Hi I'm designing for a PT 100 RTD and found some good information but am curious if anyone knows the difference between some of the current sources I'm seeing.
1. The file link I've attached is a white paper using the MCP609 and looks good but seems more complex than it needs to be. Why they don't just use the simple op-amp follower with a darlington bjt output driving a sense resistor on the emitter side? Is it just to avoid using a transistor or are they getting much better regulation using 2 op-amps?http://ww1.microchip.com/downloads/en/appnotes/00687c.pdf

2. Using the op-amp follower with a darlington bjt output driving a sense resistor on the emitter side technique it seems to me that the base current is unaccounted for if using the collector side for the DUT. You'd be off a the few milliamps it takes to activate the transistor. Am I missing something on that because no one seems to mention it?

I'm not an everyday circuit designer so I don't have all the little tricks in my head regarding op-amps anymore...

 

The circuit needs a current source for a grounded load (the PT sensor) such as the Howland circuit shown can do. The simple op amp follower with a transistor can't do that.

Yes, the BJT base current will cause an error (1/Hfe) in the constant collector current value. That's why a MOSFET is often used instead since it has no gate current.

 

Hi I'm designing for a PT 100 RTD and found some good information but am curious if anyone knows the difference between some of the current sources I'm seeing.
<snip>
2. Using the op-amp follower with a darlington bjt output driving a sense resistor on the emitter side technique it seems to me that the base current is unaccounted for if using the collector side for the DUT. You'd be off a the few milliamps it takes to activate the transistor. Am I missing something on that because no one seems to mention it?
<snip>

Your RTD will exhibit self-heating errors if you run more that a few milliamps through it. A pure Darlington (with no internal base-emitter resistors) will typically have a current gain (beta) of 10,000 or more, so the base current error will be less than 0.01%. You won't have a few milliamps of base current.
As Carl said, use a small MOSFET if you want to have zero error.

EDIT: A PNP current source can drive a grounded load.
The Howland (op amp) source is good, but it requires two pairs of matched resistors. It still might be the best way to go.

 

Thanks guys, The FET seems like a good solution too. I had come up with an idea to eliminate the error from the BJT using the attached circuit (without a Darlington) to ensure the exact design current, 1mA, passed through the RTD (the pic shows a thermistor because I didn't have the circuit symbol for RTD).
I'd be curious if you see any issues with it.
Using a 5V LDO and a 100Ohm RTD between 0-100C (delta R approx = 100 to 140 Ohms) I should never saturate the opamp over the RTD range so I'm feeling like this is a nice simple option for good regulation as long as a precision resistor is used for Rsense. I figure the LDO to supply the op amp and voltage reference ckt will minimize any supply fluctuation for a better regulation but I haven't done math to calculate a full scale error base on likely Vcc change yet.
Thanks for the input!

 

Thanks guys, The FET seems like a good solution too. I had come up with an idea to eliminate the error from the BJT using the attached circuit (without a Darlington) to ensure the exact design current, 1mA, passed through the RTD (the pic shows a thermistor because I didn't have the circuit symbol for RTD).
I'd be curious if you see any issues with it.
Using a 5V LDO and a 100Ohm RTD between 0-100C (delta R approx = 100 to 140 Ohms) I should never saturate the opamp over the RTD range so I'm feeling like this is a nice simple option for good regulation as long as a precision resistor is used for Rsense. I figure the LDO to supply the op amp and voltage reference ckt will minimize any supply fluctuation for a better regulation but I haven't done math to calculate a full scale error base on likely Vcc change yet.
Thanks for the input!

How are you going to measure the voltage across the RTD when it is not grounded? You could use an instrumentation amp, but that seems like an unnecessary expense.
If you use that scheme, you don't need the transistor. Most op amps can source way more than 1mA.

EDIT: There are voltage reference ICs, such as LM336, that are more stable with temperature than LM2936. LM336 is just an example. There are probably better ones available.

 

"How are you going to measure the voltage across the RTD when it is not grounded?"
-I was planning on just using a differential op-amp ckt to get the voltage across the RTD. The op amp might need to be an IN-Amp for the low offset and bias current, it is an extra part but I still need a filter and amplification ckt anyway so I'm not sure that's a negative. (of course that white paper suggests a Quad precision op amp which may be the best way to go)

"If you use that scheme, you don't need the transistor. Most op amps can source way more than 1mA."
-Yeah, I know this but with the Microchip ckt I would need at least 4 matched resistors (which I'm not sure how to do) and a precision Rsense whereas with the design I'm considering I only need the sense resistor to be precision (and really if I calibrate it with the trim pot it wouldn't really need to be precision either).

"EDIT: There are voltage reference ICs, such as LM336, that are more stable with temperature than LM2936. LM336 is just an example. There are probably better ones available."
-Ok cool, I hadn't even started evaluating the voltage reference yet and just picked that out of the parts list for illustration purpose. I'll check out your suggestion. Actually the way I have it drawn that would have to be an LDO to source enough current to drive the op amp and resistors without effecting the output. I might have to include a separate voltage reference for just the opamp input.
I wonder at this point if it's worth just going back to the Howland ckt and trying to figure out how to get precision matched resistors...

Thanks again for the suggestions, I'm really having a lot of fun getting back into circuit design.

 

Here is a simple current source that will allow you to use one amplifier for the current source, and the other one for the gain stage. If you want to eliminate the offset due to the 1mA current, then more design work will be required.

You could use a different reference voltage IC, but you might have to change the calibration scheme.

 

Here is a simple current source that will allow you to use one amplifier for the current source, and the other one for the gain stage. If you want to eliminate the offset due to the 1mA current, then more design work will be required.

You could use a different reference voltage IC, but you might have to change the calibration scheme.

Yeah I think I'm headed this way because I just checked Digi-key for precision surface mount resistors at .01% tolerance and they are $14.76 in singles. The matched quad precision array at +/-.02% tolerance is $19.86. That'll buy a lot of op-amps, transistors, voltage references, and LDO's!!!

This circuit will work fine but I think I'll use a 5V LDO to separate the opamp and voltage reference power from the fluctuating 12V power supply I'm using. Also found the Max6037A for a precision voltage reference which appears to be pretty good too. The combination of LDO and reference will create a very stable current.

One more question:
How do you determine that Pchannel fet will turn on fully with only with only 5V to work with on the gate? I've used fet's before that wont fully turn on until the gate reaches about 6-8V differential...

 

One more question:
How do you determine that Pchannel fet will turn on fully with only with only 5V to work with on the gate? I've used fet's before that wont fully turn on until the gate reaches about 6-8V differential...

Look at Vgs(th) in the datasheet. The maximum is -2V @ 1mA. Since the source will be at 2.5V, the gate will never have to go below 0.5V. Typically, it will be higher than that. If it were a problem, you could simply make R1 a lower value, and reduce the reference voltage to maintain the 1mA through R1.
Also, you don't have to use a 5V supply. It can be higher.

Regarding this comment:

"If you use that scheme, you don't need the transistor. Most op amps can source way more than 1mA."
-Yeah, I know this but with the Microchip ckt I would need at least 4 matched resistors (which I'm not sure how to do) and a precision Rsense whereas with the design I'm considering I only need the sense resistor to be precision (and really if I calibrate it with the trim pot it wouldn't really need to be precision either).

Eliminating the transistor doesn't mean you have to go to the Microchip circuit. What I meant was, if you wanted to use the scheme with the floating RTD, and measure the voltage with an inamp, you could just eliminate the transistor, leaving the circuit the same otherwise.
See the attachment.

 

Regarding this comment: Eliminating the transistor doesn't mean you have to go to the Microchip circuit. What I meant was, if you wanted to use the scheme with the floating RTD, and measure the voltage with an inamp, you could just eliminate the transistor, leaving the circuit the same otherwise.
See the attachment.

Ahh yes, I thought about that last night and was like, why do I even need the transistor if I'm only sourcing 1mA....I just had to catch up
Now to build/test it...