Compiled by G31RM Issue number 25 January 1994 HAPPY CHRISTMAS AND NEW YEAR Membership Since I suggested contributions last time the membership has dropped from 82 to 33 -1 expected this to happen! Publicity I have received a copy of the 1994 Arriateurfunk Almanach from Nils DKBOK which contains a short article I wrote for him on CCW. Bill de Cade has written some notes for QEX which he hopes will be published soon and he is also hopeful of getting something into QST. I wrote a review of COHERENT based on Bill's notes for Practical Wireless. COHERENT operation The following amateurs now have the program: DJ7HS DJ7PO DK80K G3CCH G31RM G3XHQ HB9DDO KOLR KD41DY OE6WTD VE21Q VE30XX Can I ask these operators to let me know when they can operate and on what frequencies. If there is sufficient response I will issue a special list to all who have the program. Output power The earlier designs for CCW all tended to concentrate on fairly low power. I may be wrong but I get the impression that this was because it was comparatively difficult to generate more than the odd watt or two while still maintaining the necessary stability. This gave them an excuse to emphasize the improvements in signal to noise ratio possible with CCW. Much of this does not apply now that very stable transceivers are available together with the fact that COHERENT can tolerate something a little short of perfect stability. This leads to the suggestion that it will probably be better to use more power and still have the benefits of CCW giving a much better chance of perfect (or at least much improved) copy. Copying CCW Some years ago there were several designs for morse regenerators. The usual idea was to use a phase-locked loop, often an LIV1567, which decoded the morse into pulses which were then used to key an audio oscillator. I tried some of these and found that the main problem was the delay before the loop locked an every new mark. I also did not like the quiet background, it just didn't seem right not to have other signals or noise present even at a low level. Something like this is present when copying CCW. I first tried wearing two sets of headphones, then one pair of headphones with CCW to one ear and the receiver output to the other. The next development was to add an audio mixer similar to that described in the 1991 ARRL handbook, page 28-7. Keeping the receiver output in the background made things feel a little more natural. However, if the CCW signal is fairly readable without using the sample and dump filter or COHERENT things tend to get a little confusing! I am wondering if it would be possible to include a notch filter in the receiver output to reduce response at 800 Hz. Audio output from the computer Audio from the COHERENT program is via the internal speaker. This may suit some people but I much prefer headphones. When I tried to get the audio out through a rear socket on the computer I discovered that both sides of the speaker have a d.c. potential above earth so a simple output jack socket with one side earthed was no good. In the end I used a stereo socket with both lines clear of the earth and an insulated screened two core microphone cable. The other end of this can be connected through capacitors to wherever it is needed. This may not be the approved way but it works. In fact I do not use the capacitors but have a small isolating audio transformer which connects into the audio mixing circuits used to connect all audio outputs to one pair of headphones. This has the advantage that a capacitor can be added across the secondary improving the note. Tuning a CCW station Bill's program requires that the audio output from the receiver be within ñ5 Hz of 800 Hz. How do we tune to this degree of accuracy? The old way was for stations to agree on a frequency in advance. This was alright when experimentation was taking place. Now we have a much easier method of sending and receiving and, hopefully, there will be more stations around, something closer to normal CW methods of operation is desirable. Using higher power it will probably be possible to hear the sync train being sent in the normal receiver channel so it will then be necessary to be able to set the beat note correctly. Ideas on how to do this will be welcome. If you have a good ear for music you may be able to do it by just listening. I use the BARTG Toni-tuner which is more than adequate but a little more complicated to build than should be needed. Perhaps it could be simplified to"cover a small range around 800 Hz. Perhaps the signal could be mixed with an 800 Hz tone and tuning carried out until the notes are the same. As another suggestion perhaps the LM567 could be used to detect the sync train and indicate when tuning is correct. Something fairly simple is needed to ensure correct initial tuning. Has anyone any more suggestions? If so, please try them out and let me know. COHERENT After reading my first attempts at writing a CCW program in issue number 23 Bill de Cade wrote telling me how he thought I should go about it. The following paragraph is from his letter:- When receiving CCW (or any kind of CW for that matter) - the bottom line is that we want to measure the AMPLITUDE (not the frequency) - at 800 Hz, and compare this to the amplitude (also at 800 Hz) when the signal is known not to be present (i.e. to the amplitude due to the background noise alone when it is passed through the same filter) - in order to make the best guess as to whether the key was up or down at the transmitter end during that particular keying window. Furthermore, to be useful at all, we need to measure the amplitude of our specific 800 Hz component accurately even though there are sure to be many other (presumably much stronger) signals in the receiver's passband at the same time. It should go without saying that as soon as we look at the analog signal in a yes/no digital way, i.e. by running it through a limiter (or a zero crossing detector, which amounts to the same thing) - ***which is what 1, G31RM was doing*** - then we are throwing away virtually all the amplitude information in the signal (we could still say whether the signal was above or below the axis i.e. whether its sign was plus or minus - but that's about all). The only information we are left with is the frequency of the strongest signal in the passband at the time, and that isn't likely to be of much use to us in trying to recover some CCW signal in there somewhere below it. The beauty of the Petit system is that it allows us to estimate reasonably accurately what the amplitude is at 800 Hz, without too much interference from other nearby signals: without any interference at all from other signals spaced exactly +/- 10 Hz away, regardless of their amplitudes. The old Petit analog filter is getting long in the tooth. It's expensive, diff icuit to tweak up, and even though its performance is quite reasonable in the circumstances, it's really nothing to write home about today. If you want to replace it with software you can use the little Sigma-Delta board and some software running in a PC. The S-D board samples at 7,200 samples per second, and digitizes each sample to 7-bits resolution. A 100-msec window would consist of 720 samples to be processed, representing 80 cycles of audio at 800 Hz. In other words, 9 samples spaced evenly over each cycle of the audio waveform, repeated 80 times over. The existing FFT algorithm computes the amplitudes of many frequency components (typically 256 of them with a 512-point FFT) - in a reasonably short time. But we are only interested in the amplitude of ONE frequency component, namely; 800 Hz. To calculate its amplitude based on our 720 samples is a piece of cake! It is easy to do it in the background, without ever having to turn off the interrupts. If we use all 720 sampled data points, we will be able to compute the amplitude of the 800 Hz component to a precision of 7 + 9 = 16 bits. This is an accuracy of one part in 32,768 (around +/- 0.003 percent error), with a dynamic range on the order of 90 dB. My guess is that these specs would be better than the best you could possibly get out of a Petit analog filter. The frequency response curve would be EXACTLY THE SAME as that of the Petit filter, that old familiar comb-notch job with the main lobe at Boo Hz and nulls at 790, 810, etc. Yes it is true that nearby signals would contribute errors in the estimate of the amplitude of the 800 Hz component, but NO MORE and NO LESS than you'd get with Ray's circuit. In other words, the performance using software and that little Sigma-Delta board would be much better, the cost would be much less, and there is nothing to tweak up to make it work property. After reading that I decided that advanced software programming was not for me and that I should stick to hardware! There was a second letter written a few days later during which time Bill had written the COHERENT program. That really did convince me. I will have more from Bill's letters next time. 2 Briarwood Avenue, Bury St. Edmunds, Suffolk.