ADS124S08: ref input impedance, offset and drift

MaSe

Part Number: ADS124S08

Dear TI-team,

having read the datasheet I was wondering what measurement error is introduced when buffering / not buffering REFXY inputs - especially from a long term drift perspective.

Regarding unbuffered:
- I understand the µV/V number and the rough 250kOhm approximation mentioned in the datasheet (and in another thread).
- How does this drift with temperature and time?
Regarding buffered:
- I'm assuming this is done through a voltage follower opamp.
- Is there anything you can say regarding offset voltage and it's drifting with temperature and time?
- Is this buffer opamp chopper stabilized as well?

Background: I'm trying to implement a ratiometric measurement with a ground side reference resistor. Since I'm planning to perform series calibration I'm concerned more about long term and temperature drift than with the error itself.

Side question: In a separate thread you were targeting having 1000h drift data to be included into the datasheet ready by Q3/2019. Any update on this data? (Or have I missed it somewhere?)

Thanks for the great support you're providing in this forum. It already helped me in several places to fine tune my understanding of the datasheet. Also: excellent content and readability marks for the datasheet itself - one of the reasons I like TI-products. :)

Regards,
Matthias

over 5 years ago

0 Bob Benjamin over 5 years ago

TI__Guru** 113765 points

Hi Matthias,

Welcome to the E2E forum! And thanks for searching and viewing previous threads. Without the reference buffer enabled, the reference is being sampled periodically in a switched cap configuration. This sampling creates a bias current and an impedance is derived from this dynamic sampling. There could be some cap aging over time, but this should not severely impact the result. Higher temperature will also effect the operation as semiconductor devices become more leaky.

Buffering helps up to a point. As with most op amps the drive capability may come close to the supply rails but will have some limitations where the output will become non-linear near to the supply rails. This limitation is 50mV. So it is suggested that when the reference is at the supply rails, the buffer should be turned off.

With the reference buffer turned off, how will this affect the measurement? It should not if the reference input is connected directly to the supply rails as there should be sufficient current to drive the input stages unless input filtering resistors are so large that it will impact the measurement.

When making a ratiometric measurement with a low-side reference resistor, REFN can connect directly to ground (unipolar supply) and the REFP will be well within the operating range of the reference buffer. This should allow for proper input filtering at the REFP input without impact of a voltage drop across the reference resistor with the buffer enabled.

Long term effects are not likely to be a concern with a ratiometric measurement. Temperature drift will always impact offset and gain. The offset can be minimized with periodic SFOCAL (self-offset calibration) command. The reference resistor will dominate the gain error. The degree of precision for the reference resistor will be determined by cost. Some of these very low drift resistors can cost more than the ADC itself.

As far at the 1000 hour drift data I will need to ask a colleague who is currently on vacation.

Best regards,

Bob B

0 MaSe over 5 years ago in reply to Bob Benjamin

Prodigy 40 points

Hi Bob,

thanks for the quick and detailed reply. I am planning to use global chop mode and tie REFN to ground wihtout buffer. Based on your answer I would enable the REFP buffer.

Your comment about the reference resistor is spot on: the one I like costs 30€ a piece. Let's see if I can get approval for that one... :)

One concern I didn't seem get across in my original post though: I'm wondering about the offset voltage temperature and ageing drift of the input buffer (drawn in red in the picture below)
I understand that this offset voltage will have much less of an impact on the conversion result compared the PGA offset (because it affects the post amplification signal only).
But if it's large enough, it could still be an issue for my design. And as far as I understood, global chop or SFOCAL does not help with that input buffer offset. I guess I was hoping that this buffer is either chopper stabilized as well or has a low drift value in the few µV/°C and 10µV per 1000h range. I guess the "dominated by the reference resistor" part of your answer already suggests that this value is comparatively low, but is there anything else you can tell me here?

Thanks and have a good day,
Matthias

0 Bob Benjamin over 5 years ago in reply to MaSe

TI__Guru** 113765 points

Hi Matthias,

My understanding of the design is that the reference buffer is also a chopper.

Best regards,

Bob B.

0 MaSe over 5 years ago in reply to Bob Benjamin

Prodigy 40 points

Thank you Bob!

This answers my question.

If they ever want to make a new status above "guru", they should name it "Bob". :)

Greetings from the Baltic Sea to Texas,

Matthias

Data converters

Data converters forum

ADS124S08: ref input impedance, offset and drift