Don't you hate it when you build a project, it works, then you put it into a case and find that it is no longer working as well as it was?
I built the VCXO-AXE 30m WSPR tramsmitter and it worked perfectly. I then put it in a nice case and began testing it with various portable antennas. At the start I received several spots of my transmissions but after a while the spots became few and far between until hours would pass without any report of my signal.
At first I thought this was due to poor conditions or the fact that the antennas I was testing were inefficient. Eventually I investigated by putting the TX on a dummy load and listening for it with my K3. I found that although I had initially set the VCXO-AXE up to transmit 40Hz above the bottom of the 200Hz wide WSPR band, the frequency had drifted low so that my transmission was now right on the bottom edge and possibly below it on many receivers since I couldn't guarantee my K3 readout was dead accurate.
It was easy enough to set the transmit frequency back to what I wanted it to be. But I still didn't receive any spots of my signals.
I looked a bit closer and noticed that although my signal started off as a horizontal line for about the first two-thirds of the transmit cycle, the last third curved down a little. The WSPR software was reporting a drift of -3Hz, which was typically what I had been receiving from those stations that had spotted me. There is a thread in the forum section of the WSPR website which suggests that few signals which drift more than 3Hz are reported. So it seemed probable that drift of my signal was preventing it from being spotted.
Before I put the transmitter into a box the frequency had been perfectly stable after the initial warm-up period. So it was obvious what the trouble was. I drilled a small grid of ventilation holes in the back of the case just above the PA FET and attempted to increase the thermal inertia of the VCXO chip by placing a blob of Blu-Tack over the top of it. This made not the slightest difference. I was still getting -3Hz drift reported on the decodes of my transmissions.
After further investigation I found that the worst heat generating component was in fact the 5V voltage regulator, which also happens to be fairly close to the VCXO chip. So I drilled a further set of holes above that. I also fabricated a heat shield from a piece of cardboard sandwiched between the PCB and the back of the case with Blu-Tack to try to induce most of the heat to go out through the holes instead of heading in the direction of the oscillator. This did make an improvement. I'm now mostly seeing reports of -1Hz drift or even none at all. And I'm getting a lot more spots of my WSPR transmissions!