Opamp between VRef and ADC

[Signal32]

Hello,

Refering to this ADC sample board: https://www.aliexpress.com/item/ADS1256-24-ADC8-road-AD-precision-ADC-data-acquisition-module/32810438858.html



The schematic shows that they put an OPA350 in between the ADR03BR and the ADS1256 ADC.
I'm wondering what are all the advantages/disadvantages of doing this (only for this particular scenario / board).
Checking the datasheets, the ADR03BR has noise of 10uV p-p and the OPA350 has 4uV RMS input noise (=12uV p-p ?), so that doesn't appear to be an advantage.
A disadvantage I can see is the OPAMP input offset voltage which will add to the absolute inaccuracy of the VREF.

Anyone know what's the main reason they put the OPAMP in there ?
Also, if anyone can translate the chinese text that may help.

Thanks

[Vtile]

The OP seems to be in the combined buffer and low pass filter configuration, I'm sure you get better answer from people who really know what they are talking asap.

[b_force]

It's very common to do this.
In fact, the annoying part is that most manufactures won't tell you what the maximum loud can be on the VREF output pin.
It depends what kind of ADC you're making, but just a few mV difference is in most cases not really an issue.
(the rail-to-rail voltage of the opamps is less accurate for example)

A 10kOhm resistor is a bit much though, probably does more harm than good (noise wise)
I would pick something like 470-2kOhm or so with a 100nF or something.
Not really that important.

As for the opamp, if I have a spare one somewhere (with a DC offset that's not to bizarre), I would use that.
Otherwise the smallest/cheapest opamp you can find with a DC offset that's decent.

Of course you can also just take the rail voltage and make a divider or so.
The disadvantage of that is that you have extra load on your voltage rails, or with higher values you will create noise.
Plus you have your VREF pin anyway (and with just a buffer it even tracks that voltage)

[suicidaleggroll]

It all appears to be worse than pointless to me.  It looks as though they're trying to aggressively lowpass the output of the reference, presumably to clean it up, but then as you said they went ahead and put an op-amp with even worse noise performance than the reference to buffer it after the filter, defeating the entire purpose.  Maybe they based the design from another board that used a much higher performance op-amp there, then just swapped in something cheap without bothering to look at why the op-amp was there in the first place.

[Simon]

The opamp is probably supposed to buffer the pin but you need a real good one. It is not uncommon for a voltage reference to have a weak output but it is also not uncommon for it to be fine to put a low pass filter straight on the output of the reference and that in itself will buffer.

[Damianos]

The answers are in the data-sheets of the relative devices. They are not novels but technical documents, so they must be studied and not simply looked up from a distance!

By the way:
- if we compare lemons with apples what we expect for the result to be?
- what is more important the mentioned offset voltage or the stability of it? Tip: there is a calibration procedure...

Regards,
Damianos

[Kleinstein]

It somewhat depends on the application how good or bad the OPA350 is. At higher frequencies it is reasonable low noise, but the low frequency part (e.g. below 100 Hz) it is really bad. For a audio type application the filtering would be acceptable, for an DC application it would be a bad idea. Changing the OP without looking for the details might be a reason for this choice.

You just can't directly compare the 4 µV RMS for the 100 Hz - 400 kHz range at the OP with a 6 µV (peak to peak) for  0.1 -10 Hz of the reference. These are totally different ranges.

 I have not found a detailed noise curve for the ARD03, but from the numbers given the OPA350 might actually be lower noise over most of the frequencies. At 1 kHz it is something like 230 nV/ Sqr(Hz) for the reference compared to about 15 nV/Sqrt(Hz) for the OP. So quite an improvement. Having 10 K at the filter might be OK to avoid very large caps. The noise is still small compared to the reference.

Anyway it might have been a better choice to start with a lower noise reference. the ADR03 does not look like a good match to the ADS1256.

[David Hess]

The operational amplifier is needed to buffer the RC filter at the output of the reference.

The filters cutoff frequency is 1.6 Hz so high frequency noise from the reference is attenuated above that.  So we care about low frequency noise from the reference up to 1.6 Hz and low and high frequency noise from the operational amplifier up to about 20 Hz because of the output filter.

Figuring out the real output noise would be complicated do to lack of specifications but it could be done.  Temperature drift will probably matter more anyway and the operational amplifier is likely going to degrade the reference's temperature drift by about 4 times.

[Signal32]

It all appears to be worse than pointless to me.  It looks as though they're trying to aggressively lowpass the output of the reference, presumably to clean it up, but then as you said they went ahead and put an op-amp with even worse noise performance than the reference to buffer it after the filter, defeating the entire purpose.

That's what I was thinking, I was hoping there was something I'm missing.

The opamp is probably supposed to buffer the pin but you need a real good one. It is not uncommon for a voltage reference to have a weak output but it is also not uncommon for it to be fine to put a low pass filter straight on the output of the reference and that in itself will buffer.

Yes, the VREF has 10mA output and the ADC has 0.1mA input requirements, so there's no reason not to put a filter directly on the VREF output.

The answers are in the data-sheets of the relative devices. They are not novels but technical documents, so they must be studied and not simply looked up from a distance!

By the way:
- if we compare lemons with apples what we expect for the result to be?
- what is more important the mentioned offset voltage or the stability of it? Tip: there is a calibration procedure...

Not sure I understand what you mean, if you have a specific reason pls do share.
As mentioned stability appears worse for OPAMP and temp ppm is definitely worse when using opamp.

You just can't directly compare the 4 µV RMS for the 100 Hz - 400 kHz range at the OP with a 6 µV (peak to peak) for  0.1 -10 Hz of the reference. These are totally different ranges.

Thanks, that may be a reason.

The operational amplifier is needed to buffer the RC filter at the output of the reference.

The filters cutoff frequency is 1.6 Hz so high frequency noise from the reference is attenuated above that.  So we care about low frequency noise from the reference up to 1.6 Hz and low and high frequency noise from the operational amplifier up to about 20 Hz because of the output filter.

Figuring out the real output noise would be complicated do to lack of specifications but it could be done.  Temperature drift will probably matter more anyway and the operational amplifier is likely going to degrade the reference's temperature drift by about 4 times.

Thanks for doing the math

[Simon]

The operational amplifier is needed to buffer the RC filter at the output of the reference.

Not necessarily, according to the poster after you it has a 10mA capacity. The reference chip I just put into a design recommended a filter on it's output, so now I have a nice little bulk capacitor and that feeds straight into the AVR pin. why put an opamp in?

[David Hess]

The operational amplifier is needed to buffer the RC filter at the output of the reference.

Not necessarily, according to the poster after you it has a 10mA capacity. The reference chip I just put into a design recommended a filter on it's output, so now I have a nice little bulk capacitor and that feeds straight into the AVR pin. why put an opamp in?

If only the low impedance filter is used and assuming the reference can drive it, then the noise cutoff increases to 20 Hz instead of 1.6 Hz.  The high impedance filter cannot drive the ADC directly because the 10 kilohm resistor's temperature coefficient combined with the 35ppm/°C temperature coefficient of the ADC's 18.5 kilohm reference input will result in massive gain drift with temperature.  The ADC reference must be driven with a low resistance source like in this case, 50 ohms.

The ADS1256 datasheet recommends a very similar filter for the reference input using the OPA350.

[Kleinstein]

Having a filter behind the reference is common for a fast ADC, and the filter circuit is also not that unusual. Just the choice of OP with the OPA350 is a little odd as this OP has a low of 1/f noise and this noise in the LF region - maybe even worse than the reference. In the LF range even an LM358 might be better.

[David Hess]

Just the choice of OP with the OPA350 is a little odd as this OP has a low of 1/f noise and this noise in the LF region - maybe even worse than the reference. In the LF range even an LM358 might be better.

Did you mean high 1/f noise?  I read the OPA350 datasheet and as is typical with CMOS operational amplifiers, it has a horrible 1/f voltage noise corner frequency of 40 kHz which is why TI lists 100 Hz to 400 kHz noise.

This operational amplifier is not intended for low frequency DC precision.  It would be good for sampling ADCs and the application notes mention this but not for low frequency delta-sigma converters.  It's offset voltage drift is likely triples or more the ADC's gain drift with temperature.

I really think the OPA350 was misapplied in this application.  I suspect a super-beta input bipolar operational amplifier, which coincidentally TI no longer makes as far as I can tell, would be better.  There are some JFET input operational amplifiers which might also be suitable except at high temperatures.  Or a chopper stabilized amplifier could be used with the same restriction.

[Damianos]

The answers are in the data-sheets of the relative devices. They are not novels but technical documents, so they must be studied and not simply looked up from a distance!

By the way:
- if we compare lemons with apples what we expect for the result to be?
- what is more important the mentioned offset voltage or the stability of it? Tip: there is a calibration procedure...

Not sure I understand what you mean, if you have a specific reason pls do share.
As mentioned stability appears worse for OPAMP and temp ppm is definitely worse when using opamp.

Did you understood how the reference inputs of the device operate?
Do you realize that when the clock changes, it also change the effective impedance of the mentioned inputs?
Do you understand how the calibration cycle works? It draws current from the inputs. The capacitors are there also to help stabilise the voltage, because this current is pulsed.
What are the consequences if you are loading with current a voltage reference? Did you studied the datasheet of the ADR03? What happens with the power dissipation of it?
...

Where have you seen that this Op-Amp is "definitely worse", than the used voltage reference, in temperature drift? Which datasheets are you using, and which characteristics are you comparing?

The part of the circuit, that is presented, is almost a copy of what the manufacturer of the device suggests. Are the engineers of TI competitors of themselves?

Example pages, with relative discussion, in ADS1256 datasheet: 16, 17, 28, 29

Regards,
Damianos

[Signal32]

Where have you seen that this Op-Amp is "definitely worse", than the used voltage reference, in temperature drift? Which datasheets are you using, and which characteristics are you comparing?

It unfortunately doesn't work like that. If you use an Op-Amp the final temp drift error will not be the temp drift error of the OP-AMP, it will be the error of the VREF PLUS the error of the OpAmp. Since the Op-Amp doesn't have a zero temperature drift "temp ppm is definitely worse when using opamp."

The part of the circuit, that is presented, is almost a copy of what the manufacturer of the device suggests. Are the engineers of TI competitors of themselves?

No, the TI engineers specifically say that you should use such a design if the voltage reference doesn't provide enough current for driving the ADC directly (which totally makes sense).
They don't say what you should do if the Reference does actually provide enough current (this situation).

[Kleinstein]

The filter circuit is good, just the choice of OP was bad. This can happen when you only look at the initial part of the data-sheet. The OPA350 is not great when it comes to offset drift, but also not that bad: a 4 µV/K drift corresponds to less than 2 ppm/K for the output voltage. However the noise looks like the bigger problem.

Estimating from the noise curve the 0.1-10 Hz noise for the OPA350 should be in the 15-30 µV_pp range. Thus about 3 time higher than the noise of the reference (6 µV_pp typ).

There are better choices for the OP, like OPA234. There is no need to have super low bias OP, with a 10 K source and not so good reference. It also depends on the supply voltage.

[Damianos]

Where have you seen that this Op-Amp is "definitely worse", than the used voltage reference, in temperature drift? Which datasheets are you using, and which characteristics are you comparing?

It unfortunately doesn't work like that. If you use an Op-Amp the final temp drift error will not be the temp drift error of the OP-AMP, it will be the error of the VREF PLUS the error of the OpAmp. Since the Op-Amp doesn't have a zero temperature drift "temp ppm is definitely worse when using opamp."

The part of the circuit, that is presented, is almost a copy of what the manufacturer of the device suggests. Are the engineers of TI competitors of themselves?

No, the TI engineers specifically say that you should use such a design if the voltage reference doesn't provide enough current for driving the ADC directly (which totally makes sense).
They don't say what you should do if the Reference does actually provide enough current (this situation).

They are summed algebraically, so the result may be worse or even better! ...

In the evaluation module of the ADS1256 there is a better reference, that is buffered using the mentioned Op-Amp!
What about the dV/dI with and without a buffer, for this specific reference?
You are worrying for the temperature drift, which is in the region of a few microvolts, and disregard this parameter, which is much more than a hundred microvolts? Note that the first can be calibrated (even partially); the last is the reference for the calibration.

The selection of the reference and the buffer is a compromise between cost and performance. It is a little difficult to make a six-digit voltmeter, which is accurate to the last bit, with an attractive price for any hobbyist!

[Damianos]

The filter circuit is good, just the choice of OP was bad. This can happen when you only look at the initial part of the data-sheet. The OPA350 is not great when it comes to offset drift, but also not that bad: a 4 µV/K drift corresponds to less than 2 ppm/K for the output voltage. However the noise looks like the bigger problem.

Estimating from the noise curve the 0.1-10 Hz noise for the OPA350 should be in the 15-30 µV_pp range. Thus about 3 time higher than the noise of the reference (6 µV_pp typ).

There are better choices for the OP, like OPA234. There is no need to have super low bias OP, with a 10 K source and not so good reference. It also depends on the supply voltage.

The selection of the Op-Amp was done by the engineers of TI (in reality of BB). Their target was to make a low cost implementation for this ADC. So they selected a cheap reference and buffer and filtered the "result" to reduce the noise.
I think they chose this model because, among other things, it has low output impedance and fairly quick response, to cope with the pulsing load. The estimation of noise that you make is a bit ' forgiving ', they refer a value about 42 ?V p-p.
They give a little description in the datasheet of the relative evaluation kit.

The same ADC is used in commercial products with different reference and buffer and, of course, with way higher price than 20$! If I remember correctly I have seen somewhere that is used in the Rigol DM3068.