Multiplexing microvolt signals and appropriate signal conditioning

[alek_a]

Hi

We have 22 signals in our system. We need to be able to measure any subset of these 22 signals. Usually, we perform a number of 4-point IV measurements of separate devices on a SUT (sample under test). Any two of the 22 signals can represent the voltage pair of a 4-point IV measurement, depending on how the SUT is bonded. The signal level of the voltages is in the 100s of uV range and I need an accuracy as best as possible, thus unit-digit or even sub-uV. Up to now, we had all 22 signals connected to a box (selfmade) acting as a distribution point, where a connector (SMB) is available for each signal. We then manually choose which two of 22 the signals will be connected to a differential low-noise instrumentation amplifier (also selfbuilt), according to the SUT bonding schemes. The amplifier, with a gain of typically 1000, outputs the signal to an acquisition card. The current (excitation) is separately read out.

There is a problem with this. We need to manually switch connectors on the distribution box to select another device on the SUT and feed it to the amplifier. This is not good for a number of reasons that I cant go into here.

Therefore, I would like to multiplex these 22 signals to several outputs (for future expandability) that will be hard-wired to the signal conditioning. There will be no manual connecting/disconnecting of these signals in this way. The first question is: can a switch matrix or multiplexer be used to route microvolt signals reliably? If yes, then what kind: electromechanical relay, reed or a solid-state? Is the thermal EMF in the mux/matrix specs just an offset or is that noise?

The second question is: which acquisition card/signal conditioning solution using only NI products (thus no selfmade and legacy equipment) from the same line as the matrix/mux can achieve 1uV accuracy? For us, offset is not a problem since we measure the device in 2 quadrants and later cancel it in software but noise (and gain accuracy) is important. Our amplifier has a noise floor of 3nV/sqrtHz and a bandwidth of 15kHz. The system to replace it must be comparable. I have a choice between:

• the SCXI based ones, isolation instrumentation amplifier modules 1121 (4ch) or 1122 (16ch), with a noise floor that I calulate to 20 and 30nV/sqrtHz respectively. These already have some muxing capability btw.
  
• the PXI 628x with a noise floor that I calculate to around 10nV/sqrtHz
  
• the PXI 446x with the lowest noise floor of 8nV/sqrtHz as given in the datasheet
  

The universal 24bit 4-channel module NI 9219 that works with 2 samples per second has a very low total noise but at a very low sample rate and that I cant consider comparable to what we have now. Besides, the system must be able to have few kHz bandwidth for dynamic measurements in the future.

Are there any other ideas?

Greetings
Aleksandar Andreski

[RikPrins · Accepted solution]

Hello Aleksandar,

Looking at the complexity of you application I think it would be better if your contact our office directly and discuss your application with a Sales Engineer. That way I think we can be of better advise. If needed an appointment can be made so a Field Engineer can discuss your application with you onsite.

The phone number of our (Dutch) office is +31 (0) 348 433 466

If your are based in an other country please find the phonenumbers of all our offices by filling out the form on:

http://www.ni.com/niglobal/

Regards,

RikP - National Instruments Applications Engineering