I can't say I've seen any problems on any of the sixteen KATADCs that have been operating on KAT-7 for the past 3 years.
Your attached spectrum looks mirrored about Fs/4. Is this normal for your system (you flip left and right halves to assemble your band; complex sampling?)? Or is this just the way you did your Python FFT? If so, your attached spectrum looks ok to me for a sloppy highpass filter. Can you more accurately describe the problem, or attach a "known good" spectrum? What should I be looking for? Are you concerned with the rolloff down to DC? I assume you've checked the clock for power and stability? Cabling and connectors are usually my first suspects in known-good designs. The KATADCs are natively interleaved (multicore) devices. Could it be that one of the cores or the alignment's not right? (check every fourth sample in your snap output). Do you have to wait for it to cool down before it works, or does it come right if you try to reprogram the FPGA a few times consecutively? If it works after a few consecutive resets, then it could indicate a digital alignment problem. If you have to wait for it to cool down before it works, then I suspect a bad connection somewhere (analogue clock, analogue signal, PCB fault or power supply problem). Jason On 22 Dec 2013, at 01:51 , Gary, Dale E. wrote: > Hi All, > > From time to time, one of our KatADC boards (on ROACH2) shows an anomalous > spectrum. The bottom plot of the attached file is the result of 100 "snaps" > of 8192 time-domain samples right out of the ADC, each FFTd in Python and > added. Normally the spectrum is nice and flat, but intermittently the KatADC > board gets stuck in this mode, and stays there. The only solution seems to > be to shut down the ROACH and let it sit for awhile, which would indicate a > thermal problem. However, the KatADC temperature does not seem that high--it > was about 68 C when I snapped this. The other KatADC in the same ROACH2 is > fine. Has anyone seen this, and is there a solution? > > Thanks, > Dale > <katadc_prob.png>
