THS4551: Output offset problem when doing ESD test

Hello Vincent,

   Thank you for the detailed setup; it helped clear a lot of initial questions. Can your customer measure the differential input voltage of the THS4551 (at pin 8 and pin 1 of U203) at both before applying the ESD test and after applying the ESD test?

Thank you,

Sima

Hi Vincent,

the input circuitry is incomplete. Resistors for providing a current path for the input bias current of OPA2197 to signal ground are missing.

Kai

Hi Vincent again,

R250 between the inputs is no good idea. But a 1M resistor from each input (V1+ and VI-) to signal ground (AGND) is.

By the way, have you connected VI- to AGND during the test or have you let it float?

Kai

Hey Vincent, 

I was going to take a look at the MFB for noise improvement, but then I noticed that 44kHz RC pole at the inputs of the buffer amps using 25kohm R's - that kills your noise right there so nothing to do in the FDA stage. 

But, I had the MFB LG test files from the THS4551 datasheet prep, so I modified one of those to this MFB for a phase margin test - oops, looks pretty oscillatory - maybe at 50MHz -40deg past 180deg phase shift - that is the 4.7nF feedback C doing this - this later article talks about these issues a bit. If I got the RC values wrong let me know and I can change easily, And, I was using the original release TINA model as it has a more accurate open loop Zol (maybe). But I could be making a mistake here as well, 

Also, the filter is only 25kHz, could do this pretty easily with the THS4531A and save power - that part is a lot higher input noise, but again it doesn't matter with the noise coming out of the buffer amps. 

here is the the THS4551 LG test Tina file, V7

THS4551 MFB LG original model 25kHz.TSC

Hey Vincent, 

I went ahead and pursued this a bit with the THS4531A updated 2018 model (oddly, the original THS4531 still has a very old 2012 model I think - would be easy to update that online to the improved THS4531A model relaxing the model offset voltage which is the only change). Anyway, just dropping in the THS4531A sill had phase margin problems - this is a combination of the lower build out R to the load C and the interaction of the open loop Zol with the feedback C. 

To get this phase margin up into the 30deg region, I made the changes shown here - 

And then this last file - again, I am not 100% I have these RC values right or that this is your root issue, but normally - you don't want to waste any time with mystery symptoms until the phase margin is good. I did not try the closed loop response with the THS4531A but would expect the filter shape to be fine, 

THS4531A MFB LG original model 25kHz.TSC

Hi Team,

Thanks very much for the comments and suggestions. There is some information updated.

1. Customer put the board in temp. chamber and set Ta to 55℃, can replicate this issue.

2. Put the board running at room temperature for a while, say 5 hours, customer also can replicate this issue.

It seems like the design has problem originally, ESD test or high temperature test just speed up the failure to happen. I suspect that this issue may relate to common mode shift in OPA2197 circuit. The 20mA current generator doesn't have ground wire to connect AGND, and OPA2197's input side is floating. Below diagram shows how customer connect current gen. to the terminals of PCB.

I will collect information you guys asked above and get back to you.

Thanks very much.

Vincent Chen

Hi Team,

After did some experiments then we almost can confirm that the issue is caused by input common mode shift.

1. Remove OPA2197 and ADS8912B, apply 5V differential voltage at input side of THS4551 and THS4551's output is 1.78V that is a normal value. Keep it for hours and output can keep at 1.78V. There is not problem on THS4551's circuit.

2. Remove R204 and R205 to test OPA2197 circuit only. Short VI1- to AGND and apply 5V voltage on V1+ and VI1- as a differential input voltage. There is no longer to have output offset issue.

3. Test whole system and let it running for hours and then appeared output shift issue, then short AGND and VI1- the issue was gone. The THS4551's output voltage became to correct value of 1.78V.

Remove R250(1M resistor) and put 1M resistor from V1+ and VI1- to AGND that is the purpose to keep differential voltage input to have a common mode level close to AGND, is it correct?

This approach seems workable but when customer used multimeter to probe voltage at any point of whole system, it will affect the result a little bit. We are trying to make sure this approach is OK. Is there any other way to let common mode close to AGND and has less side effect? 

Besides, put 1M resistor from V1+ and VI1- to AGND has side effect or not?

Thank you very much for your support.

Vincent Chen

R250 between the inputs is no good idea. But a 1M resistor from each input (V1+ and VI-) to signal ground (AGND) is.

Hello Vincent,

  Kai pointed out that the customer's input circuitry is missing a connection to ground which would give the circuit a DC bias current return path. At the moment, your input is floating at DC as you mentioned. This is most likely the cause of this offset error. Here is the Tina-TI simulation demonstrating Kia's suggestion as well as a method of including both connections, but Kai's comment seems to be the best fix. After this fix, I would suggest reviewing Michael's comments to increase the stability of their circuit at the FDA end. 

Thank you,

Sima

Vincent_OPA2197.tsc

Hi Sima,

Thanks very much for the advice.

Since customer followed app. note design,SBAA267A, to achieve high input impedance. The app. note didn't show any connection to ground but has very good performance. Therefore, engineer followed this reference design for PLC AI module of differential input application.

Although adding 1Mohm resistor to ground can fix common mode issue but the side effect is thermal noise, right? The resistor's noise density seems large than that of OPA2197 and THS4551. Could you please help comment my concern?

Thanks very much.

Vincent Chen