OPA4141: Popcorn noise

Morning Alex, 

Some parts do get popcorn screening but as you note the long cycle ones are very expensive to screen out. 

The easiest way to ensure no 1/f or popcorn noise is to use a chopper op amp. 

Hi Michael,

As for me - it is impossible to use choppers in biopotential measurements systems. At least I tried OPAx388 and OPAx188 and OPAx180 - all of them demonstrate very high noise in real situation with high impedance source and long cable (10-100kOm and 1m wire). I`m not sure why - my suggestion is wether it is high EMI susceptibility or something with charge injection from input switches. And as I see now - there is no "medical instrumentation" tag in OPAx388 datasheet (it was some time ago).

I even posted about this problem here before and deсision was "choppers are just not well suited for high-impedance sources".

UPD:

...and here is fourth (from 21) failure chip. It spikes very, very rare, but with groups of severe spikes.

Well yes, that is the chopper input current getting gained up by the source impedance - if you have not, you might try a filter cap right on the V+ input pin to ground. Your required signal BW is probably low, and the chopper F is high - giving it  a low impedance path at V+ to ground might help

It is not a good way to place capacitor at input because it introduce small phase shift at 50/60Hz thus decreasing common mode rejection. At least, my tests shows that this capacitance need to be too high to keep chopper-induced noise low... May be I`m doing something wrong, but anyway I did not seen any really working reference design for biopotential measurement using choppers, but a lot of promising scientific(!) papers.

Btw, I`m still using two x388 in this project to condition OPA4141 output and them works well.

Hi Alex,

can you show a schematic of your circuit which uses the chopper amplifier?

Kai

Shematic is almost same as in my old post, except I tried to put caps near +In of both amps and changed 10kOm input resistors to 100kOm. As you can see on the picture below, noise is much higher than with OPA4141 even with 1nF on +In pin. With OPA4141 noise is almost at theoretical minimum for 100kOm. Please note that this is a simultaneous record aquired at a single board. I just replaced one of my OPA4141 with OPA4188 chopper and add caps from +In leg to VCM. Without caps on +In pins noise is 4-6 times greater,  about 4-6uV rms.

1nF at input means only 2,65MOm impedance value for 60Hz... It gives very bad CMRR when electrode resistance differs even by 10kOms.

Hello Alex,

The OPA4141 was introduced ten years ago and is fabricated on an older TI JFET/bipolar process. Even thought the differential input devices are JFET, there are bipolar transistors in the input stage as can be seen in datasheet Figure 1. It is highly unlikely that the JFETs are exhibiting popcorn noise, and much more probable that the bipolar transistors are responsible for enerating it. 

It is good you tried the OPA4188 auto-zero op amp in the application even though it didn't work out as well as might be thought possible. Although auto-zero op amps such as the OPA4188 don't exhibit 1/f noise, they are noise sensitive to the impedances that they connect to. There is charge transfer current noise generated by the internal switching and in high impedance circuits can result in that noise being converted to voltage noise. That may be why the noise was higher with that particular op amp.

I suggest trying the OPA4192 in your EEG application. It is a very modern, precision CMOS op amp having moderately low noise. It is processed on our modern CMOS process which is very unlikely to have popcorn noise. It does not use any chopping, or auto-zero methods to achieve the very low offset and drift. The noise isn't as low as what you would observe with an OPA4141 having no popcorn noise, but you may find its overall noise performance is acceptable for your application.

www.ti.com/.../opa192.pdf

Regards, Thomas

Precision Amplifiers Applications Engineering

Hi Thomas!

Thank you very much for this explanation. Now we are thinking about two devices: more expensive but ultra low noise version  with OPA141 and low-cost with some different amplifier. Because our volume is not very high, may be it is acceptable for us to check every amplifier for popcorn, even for  several hours. Just need a few test pcb`s with SOIC14 sokets. But the main question long-term probability function of popcorn cases (sorry for my english), may be you know some studies about that? 

Thank you very much, guys! Of course, it would be nice for consumers that manufacturers indicate information about susceptibility to popcorn noise in the datasheets (as they draw histograms for offset voltage)...  I think that it cannot be said that a typical analog engeneer (not to mention hobbyists), even with good experience, knows a lot about this type of noise and understands what happens when faced with it.

You bet Alex, 

Having worked on popcorn screening in the past, it did occur to me that wouldn't it be nice to see comprehensive list of screened part numbers and the length of screening (and as long was we are dreaming, perhaps an estimated test escape rate?) Oh well, since popcorn is not supposed to exist at all, very much a sweep it under the rug issue (not just TI, ADI as well - I have plenty of test data showing popcorn on some of the ADI parts we tested from them - if you ask, doesn't exist and we clearly had that popcorn in our test fixturing). 

And, from an operational perspective, popcorn screening is like several orders of magnitude longer test time - every time you move assembly and test sites, some eager new hire wants to strip that out as a superfluous test line and can easily generate a bonus for himself eliminating it. 

One of the recent AudioXpress articles I did was showing a popcorn preamp updated to a chopper implementation (essentially a high gain active filter from 0.1Hz to 10Hz) - nice to use those as you know you don't have popcorn in the test fixturing.