Fullscreen HTML5Med Res (??MB)HTML4

REMOTE CONTROL BY BOB YOUNG How pulse code modulation decoders were developed Of all the units that comprise the modern radio control set, the decoder has probably undergone the most circuitous development. Where many discrete components were once required, it can now be done in one or a couple of ICs. As I have pointed out in the past, when the new generation of PCM sets were launched onto the marketplace, accompanied with the usual hoopla, much was made of the fact that PCM was fitted with "Fail-Safe". Amid many gasps, oohs and aahs, the unsuspecting modern generation of the modelling fraternity eagerly embraced PCM, as do all fraternities eagerly awaiting their next technological "fix". This appears to be the problem when dreams become reality, the reality fades quickly and a jaded technological palate seems to require constant boosts of new technology. If only people stopped to soak up the wonder of all of this technology we are surrounded with, they would be much more satisfied with what they have. However, not all of the modelling fraternity eagerly embraced PCM. The CLOCK RESET CH1 OUT CH20UT CH3OUT CH4OUT fl____________________.r _Sl___________________ - -n~ - - n _ __ _ ____. .....__ Fig.1: repeated from last month, this diagram shows the essence of the serial to parallel conversion performed in the decoder. The serial pulses from the receiver are converted into control pulses for the servos. 86 SILICON CHIP greenhorns may have, but us old timers recognised it for what it was; a recycled version of the system that by 1964 had become known as "that circuit which neutralised the controls on the way to a crash". PCM has become one of the sad stories in the development of R/C systems. However, I digress. The modern digital proportional system grew by a tortuous process and progressed through all sorts of developmental periods, amongst which were pulsed multi-tone systems. These systems came close to giving good results but the technology for tone decoding did not exist in those days (late 1950s). I still look at the multi-tone system with interest and with the new technology, I feel there is potential for a much more interference free system here. However, that is for the future; our story is about the past. In the early 1960s, Don Mathers and Doug Spreng developed the first really successful digital system and it completely revolutionised model radio control. Gone were the days of constant retuning, bulky audio filters and poor response times. The Mathers and Spreng system delivered tuning-free control with a maximum response time of 16 milliseconds and almost perfect proportional control. The greatest benefit howe.ver was that all controls were simultaneous . Gone were the days of manually pulsing alternate controls. The modellers of the day were in raptures. Here was genuine progress and the marketing men had a field day, and for once I did not mind. I was in there helping them. Real progress I am all in favour of. However, the first generation de- however, and the second generation sets allow the owner to select FailSafe or leave it off. They even allow choice of PPM or PCM modes. The Germans have a quaint way of expressing the situation; they state that Fail-Safe PCM allows you to "crash like a gentleman". If you have one of these modern sets, select PPM and no fail-safe. The results are well worth the effort. Personally, I feel that a more productive approach to PCM and microprocessor systems would be in error detection and correction as in CD technology, leaving Fail-Safe out completely. This would represent, to my mind, a valid application of technology. Circuit techniques The Bonner Digimite was quite advanced for its time but incorporated the FailSafe concept. If the incoming pulse train was corrupted, the decoder was shut down and a neutralising DC level was sent to the wiper pads in the servos. The throttle was set to low and controls to neutral. coders had the dreaded "Fail-Safe" built into the system. In retrospect, the "Streakers Defence" applied in this case. It did seem like a good idea at the time and I can remember being very impressed with the concept. In those days many of us were still flying what were virtually free-flight models with a very high degree of natural stability built in. If the radio failed, and they did fail more often in those early days, the model kept flying by itself. The motor was cut by the Fail-Safe and all one had to worry about was the odd thermal which carried off many a model, never to be seen again. Keep in mind also that our old friend "V2 " is a component of kinetic energy and cutting the throttle is very important as it reduces crash damage considerably. It still does and I am all in favour of Fail-Safe throttle even now. My real objection to Fail-Safe is in the locking out of the controls once Fail-Safe is activated. However, and here is the crux of the story, by 1964 modellers had discovered the neutrally stable aerobatic aircraft and now we have a vastly different story. Neutralised controls on this type of model are a death sentence, hence the epithet above. These things flew as straight as an arrow in the direction of the last control command and once control was locked out, crash they certainly did. It did not take long for the designers to wake up to the fact that some control was better than none, thus Fail-Safe was consigned to the rubbish bin and models were allowed to fly through interference or weak levels of RF until control was regained properly. The odd snatch of control available in the noisy periods was often enough to keep the model flying. The problem with Fail-Safe is that once the decoder decides that the signal is unreadable and shuts itself down, all control is lost until the decoder deems it proper to restore it. I personally dislike having electronics decide. for me when I can or cannot have control of my model. Fail-safe is an invalid concept in theory, proved itself invalid in practice and was quite correctly consigned to the rubbish bin; that is, until the new generation of university trained designers were turned loose into the practical world and dragged it out again in the form of the PCM set. Even these people learned quickly, The first really successful commercial proportional system appeared around 1964 in the shape of the Bonner Digimite and featured, for those days, absolutely revolutionary ideas. They were heady days for modellers and I can still remember the excitement generated by the American advertising. Bonner was a leading manufacturer of servos for use in reed receivers and when he went to a proportional system he spared no expense to make it the best available and it was. It gave good service to many people for many years - quite an achievement for a pioneer set. It did, however, feature Fail-Safe which was a pity, but there was little known about this concept in those days. Bonner followed the Spreng and Mathers concept in which a serial stream of pulses are transmitted to the receiver and then passed to a serial to parallel converter for decoding. Fig.1 (repeated from last month) shows the timing sequence for this serial to parallel conversion. The system is termed Pulse Position Modulation (PPM) and in this system a master clock in the transmitter triggers a cascaded series of pulse generators. Each of these generates a pulse, the width of which is controlled by the potentiometer coupled to the transmitter control stick. Convention has this width at 1.Sms for neutral; lms for minimum and Zms for maximum pulse width. Depending upon the number of channels (controls), a set will have APRIL 1991 87 CH2,3,4,5 CH1 .005 CH6 1k 560k 27k I I cc 01 2 I cc o-:al"·-+-'WV.-----+-1. .005 .005 I .005 4.7k 10k I I 10k 10k I I OS I OS OS 1N914 1N914 1N914 I HAST STAGE ONLY I REPEAT FOR EACH CHANNEL Fig.2: this serial to parallel decoder is based on BRY39 silicon controlled switches, one for each servo channel. The circuit was simple yet reliable & was used for many years until IC decoders appeared on the scene. from two to eight pulse generators. These serial pulses are then converted into a marker pulse stream in which a 350µs marker pulse marks the beginning and end of each pulse. Transmission is continuous, with a new frame transmitted every 16-20ms (new clock pulse). Thus, in a 4-channel set with a frame rate of 16ms, full deflection on all controls will result in a data stream 4 x 2ms long followed by a resting period of 16 - 8ms or 8ms. This rest period is used as an identification or synchronisation pause. This serial data stream may be transmitted by NBFSK or amplitude modulation and the receiver passes on a duplicate of this data to the decoder. The decoder is essentially a serial to parallel converter and there is a wide varity of approaches 'to the decoding process. However, all follow the Spreng and Mathers concept of reconstituting the serial stream to the original number of variable width pulses which are essential for the operation of the proportional servo (described last month). Bonner used an 8-transistor array in a Johnson ring counter, with a diode matrix for decoding. Frame validity was verified by a separate counter. Bonner went the whole hog and came out with an 8-channel set and decoding alone took 15 transistors. If the incoming pulse train was corrupted, the decoder was shut down and a neutralising DC level was sent to the wiper pads in the servos. The throttle was set to low and the controls to neutral. As you can imagine, this was quite an elaborate circuit for discrete corn- A_ _ _ _ _ _ _ _ _ _ _ _ _.___ _ _ _~_ _..._ _ _ _ _ _.___ _ +4.8V 14 B 12 68k .,. IC2 74C164 6 10 11 12 13 .,. Fig.3: this circuit used a 74C164 serial to parallel decoder & is still current today. However, many manufacturers are now going over to custom ICs which enable them to incorporate more features. 88 SILICON CHIP ponents and the set was a little bulky. The problem with this system was that corrupt data was often encountered and the set immediately went into lock-out until valid data was again established, thus denying the pilot access to the controls. Quite often this corrupt data was only present just long enough to activate the Fail-Safe. Once activated, the pilot had to wait out the time-out period, thus the FailSafe exaggerated what was in reality a very minor glitch - one that the pilot may not have noticed under normal conditions. He certainly noticed the throttle come off and the controls fly back to neutral and stay there, even if it was only for a very brief time. But at 100km/h, even a very brief time can be disastrous. The lessons learned from this system and many others were quickly picked up , and the second generation systems came out without Fail-Safe , the concept being that it was better to let the model fly through the corrupt data. The era of truly reliable radio control modelling had begun. Bonner never learned the lesson and stuck with Fail-Safe and gradually faded from the market. Decoder development followed quickly from there on and there were many ingenious circuits, all aimed at improving reliability and reducing component count. This in turn gave smaller size and lighter weight. One very popular circuit was that shown in Fig.2. The heart of this was the SCS (silicon controlled switch). One switch was used per channel and the pulses just simply stepped through the counter. The output was a positive pulse whose width equalled the distance between the leading edges of the clock train. Simple and reliable , this circuit was used until the IC decoder began to appear. The IC decoder had a patchy beginning and was usually cobbled together out of a number of chips. JK flipflops were a popular item and in 2-channel sets the 4013 dual-D flipflop is still used extensively. The problem with using several IC chips is that they are expensive in regards to board space and size is always a problem in R/C receivers. This was particularly true in the days before surface mount devices. Three 14-pin DIP packs would use all of the space available and thus IC decoders were seen to appear in one manufacturer's equipment while others stuck with the SCS or discrete decoders. The most usual reason in those days for going to ICs was just simply to hop on the bandwagon. Then the serial to parallel decoder made its very welcome appearance. Here was a single chip solution and chips such as the 74C164 would give 8 simultaneous controls in a single chip. Fig.3 is a circuit of a decoder using this chip. It is part of the circuit featured on page 111 of the December 1990 issue. The 74C04 is used as an This modern radio-control transmitter uses pulse code modulation (PCM) techniques to provide simultaneous 7channel control. Other features include channel mixing, dual rate control, trim adjustment & servo end point adjustment. audio amplifier/shaper/inverter. Briefly the circuit in Fig.3 works as follows . Transistor Q5 is used as a small signal amplifier with a slicing action. Thus, low level noise is eliminated and the amplified pulse train is passed on to the 74C04. This acts as Mailbag - continued from page 5 "Much more expensive to repair": how do you look at the modulation on an FM Tx? Answer: with great difficulty or use a modulation meter. Quoting my own case, having manufactured AM sets for many years, suddenly my test equipment was inadequate. There followed a spending spree on new test equipment which included a modulation meter, a more elaborate signal generator and a more accurate frequency counter, amongst other things. Then followed a stock of more expensive crystals and other components, and believe me broken crystals are a big item in R/C receivers. Add to this a more elaborate alignment procedure. Need I go on? "Much more expensive and sometimes more difficult to change crystals in": this is a very important practical consideration. Quoting from the Futaba (Aust.) price list again: AM crystals, $24.95 per pair; FM crystals, $38.50 per pair. Because of the narrow bandwidth of some FM sets, crystal tolerances can put the frequency outside the passband with a loss of range occuring. Moral - always check the range after crystal changes. Complexity covers more than component count and includes alignment which we have already covered. The "simple AM receiver" was included as a bit of history to illustrate the development of narrow band spacing without elaborate ceramic filters. I did point out that it was a 20-year old design. A modern AM receiver using ICs in the RF section (as do the FM examples you gave) would present a much lower component count. The question is would they work any better and the answer is probably not. Capture effect was the weakest statement in the article and probably should not have been included. But an inverting and squaring amplifier to provide the necessary pulse information to the 74C164. The output of pin 2 is the clock stream and is applied to pin 8 of the 74C164 . Pin 4 of the 74C04, D3 and the .033µF capacitor comprise the identification network or sync separator. Pin 6 of the 74C04, D4 and the 2.ZµF capacitor comprise the chip enable network. This network filters the incoming pulse train and sends the chip enable (pin 9) high. Thus, the chip is disabled if there is no incoming pulse train. This prevents random noise from damaging the servos by driving them hard up against the end stops. The timing still follows the broad outlines of Fig.1. This type of decoder is still viable today for PPM, however the larger manufacturers have begun to move over to custom ICs which incorporate several features not found in off-theshelf decoding ICs. Such features as voltage regulation and noise filtering are often built into the decoder chip, thus reducing even further the component count. The really significant development in decoding, however, has been in the area of the microprocessor and here there is great promise for the future. SC as you point out, it is laboratory demonstrable and I do seem to remember a series of editorials in RCM&E some years ago on "Pirated models due to capture effect". Actually, I thought I was being very fair in the article as I gave pros and cons for both systems. The remarks on oscillator design were virtually a direct quote from my friend who checked the article. I had forgotten to mention the supply rail stabilisation and he pointed this out to me. All of these remarks I heartily endorse. I did not say design was impossible but that NBFSK placed "more stringent demands on the designer". I stand by these remarks. All of the NBFSK oscillators I examined had zener stabilised supply rails and heavy decoupling. I have yet to see an AM transmitter with a zener stabilised oscillator. Bob Young, Silvertone Electronics, Riverwood , NSW. APRIL 1991 89