ADS1278: Cross-coupling or offset issue in differential inputs of ADS1278 with THS4521 buffers

Hello Javier,

Based on your accuracy requirements, I do not think this approach will work.  The differential input impedance in this case is 2*Rg=40kOhm, but the common mode input impedance is approximately (Rg+Rf)/2=12.5kOhm.  This common mode impedance will have additional current relative to the common mode voltage.

In your above example, I assume the debugging setup is isolated from the power supply used for the ADS1278.  (There is no ground connection between the 15V source and the ground of the ADS1278.)  In this case, if you only connect one channel, then the only error will be due to the input differential impedance of that channel, or 40kOhm in parallel with your 100ohm resistor.  However, once you connect a second channel, you now have a common mode path between the channels and the ADS1278 ground.  In this case, much more current will flow.  I would need to setup a SPICE simulation to get exact numbers, but I think if you connected an input channel to R1 and a second channel to R5, you will now see an approximate common mode current of 15V/(CH1CM_Rin+CH2-CM_Rin)=0.6mA.  0.6mA of error current through the 100ohm resistor will result in about 60mV of error voltage.

If you are measuring across a much lower impedance source, such as a string of batteries, then the error voltage will reduce quite a bit.  You can verify this by using 10Ohm resistors (will need to be higher power rating) instead of the 100ohm resistors and do the same tests.

If you are measuring voltages on a resistor network similar to your setup, you will need to use either a difference amplifier with a much higher input impedance, or for best performance, an instrumentation amplifier.  Unfortunately, with the THS4521, you really cannot increase the resistor values higher than what you already have and maintain good DC (input bias currents will create additional errors) and AC performance (input capacitance will interact with the feedback resistance, causing stability issues).

Assuming your common mode voltage is in the range of +/-10V, and you have +/-15V supplies available, you could use the INA851.  This is a differential output instrumentation amplifier that can directly drive the ADS1278 inputs, removing the need for the THS4521.

Another possible option is to use the INA149.  This is a difference amplifier that supports much higher common mode voltages, and uses much higher input resistors.  The resulting error current will be much lower than the configuration you created with the THS4521.  If using the INA149, you will need to also use the THS4521, since the output stage of the INA149 will not be able to drive the ADS1278 input and maintain high accuracy.  Also, the INA149 can operate from a single 5V supply, but the output will be limited from 1.5V to 3.5V in this case, which will limit the input difference voltage as well.

If you can provide additional details on what you are actually trying to measure (source impedance, common mode voltage range, differential voltage range), I can refine my recommendations.

Regards,
Keith Nicholas
Precision ADC Applications

Hello Keith,

fist of all thanks a lot for your time and ideas, by reading it I realize why I like to use parts from TI. I have yet to see another forum where I can get such a level of support. Regarding my application, it is basically meant for measuring voltage drops among semiconductors (typically Vds in SiC MOSFETS, Vce in IGBTs or Vf in Diodes) which are placed in series in several branches which are connected in parallel (up to 4) and each one enabled/disabled via a switch (MOSFET) for a certain period of time. It is basically power cycling:

Below I1, I2 and I3 we would have up to 4 devices under test connected in series. We have as well applications where the target would be to measure leakage currents from semiconductors. Precisely to the requirements actually for each DUT, depending on the test:

1- We would have to be capable of reading signals with 10Vpp differentially for each channel (DUT), for a complete branch the requirement would be in worst case 45Vpp (we could limit it to 40Vpp if necessary). In our measurement system we have several isolated ADC blocks with ADS1278 but it would be important to be able to cover one complete branch with just one block. This would mean, that between channels of the same ADS1278 the voltage difference could get up to almost 45Vpp (common mode voltage range).

2- Regarding input impedance of each channel, ideally we shall be able to increase it to >1Gohm, although here we have some flexibility if we have to balance among requirements.

3- Regarding the supply of the front-end, it would be important that we manage to use the single supplies of the ADS1278, f.i. the 5V one, as it is the case currently for the THS4521. Regarding the voltage we have some flexibility though (if really necessary), as before the LDOs that supply the ADS1278 we have a flyback with output voltages of around 8V, but I could increase them easily and place an additional LDO if necessary. The important thing is to cope with 1 single supply if feasible to reduce complexity.

4- Regarding the package, the VQFN of the INA851 would be an issue. We would like a package which is more hand-soldering friendly for rework/development, such as the SOIC or VSSOP.

5- If possible I would as well try to avoid using a combination of two amplifiers, such as THS4521 and INA149, but a single one.

6- As we will have to use many amplifiers for each board, the BoM costs shall be optimized as much as possible (f.i. INA826 or INA821?). If we can find some parts which are cost effective we could leverage some of the requirements.

Given these requirements what would you think it would be the most suitable TI part?

Another question. Do you think that there is a way to increase the input impedance (common and differential) of the THS4521 at the expense of noise so that we come from the 60mV difference of the setup that I have sent to you to something in the area of 1mV? The question here is really if we can fix short term somehow these first boards to run some tests.

Thanks again for your valuable support,

Javier

Hello Javier,

You may be able to increase the input resistors to 400k; according to simulations it looks like this should work.  However, in order to support up to 45Vpp of common mode, you will need to increase the attenuation (reduce the gain).  This will reduce the differential voltage at the ADC inputs, resulting in lower SNR, but this may still be acceptable.  In any case, with such a high common mode voltage, in order to reduce common mode errors, the tolerance for the resistor networks will need to be 0.05% or lower.

Below is a suggested circuit to try out.  The offset voltage will increase as Rf increases due to the mismatch in input bias currents, but you should be able to calibrate this error.

If you need 1Gohm input resistance, then the only way to achieve this is to use high-voltage input buffers, which will require a supply voltage greater than the common mode input voltage.

Since you need to support common mode voltages up to 45V, the instrumentation amplifiers are not going to work.  In this case, creating a discrete solution using the THS4521 is probably your best choice.

Regards,
Keith

Hello Keith,

thanks, I will follow your advice and try the modifications on the THS4521 circuit. Regarding the 45Vpp, the majority of our applications would just need 20Vpp. Would an INA821 or an INA826 fit in this case?, which supply would we need for them?

Would it be possible that show me the reworked THS4521 setup for the 20Vpp as well? How do you calculate the maximum input common mode voltage of the THS4521 and the related attenuation factor to comply with it?

Regards,

Javier

Hello Arijav,

My below comments are based on the assumption that the ADS1278 board ground can be connected to the low side of the DUT, as shown in the figure below (required for proper INA82x input bias).  Also, I assume the voltage on the DUT relative to this system ground will aways be positive.

In this case, you could use the INA826 with a single supply that is 1V or more greater than the maximum input common mode that needs to be measured.  If the maximum voltage on the top side of the DUT (emitter of the IGBT) is 20V, then the supply voltage for the INA826 would need to be 21V or higher (up to the maximum rated voltage of 36V).

In this configuration, the INA826 would be able to measure a differential voltage from 0.1V up to 10V.  If the INA826 was powered with the maximum 36V rating, then it could support measurements up to 35V on the top side of the DUT (emitter of the IGBT) relative to the ground shown above.

I ran a SPICE simulation to determine the minimum attenuation needed to meet the input voltage requirements of the THS4521.  Re-running the simulation with a maximum voltage of +20V results in a gain of 2x the previous recommendation, or Gain=0.05.  Your initial design, with Gain=0.25, common mode is limited to about +7.5V.

Regards,
Keith

Dear Keith,

I have setup in Tina the same simulation that you have prepared and now I understand better how to tune the current circuit based on the THS4521 for my needs, many thanks.

For the next revision I am still wondering which is the best path. I have 2 options considering a maximum Common Mode voltage of 25Vpp:

1- INA149 + THS4521: In this case I might have to go to 2 rail supplies but it could be a good solution. Is there any application note where I can see a design both with the INA149 as well as the THS4521 with an ADC? In this case I would use the THS4521 with a gain of 1, correct?

2- INA821/826: Actually in my application I do not have a ground connection from the ADS1278 block as you have shown in your picture. I would prefer as well to avoid it if it is not 100% necessary. If I understand you well this is a must have for these instrumentation amplifiers, correct? Are there as well application notes where I could see the design for an INA821 together with an ADC? Specifically for the INA821 I will run some simulations but I have already used your Analog Engineering Calculator:

If I do not connect the Grounds my assumption would be that I would be able to have input common mode voltage differences of up to 20Vpp for +/-18V supplies. This would be even if not perfect quite good for many of my applications. Is it really necessary to connect the grounds?

Thanks and Regards,

Javier

Hello Javier,

I am not aware of any specific application notes that show the combination of the INA149 and a differential amplifier.  However, the INA149 is a fixed gain of one, so you would still need to attenuate by 4x for a 10V differential input when using the ADS1278 with a reference of 2.5V.

Here is an application note that shows using the INA149 with a +/-10V ADC input.  Using the THS4521 with a gain of 0.25 will then provide the same input signal range.

https://www.ti.com/lit/an/sbaa337/sbaa337.pdf

Regarding the use of an instrumentation amplifier, you can use one of these devices without a direct ground connection to the DUT, but you need to use biasing resistors in the 1Meg to 10Meg range.  The below application note discusses this approach.

https://www.ti.com/lit/an/sboa503/sboa503.pdf

If interested in more background information, we have a training series, TI Precision Labs, with a lot of material.  There is an entire section that covers the use of Instrumentation amplifiers.

https://www.ti.com/video/series/precision-labs/ti-precision-labs-instrumentation-amplifiers.html

Regards,
Keith

Thanks Keith, one last question. When using the INA821 I will have to attenuate the signal with a 0.25 factor. Do I have to use the THS4521 additionally to the INA821 or is there a more effective way to do so?

Due to the topic of the attenuation I am actually rethinking the use of the INA851, even if it is not that cost effective.

Hello Javier,

In the case of the INA821, the minimum differential gain is 1, so for a 10V differential input, you will need an additional attenuation of 0.25.  The most convenient place to do this is in the THS4521 stage.

One other possibility; since you need a differential gain less than 1, another option is to use a PGA.  The PGA855 can be set to a gain of 0.25, and can directly drive the ADS1278 inputs using the differential output.  However, similar to the INA851, this part is only available in a QFN package.

Regards,
Keith

Thanks Keith, I have decided however to move forward with the INA851. Would you suggest me to use for each channel of the ADS1278 the same circuit as in figure 9-9 of the datasheet and adding the 1 to 10M biasing resistors between Vin- and GND and Vin+ and GND or shall I modify something? My target is to sample the channels at 100KSPS:

The VOCM pin for the 8x INA851 can be driven in the same way as with the THS4521 using an OPS2350 and the VCOM output of the ADS1278, right? In my current design I have a 100K resistor and a 100nF capacitor instead of the 10Ohm and 150pF shown in the datasheet.

I have as well modified the current design with the THS4521 using 200k + 10k Resistors. Unfortunately I experience exactly the same behavior as before. If I read the voltage of R1 when R5 is connected the voltage jumps around 180mV this time (probably more than before because of the attenuation factor). Therefore this issue at least for the THS4521 does not have to a relationship with the resilience to higher common mode voltages. Do you have any clue what could be the root cause for it?, I would not like to go to a new design with the INA851 and realize that I have still the same issue...

Regards,

Javier

Hello Javier,

My previous reply is still a simplified schematic.  The configuration that you show above will work.  However, I recommend connecting the 1M (or 10M) resistors directly to the inputs (INA_AINN1 and INA_AINP1) to reduce any gain error. 

Yes, you can also use 1M resistor differential and 10M for bias, as suggested in the application note to improve measurement accuracy (reduce offset). 

Also, I would suggest reducing the capacitor values by 10x and increasing the resistor values by 10x to improve input-over voltage protection.  You have a TVS diode to protect the differential; you may also need a TVS to the system ground as well.

Following all of these recommendations, the circuit will look like the below image:

Also, as long as the entire measurement circuit operates off of an isolated power supply relative to your DUT (bipolar supplies for the INA851, 5V,3.3V,1.8V supplies for ADS1278), then you can use +/-18V supplies and the bias voltage will then simply be connected to the circuit ground.  This should allow a total voltage across 4 DUTs of about 28V and still meet all of the input and output voltage requirements of each INA851.

The VCOM input from the INAs (in total 8x per Block) go to the 2.5V Output of the ADS1278 buffered by an OPA2350. Do you see any inconvenience in this case? 

Yes, using OPA2350 to buffer the VCOM signal and then routing to each INA851 should work well.

Regards,
Keith