ADC reading falling overtime when measured

Thread Starter

Vindhyachal Takniki

Joined Nov 3, 2014
594
1. I am interfacing sensor thermopile and thermistor , whose data-sheet is attached.

2. Here is schematic of 24 bit adc which is used to measure thermopile mV and thermistor mV.

1593848144849.png


3. Issue is thermistor tempreature either keep on falling or increasing even if room tempreature is stable. It increase/decrease gradually in steps of 0.01 in either way sometimes up, sometimes down. I have pasted external fluke meter sensor to measure ambient its stable.

4. Measured by CRO the REFPN remians stable also

5. Is the filter capacitors? C1,C2,C3 and C8,C9 & C10? and resistors of 47ohm in series?
 

Attachments

DickCappels

Joined Aug 21, 2008
10,139
From the schematic C8 and C10 are in series and not connected to R9 (there is no connection dot). I don't think that is your problem, but you might want to take care of it anyway.
 

MisterBill2

Joined Jan 23, 2018
18,061
A 24 bit A/D converter has incredibly fine resolution. But we have no idea of the voltage range it has, and so we have no clue. so let us know the full scale range of the 24 bits. That is rather important.
My first question is how long after the power is applied to the circuit are the readings taken? Do the readings continue to change as time passes?
My suggestion is to also monitor the supply voltage because it has the same effect on the reading as the temperature has. It may easily be that it is the supply that is changing.

And if you can have the computer record the temperature every ten seconds for an hour you may see some pattern that will help find the source of the change.
 

KeithWalker

Joined Jul 10, 2017
3,050
From the schematic C8 and C10 are in series and not connected to R9 (there is no connection dot). I don't think that is your problem, but you might want to take care of it anyway.
Yes, I think the junction between C9 and C10 is supposed to be connected to R9. I have known faulty capacitors to cause strange problems in high impedance circuits
 

MisterBill2

Joined Jan 23, 2018
18,061
We still do not have an explanation as to the character of the drift, or even what readout is used to see it. There is a small amount of current flowing in the sensor so it might possibly be related to heating.
And for a 24 bit system the power connections, including grounds, need to be handled much less casually.
Also, I see no connection to the A/D common, and since many differential inputs are actually two separate single ended inputs that common side connection is very important. I know because early in my career I had to deal with exactly that issue.
There needs to be, at each of the inputs, a high value resistor tied to the inputs common side. The resistor should be about 10 times the source impedance, or in this case use similar 1 Meg resistors. The problem might be as simple as common mode drifting.
 

Thread Starter

Vindhyachal Takniki

Joined Nov 3, 2014
594
suppose initially i turn on device and it shows 24.72C.
Then it go up/down in 0.01 steps and in 15 minutes it changes by 0.5C

while calibrator device remains same tempearture.
 

MisterBill2

Joined Jan 23, 2018
18,061
WE are shown only one small part of a system and being asked why there is a change, when the output depends on all of the parts within the system.

What is the readout device? All that we have been shown is an input connections to a thermistor and a "thermopile", which may be a string of thermocouples, or possibly something else. I doubt that the A/D output bits are being read directly. So there are a lot of additional parts of the system that we do not see. In addition, are both temperature reading changing? Are the changes the same?

One more thing is that a thermopile is a voltage-generating device and so it is not appropriate to couple it with an external power source.
Another thing is that in most temperature measurement systems every junction in the wiring can be acting as a thermocouple and generate a small voltage. I see no hint that the TS has made any effort to compensate for that.

In any system with 24 bits of resolution every element matters, and every element can have an effect on the output data. If the 3.3 volts supply varies at all, the current will vary, and so the output after the amplification can also change. And still, we have not been told what the full-scale voltage input range is. That is a rather important thing, since it affects the voltage per bit sensitivity.

And any voltage readings taken with a CRO (Cathode Ray Oscilloscope), used to monitor the power supply, are not likely to be closer than 1% uncertainty. I would never expect any better.
 

KeithWalker

Joined Jul 10, 2017
3,050
How long are the wires between the thermistor and the ADC? If they are more than a few centimeters, you should be using a shielded twisted pair.
What trouble shooting have you done so far? There are a number of things that you can do to isolate the source of the problem:
Measure the in-circuit voltage across the thermistor with a good 4digit+ voltmeter. If the voltage drifts with the circuit indicated temperature, then the thermistor, R6, R7 or the supply is faulty.
Subtstitute a 100 Kohm resistor for the thermistor. If the problem goes away, the thermistor is causing the problem.
Replace R6 and R7, in turn and disconnect C8, C9 and C10, one at a time. Each time, check for drift in the temperature reading.
If any of the steps taken stop the drift, then you know what the source is. If they don't, Then you have either a faulty 3.3V supply, a design problem with the ADC circuit or a faulty ADC.
Regards,
Keith
 
Last edited:

DickCappels

Joined Aug 21, 2008
10,139
It would be very helpful to know the range of voltages that you expect to be fed into the A-to-D converter and what the range of expected values will be on the output.<==Maybe very important.

What are you using for an A-to-D converter?

Related to MisterBill2's concern about drifting common mode voltage I noticed that example schematic in the datasheet for the thermopile they used a very low drift and low noise voltage reference. Any drift or noise in your power supply and resistor-divider will affect the output, seeing the specs for the A-to-D converter would help in evaluating this.

What are you using as a radiation source for your testing? Wavelength or spectrum and magnitude information if you have it.

3. Issue is thermistor tempreature either keep on falling or increasing even if room tempreature is stable. It increase/decrease gradually in steps of 0.01 in either way sometimes up, sometimes down. I have pasted external fluke meter sensor to measure ambient its stable.
To repeat a question asked in post #2 by @crutschow question in post #2:
".01 of what?"

There is a thermopile inside the package from which to take measurements and a thermistor that can be used to calculate correction for the ambient temperature. You referred to the thermistor drifting -do you mean the thermopile is drifting?
 

MisterBill2

Joined Jan 23, 2018
18,061
It would be very helpful to know the range of voltages that you expect to be fed into the A-to-D converter and what the range of expected values will be on the output.<==Maybe very important.

What are you using for an A-to-D converter?

Related to MisterBill2's concern about drifting common mode voltage I noticed that example schematic in the datasheet for the thermopile they used a very low drift and low noise voltage reference. Any drift or noise in your power supply and resistor-divider will affect the output, seeing the specs for the A-to-D converter would help in evaluating this.

What are you using as a radiation source for your testing? Wavelength or spectrum and magnitude information if you have it.



To repeat a question asked in post #2 by @crutschow question in post #2:
".01 of what?"

There is a thermopile inside the package from which to take measurements and a thermistor that can be used to calculate correction for the ambient temperature. You referred to the thermistor drifting -do you mean the thermopile is drifting?
OK, so finally we have a statement that the thermistor reading is drifting relative to the thermopile reading. Since I see no connection to the differential analog input common I will presume that the two differential inputs are free to vary up and down. That will always cause a drift in readings for any isolated differential input.
In addition, we still have no mention as to what the A/D input full scale range is, nor what the system is that displays the temperature.
One more thing is the statement that the 3.3 volt supply is read with an oscilloscope, (CRO), and I know that those readings have 1% rated accuracy. Not 24 bit accuracy.
So we only have a small portion of the circumstances, while a description of the whole measurement string is needed. So far all of the suggestions have been best guesses, based on very little information. Is the TS unable to answer the questions asked? Or unwilling to answer them? Or totally not understanding what we are asking?
 

Thread Starter

Vindhyachal Takniki

Joined Nov 3, 2014
594
I have removed C8 in #1. But still varation is there. But when i connect fixed 100K resistor there, I measure 25C without any fluctuation.

1. One reason could be this, I was reading this. Currently minimum current which will pass thorugh my thermistor at 25C is = 3.3V/(100K + 100K + 100K) = 11uA
But i was reading this pdf , here it says to have max 1 to 5uA .

1594565847674.png


2. Can this be reason?
Thermopile, has some sort of lens in front and whatever infrared falls onto it, it concentrates on inside. There we have thermopile and thermistor? Could it be reason, when heat gets into inside, it also heats up the thermistor? This is very random thought though. Dont know if this could be issue?
 

MisterBill2

Joined Jan 23, 2018
18,061
If you have no other power into the assembly with the thermistor, then the only heating is from the measurement current. BUT the claim of drifting is that it is both up and down. Internal heating would always cause drift in the same direction.
Then there is no drift with a 100K resistor, where is it connected? Try it with the same circuit except that the thermistor is replaced by that 100K resistor, but with all of the same wires and wiring as with the thermistor. If there is no drift then it is the thermistor, if no drift then it is with the rest of the circuit.
 

KeithWalker

Joined Jul 10, 2017
3,050
The thermopile is a number of thermocouples connected together in series. The measuring junctions face the lens. The reference junctions are thermally connected to the base of the housing. The output of the thermopile is the sum of the thermocouple voltages which is related to the temperature difference between the lens and the base. The thermistor is mounted on the base and measures it's absolute temperature. The sum of the two measurements give the absolute temperature at the lens.
Any current passing through the thermistor will warm it up The base has a relatively large thermal mass compared to the thermistor so there will be a lag in temperature changes between the two. That will cause a small drift in the measured temperature, especially on power-up.
If you want to make such accurate measurements, you have to either wait for the temperature of the device to stabilize and/or use an integration factor to correct for changing ambient temperatures. You can reduce the drift by using a lower thermistor current and Increasing the mass of the base using a heat sink but you can not eliminate it completely.
Regards,
Keith
 
Last edited:

MisterBill2

Joined Jan 23, 2018
18,061
It should, eventually, as long as the ambient temperature doesn't change.
Keith
Even if the ambient changes, if the measured value tracks the ambient rather than drifting away from the ambient then the drift would be a temperature challenge.. Do you have a program running to record temperatures? Can you record the temperature reading once a minute for a few hours? AND, could that same program record the power supply voltage once a minute at the same time. We really do not have much information about the rest of the system.
 
Top