PXI-4351 with TBX-68T thermistor errors

[elroy]

I am attempting to accurately measure temperatures using my PXI-4351 DAQ with TBX-68T and multiple thermistors (9 thermistors).  When exposing all thermistors to the same temperature (room ambient), I noticed that the temperature displayed at each subsequent channel decreases by approximately .1C .  Thus, the temperature displayed at channel 11 is almost one full degree C lower than that measured at channel 2.  To trouble shoot the problem, I have installed a fixed resistor in place of one of the thermistors and moved it from channel 2 to channel 11.  It appears that the voltage measured across this fixed resistor increases slightly as I move it from channel 2 then channel 3, then 4 etc. FYI, I am using the 25 microamp constant current source.  I have also verified that I am not exceeding the total resistive load limit of 600 Kohm.  Does anyone have any idea what might be the source of this voltage measurement error?  Any suggestions would be greatly appreciated.

[Henrik_Volkers]

Shot in the dark: Last time you made a auto/self calibration?

Message Edited by Henrik Volkers on 05-25-2010 09:48 AM

[elroy]

Henrik  ....Thanks for the reply.  The PXI-4351 is a legacy card and I do not believe that it has a calibration utility.  I looked in the MAX and in the device specific VI library but have not located anything that provides a means of calibrating out the offset.  In preparation of this measurement task, I did have my PXI-4351 card calibrated by a NI certified calibration facility.  I also had my thermistors calibrated with individual Steinhart Hart equation coefficients provided.

One of the Electrical Engineers on our staff has suggested that maybe the voltage error is a result of multilplexing the channels faster than they can provide a stabel reading.  Any thoughts with regard to that theory?

[Kevin_Price]

I've been using a PCI-4351 on and off for a few years now, but haven't really touched the core code in quite a while.  I vaguely recall that it is pretty slow at sampling multiple channels, that being the trade-off for 24-bit resolution.  As I recall the primary programming control over the sampling rate is based on setting a line filter frequency at one of several presets like 60 Hz, 50 Hz, 10 Hz.  And I *think* I recall seeing a similar channel-to-channel influence at the higher line freqs / higher sample rates.  I got more favorable results with the 10 Hz setting, but the effective sample rate was also much slower.

I eventually decided that instead of asking for the *next* temperature sample and having to wait for it, I would ask for the most recent temperature sample (from an unknown fraction of a sample period in the past) and receive it immediately.  This made the software more responsive, while also allowing me to painlessly use the 10 Hz (slower, better quality sampling) setting.  The only downside was that I was always finding out what the temperature used to be just a little while ago.  For our tests, that wasn't a bad trade-off.

In summary, it may help to use slower sampling based on a lower line filter frequency.  That may in turn make it advantageous to do a buffered acquisition and then explicitly ask for the most recent past sample(s) rather than the next future ones.

-Kevin P