During the BARTG test I had some weird moments after I QSY'd from 20m to 40m -- DM780 would key the K3's PTT (via HRD IP Server) but there would be no audio into the Line In port, and thus no RF output. Sometimes rebooting the K3 would work, other times quitting and re-launching DM780 did the trick. I'm not 100% certain of the cause but I'm leaning towards software rather than hardware.
Then I noticed low RF output at certain audio frequencies, and this only in AFSK-A reverse (USB) mode; no trouble in normal (LSB) AFSK-A mode, nor in DATA-A normal and reverse. The output power would be OK when the center frequency was set low on the waterfall (+/- 400 Hz, for example) but would drop off at higher Fc setings. And the cutoff point seemed to vary -- sometimes I could get full 100W RF out at around 1500 Hz Fc (I typically use 1530 Hz as Fc for RTTY), other times the output would drop above 900 Hz and be effectively zero at 1300 Hz and above. After some messing around with soundcard and line input levels I seemed to have gotten it somewhat sorted out, but as I did all that in the middle of a contest while in semi-panic mode I didn't take too many notes.
I did some basic tests today by transmitting RTTY diddles at 100W into the MFJ-267 dummy load/wattmeter while in AFSK-A REV mode; mic gain is adjusted for 3 flickering bars of ALC as per the advice of Elecraft. I got full 100W reading on the meter at 2200 Hz Fc, but it steadily decreased as I inched upwards -- at 2300 Hz Fc the output was down around 60W, and by 2400 Hz the wattmeter needle barely moved. But when I switched from REV (USB) to NOR (LSB) I get full output right up to the top edge of the waterfall (2900 Hz Fc) -- pretty much what I experienced yesterday during the RTTY test.
I then shut the whole system off for a while and repeated these tests an hour or so later to see if things changed from a "cold" state: starting out my output at 2200 Hz was down around 10 to 15W. The output then slowly rises as PA temperature (as measured by the K3's built-in PA TEMP meter) reaching full 100W at 43 deg C. But after cooling off (PA: 35 deg C, front panel: 32 deg C) power output remains at full 100W! This is driving me absolutely bat-shit crazy.
I'm not really concerned that I won't be able to transmit RTTY at high center frequencies as my operating style is to find and click on a signal in the waterfall, center it at 1530 Hz, then switch to a narrow filter setting; I'm more troubled by how the cutoff frequency changes, sometimes below 1530 Hz, and apparently is a function of temperature. It doesn't appear to be a PA problem -- the same power drops occur when bypassing the PA unit and running at 10W, and I'm getting full power in all other modes -- so I'm going to guess it's an issue with the DSP board and its handling of audio at the Line In port. More tests to follow, and when I have a better set of test results I'll contact the boys in Aptos.
Update 26-Jan-2010: I believe it has been solved -- the only things I didn't check until today were the filter parameters in the CONFIG menu. Sure enough, I had FL3 set to 2.4 kHz even though the filter is actually a 2.1 kHz filter. Unsure how or why this caused the problem, but since setting it to the correct setting I am now able to get full output power at the mark frequencies indicated by the K3's MARK-SH setting (2125, 1445, 1275 or 915 Hz). Not sure how the FL3 parameter got changed; perhaps inadvertently while running K3-EZ, or perhaps it was set improperly during my assembly and setup back in January '08. In any case, all's well that ends well. My apologies to Wayne, Eric, Lyle, Don, Greg, etc. for bothering them with what turned out to be user error. I will now sit in the corner wearing my "Dumb Ass" hat as penance.