ADS1256: ADS1256 input voltage range

Hi Bryan,

Thanks a bunch. I was guessing along the lines you just explained and you have once more provided a very good answer, but please let me ask a few related follow-up questions in the same thread. Hopefully they also have some value to others as general design advice:

As I mentioned in a previous question, I am modernizing an old application based on the ADS1256 and the original goal was to maintain the analog part as it was in the previous version of the product. Having started on the ADC code conversion I now need to deal with old design choices of which my gut feeling doesn't approve, but to get a mandate to change things I need to present input and recommendations from an expert.

The original question in this thread partly relates to a more general question whether it is wise to measure the AVDD analog 5V supply and use the common analog rail also for power supply to external accessories and analog sub-circuits that generate measurement signals back to the ADC.

You can avoid over-ranging by using an adjustable LDO set just below the minimum 2.5V precision reference voltage times two (in my case between 4.86V and 4.93V). However, the design I inherited also incorporates an alternative ADC voltage reference based on a 10k, 10k 0.1% resistor divider from AVDD through an OPAMP follower with a +-2mV maximum offset. Due to tolerances, it isn't possible to generate a nominal AVDD/2 reference without having to deal with over-ranging in roughly half of the produced units. A slightly asymmetric 10k1, 10k resistor divider would avoid over-ranging, but still be subject to rail fluctuations reaching the different parts of the device in a non-uniform fashion, greater temperature drift, and other things.

I don't have a clear picture of the design of each and every external accessory for the product in question, so I have started with a new reference and calibration procedure instead. Could you please comment on it?

1. Run SELFCAL with buffer, gain 1, and the unit's 2.5V precision reference (that is currently not exactly known) to get OFC and FSC for a single-ended starting point.

2. Connect e.g. AD3 to AGND and run a measurement sequence on AD3-AINCOM. This will get system offset correction information for adjustment of OFC. Repeat for AD4.

3. Connect AD3 to an external 2.5V precision reference (that is exactly known) and run a measurement sequence on AD3-AINCOM. This will get a correction factor for adjustment of FSC. Repeat for AD4.

4. Connect AD3 and AD4 to the external 2.5V precision reference and run SYSOCAL for AD3-AD4. This will get OFC for differential measurements.

5. Connect AD3 to the external 2.5V precision reference and AD4 to AGND and run a measurement sequence for AD3-AD4. This will give input to adjustment of FSC without running SYSGCAL (since it requires a full-scale input signal that's too big for the buffer and would not be constant).

Repeat for any combination of buffer/non-buffer, gain, data rate, and so on.

A few questions regarding (5.):

* Does differential measurement on ADS1256 give better results on input signals that are symmetric around mid-range 2.5V, e.g. instead connecting AD3 to an external 2.8V precision reference and AD4 to an external 2.048V precision reference?.

* Since the datasheet maximum absolute input voltage with buffer is AVDD - 2.0V, is connecting an external 3.0V precision reference to AD4 a no-no if AVDD is minimum 4.85V? 3.0V references are more common than 2.8V and have better specs.

Also:

- Do measurements near 0V suffer from degraded linearity and noise performance? (This has two implications - being near end-of-scale and being more negative compared to the 2.5V unit reference than AVDD between 4.86V and 4.93V is positive compared to the 2.5V unit reference.)

- Similarly, do measurements near (but not exceeding) full-scale suffer from degraded linearity and noise performance? Either way, my gut feeling says I should limit the input range to [0, 4.096] (so much below AVDD that the external accessory powered from AVDD is certain to be able to produce an output signal in the range [0, 4.096]. (Or perhaps [0, 2.8] so that the buffer could always be used.)

I'm grateful for your input.

Best regards

Niclas

Hi Niclas,

Just as a head's up: this is a lot of information to digest and respond to, so it might take a few days to get back to you. But I don't want you to think that I forgot about you, just give me some time.

-Bryan

Hi Niclas,

I was able to read through your post and my feedback is below in bold:

1. Run SELFCAL with buffer, gain 1, and the unit's 2.5V precision reference (that is currently not exactly known) to get OFC and FSC for a single-ended starting point.

I am unclear here when you say the reference is not known, after you just said it was a precision 2.5V VREF? Performing SELFCAL determines the offset and gain error of the ADC by itself, and the ADS1256 always takes a differential measurement (AINP – AINN). In other words this isn’t really a “single-ended" measurement if that’s what you are worried about.

2. Connect e.g. AD3 to AGND and run a measurement sequence on AD3-AINCOM. This will get system offset correction information for adjustment of OFC. Repeat for AD4.

I am guessing AD3 and AD4 are the inputs to your system (maybe the terminal blocks) and then you have some filtering, etc., after these inputs that then feeds into the ADC? Do you have a schematic to share that I could review? Please note that AINCOM is not automatically tied to AGND, you will have to tie this pin to GND manually if it is not already. AINCOM is just another input like AIN0, AIN1, etc. Also, we recommend shorting the inputs to mid-supply (buffer off) for an offset calibration, not tying them to GND.

3. Connect AD3 to an external 2.5V precision reference (that is exactly known) and run a measurement sequence on AD3-AINCOM. This will get a correction factor for adjustment of FSC. Repeat for AD4.

4. Connect AD3 and AD4 to the external 2.5V precision reference and run SYSOCAL for AD3-AD4. This will get OFC for differential measurements.

5. Connect AD3 to the external 2.5V precision reference and AD4 to AGND and run a measurement sequence for AD3-AD4. This will give input to adjustment of FSC without running SYSGCAL (since it requires a full-scale input signal that's too big for the buffer and would not be constant).

Repeat for any combination of buffer/non-buffer, gain, data rate, and so on.

So it seems like you are trying to perform a system offset and gain calibration per channel using both single-ended and differential measurements. Admittedly, this seems like overkill, but I am not entirely sure what the application is here or how accurate it needs to be. Do you know if the error external to the ADC e.g. the signal conditioning circuitry, is significant enough to warrant this lengthy calibration procedure? In other words, if you just calibrated the ADC errors would that be sufficient for your system, or are these other steps actually necessary? I also wouldn’t perform a self-calibration if you are also going to perform a system calibration, as the system calibration will also calibrate out the ADC offset / gain error.

Will you need to take both single-ended and differential measurements in this system? It might help to know more about what your circuit is actually supposed to be doing, since all I know now is that the ADS1256 is being used in some capacity. But the more info you can provide, perhaps there are better solutions to some of these challenges.

I do agree that you should perform a new calibration if you change the data rate, PGA setting, or buffer, as per the datasheet (pg. 26).

A few questions regarding (5.):

* Does differential measurement on ADS1256 give better results on input signals that are symmetric around mid-range 2.5V, e.g. instead connecting AD3 to an external 2.8V precision reference and AD4 to an external 2.048V precision reference?

Not necessarily. The ADS1256 PGA is not a true amplifier that has nonlinear regions, CM requirements, etc., so your common-mode can be anywhere within the allowable input range. You just need to make sure you do not violate the absolute input range requirements depending on your buffer configuration (as well as other applicable datasheet specs)

* Since the datasheet maximum absolute input voltage with buffer is AVDD - 2.0V, is connecting an external 3.0V precision reference to AD4 a no-no if AVDD is minimum 4.85V? 3.0V references are more common than 2.8V and have better specs.

Correct, you cannot apply a 3V signal if AVDD = 4.85 V and the buffer is on. The buffer inside the ADS1256 is an amplifier such that it has headroom requirements and nonlinear regions of operation. That is why you must stay between AGND and AVDD-2V

- Do measurements near 0V suffer from degraded linearity and noise performance? (This has two implications - being near end-of-scale and being more negative compared to the 2.5V unit reference than AVDD between 4.86V and 4.93V is positive compared to the 2.5V unit reference.)

Not necessarily. As long as your input signals are meeting datasheet specs, you should expect datasheet performance. If you apply input signals too close to the limits however, any variation on those inputs (or the supplies, VREF, etc.) could force your signals outside the allowable range, where linearity, noise, etc. would be degraded and not guaranteed anymore

- Similarly, do measurements near (but not exceeding) full-scale suffer from degraded linearity and noise performance?

See my answer to the previous question

Either way, my gut feeling says I should limit the input range to [0, 4.096] (so much below AVDD that the external accessory powered from AVDD is certain to be able to produce an output signal in the range [0, 4.096]. (Or perhaps [0, 2.8] so that the buffer could always be used.)

As per my previous responses, you probably want to provide a little bit of headroom for the voltage inputs in case of supply droop, ground noise, etc. Especially when the buffer is on, I would stay a few 100 mV away from the inputs just in case. But again, the device will work anywhere within the specified voltage range, it is just best practice to give your circuit some margin.

-Bryan

Hi Bryan and thanks again,

The product has an internal 2.5 V precision reference with a datasheet accuracy of +- 0.06% which makes its exact value initially unknown. All we know at the start is that it's in the range [2.4985, 2.5015] V. We input a known (e.g. 2.4999864 V) DC signal and use the measurement result as a correction factor to the FSC value obtained from the SELFCAL done at the beginning of the calibration/adjustment sequence.

The old design has a calibration/adjustment sequence that extracts a number of "sets" of OFC and FSC data that are then stored in RAM. Depending on the type of actual measurement, the corresponding OFC and FSC "set" is read from RAM and written to the ADS1256 before each measurement series. My intention is to continue in this fashion but revise and correct the procedure.

In (2.) I am not referring to a SYSOCAL but to a regular measurement series that is averaged and used to correct the SELFCAL OFC obtained in (1.) in order to generate a home-brew SYSOCAL OFC.

Thank you for emphasizing that the SYSOCAL should be done at mid-supply. The datasheet says "a zero input differential signal", which could be misinterpreted as 'connect both inputs to zero volt'. The datasheet doesn't seem to recommend regarding buffer off vs. buffer on, so could you please expand on this? My plan is to generate one home-brew SYSOCAL OFC with buffer and one without buffer (per used gain and per used data rate (currently only 1000 SPS)).

I cannot share the schematic, but the ADC circuit contains two 8:1 analog multiplexers connected to AD3 and AD4 respectively. The voltage reference to the ADS1256 is switchable either to the internal 2.5 V precision reference or to a 10k 10k 0.1% resistor divider from AVDD with a unity gain OPAMP buffer that has an offset error of +- 2 mV. (And the analog switch might have an offset error.) AINCOM is tied to AGND at the ADS1256.

The old version of the product has a series of external accessories based on resistive temperature sensors. They are powered from AVDD (which causes a number of potential error sources, e.g. LDO loading, cabling voltage drop, and susceptibility to whatever else is going on in the system that might affect AVDD). The new version of the product must continue to support these accessories. The 'sensor elements' in these accessories are probably quite 'rough', which is dealt with in a different type of multipoint adjustment procedure that is not relevant to my questions regarding the ADS1256. In other words, at this point I'm assuming that the the accessory multipoint calibration compensates for the old version's less than perfect ADS1256 calibration/adjustment.

However, the new version also has added connectivity (a great increase in the number of digital and analog IO), so I am also trying to lay a proper foundation for future accessories with greatly improved measurement quality based on essentially the same schematic for the analog block. Some of these new connections will enable differential measurement (through the two analog multiplexers into AD3 and AD4). Others will be single-ended directly into the ADS1256. There are differences in filtering on the various inputs. The exact need for calibration/adjustment/compensation will not be known until testing has been made on the production version. At this point I am trying to include everything that might be needed and later exclude the unnecessary calibration steps, which I believe will be easier than making a minimum procedure now and later extend it.

You are perhaps somewhat contradicting yourself regarding whether differential measurements should be mid-range centered: ("we recommend shorting the inputs to mid-supply (buffer off) for an offset calibration" vs. "so your common-mode can be anywhere within the allowable input range".)

My plan is to build the following external voltage references for calibration/adjustment/testing:

A. MAX6325CPA  Single-ended 2.5V with trim (and differential mid-supply for SYSOCAL)
This is sufficiently close to the high end of the buffered range (2.85 V in my case).

B. MAX6341CPA  Single-ended 4.096V with trim
This is the precision reference closest to the high end of the un-buffered range (4.85 V in my case).

C. MAX6325CPA  Differential 2.5V with trim in series with a 0.2 V voltage source
This gives a differential input close to the high end of the buffered range but with 0.2 V and 0.25 V margin to the range ends.

D. MAX6341CPA  Differential 4.096V with trim in series with a 0.4 V voltage source
This gives a differential input close to the high end of the un-buffered range but with 0.4 V and 0.354 V margin to the range ends.

I do understand that the ADS1256 is inherently differental, so the above "single-ended" refers to measurements with AINCOM as AINn.

The input from you is tremendously helpful and it has given me the expert support I need for a number of changes/improvements to the calibration/adjustment procedure in the new version of the product. If you can find the time to comment on the above, I think I will answer all the questions that I currently understand that I have.

Thank you and best regards

Niclas

Hi Niclas,

So it sounds like what you are trying to do is just have a full-scale calibration method for each input path and configuration system for your settings. You will then have a lot of different combinations to test (2x VREFs, buffer on/off, SE vs diff, etc.), but if you want the best accuracy this is the best way to accomplish it. I would say that you probably don't need a separate calibration scheme for single-ended or differential measurements, assuming the same input path is used, but it seems as though you will have different input paths for the different input types (SE vs diff), so that makes sense. I also hope that after your initial testing you are able to determine the largest sources of error and correct for those without requiring all of this calibration for your final product.

I mentioned applying a mid-supply voltage for the offset measurement only in the case when buffer is off, because technically with the buffer on a mid supply voltage is still 2.5V. However, the limited input voltage range with buffer = on means that the offset voltage should be at mid range, or 1.5V (assuming a 5V supply), since the input range there would be 0-3V. I hope this did not lead to too much confusion.

In general, I would recommend keeping your input common mode signal near mid supply, as many ADCs have a true INA gain stage where VCM reduces as the gain increases. You can see this for example on the ADS1261, which is the next-generation version of the ADS1256. However, the ADS1256 does not have this requirement since it uses a different amplification scheme such that the VCM can be anywhere within the ADC's absolute input range. My comment was more to highlight that while this can be done for the ADS1256, it may not be possible for other ADCs. So if you adopt this as a best practice now, it will benefit you later I believe.

I am not sure if I answered all of your questions, so let me know if I missed something from your last post.

-Bryan

Hi Bryan and thanks once more,

I think you have understood and answered my questions very well.

I agree that the calibration procedure I am starting off with is a bit elaborate, but after testing it I will know which steps are really necessary. In the scenario where all steps are necessary they could perhaps be automated. The described calibration board will initially be used to measure the quality of the HW design.

I have a quick follow-up on the 'mid-range' aspect:

SELFOCAL: "The analog inputs AINP and AINN are disconnected from the signal source and connected to AVDD/2."

Since this can be (should be) done both with and without buffer, connecting the two inputs to exactly mid-range can't be critical for the ADS1256 (?). I mean, SELFOCAL will only be mid-range without the buffer. (Or is this answered by the 3rd paragraph of your reply?)

Best regards

Niclas

Thanks!