segment is passed to a segment analysis unit 32, which matches the message identifier in the segment header with the contents of a message identifier register 33. The action taken depends on whether or not there is a match.

If there is no match, then the segment is the first segment of a new message. In this event, the segment analysis unit 32 passes the message identifier to register 33 to store it therein, and sends a signal to the acknowledgement generator unit 34, which generates an acknowledgement (including the message identifier and length value from register 33) which is sent back to the transmit ting unit 10. It also allows the rest of the segment - that is, the message code words in the segment - to pass into the store 30, in which they are accu mulated as the first code words of the message. The unit 32 also sets a length comparison store 35 to the length value in the received header.

If there is a match, then the segment belongs to a message which is already being received, and the segment analysis unit 32 sends a signal to an acknowledgement generator unit 34, which generates an acknowledgement (including the message identifier from register 33) which is sent back to the transmitting unit 10. Unit 32 then compares the length of the segment with the length value in the length comparison stor-e 35. If the segment length matches the length value in the length comparison store. this indicates that the segment has already previously been received, and no further action is taken.If, however, the segment length does not match the length value in the length comparison store, this Indicates that the segment is a new segment. In this case, unit 32 there- fore changes the length value in the length comparison store 35 to the length value of the new segment end also allows the rest of the segment - that is, the message code words in the segment - to pass into the store 3O, in which they are accumulated with the code words from previously received segments of the mes- sage.

If message interleaving on the transmission medium 12 is allowed, then the apparatus must obviously be elaborated so that it responds only to segments forming part of the message being transmitted between the two stations shown and ignores segments of other messages. In particular, the receiving station must inspect any incoming segment which is not part of a message which it is already receiving, to determine whether it is the first segment of a new message directed to that receiving station.

It has been assumed in the above description that the first word of a segment is a header containing only system information. (This may include various types of information not discussed here.) In practice, however, the number of bits required for such information may be much less than the length of a code word, and part of the header can then be used for user (message) information.

It is clear that the present technique is not limited to the use of 2 alternated segment lengths, but can be extended to use 3 or more segment lengths.

used in repeated sequence, or indeed to any variation of segment lengths such that every segment differs in length frorn its neighbouring segments. If n different segment lengths are used, then the number of segments which can be sent by the transmitting station before it has to receive an acknowledgement is n-l.

The use of a large n allows the transmitter to continue to send segments while awaiting an acknowledgement. So if an acknowledgement is not received (either because of a segment loss or an acknowledgement loss), then it can continue sending further segments while the acknowledgement receipt time-out timer is running. Alternatively, the receiver can delay sending acknowledgements until it has received several segments.

The segmented system can include "RLA"s - that is, requests for acknowledgement sent by the transmitting station if it does not receive an acknowledgement of a segment within a reasonable time. (RLA = Repeat Last Acknowledgement.) An RLA can usually be a very short transmission, e.g. a single code word. If the receiving station has received the segment, it will have sent an acknowledgement which has become lost, and will retransmit the acknowledgement.

This saves the transmitting station from having to retransmit the entire segment.

(Of course, if the segment was not received by the receiving station, the RLA will not elicit an acknowledgement, or will elicit a NAK acknowledgement, and the segment will have to be sent again.) RLAs can be used in the present system; it will be realized that this is likely to be more useful if either (a) n is fairly large, since the time delay between sending a segment and sending an RLA for it can be usefully occupied by the sending of further segments, or (b) the system is a multiplexed one, when the chance of intereference is minimized by the use of RLAs by mobile stations at specified times.

As noted above, 8 segmentation system may include error- detection and correction means in addition to the acknowledgement technique for dealing with completely lost segments. In one such known system, a segment may be received correctly (in the sense that it is recognized as a segment and is processable by the receiving station), and each code word in it is individually checkable for possibly uncorrectable errors. The result is that some (usually most or all) of the code words in- the segment are accepted as correct, but some are rejected as faulty, and these are retransmitted in a later segment (usually the next). The advantage of this is that when a segment contains a few faulty code words, only those code words rather than the whole segment have to be retransmitted.

For this code word restoration technique, the faulty code words must be identified to the transmitting station. This is conveniently done by including an EFLAG (error flag) field in the acknowledgement which has one bit for each code word in the segment; each bit is set to say 0 if the code word was received correctly, and to 1 if it was not. The transmitting station, when it receives the acknowledgement, includes the incorrectly received code words in the next segment. When the receiving station receives this next segment, it fills any gaps in its sequence of code words in the message it is building up from that segment, and then adds the rernainine code words of that segment to the end of the sequence.

The r-epeated code words in the segment must of course be identifiable by the receiver. This can be achieved by locating them at any convenient prearranged position in the next segment - conveniently at the beginning of the segment (after the segment header, of course), Alternatively, the segment can contain some form of marker indicating the repeated code words.

This code word restoration technique can clearly be utilized in the present systern. The technique is most easily operated if n = 2. If n 2, 2, then additional housekeeping information is required so that the receiving station can reliably distinguish replacement code words from new code words and insert them into the proper locations in the code word sequence. There is little point in having n > 2 unless the transmitting station can send up to rrl unacknowledged segments, and it is then also necessary for the acknowledgements from the receiving station to identify which segments they are acknowledging.

This can be done by including e segment length value field in each acknowledgement. Such a field would need 5 bits for a maximum segment length of 32 code words, since it must be able to indicate the length of a final short segment as well as the alternation between 31 and 32 code words for the segments forming the main body of the message. However, it can conveniently be done by expanding the functionality of the EFLAG field by 1 bit. For this, the length of the EFLAG field is increased by 1 bit, and the end of the segment (as received) is indicated by a 1 immediately following the bit for the last code word in the segment, with any following bit positions filled by Os. At the transmit ting station, the final 1 in the EFLAG field of a received acknowledgement and any following Os are stripped off.The number of bits remaining is the length of the segment being acknowledged, and this identifies the segment to the transmitting station.

Thus taking the maximum segment length as 6 words (for ease of illustration), the EFLAG field would normally be 8 bits, xxxxxxxx (where x = 0 for a correct code word and 1 for an erroneous one). Expanding this by 1 bit, the EFLAG field for an 8-word segment would be xxxxxxxxl; for a 7-word segment, xxxxxxxl0; and for say a 4-word final segment, xxxx10000. It will be evident that the length of the segment is correctly indicated whatever the values of the "x" bits.

Other methods can of course be used for indicating which code words in a segment have not been received correctly. In mobile radio systems, a common cause of the loss of code words is fading, and this normally affects a continuous group of code words. A missing group of consecutive code words can be indica- ted by specifying the positions of the first and last code words in the group, or the position of the first and the length of the group. If the segment length is substantial (e.g. 32 or 62 code words), this requires considerably fewer bits than the EFLAG technique.

This method can of course only cope with non-consecutive missing code words if the acknowledgement contains room for specifying more than one group of missing code words. If the number- of missing groups exceeds the capacity of the reporting acknowledgement, then either the whole segment must be treated as lost and thus needing to be retransmitted, or the code words between the missing ones must also be treated as missing.

In the forms of the present system discussed up to now, it has been assumed that, except for the last segment of a message, the segment lengths follow a fixed sequence - 32-31-32-31-... in our particular example. It was noted that if n = 2, then successive lengths must be different, but we did not discuss that possibility. In the absence of errors, there is indeed no point in deviating from a fixed sequence. If the code word restoration technique is used, however, then the possibility of such deviation becomes significant.

There are broadly two techniques for dealing with errors using this technique. One is to insert the repeated code words in the next segment to be sent (preferably at the beginning of the segment). The other is to keep the segments strictly independent, so that if there are errors in a segment, these are corrected before the next segment is sent.

In the above description of errors, the first technique has largely been assumed. This technique maximizes the speed of the system in getting the entire message passed to the receiver, but requires a large receiver buffer capacity, because it is possible for the same code word to be in error on several successive segments, so that by the time it is finally received correctly, it will be a large lumber of code words behind the most recent one received.

The second technique reduces the r-equirernents on the receiver, but results in slower overall transmission rates. With this technique, if there is an error, the words in error are retransmitted and no further words are transmitted until the error(s) have been corrected. Since the number of words in error in a segment is generally small, this generally involves the transmission of a relatively short segment. Because the length of this segment is less than that of the previous segment (of 32 or 31 words, in our specific example), the receiver knows that this is a new segment, sent in response to the error field in the last acknowledgement sent.

It is of course possible that there may be er-rors in the reception of this short correction segment. If that happens, then the receiver inserts suitable information in the error field of its acknowledgement, asking for the resending of the words which are still in error.

It is also possible for all words in a segment to be in error. One way of dealing with that is for the receiver to ignore the segment entirely and send no acknowledgement. The transmitter will then retransmit the segment after a timeout, and as the receiver took no action when it previously received that segment, it will sccept the segment as the next one it is expecting. (Actually, it does not matter if the receiver does send an acknowledgement of a segment in which all code words are received in error.The transmitter will retransmit it as before and the receiver will know that the next segment it should receive should be the same length as the one which it has just received.) In the above discussion of the known segment numbering system, it has been assumed that the segment length is fixed (except for the last segment of a message).

One known modification of such a system involves dynamic variation of segment length, depending on such matters as mean message length, noisiness of the transmission channel, etc. The same principles can obviously be applied in the present system, provided of course that such adjustment of segment lengths does not conflict with the variation of segment lengths between neighbouring segments of the present system.

Another known modification of such a system involves control of the segment lengths by the receiving station. This is achieved by the receiving station including a segment length control value in the acknowledgement sent back to the transmitting station. This technique can for example be used in conjunction with the code word restoration technique discussed above, with the receiving station normally calling for a segment of maximum length but, in the event of some code words in the last segment being faulty, calling for a short segment containing only those code words.

It is evident that the variation of neighbouring segment lengths in the present system can similarly be controlled by the receiving station, either alone (so that the lengths are strictly alternated) or in combination with the calling for short segments when a segment has been received correctly but one or more code words in it are faulty.

The overall functionality of the present system is, as far as its general operating characteristics which it presents to the user are concerned, largely equivalent to that of the known segment numbering system.

Claims (12)

Clai ms

1 A message transmission system comprising transmitter means for segmenting a message into segments which are transmitted individually, each with a message identifier, and receiving means for receiving and acknowledging the segments and reassembling them into the message, characterized in that the transmitter means selects the lengths of the segments in accordance with a predetermined pattern of different lengths, the receiving means assembles the segments into the message only insofar as the lengths of the received segments are in the predetermined pattern, and the sending of a segment of the same length as a previous segment is inhibited until the transmitter means has received an acknowledgement that that previous segment has been received by the receiver- means.

2 A system according to claim 1 wherein the predetermined pattern of lengths comprises two lengths, the first a power of 2 and the second 1 less.

3 A system according to claim 2 wherein the power of 2 is 32.

4 A system according to any previous claim including means for generating and responding to a request for an acknowledgement of a previously transmitted segment.

5 A system according to any previous claim including means for generating an error flag field in an acknowledgement indicating erroneously received words in a received segment and responding thereto by resending those words in the next segment.

6 A system according to claim 5 wher-ein the error flag field is of fixed length and comprises a sequence of variable bits corresponding to the words in the segment, followed by a 1 indicating the length of the segment and a pattern of Os filling any remaining bits.

7 A system according to any previous claim wherein the lengths in the predetermined pattern of lengths are dependent on the mean message length and/or- transmission conditions.

8 A message transmission system comprising transmitter means for segmenting a message into 'segments which are transmitted individually, each with = message identifier, and receiving means for receiving the segments and reassembling them into the message, cheracterized in that the transmitter means selects the lengths of the segments in accordance with a predetermined pattern of different lengths, and the receiving means assembles the segments into the message only if the lengths of the received segments are in the predetermined pattern.

9 A system according to claim 8 wherein the predetermined pattern of lengths comprises two lengths, the first a power of 2 and the second 1 less.

10 A system according to claim 9 wherein the power of 2 is 32.

11 A system according to any one of claims 8 to 10 wherein the lengths in the predetermined pattern of lengths are dependent on the mean message length and/or transmission conditions.

12 Any novel and inventive feature or combination of features specifically disclosed herein within the meaning of Article 4H of the International Convention (Paris Convention).