Yes Peter, Mine moves down when the TX comes on too. I can't say exactly by how much. I had thought it was my not wonderful power supply, but it may be some kind of loading issue on the oscillator. To be frank, wspr asks a lot of an oscillator. If you were on cw, you'd hardy be likely to notice a few htz change. Here it counts, especially since the micro controller is varying the frequency by 1.4 htz to create the wspr tones and trying to hold the oscillator stable, though I don't think it does that when wspr is actually transmitting. Hans wrote me some interesting words the other day when I was discussing the way my fsk width changed. I hope he won't mind me quoting them here - it will save him the trouble of repeating them, and I know he has a lot of demands on his time. Here is his note to me on changing width of fsk when frequency alters:
HANS QUOTE STARTS:
"Firstly it is important to realise that a graph of frequency (vertical axis) against varicap diode voltage (horizontal access) shows a curve, not a straight line. E.g. see this image??from my varicap diodes page??.?
The FSK applied by the microcontroller is a reverse voltage to the varicap (the 5mm reverse-connected Red LED In the kit). The microcontroller assumes that the frequency/voltage line is a straight line. It uses the FSK per Hz ("FSK Adj.") setting which you have entered in the configuration menu. However, this setting is only valid at one particular reverse voltage (i.e. one particular frequency). So if the frequency drifts far from the one you have set, then the actual FSK Adj value would differ from the value you set. This would case the FSK size to grow or shrink.
If your FSK Adj value is large, and your oscillator frequency is a long way from the actual target value you have set in the configuration menu, then the GPS will steer the frequency a long way, and this will cause the operating point to move along the line to a point where your FSK Adj value is now incorrect to generate the amount of shift that you need.
What FSK Adj. value do you use?
I think your problem could be resolved in one of two ways:
1) Put the kit in the TEST mode, with the GPS connected. In this mode, the display behaves as a frequency counter. The GPS doesn't lock the oscillator, it is just used to accurately measure the frequency so that you can tune it accurately. Then use the trimmer capacitor to tune the frequency as close as possible to the target frequency value you have set in the configuration menu. Now change the "FSK Adj" value experimentally to get the right amount of shift on Argo. Remember that in TEST mode it shows a square-wave, set the Speed to 003 for example and then you'll get 3 seconds low, 3 seconds high. With FSK Hz set to 4Hz, you should see a 4Hz shift if your FSK Adj value is set correctly. So the trick is to get a correct value of FSK Adj, to get the correct shift, and on the correct frequency (very close to the frequency you set). Then when you move to WSPR mode, the microcontroller will only have to make small adjustments to correct any drift occurring. The position on the curve I described will not move far from where you set the frequency up, and so the error in FSK will be small. In other words: the microcontroller + GPS is intended to be able to slightly correct the frequency due to drift. It is not intended to completely fix any situation in which the oscillator isn't well-tuned in the first place. Do you see what I mean?
2) If all that still doesn't work, then you could try increasing the value of C3 from 22pF to 47pF (or just put another 22pF in parallel with the existing one). This would let the microcontroller have a more aggressive control of the frequency drift. That gives it a greater tuning range, but with less precision. Anyway it would mean that the GPS would be able to correct a larger frequency error, without moving the operating point on the curve as far.?
Generally I think that if you set it up carefully as in 1) above, then that may well solve your problem. This is how I would tackle it, anyway. Please give it a try and let me know how it goes."
END OF HANS'S remarks.
That explains why it happens - and following the instructions put it right. BUT - the solution only lasts until the oscillator moves again, which in the case of mine, it certainly will do.?
In your case, putting the thing in a box may also change the frequency slightly from where it was set. We are talking very small changes here and the environment is important for that.
I am compounding the challenge for my kit by running it at half a watt output (9.2volts on two BS170 fets). This generates extra heat even though I made a decent sized heat sink for them. I tried to get rid of the drift by clipping a crystal heater onto the 10.140 crystal. I have two old 16000mhz down convertors which use have two in each of them. These are old Murata BM500N crystal heaters in them. They are meant for twelve volts but I applied the 9.2volt pa voltage to them and they warm the crystal. However, they don't have a lot of effect on the frequency drift of -2 to -3 that I am getting on my 'hot' kit. I suspect that the frequency changing effects may be on changing capacitance by heating the board through he legs of the output fets. It is warm underneath them and it would only take tiny changes in capacitance to move the frequency by one or two parts in ten million which is all we are talking about. As I said, I have voluntarily compounded the issue by running it hotter and we really are talking tiny frequency changes. Nevertheless - I get loads of decodes from far and wide and get into VK about once every three days depending on the condx. It's a great little kit, even if it is a bit hands on at times. Keeps me out of mischief anyway and for next to no outlay too...
:)))
Regards and I hope Hans is OK with my quoting his explanation.?
73s
Tony G0BZB
