OPA189: OPAMP choice for ADC

Hi Niclas,

Q1. Is this a reasonable assumption for four such dividers in parallel fed by a voltage reference or an OPAMP unity-gain buffer?

If REF70XX reference voltage is not available, you may try REF5025, which is capable to drive up to 25mA in current. Can you show me the image of the parallel resistor dividers in how to you are going to divide it down the reference? What errors are you are looking for your reference? I am not clear what is your reference voltage after voltage divider or gain up the reference voltage. 

Q2: 2. Could you recommend a suitable OPAMP unity-gain buffer for the resistor dividers?

OPA189 will meet your need. I see the OPA189 has very good performance specification than others in the list below. 

FYI, we do have THP210 SE to FDA now. If you need differential input in your ADC, may be it can be a good option. 

Q3. Should you always incorporate a feedback resistor for the unity-gain buffer?

OPA189 has already implemented bias current cancellation circuit in the IC. You do not need to add 100-300Ω feedback resistor in the unity gain buffer. Unless your circuit has some input impedance connected to the non-inverting input pin, you may consider to place an equivalent impedance at the  unity gain feedback loop to cancel the imbalance of the input bias current. 

Q4. Should there be an RC network at the output of each unity-gain buffer?

If your input signal has high frequency noise components, then RC network at the output of buffer may be helpful. Please make sure that your RC pole is a decade (at least ) higher than your ADC sampling frequency. Otherwise, it may attenuate the input signal.

Q5. For the instrument's 2.5V reference, it is sometimes a good idea to use an OPAMP buffer. The OPA350 is commonly used as an example, e.g. in the ADS1256 EVM, but are there newer, higher-performance OPAMPs, that are currently recommended?

The REF70xx series has slight better performance specification than REF50XX series, but they are all capable to source/sink 25mA. OPA350 can source/sink up to 40mA in current. If the voltage reference has very light load (say 10% of 25mA), then it is not necessary to buffer with OPA350 op amp. OPA350 may have up to +/-150uV in a typical Vos specification, up to 0.5 to 1mV at higher temperature. I will not buffer the reference if you do not have to. 

If you have additional questions, please let us know.

Best,

Raymond

Hello Niclas,

In addition to my colleague Raymond's excellent answers here is some addition discussion about your questions. Hopefully, between Raymond and my responses we add to your understanding.

Questions:

1. I assume that about 10k ohm over each resistor divider is a good compromise between low load on the 2.5V reference and low source resistance. Is this a reasonable assumption for four such dividers in parallel fed by a voltage reference or an OPAMP unity-gain buffer? (1 mA load in total.)

Certainly the load you connect to the output of a reference needs to conform to the ability for the reference to supply the load current without being loaded down to the point where it doesn't provided the required reference voltage characteristics. Most op amps will easily drive a 1 mA load and often considerably more. Op amps are often applied as reference buffers and provide a low impedance source. However, if you plan to add a large capacitance to their output they may become unstable and compensation is required to stabilize them.

2. Could you recommend a suitable OPAMP unity-gain buffer for the resistor dividers? If I understand correctly, such OPAMPs must be designed to work with high input impedance, which rules out some zero-drift, low-noise OPAMPs? I am currently looking at OPA189/OPA2189, but are there other ones better suited for this?

The OPA189/OPA2189 is an zero drift design based on an advanced chopper architecture. It features exceptionally low voltage offset and drift, but since it utilizes switching in the signal path and there can be current noise spikes present due to charge transfer among the internal switch elements. That current noise can be converted to noticable voltage noise when there are high resistances in which the Op amp's input currents flow. A 10 k resistance is in the region where it might become an issue. My preference is to use one of TI's CMOS e-trim op amps such as the OPA192/OPA2192. They too exhibit very low offset and drift, and don't rely on switching within the signal path to achieve those attributes. Resistances of 10 k, or higher, will not be an issue for the OPA192/OPA2192 except for the thermal noise the resistors produce naturally.

3. Should you always incorporate a feedback resistor for the unity-gain buffer? Would 100 to 330 ohm be good as a rule of thumb or should it match the input impedance from the resistor divider?

Most of the time it is not necessary, or desirable to add a resistor in the buffer feedback loop. In some cases adding the feedback resistor my help improve stability, but most often it degrades the phase phase margin and could lead to instability. It really comes down to the particular op amp open-loop gain and open-loop output impedance (Zo) characteristics, and the load on the output as to when it is appropriate to add a feedback resistor.

4. Should there be an RC network at the output of each unity-gain buffer?

I assume you are talking about a series RC connected from the Op amp output to ground? That is called a Zobel network and is added in some cases to improve the phase margin and stability characteristics with certain types of loads added to the Op amp output. Most often this is not necessary. It tends to be used with power op amps driving specific loads that are not only resistive but reactive as well. You probably don't need this type of network for your application.

5. For the instrument's 2.5V reference, it is sometimes a good idea to use an OPAMP buffer. The OPA350 is commonly used as an example, e.g. in the ADS1256 EVM, but are there newer, higher-performance OPAMPs, that are currently recommended? The REF70xx series is not yet available and the REF60xx is not a preferred choice.

The OPA350 is a terrific buffer and will drive almost any level of output capacitance and remain stable. The next closest in performance to the OPA350, and a more modern Op amp, is the OPA320. It is often used to drive the reference pin, and input pin of an ADS.

Please don't redirect me back to the Data converters forum.

Well, it depends on how nicely you treat us!

Regards, Thomas

Precision Amplifiers Applications Engineering

Thank you both very much for excellent replies.

Could I please also ask you to comment on the planned approach? Are there better and/or simpler ways to achieve this, e.g. using an MDAC programmable attenuator?

-

The product is a refresh of a handheld instrument with the following HW abilities:

The ADS1256 24-bit delta-sigma ADC can operate with or without input buffer and with single-ended or differential input.

Without buffer its maximum range is 0V to 5V ('minus' rail margin and in our case fed from a slightly lower supply voltage (4.85V).

With buffer its maximum range is 0V to [supply voltage - 2V], so 0V to 2.85V 'minus' (GND) rail margin.

It has a PGA that can amplify the input signal in 'binary steps' from 1:1 to 64:1. It probably makes more sense to use the PGA for differential measurement.

-

The MCU 12-bit SAR ADC can operate with single-ended or differential input. It has no PGA and AFAIK it doesn't incorporate an input buffer. We might add an external differential PGA.

-

Both will be running off a 2.5V reference (a common one or two separate ones TBD). In the current design it is a MAX6126 but it will be replaced by a REF7025 once it becomes available. This reference voltage will possibly also be distributed to one or two Wheatstone bridges, in which case an OPA350 (or an OPA320?) are strong candidates.

---

The current generation of the instrument mainly does (GND-based) S-E measurements but my goal is to change all or most to differential measurements. For this instrument we need an external calibration apparatus that can provide all applicable calibration reference voltage levels to what is a total of 5 connection 'points' in 3 connectors:

1.  Two inputs to the ADS1256 with ADS1256-GND
2a. Three inputs to the ADS1256 with ADS1256-GND
2b. Two inputs to the MCU SAR with system (digital) GND
3a. Three inputs to the ADS1256 with ADS1256-GND
3b. Two inputs to the MCU SAR with system (digital) GND

The ADS1256 can handle overranging but its performance in this area is not specified. For this reason the calibration voltage levels must be less than full-range (but at least 80% of full-range). I assume that the MCU SAR can not be overranged. Henceforth I will state the nominal voltage levels (that are more pedagogic) but please keep in mind that they must actually be slightly lower.

As we only need to calibrate one (pair of) input at a time, it seems practical to fan-in from 16 voltage level inputs to 2 'rails' and then fan-out to 16 outputs (for connection to the instrument's connectors). A total of 5 dual-coil relays (EA2-12TNJ) will be used to switch calibration board GND to instrument connector GND. This relay is used in Keithley 2000 DMMs so it should be good enough for us. TI's TMUX1108 would have been a fine choice but the MAX14661 2:16 switch-MUX is more versatile.

So, for calibration I plan to use the following voltage levels:

1. 4.096V reference elevated by 0.25V (by an LT3021 LDO) (diff)

2. 2.5V reference elevated by the same 0.25V (diff)
   (It shouldn't matter that these two near-full-range levels are on an LDO (which is less stable than a precision reference) as it should be common mode?)

3. 2.5V reference from GND supplying a number of resistor dividers (currently buffered by an OPA2192)
   Leg 1: four sets of '3 resistor dividers' generating 1.25V, 0.625V, 0.3125V, and 0.15625V (0.078125V didn't make the cut) centered on 1.25V (diff)
   Leg 2: three sets of '2 resistor dividers' generating 1.25V, 0.3125V, and 0.15625V from GND (S-E) (0.625V is available from the 'diff' leg)

4. 4.096V reference from GND (S-E)

5. 2.5V reference from GND (S-E)

I initially designed for nominal values which in many cases limits your freedom with respect to total divider resistance.
BTW, TI has an online resistor divider calculator that comes in very handy if you need the nominal values:
www.ti.com/.../volt_div3.htm

But, when designing for less than nominal voltage levels we will be able to pick resistances that fit our need. I thought that four 10k-ish dividers from 2.5V (1mA load) would be a reasonable trade-off between low load and susceptibility to noise, but maybe I should choose lower resistances in order to reduce thermal noise.

---

Would you also know at what voltage level thermo-electric effects will have a real impact?

Best regards
Niclas

Hello Niclas,

That is certainly a lot of information to consider. You are working from a proven design so that provides you with a basis to work and improve upon. Since Raymond and I are not an expert on your system design, it is a bit difficult for us to provide specific recommendations to improve the system's performance. However, if there are specific changes you want to incorporate that require a precision Op amp then certainly that is where we can best assist.

You pose the question at the end "Would you also know at what voltage level thermo-electric effects will have a real impact?" Are you talking about the minute thermocouples that can be inadvertently formed by a temperature gradient across the point where two dissimilar metals make contact? Some years ago one of my colleagues did an investigation into the effects on voltage offset contributed by PC board thermocouples. I am attaching that information which is a portion of the total presentation.

Regards, Thomas

Precision Amplifiers Applications Engineering

Parasitics in Precision PCB Layouts_Feb2013.pdf

Hi Thomas,

Thank you for the presentation on parasitics.

There is another way to phrase my question about ADC calibration: Does TI have an application note or some other document on ADC calibration best practice? TI is the leading manufacturer. Surely you have lots of wisdom to share on this topic?

It is probably 50% Amplifiers and 50% Data converters, so maybe a joint response would be fruitful?

Regards

Niclas

Hi Niclas,

Here are some ADC information that may help. 

https://www.ti.com/lit/an/spraad8a/spraad8a.pdf?ts=1616532568059&ref_url=https%253A%252F%252Fwww.ti.com%252Fsitesearch%252Fdocs%252Funiversalsearch.tsp%253FsearchTerm%253Dadc%2Bcalibration%2Bnote

https://training.ti.com/ti-precision-labs-adcs-understanding-and-calibrating-offset-and-gain-adc-systems

https://www.ti.com/europe/downloads/f2810_12_calibration_10.pdf

If you have additional questions, please let us know. 

Best,

Raymond