PXI 2527 & PXI 4071 -Questions about EMF considerations for high accuracy measurements and EMF calibration schemes?

[Kjell-E]

Hi!

I need to perform an in-depth analysis of the overall system accuracy for a proposed system. I'm well underway using the extensive documentation in the start-menu National Instruments\NI-DMM\ and ..\NI-Switch\ Documenation folders...

While typing the question, I think I partially answered myself while cross-referencing NI documents... However a couple of questions remain:

If I connect a DMM to a 2 by X arranged switch/mux, each DMM probe will see twice the listed internal "Differential thermal EMF" at a typical value of 2.5uV and a max value of less than 12uV (per relay). So the total effect on the DMM uncertainty caused by the switch EMF would be 2*2.5uV = 5uV? Or should these be added as RSS: = sqrt(2.5^2+2.5^2) since you can not know if the two relays have the same emf?

Is there anything that can be done to characterize or account for this EMF (software cal, etc?)?

For example, assuming the following:

* Instruments and standards are powered on for several hours to allow thermal stability inside of the rack and enclosures

* temperature in room outside of rack is constant

Is there a reliable way of measureing/zeroing the effect of system emf? Could this be done by applying a high quality, low emf short at the point where the DUT would normally be located, followed by a series of long-aperture voltage average measurements at the lowest DMM range, where the end result (say (+)8.9....uV) could be taken as a system calibration constant accurate to the spec's of the DMM?

What would the accuracy of the 4071 DMM be, can I calculate it as follows, using 8.9uV +-700.16nV using 90 days and 8.9uV +- 700.16nV + 150nV due to "Additional noise error" assuming integration time of 1 (aperture) for ease of reading the chart, and a multiplier of 15 for the 100mV range. (Is this equivalent to averaging a reading of 1 aperture 100 times?)

So, given the above assumptions, would it be correct to say that I could characterize the system EMF to within  8.5uV+- [700.16nV (DMM cal data) + 0.025ppm*15 (RMS noise, assuming aperture time of 100*100ms = 10s)] = +-[700.16nV+37.5nV] = +- 737.66nV? Or should the ppm accuracy uncertainties be RSS as such: 8.5uV +- sqrt[700.16nV^2 + 37.5nV^2] = 8.5uV +-701.16nV??

 As evident by my above line of thought, I am not at all sure how to properly sum the uncertainties (I think you always do RSS for uncertainties from different sources?) and more importantly, how to read and use the graph/table in the NI 4071 Specifications.pdf on page 3. What exactly does it entail to have an integration time larger than 1? Should I adjust the aperture time or would it be more accurate to just leave aperture at default (100ms for current range) and just average multiple readings, say average 10 to get a 10x aperture equivalent?

The below text includes what was going to be the post until I think I answered myself. I left it in as it is relevant to the problem above and includes what I hope to be correct statements. If you are tired of reading now, just stop, if you are bored, feel free to comment on the below section as well.

************************ 

The problem I have is one of fully understanding part of this documenation. In particular, since a relay consists of (at least) 2 dissimilar metal junctions (as mentioned in the NI Switch help\Fundamentals\General Switching Considerations\Thermal EMF and Offset Voltage section) and because of the thermo-couple effect (Seebeck voltage), it seems that there would be an offset voltage generated inside each of the relays at the point of the junction. It refeers the "Thermocouple Measurements" section (in the same help document) for further details, but this is where my confusion starts to creep up.

In equation (1) it gives the expression for determining E_EMF which for my application is what I care about, I think (see below for details on my application).

What confuses me is this: If my goal is to, as accurately as possible, determine the overall uncertainty in a system consisting of a DMM and a Switch module, do I use the "Differential thermal EMF" as found in the switch data-sheet, or do I need to try and estimate temperatures in the switch and use the equation?

*MY answer to my own question:

By carefully re-reading the example in the thermocouple section of the switch, I realized that they calculate 2 EMF's, one for the internal switch, calculated as 2.5uV (given in the spec sheet of the switch as the typical value) and one for the actual thermocouple. I say actual, because I think my initial confusion stems from the fact that the documenation talks about the relay/switch junctions as thermocouples in one section, and then talks about an external "probe" thermocouple in the next and I got them confused.

As such, if I can ensure low temperatures inside the switch at the location of the junctions (by adequate ventilation and powering down latching relays), I should be able to use 2.5uV as my EMF from the switch module, or to be conservative, <12uV max (from data sheet of 2527 again).

I guess now I have a hard time believeing the 2.5uV typical value listed.. They say the junctions in the relays are typically an iron-nickel alloy against a copper-alloy. Well, those combinations are not explicitly listed in the documenation table for Seebeck coefficients, but even a very small value, like 0.3uV/C adds up to 7.5uV at 25degC. I'm thinking maybe the table values in the NI documentation reffers to the Seebeck values at 25C?

[Henrik_Volkers]

Seebeck EMV needs temperature gradients , in your relays you hopefully have low temperature gradients ... however in a switching contact you can have all kind diffusions and 'funny' effects, keeping them on same temperature is the best you can do. 

Since you work with a multiplexer and with TCs, you need a good Cold junction ( for serious calibrations at 0°C ) and there is the good place for your short cut to measure the zero EMV. Another good test is loop the 'hot junction' back to the cold junction and observe the residual EMV.  Touching (or heating/cooling) the TC loop gives another number for the uncertainty calculation: the inhomogeneous material of the TC itself..

A good source for TC knowledge:
Manual on the use of thermocouples in temperature measurement,
ASTM PCN: 28-012093-40,
ISBN 0-8031-1466-4

(Page1): 'Regardless of how many facts are presented herein and regardless of the percentage retained,

                all will be for naught unless one simple important fact is kept firmly in mind.

                The thermocouple reports only what it "feels." This may or may not the temperature of interest'

Message Edited by Henrik Volkers on 04-27-2009 09:36 AM