Time Division Multiplex in Amateur Radio Steve Sampson, N5OWK, March 1992 Public Domain (p) 1992 Abstract The current practice of one radio frequency for every communications channel is a waste of resources. Amateurs should begin experimenting with new modes which share a frequency among many stations. If a channel is not carrying more than one conversation or data transmission, it will be too inefficient for future spectrum requirements. Radio systems ten years from now should be designed to compress as much information into each frequency channel as is practicable. Introduction Any Amateur knows that when 100 people try to call one station in a pileup, that chaos reigns. There are systems however, where this is not true. One hundred different communication channels on one frequency are possible. Rather than all 100 channels going simultaneously, each is given a time slice to conduct its business. This is referred to as Time-Division Multiplex (TDM) or Time-Division Multiple Access (TDMA). In this paper I will discuss a hypothetical 9 channel four time slot system, more for simplicity of design than anything else. In my petition to the FCC (dismissed) I also wanted the multiplex system to carry data as well as voice. So I will propose here that we discuss a system that combines data and voice over the same channels. Multiplex Theory A good example of TDM is your normal conversations on the local repeater. As one station finishes, the next begins, and so on. The time division however, is random in both length and start time. Only one user can transmit at a time (more can transmit, and usually do by accident, but only one person is intelligible). The person transmitting is said to be occupying the total bandwidth of the channel. Since the channel is busy even when the operator is silent or composing another thought, it is also a slow and inefficient system. The advantage of course is that simple technology is very inexpensive, and having more spectrum than users in most of the country, allows us to get away with it. TIME SLOTS If we could give each operator 10 milliseconds and then switch to the next user, we could time share the channel. Computers do this every day. They can have 100's of users who think they have the computer to themselves, but each is actually being given a time slice of the total power of the machine. As a further improvement, these computers wait for the operator to strike a key or request output before they even provide a time slice. So 99 terminals sitting at a desk while everyone is at lunch, demand no computer time, while the last terminal in the basement used to play chess by the janitor gets all the time. As these workers come back from lunch, the janitor gets less and less time. Worse-case is 1000 milliseconds (100 users times 10 millisecond slices) or one second before his next time slice. DIGITAL AUDIO We live in a time when many things done by analog electronics in the past, are now done by digital electronics. The analog radio transmitter and receivers are connected to digital processors, or merely transport digital information. Many radio stations use Compact Disks that store digital information, and convert this to analog for broadcast. What they are really doing is wasting their high resolution recording by converting it to a stream of low resolution analog audio. They could just as easily transmit this digital directly to the user, and have the user make the conversion with much greater fidelity. COMPRESSION Music is very hard to compress, because it has information during the whole transmission. Voice on the other hand has great pauses and vocal repetitions. The sentence "I want a cookie" can be compressed several ways based on local dialect, but in every case would take a fraction of the normal time to transmit. This is the basis for getting the voice to fit into the TDM time slice. The audio is converted to digital, analyzed, and redundant parts are compressed. This is then transmitted during an assigned time slot. Your telephone for example, is multiplexed along with many other users. The analog voice over the twisted pair is converted to 14 bit digital samples. These are then quickly coded into 8 bit Pulse Coded Modulation (PCM) samples. This is the first step in compressing the information. Each sample is then converted to serial bits and dropped into the correct time slots. These are then recombined at some distant switching center. Your voice is first passed through a band-pass filter that removes all frequencies except those necessary for voice. (300 to 3500 Hz). The analog voice is then sampled at 8 kHz, producing one 8 bit digital word every 125 microseconds. Since 8 bits times 8 kHz is 64k, this is commonly referred to as a 64 kbps interface. Notice that it is just a bit greater than the authorized 56 kbps in Ham radio. But you don't want to use these raw data rates for voice communications anyway. SYNCHRONIZATION A good data rate to begin experimenting is 9.6 kbps. Since 9.6 kbps divided by 8 bits is 1200 bytes, I propose we design a TDM system that converts audio to 8 bit bytes and has 4 time slots. This produces 300 bytes per slot (some of which may be used for synchronization or guard bytes). Each radio is assigned by the operator to a time slot and mode. When the radio is first turned on it looks for a data clock on channel 0. This is usually transmit by a master time station at a high elevation (an encoded time burst every couple of seconds). The sync clock identifies the start of the cycle (using the masters encoded callsign or a tone). The radio will then automatically switch to the Slave mode. Alternatively the users can select one station to be the master, and that radio will then transmit the sync clock. At this point further radio configuration is selected by each operator and transmissions begin. The TDM cycle is based on 1200 bytes and lasts one second. It is repeated again and again. Individual stations wait for their time slot period to transmit. The first station to transmit in a time slot sets an activity indicator and owns the slot. The slot is relinquished after two (or more) cycles of silence unless the activity hold switch is enabled. This last feature can allow roundtable type communications or hold the channel slot for direct communications between two Amateurs. As a further safeguard, the activity hold should timeout after a few minutes when no use is detected. Activity hold causes a transmission of the users call sign with no information. The Operator Controls In commercial systems, the radio user is not given much control over where their transmissions will go. The Amateur however, needs to have access to all of the radio options. They want to be able to select a frequency, select a net, or group of users. The control panel of a multiplexed radio should therefore have all of these options available through keyboard control and stored in EEPROM (Electrically Eraseable Programmable Read Only Memory) so you don't have to type it in every time you turn on the radio: I. Callsign Entry II. Master Volume/Power switch III. Master/Slave Entry and indicator IV. Microphone Enable (M1, M2, M3, M4) V. Microphone with Push-To-Talk switch VI. Frequency Channel (1 - 9) A. M1 Select Mode (Multiplex Slot 1) a. Tx/Rx Simplex b. Tx Duplex c. Rx Duplex 1. Activity Indicator/Hold switch 2. Volume 3. Remote computer/audio Jacks B. M2 Select Mode (Multiplex Slot 2) C. M3 Select Mode (Multiplex Slot 3) D. M4 Select Mode (Multiplex Slot 4) Since these radios will be digital, no squelch control is needed. There will be no output to the speaker unless the data is decoded. When activity is detected on the time slot, no further transmissions are allowed. The time slot activity indicator will be released two cycles after end of transmission unless retriggered. This feature is designed to prevent interference and loss of time slot by another station. Our hypothetical radio is a 70cm band radio. It operates on nine channels: 0. 446.050 (Synchronization Channel) 1. 446.075 2. 446.100 3. 446.125 4. 446.150 5. 446.175 6. 446.200 7. 446.225 8. 446.250 9. 446.275 Each channel is run at 9.6 kbps and has four time slots, or 1200 bytes per second. The cycle consists of 9600 bits, or 2400 bits per multiplex channel. A good first experiment will be to use a 22 byte guard, 256 bytes data, and a final 22 bytes guard. M1. 0 - 299 300 bytes 250 ms M2. 300 - 599 "" "" M3. 600 - 899 "" "" M4. 900 - 1199 "" "" Each multiplex channel transmits for 300 bytes (250 milliseconds), and then must wait for 900 bytes (750 milliseconds) before transmitting again. Using this example, four Amateurs can conduct individual conversations or data transfers, or alternatively one Amateur can use all four time slots for multiple connections. DUPLEX A good example is a full-duplex connection between two Amateurs. Each Ham will select a frequency and time slot for transmission, and another pair for reception. At this point they may begin talking as if on a telephone. Another example is a file transfer between computers. As one computer transmits a packet of data, the other computer will either ACK (Acknowledge) or NAK (Negative Acknowledge) the packet. The computer can load the circuit with packets and the remote computer will ACK/NAK without waiting for the sender to stop and listen. If you wanted to play Flight Simulator over the air, both systems could transmit aircraft position simultaniously. Another example is Direction Finding. A network of listening stations is set up to report bearing, amplitude, and time of detection (based on the masters time). These reports are then sent at the assigned time slot. Each computer can then produce a probable location of the transmitter, as well as filter out multipath reports by processing the report history. SIMPLEX In the half-duplex mode, each time slot is used round-robin. As one Amateur finishes their transmission, the next proceeds with theirs. The radios probably should be equipped with a tail tone when using voice mode. The Activity indicator is dropped after each cycle in the data mode. For example, on channel 1 there are four multiplex programs being conducted. On M1 a simplex voice weather net is in progress, M2 has a simplex digital packet weather roundtable, M3 has a simplex voice swap net, and finally M4 has a simplex digital packet BBS. RADIO SET UP How must the radio work in order to participate in all of these events? The operator first selects each frequency, and then enables the multiplex channels that they are interested in using. Usually one frequency is used for uplink, and another (spaced away from the other) is used for downlink. The time slot should be different also to prevent transmitter desense of the receiver. The decoded audio is then mixed to the speaker and also output via a rear panel jack. The rear panel jacks are for interfacing with a computer or standard audio levels. The radio will include all the modems necessary to operate with another station. The use of an external TNC (Terminal Node Controller) will not be required. When the Push To Talk (PTT) is enabled, the audio is quickly digitized and compressed, and then is output during the next appropriate time slot. Compression is achieved through a vocoder algorithm. There is also a digital comparison which only passes voice above a certain threshold level. The vocoder algorithm should be a standard 9.6 kbps LPC one. For digital packet, the whole time slot is filled with callsign, addressing, and data information. Conclusion This paper introduced and outlined a TDM system that consists of 9 channels with 4 time slots each, resulting in 36 possible communication events in 225 kHz. The technology is available today to accomplish and implement this design. While more expensive at the outset than current systems, the time is fast approaching when we can no longer afford the one channel one user system. This proposal offers twice the events per Megahertz as compared with current Narrow-Band FM (NBFM) systems (15 events at 5 kHz Bandwidth with 10 kHz guard bands each side, to 36 events TDM with 25 kHz Bandwidth). Final systems could offer either more time slots per channel, longer cycles, or even reduced bandwidth, such as 5 kHz channels. This is all hypothetical however, as I don't really know if nine 25 kHz channels butted up against each other can actually be manufactured. But even with guard bands of 5 kHz between these channels results in more communications events than NBFM. Coments, critiques to: ssampson@sabea-oc.af.mil