[Thermistor & Sampling Rate] effect on measured voltage?

[Nocturn]

Hello,

I don't know if this question belongs really to this board because it is more specified on the DAQmx-cards.

Well I am using the PCIe-6353 and the USB-6363 to acquire signals from different thermistors and a PT100.

All sensors are measured in differential mode with 1MOhm bias-resistors and are fed by an external constant

current source ( - those can be feed with 5V by the DAQmx-card or by an external source [makes no difference]).

In a test-setup I used very short shielded twisted-pair cables to wire the sensors to the connector-boards.

The problem is that the measured sensor-voltages (same ambient conditions) will differentiate at different

sampling rates.

For example:

# when I start a measurement with 100 samples @ a sampling rate of 1k I get some voltages about 0,110V

# when I start the same measurement with 1000 samples @ a sampling rate of 10k the voltage will be 0,112V

 (=2mV offset to the previous one)

... the offset affects the whole signal and  increases with rising sampling rates (6mV @ a sampling rate of 100k).

Is it possible that the sample-clock-voltage affects the DAQ-channels? Or may it be that there are some stray capacitances

or inductivities even while using short (<10cm) STP cables?

Is there a way to handle this?

Many thx in advance!

[PackersFan]

Are all your signals in this voltage range?

As you speed up your sampling rate, the multiplexer on your board will be running faster and therefore your chances of experiencing ghosting increase.

Don't know if this is your problem, but certainly possible.

[johnsold]

As you change from 1 kHz to 10 kHz you see about a 2% error.  If you are seeing this on a 10 kohm thermistor, the shunting impedance is about 500 kohm. At 10 kHz this is the reactance of 32 pF. That is the capacitance of about 30 cm of coaxial cable .  It not unreasonable that your shielded twisted pairs could have 3 times as much capacitance.

If you plot all the samples on a graph, are they all the same (for a given sampling rate) or do you see periodic or random time variations between samples?

Lynn

[Nocturn]

Thank you very much for your quick response!

@Chris: Yes, almost all the thermal sensor-signals are in this voltage-range and have the same effect on higher sample rates. 

@Lynn: They are almost all the same with little periodical changes due to convection (air).

------------------------------------------------------------------------------------ 

There is also a slight (and steady) difference between the outer two sensors of the thermistors (ntc) and the inner three.

- These five are series-connected. I assume that this behavior is an geometrical emc-issue... 

I have tested yet another sensor (photoelement, connect by coax) with different sample rates and samples read.

When I use a count of less than 12 or 10 samples read at different rates I measure a steady and very small noise on the signal.

The noise will increases with the number of samples read - unattached by different rates.

When I read more than 12 samples (or much clearer with more than 100 samples), then the rate affects also the signal by an offset

depending on the rate itself (as described previously).

If it is really the switching of the multiplexer, or some (dis-)charging stray capacitances which affects the line, then it's probably problem

with the wave impedance of the cables. - The waves are reflected at the cables end and interfere. - With every new sample read another wave is reflected

and interferes with the previous one (when in phase).

I will give it a try by changing the STP to coax and use an appropriated in- and output impedance.

Regards,

NoC