THS4541: Experiencing drift with ths4541 and OPA857

Hey Josh, 

You are correct that zero drift (or chopper) amplifiers eliminate drift (and 1/f noise!) but they are limited to approx less than 10MHz GBP - so no, there are no options like that for you

IF you included a ckt, I do could more - TINA file preferably. The THS4541 datasheet is one of mine so there are sometimes ways to improve these things. 

Hi!

Thanks a lot for your help. 

The data from the PSD is sampled at max. 100kHz in my application. A 10MHz GBP should be more than sufficient or am I missing something here?

Attached, I am sending you the circuit design. Unfortunately I dont have TINA-files. Sorry.

Josh

1.sch.pdf2.sch.pdf3.sch.pdf

what you sent does not have much in it, an if you rmax is 100kHz these parts do seem a bit fast, 

Hi Michael,

Did you notice that I have attached three pdf-files? The first one shows the PSD element with its four transimpedance connections using the OPA857. After that the voltage signal is further amplified using the differential amplifier of type THS4541 (shown in the second pdf-file).

The purpose of this circuit is to exactly determine the position of a small light spot along the PSD. The problem is that all 8 amplifiers show a temperature-related drift, which is interpreted as a false movement of the light spot.

Since the whole setup is compact, I cannot use any kind of external heat compensation. That's why I was hoping to find less-sensitive amplifiers. I am also monitoring the temperature of the board while operation in the intended field of application. It varies between 45°C-55°C.

Josh

1. The key thing for the Zt stage is the total source cap - is that 1.5pF all there is or should we consider the diode cap also? Also those output R's are ambiguous, 33ohm? or 383ohm?

2. Essentially the task here is to assess min required BW in total and then each stage - you said 100kHz what min BW in the amplifier channels is required - you are using very fast parts right now? You do not necessarily need a diff out Zt stage, the FDA will make that conversion where perhaps the THS4551 would be enough speed and it is better drift. 

Hi!

The terminal cap of the PSD is 150pF (typically, max 300pF). I don't know if or how this should be considered. Any ideas here?

These R's are 383ohm each. Sorry for that.

At the end I want to track the light spot on the PSD at 50kHz, so the sampling frequency (or amplifier channel BW) should be at least 100kHz. Right now I use a gain for the FDA of about 5.2.

I have looked into the THS4551. Thanks for that hint. From my understanding it indeed looks more appropriate for my purpose. It also has lower current consumption and I hope this will also reduce the heat development.

Ok, so your total gain is 104kohms and we need to split that between Zt and FDA stages. I will set this for 50kohm input Zt and 2.04V/V FDA gain with a 3rd order filter there to define BW as the Zt stage will vary with source C. 

I am not sure you really want to go chopper in the Zt stage due to sampling noise, but go ahead and try the OPA4388 quad 10Mhz part, 

For 0V outputs, we need some headroom on all outputs, use the LM7705 for all negative supplies (1 unit) giving a -0.23V reference)

I kind of targeted the input stage at >100kHz to set the response in the next stage, here that is, 

For the FDA stage, set it for a gain of 2.05 for now (can easily put much more there if need be). Then target a 3rd order Butterworth at 120kHz requires a 2nd order Fo = 120kHz and Q=1 with an output RC at 120kHz. You may not have the space for the filter RC but in any case use dual THS4552. here is the 3rd order filter stage, ignore the Vmic thing, this was a file I adapted from another ckt for an audio magazine article, 

And then putting this with Zt stage you can see the overall response varies not at all now with the source C

And then the input referred spot current noise is very low, as is the nominal output DC offset, The drift will only be due to the THS4551 now, I had to use the 2016 model as the new 2019 THS4551 model is not modeling noise correctly. 

Thanks again for your effort. This really looks like a lot of work. 

You seem to have doubts about the chopper.. Well, I have no experience with that. But a stable signal over a long period of time at varying temperatures (+/-10°C max) is my highest priority. That's why I would like to try that.

I've probably been sitting in front of the computer too long and can't think straight. But I don't understand how the gain of 2.05 results from the circuit.

Yes, I should have mentioned the standard value snap routine is looking for best Fo & Q fit and the letting the gain go off one E96 standard value, 

Here is that test case matrix, The MFB gain is feedback R divided by input R, R1/R3 in this case. So nominal FDA stage gain is 2.06 in this case.