Characterization of an Amplification/ADC circuit

Thread Starter

Henry603

Joined Nov 19, 2018
69
Hello guys,

I build a small amplification circuit in order to amplify the low voltage signals of my passive sensors (the sensors have a source impedance of about 50-200 Ohm and they output a low voltage signal in the range of 0-20mV. The sensor output changes very very slowly as they measure sun irradiation and the sensors respond very slow as well, respond time >20 sec).

The hardware I build consists an instrumentation amplifier (AD8237: Datasheet) and a high resolution ADC (MCP3422: Datasheet). Please see the schematic below for a better understanding:



Now that the hardware is ready I was thinking about how to characterize the circuit. I began by getting familiar with the most important specs of the components I used (gain, offset, DNL, INL, linerarity etc.). Basically the values of interest I will need to evaluate during the characterization of my custom board.

Also I came accross this website that has a neat little overview regarding ADC characterization: Link.

The first thing that came to my mind is the use of a signal/function generator.
Please keep in mind that I want to evaluate the whole circuit and not just the ADC.

| Question |

How would you go about doing the characterization in my case? I think in my case for sure I should apply the test signal at the input of the instrumentation-amplifier as all components are going to introduce noise/offest/errors and to do a good compensation for these errors later in the digital domain (in software) I should include the whole signal chain (complete circuit) in the evaluation/characterization.

My first idea was to hook up the non-inverting input of the circuit to a signal generator and start a DC-sweep/ramp within the valid range of the sensors I will use (0-20mV). The inverting input of the amplifier would be tied to common ground during the tests (however as you see in the provided schematic I will use differential measurement in my final setup with the sensors). Any other ideas or thoughts on that? Should I even hook up two function generators for this (one on every pin of the instrumentation amplifier)?

How about impedance issues? As the inputs on the INA are high impedance (100 MOhm) and my sensors will have a low output impedance (about 100 Ohm) I guess that my signal generator should also exhibit similar conditions. I guess It would be sufficient to just select a 50 Ohm output impedance for the probe of the signal generator? Any other pitfalls?

So this is my basic idea: Hook it up to a signal generator, driving it with a DC ramp and taking digital values for some defined DC-points from the ADC (still not sure how to automate this process, are there options to connect signal generators to a computer and synchronizing the DC-ramp with the ADC readings?). I would appreciate any input from experienced people how to do it the best way, any insights on past experiences, pitfalls, best practices is welcome (as the basic idea is there but I'm still pretty clueless regarding the practical execution).

Thank you very much in advance. :)
 

crutschow

Joined Mar 14, 2008
23,363
Before you get into testing, there is a design issue to discuss.
Why do you have such large input resistors (R_r) to ground?
The input offset bias current of the INA is
upload_2019-2-10_12-2-49.png

This will cause several mV of input voltage offset across those 10MΩ resistors, comparable to the signal voltage.
Since the sensor impedance is so low, I would go with much lower value resistors (say 10-100kΩ) to minimize that offset voltage).
 

Thread Starter

Henry603

Joined Nov 19, 2018
69
Before you get into testing, there is a design issue to discuss.
Why do you have such large input resistors (R_r) to ground?
The input offset bias current of the INA is
View attachment 169934

This will cause several mV of input voltage offset across those 10MΩ resistors, comparable to the signal voltage.
Since the sensor impedance is so low, I would go with much lower value resistors (say 10-100kΩ) to minimize that offset voltage).
Hi, thank you for your input crutschow.

that is indeed smth. I could not fully understand so far, so I asked for help here:
https://forum.allaboutcircuits.com/threads/how-to-properly-provide-a-return-current-path-for-an-amplification-circuit.156215/

All members that replied agreed that a 10 MOhm resistor is the way to go in order to provide a proper return current path in my case.
I would really appreciate it in case you could take a quick look at the first few posts in the mentioned thread and tell me if the members that replied to me were wrong and did not take the current you mentioned for my INA into consideration for example.

Thank you very much.
 

crutschow

Joined Mar 14, 2008
23,363
They are correct.
I forgot that the low impedance of the sensor will swamp any offset voltage from the bias current. :oops:
But one resistor would be sufficient.

As to testing, connecting a signal generator (likely through an attenuator to get the low voltage) should work fine for the test.
If you want to synchronize the generator to the measurement, see if the generator has a pulse or sync output.
If you want precise readings for specific inputs, you might want to do a step signal rather than a ramp.
A programmable pulse/function generator should be able to do that.
 

Thread Starter

Henry603

Joined Nov 19, 2018
69
They are correct.
I forgot that the low impedance of the sensor will swamp any offset voltage from the bias current. :oops:
But one resistor would be sufficient.

As to testing, connecting a signal generator (likely through an attenuator to get the low voltage) should work fine for the test.
If you want to synchronize the generator to the measurement, see if the generator has a pulse or sync output.
If you want precise readings for specific inputs, you might want to do a step signal rather than a ramp.
A programmable pulse/function generator should be able to do that.
thank you for the clarification :)

So signal generators are not able to output precise low voltage signals (in the 10-20mV range)?
In that case I guess I can not really automate the process and only can do very few measurement points where I have to manually set the low voltage level using a homebrew resistive divider for example?

Signal generators should have about 50 Ohm source impedance, that should be ok for the simulation of my sensor (that will have about 100-200 Ohm, right)?
 

crutschow

Joined Mar 14, 2008
23,363
So signal generators are not able to output precise low voltage signals (in the 10-20mV range)?
Generally not (look at the specs).
But it's easy to make a two resistor attenuator to get a low level.
In that case I guess I can not really automate the process and only can do very few measurement points where I have to manually set the low voltage level using a homebrew resistive divider for example?
I don't understand why you can't automate it using a resistive attenuator. :confused:
You program the signal generator to the voltage you want once you've calibrated the attenuator to the desired accuracy (or use a pot to adjust the attenuation to the desired value).
Signal generators should have about 50 Ohm source impedance, that should be ok for the simulation of my sensor (that will have about 100-200 Ohm, right)?
Yes.
The input impedance of your amp is much higher than the impedance of the generator and attenuator.
The sensor impedance doesn't enter into that since it's not in the circuit when you do the test.
 

Thread Starter

Henry603

Joined Nov 19, 2018
69
Generally not (look at the specs).
But it's easy to make a two resistor attenuator to get a low level.
I don't understand why you can't automate it using a resistive attenuator. :confused:
You program the signal generator to the voltage you want once you've calibrated the attenuator to the desired accuracy (or use a pot to adjust the attenuation to the desired value).
Yes.
The input impedance of your amp is much higher than the impedance of the generator and attenuator.
The sensor impedance doesn't enter into that since it's not in the circuit when you do the test.
Thank you for the info! :)

So basically what you sugegst it to use some resistive dividers between the generator and the INA-input?
(so output a fixed voltage using the signal generator and attenuating it with the voltage divider. Then doing the measurement with the ADC and then apply another voltage by changing the resistive divider).
Did I get you right?

Also maybe you can help me with my understanding regarding the impedances.
Often datasheets of INAs say that the REF-input can NOT be driven with a high impedance source (like e.g. a resistive divider).
But in case I use a resistive divider, the input (or the load) will only see the first value of the first resistor (as additional series resistance) in the voltage divider as source impedance, right?
 

crutschow

Joined Mar 14, 2008
23,363
So basically what you sugegst it to use some resistive dividers between the generator and the INA-input?
(so output a fixed voltage using the signal generator and attenuating it with the voltage divider. Then doing the measurement with the ADC and then apply another voltage by changing the resistive divider).
No.
You use the resistive attenuator to reduce the signal by a fixed percentage and then just change the generator voltage to change the attenuator output by the same percentage.
For example, suppose you use a 100:1 attenuator.
Thus 1V in will give 10mV output and 0.1V input will give a 1mV out.
Often datasheets of INAs say that the REF-input can NOT be driven with a high impedance source (like e.g. a resistive divider).
But in case I use a resistive divider, the input (or the load) will only see the first value of the first resistor (as additional series resistance) in the voltage divider as source impedance, right?
I don't follow.
They stated not to use a resistive voltage divider at the REF input so why do you still want to do that? :confused:
The REF input establishes the output voltage for zero differential input.
If you use a resistive divider it will upset the internal gain and biasing (and possibly the common-mode rejection) of the the INA output.
So if you are only using a single supply and want the output to be referenced to other than ground, use a voltage divider at the input to an opamp voltage follower and connect the op amp output (which inherently has a very low impedance) to the REF input.
That way the resistive divider impedance is not seen by the REF input.
 

Thread Starter

Henry603

Joined Nov 19, 2018
69
No.
You use the resistive attenuator to reduce the signal by a fixed percentage and then just change the generator voltage to change the attenuator output by the same percentage.
For example, suppose you use a 100:1 attenuator.
Thus 1V in will give 10mV output and 0.1V input will give a 1mV out.
I don't follow.
They stated not to use a resistive voltage divider at the REF input so why do you still want to do that? :confused:
The REF input establishes the output voltage for zero differential input.
If you use a resistive divider it will upset the internal gain and biasing (and possibly the common-mode rejection) of the the INA output.
So if you are only using a single supply and want the output to be referenced to other than ground, use a voltage divider at the input to an opamp voltage follower and connect the op amp output (which inherently has a very low impedance) to the REF input.
That way the resistive divider impedance is not seen by the REF input.
ok, thank you.
So I guess you are talking about a BNC attenuator that can be directly attached to the signal generator?
I never worked with it before so forgive me if im asking silly questions :)

I know about that and that one can use a voltage follower as a buffer to resolve the impedance issues for inputs that require to see a low input impedance (I also do not intend to use a resistive divider here).
I just asked out of curiosity what impedance the input actually sees in case I tie it to a resistive divider directly (the value of the first resistor in the resistive divider I guess?)

Thank you.
 

crutschow

Joined Mar 14, 2008
23,363
So I guess you are talking about a BNC attenuator that can be directly attached to the signal generator?
Basically, yes.
You can use a commercial one or, since the frequency is low, you can roll your own.
what impedance the input actually sees in case I tie it to a resistive divider directly (the value of the first resistor in the resistive divider I guess?)
Wrong guess.;)
How can you include one resistor but ignore the other?
The impedance of a voltage divider is the parallel value of the two resistors.
 

Thread Starter

Henry603

Joined Nov 19, 2018
69
Basically, yes.
You can use a commercial one or, since the frequency is low, you can roll your own.
Wrong guess.;)
How can you include one resistor but ignore the other?
The impedance of a voltage divider is the parallel value of the two resistors.
By building my own you mean smth. like this?:


And yes, it's getting late...
How could I ask smth. like this :D

But thanks for staying with me...
 

BR-549

Joined Sep 22, 2013
4,936
If you are going to be doing much of this kind of work......you should consider a waveform generator.
I can set the output of my WG to 1 mv. Anywhere from 1 mv to 20 VPP.....in 1 mv increments.

And I can sweep or vary that output over that range(or any range, say 1 to 4 mv if wanted) at any frequency or waveform.

This is just one of many features that digital electronics provides. The best feature was the cheap cost.

Some have even been on sale lately. Best gadget I ever bought.
 

Thread Starter

Henry603

Joined Nov 19, 2018
69
If you are going to be doing much of this kind of work......you should consider a waveform generator.
I can set the output of my WG to 1 mv. Anywhere from 1 mv to 20 VPP.....in 1 mv increments.

And I can sweep or vary that output over that range(or any range, say 1 to 4 mv if wanted) at any frequency or waveform.

This is just one of many features that digital electronics provides. The best feature was the cheap cost.

Some have even been on sale lately. Best gadget I ever bought.
I can use this piece of equipment in the lab:
https://www.sicamax.ch/downloads/datasheet_dg1022_en.pdf

It says 2mVpp to 10Vpp for Amplitude so this generator also should be able to go pretty low (so I might be able to omit the suggested attenuators).
However, it also says 'except DC' for the waveforms specifications so I'm not sure.
I will have to try out next time I'm in the lab I guess...

It also has trigger/sync outputs for the channels. Any best practices how I could automate the measurements?

Thank you!
 

crutschow

Joined Mar 14, 2008
23,363
I can use this piece of equipment in the lab:
https://www.sicamax.ch/downloads/datasheet_dg1022_en.pdf

It says 2mVpp to 10Vpp for Amplitude so this generator also should be able to go pretty low (so I might be able to omit the suggested attenuators).
It depends upon the resolution you want for the signal.
You will get better resolution if you use an attenuator with a higher voltage out from the generator.
 

Thread Starter

Henry603

Joined Nov 19, 2018
69
It depends upon the resolution you want for the signal.
You will get better resolution if you use an attenuator with a higher voltage out from the generator.

So building an attenuator is as simple as this? :D
(As you said you do not mean a resistive divider but I can 'roll my own attenuator'. When researching attenuators its basically this small circuit, nothing more to it... Did you mean smth. else?)
 
Last edited:

crutschow

Joined Mar 14, 2008
23,363
So building an attenuator is as simple as this?
Yes.
The output vs. input signal is equal to R2/(R1+R2), so for example, for a 100:1 attenuation, you could make R2=1kΩ and R1=99kΩ.
Since 99k is not a standard value, you could parallel 105kΩ with 1.74MΩ to give a nominal value of 99kΩ within .03%.
 
Last edited:
Top