The first problems I noticed were that the AO waveforms would work fine for a while and then would start getting clipped, and when the AO software was shut down, the outputs would continue to generate a low frequency and very uneven waveform. At the same time, the AIs would start to show a periodic signal, a short-duty-cycle negative going pulse, up in the kilohertz range (I assume some internal clock signal). I initially assumed this was a software problem, or at least brought about by software, and I could clear it up by restarting the laptop. However I soon realized that the problem occurred more quickly after I had just rebooted than when I first turned it on in the morning, which is the classic symptom of thermal problems. I later confimed this hypothesis by actually blowing into the slot and watching the AI readings return to normal after about 10-20 seconds, and then I made that series of temperature measurements.
The AOs were driving 3 V sine waves into 20 kOhm impedances at 300 Hz (impedances confirmed with a VOM). While I intially thought the AOs might be related to the problem, I later found that the anomalous AI readings would appear eventually even if I just turned on the device test panel in the Measurement and Automation utility and watched the AI channels, without driving the AOs. This test sometimes showed the AI symptoms on both laptops, without any external connections to the card. The AOs _may_ have made it happen a little faster, especially when I had them at 5V, but it was hard to tell for sure. On the other hand, I could also go days without seeing the problem sometimes, apparently depending on room temperature.
I can't say exactly what internal temperature it was at when it started to show the anomalous readings, since the anomalous waveform would appear on the temperature sensor channel as well whenever the symptoms began. However, the last clear readings I saw before it flaked out were around 59-60 C, and since the main laptop heatsink was reading 4-5 degrees above where it had been when I measured the card at 56 C, basic principles of thermal physics would suggest the critical point must have been around 60-61 C.
I never saw anything running as high as 70 C in any case.
After several days of normal operation, the card once again resumed this anomalous behavior last Friday, unfortunately while I was out of the lab, so I didn't get back until it had been doing this for a while. On Monday, the grounded AI channels all seemed to be permanently fixed to a 10 V reading (temperature channel reads about 8 V, ~ 800 C...)and the AOs failed to respond, leading me to conclude that thermal runaway had finally gotten to it. I have also noted that now, contrary to my prior experiments, the card seems to thermally load the laptop heatsink enough for it to show a precipitous temperature increase of over 4 degrees after I plug it in, until the fan kicks in and starts fighting to hold it down. Of course the PC Card controller is supposed to cut off shorted cards before they try to burn the laptop too, but I wasn't going to take chances with leaving it in too long. In conclusion, this is an ex-DAQ card.
The only thing I would say I was doing that might be considered unusual was that I was routing the AI start trigger out of PFI0 (making sure that nothing else was driving it of course) and connecting that to the CTR 0 source, basically using the analog trigger to do hardware-based peak counting, which was working quite well, BTW.
Like I said though, it could well just have been a bad unit. Maybe these do work fine most of the time. However, it seems pretty clear that the design is close to the edge thermally, and I would just have preferred to know I was getting a close thermal design so I could have taken precautions, like a double slot with a fan, before I invested money in the rest of the hardware. If you know "some laptops" can cause it to overheat, I mean, it would be useful to know which ones it works in and which ones it doesn't. If NI won't say what thermal conductance or thermal rating they require, and Compaq won't tell me how much they provide, a list of compatible systems is the only thing going.
At this point, I guess I would be happy to trade it in for a similar PCI card, since I think I'm just going to get a (sigh) $900 external PCI/PC Card adapter that comes with a reassuring fan. I've already spent that much in time futzing around with this.
Thanks,
Ben Buckner
