I want to set up multiple RTDs or PRTS to measure temperatures in 4-wire mode with a data acquisition system. Seems like I could run the excitation current through all of them, as well as a couple of reference resistors, in series. Then each one I want to measure would be a differential voltage pair on my DAQ card.
One problem would be noise picked up in this big loop. I know how to measure and remediate noise, so I plan to deal with this.
Another problem would be if the mux chips have too low an open-channel impedance. I need to research this and know how to calculate the size of the error introduced.
A third problem would be the common mode interference in my differential measurements, because each channel is at a different average voltage. I’d have the ability to reverse my current (to cancel out thermal emfs) so this is something I could test, and I can predict its size based on mux specs anyway.
A forth problem is that this would require a steady current source, but I know how to shop for and estimate around this.
I figure a couple hundred millivolts per each of a dozen sensors and reference resistors still only uses the middle region of my daq range.
So,is there anything wrong with this scheme that I’m not seeing? I haven’t read about this approach, or seen it in any products - any idea why not?
I did an error analysis in Excel based on some assumptions. If you email me, I’ll email you the spreadsheet. The upshot is that, if you assume the same current flows through each PRT, and if the input impedance for each channel is 1M Ω, then you’ll get an error of about -0.1 °C at 500 °C.
The common mode voltage shouldn’t be very high. In the spreadsheet I assumed you would be using 10 PRTs. The calculated voltage across all the PRTs was less than 2 V. If we assume the constant current source is ground-referenced, then the worst common mode voltage will also be less than 2V. This is low enough that you shouldn’t have to worry about it.
As far as current reversal goes, it does improve the accuracy by canceling out thermal EMFs. All DC SPRT bridges use this technique. But I’m not sure how much it will buy you when using IPRTs.
First of all, yes - it would be nice to have a very steady constant current source. But it doesn’t have to be ultra steady. This is because you’ll want to measure the current using a precision (wire wound) shunt resistor. Secondly, there are lots of circuits out there for implementing a constant current source. One thing you want to pay attention to is the current source’s compliance voltage.
I don’t see any big problems with this approach. However, I’m not sure how the DAQ will “react” to the HI of each channel being tied to the LOW of the adjacent channel. Theoretically a DAQ with differential inputs should be able to handle this situation without any problems. But sometimes bias currents can do funny things…
>>I could run the excitation current through all of them, as well as a couple of reference resistors, in series
>you’ll want to measure the current using a precision (wire wound) shunt resistor.
Right, well, I hope to measure RTD resistance ratiometrically relative to the reference resistors, rather than calculating current as an intermediate step. This lets me adjust the whole system according to statistical analysis of the variances introduced by each measurement, and use least squares methods to minimize the total error.
The beauty of RTDs is that they turn thermometry into resistance measurement, and it’s pretty hard to beat resistance measurement by ratiometry to reference resistors.