Choosing bias resistors for differential voltage measurement from changing impedence sources

[ramcd]

Hello,

I am using NI 6225 (M Series DAQ) to read voltages from thermometers (RTDs - negative temperature coefficient type). There are 8  thermometers in series powered by a constant current source. The resistance of these thermometers vary from about 70-80 ohm at room temperature, and become as high as 12000 ohms when cold. The voltage measurement is differential and the thermometers are 'floating voltage sources' (this is my assumption, a clarification would be helpful as well) and hence bias resistors are needed at the DAQ analog input channels. I am refering to the Field Wiring and Noise Consideration for Analog Signals application note, which says that DC coupled voltage measurements from low impedence sources need only one bias resistor (10-100 kohm) to be connected between AI- and AIGND on the DAQ. If the impedance sources are high two bias resistors are needed, each between AI+, AI- and AIGND. My queries are:

1) It is not advised to connect two bias resistors, one between AI+ and AIGND while other between AI- and AIGND in case of low impedance sources? The reason I want to know this is because by thermometers start out with low impedance and end up as very high impedance sources. Changing the circuitry during an experiment would be difficult and so I would like to start by putting two bias resistors per AI channel (i.e AI+ to AIGND and AI- to AIGND).

2) Since the resistance of thermometers vary significantly, what bias resistor values to I use? It is mentioned in the application note that the bias resistors should be high enough to ensure that the signal source (thermometers in this case) remain floating and not load the signal source, but should be small enough to keep the voltage in range of the input stage of the device (I am a little confused what this range of input stage means! For a heads up, the voltage from thermometers will NOT exceed 15 mV, so the range on the DAQ is set to -200 mV to 200 mV). Are there guidelines to choose the values of these bias resistors, given the signal source impedance?

thanks

Ram

[johnsold]

Ram,

Some of your numbers do not seem to add up.

First, I have never heard of an RTD which changes from 70 to 12000 ohms.  One of Omega's RTDs ranges from 81 ohms at -200 C to 390 ohms at 850 C. That sounds more like a thermistor, although the values seem lower than what I usually see.

You should not need any bias resistors. The temperature sensors (total resistance of 8 sensors at 12000 ohms) is low enough to provide the bias path.

If you are keeping the voltage across each sensor to 15 mV or less and running constant current, then the current must be no greater than 1.25 uA (15 mV/12000 ohms). At the high temperature end a 1 degree change may be represented by about 1 ohm change in the sensor. That would result in a total voltage across the sensor of about 88 uV and a change of about 1.25 uV. The 6225 has a sensitvity specification of 5.2 uV on the +/-200 mV range and accuracy of 112 uV.  This suggests that your temperature resolution at room temperature would be about 4 degrees and tha accuracy would be ~80 degrees (not including the non-linearity of the sensor).

So, I do not think you need bias resistors. You really do not need differential connections (assuming one end of the sensor chain can be grounded). Measure the 8 voltages and subtract.  What you need to do is the calculations to see if you can get meaningful results.

Lynn

[ramcd]

Hello Lynn,

Thank you for the clarification. I am using Cernox thermometers which do have large resistance variation in the temperature range they span. I am measuring temperatures as low as 1.8 Kelvin at which these thermometers have resistance as high as 12000 ohm (check out the CX-1050, this is the one I am using), upto 100 Kelvin (somewhere around 200 ohms). The sensitivity of these sensors around 2 Kelvin is also pretty high (about 10000 ohm/K) so the temperature change as small as 10 mKelvin can be resolved. Using a constant current supply (1 uA, 3 uA and 10 uA) and NI 6225, the resistance of 8 such thermometers is measured by a 4-wire scheme.

However reading through the NI literature/application notes/6225 manual and going through the advantages of using differential measurements, I ran into confusion about the bias resistor requirements. Is it from experience you say that 12000 ohm signal sources would not need any bias resistors, or are there any guidelines? Another question then is, when are we supposed to use the bias resistors while doing differential measurements? Here is another post of mine describing a problem I earlier faced while setting up the DAQ. Putting a bias resistor helped me in this case.

Thanks

Ram

[johnsold]

Ram,

Thanks for the link to the Cernox resistors. I had never heard of those.  I do work with some people doing cryogenic measurements occasionally, so I will keep these devices in mind.

Your first post in this thread mentioned wiring the sensors in series with a current source. Is any point in that circuit connected to ground?  If it is, then you probably do not need any bias resistors or differential connections.  

The 6225 specifications give the bias current as 100 pA.  That is about 1 part in 10000 of your excitation current, so it may become noticeable at low temperatures. Bias current is only specified as a typical value and its variation with input voltage, channel, and DAQ device temperature is not defined. 

This is beginning to appear to be an interesting measurement problem. When microvolts and 100 pA can affect the results, selecting and validating a suitable measurement configuration is not trivial.  I would love to spend about a week in your lab working on this, but that is likely not an option for you or me.

Can you post a schematic diagram showing the current source, DAQ device, and sensor connections?

Lynn

[ramcd]

Hi Lynn,

A hand drawn schematic is attached. I believe that no point in the circuit is connected to ground except the wire shields (which is ofcourse not a part of RTD circuit).

I've shown only 4 thermometers but in reality there are 8 on my experiment...

[johnsold]

Ram,

I would try connecting the current source (-) to AI GND and switching to the RSE (Single-ended-Ground-Referenced) input configuration.Then you will have only one DAQ input connection at each sensor junction rather than the two you have now. Leave the shield connected to the source ground.

Are you more interested in resolution (detecting small changes) or absolute accuracy? 

Lynn

[ramcd]

Thanks Lynn,

Will try that out. I need accuracy more than resolution in my experiment. The temperature are expected to change very rapidly and significantly in magnitude, and hence measuring very small changes in not critical. Let me try doing the RSE measurement. Just for my clarification, is the AI GND on 6225 board internally connected to its electrical ground/chassis? The manual suggests that AI GND is to be used to connect bias resistors when doing differential measurements, while the ground 'lug' on the chassis be used to connect shields of sensor voltage cable. Are these two internally connected?

regards

Ram

[johnsold]

I do not know which grounds are connected. Check the manual or measure between the grounds with an ohmmeter with the power off.

Lynn

[johnsold]

The reason I asked about accuracy versus resolution is to determine whether a detailed error analysis is appropriate.  Accuracy being more important suggests that a good error analysis should be done to decide whether the system is good enough.

The manual for the DAQ device shows how to do an error analysis for its internal errors. Other things which need to be considered include the bias current error, any thermoelectric effects, the error of the current source, errors in the Cernox devices, and noise sources not included in the DAQ device error calculation.  Plus any other errors which might be indicated by a thorough overview of the entire system.

This is not something which can be done in a few minutes. Depending on how much information on some of the errors is available and how much needs to be researched, this could take several hours to several days.

Lynn

[ramcd]

Hi Lynn,

Thanks for bringing this out, will keep this in mind.

Ram