Subject: 9600BPS/3KHz-G1NTX Fast Packet Systems By: Simon Taylor G1NTX r 6484 Part 1 of 2 From CONNECT INTERNATIONAL - April, 1989. Copyright 1989 by Radio Society of Great Britain. Reprinted by permission For some time now I have been interested in the discussions going on regarding fast packet and data links using RF modems, specifically 9600 bits per second (bps) modems. There seem to be two schools of thought: 1) To use modems connected directly into transceiver IF strips and modulate the carrier directly with data. 2) To connect the data modem via the audio connections of the rig, and operate in a similar way to the technique use on 3KHz bandwidth telephone lines. A colleague of mind (G8DXZ) and myself have proved that this technique works up to 9600bps and we plan to try 14,400bps modems soon. The purpose of this article is to disscuss the latter technique and (hopefully) stimulate some interest and maybe even some more experiments with these modems. THE PRINCIPLES Telephone modems, because of the transmission medium must operate within a 3KHz bandwidth. The frequency response of the telephone line is normally quoted as being between 400Hz and 3400Hz. Most people are familiar with normal frequency shift keying (FSK) using two different tones as used in existing packet radio, but to go must faster than 1200bps within a 3KHz bandwidth requires some further thought. The first principle used is Phase shift keying (PSK) which uses one audio tone (the carrier) with phase changes introduced into this carrier which can be detected at the receiver. The advantage here is that one phase change can theoretically be introduced every cycle of the carrier and if four types of phase changes are used, then two bits can be encoded per sampling time. Secondly, amplitude changes can be added so giving more combinations and more bits encoded per sample time. At this stage, we should introduce another iece of jargon - the Baud. Baud defines the sampling time, i.e. the rate of Phase and Amplitude changes, so for example if four bits are encoded during every baude, and the 'Baud rate" is 1200, then the effective bit rate will be 4800 bps. Given below is a table showing some half-duplex modulation techniques and their data rates. Technique Bit Baud Bits per Phase Amplitude Carrier rate rate Baud Changes Changes Frequency V.29 9600 2400 4 8 1 1700 V.29 7200 2400 3 8 0 1700 V.29 4800 2400 2 4 0 1700 V.27 4800 1600 3 8 0 1800 V.27 2400 1600 2 4 0 1800 Another aspect of these modems is that of 'training'. When data are sent, they are scrambled to made sure that all of the data points are sent even with no data being sent. This makes most efficient use of the transmitted spectrum. The receiving modem will synchronise to the transmitting modem, keeping track of the phase changes as transmission goes on. This traning does take some time however, and will cause time overheads if the channel is turned around frequently. The main reason for training using these patterns is to determine the phase and amplitude restrictions of the path, and to set up an equaliser that is used to give a flat response during data transmission. The modems we have tried also employ 'adaptive equalisation' which will adjust equaliser values during data transmission for small changes in the quality of the received signal. The time taken to train may make transmission using this faster data mode an overhead rather than an advantage if only small packets of data are sent. V.29 for example, needs 270 milliseconds to train before any data are sent, which is equivalennt to about 40 characters of information at 1200 bits per second. Therefore, we should send at least this amount of data and preferably more to take advantage of the higher data rate after training. Below are some packet sizes and the times to transmit using existing 1200 bps packet versus V.29 at 9600bps. Packet Time @ Time @ Size (chars) 1200bps V.29/9600bps 20 0.133 0.286 50 0.333 0.311 100 0.666 0.353 200 1.333 0.436 500 3.333 0.686 1000 6.666 1.103 2000 13.333 1.936 Times given is seconds. From the table it can be seen that the larger the packet, the greater the advantage. It may be that this mode of transmission is not suitable for use with the existing AX.25 standard, but some sort of alternative protocol could be used (or developed) which will not transmit until it has a certain amount of data to send. Further discussion about protocols is beyond the scope of this article, I shall leave it to the national packet network... Remember that these modems are designed to operate within the 3KHz availaable on telephone lines and a larger audio bandwidth is normally used on VHF/UHF FM, so the quality on a good path is usually found to be better than that obtained via our national telephone system. THE PRACTICE There are a number of modem devices which can be used for the audio modulation part of a fast RF modem. Connection to a rig can be simply via Audio in, Audio ouut and PTT and these modem should be simple to connect to existing TNC's such as the G0BSX-2 or similar, but I have not tried this yet. So far I have tried communications using an IBM-PC directly controlling the modem and PTT without any rigid packet structure as such, but this has proved that the principle at least works on VHF and UHF FM. All of the modems I have tried have been similar in that they require CPU control via a bus which is 8080 compatible and have simple audio in and out connections. All that has been needed is a D>C> blocking capacitor between the modem output and the microphone input (some rigs may also need some reduction of the signal), and a capacitor from the earphone output of a typical rig. A relay should then be driven to control the PTT. Suitable modems I have tried include: The R96MD, this is a V.29 and V.27 modem primarily intended for FAX machines, but makes an ideal half-duplex data modem. This device is supplied on a small pCB with two rows of pins allowing it to be assembled like a large DIP device. It will opeate from 9600bps down to 2400bps, as well as at V.21 at 300bps FSK. DTMF is also provided which may be of use to some amateurs. This modem, because of it's application in FAX products benefits from a reduced cost due to it's use in massive volumes. The R96MFX and R96EFX, these are CMOS single-chip modems with similar features too the R96MD. The R96EFX is especially interesting because it has a V.27 short train feature, training in 50 milliseconds instead of the 270 milliseconds standard, and HDLC packetising and error detection built-in, so avoiding the need for external HDLC controllers. We soon plan to try toe R144HD which is a V.33 modem which operates at 14,400bps. Again the modem is designed to operate in a 3KHz telephone bandwidth, so VHF/UHF operation should not be a problem. If you would like data sheets or data books on these modems, then I can be contacted QTHR. Sending out information will not prove a problem. Also you can leave messages for me at GB3UP (G1NTS @ GB3UP.GBR.EU) Reference reading: "Quality of Received Data for Signal Processor Based Modems" application note (Rockwell 1987 Modem data book), this data book also includes data sheets on all of the modems discussed. "Rockwell Interface Guide", This gives detailed information as to the connection, use and monitoring techniques used forr these modems, (but is a cost item.) Simon Taylor G1NTX - 21st March 1989