Sensitivity of opamps to air coupled EM fields, especially of the LM4562/LME497X0 family

[pma]

However there is no doubt something wrong in your test set up to obtain such dire results out of a ne5532 circuit.

Alan, yeah I heard it many times. Did you read how many times it was repeated in a different setup, different test places and with different instruments? And hope you see it is related to instant conditions of the surrounding EM field, and efficiency of shielding of course.

BTW, I got a nice PF card from my colleagues and customers from a VHV lab. I am still doing some specific electronic instrumentation for them, time after time. You know, if the thing works in that place, it would work everywhere. But "we are fully aware", as you are saying.

Happy New Year to everyone!

 

[March Audio]

Alan, yeah I heard it many times. Did you read how many times it was repeated in a different setup, different test places and with different instruments? And hope you see it is related to instant conditions of the surrounding EM field, and efficiency of shielding of course.

BTW, I got a nice PF card from my colleagues and customers from a VHV lab. I am still doing some specific electronic instrumentation for them, time after time. You know, if the thing works in that place, it would work everywhere. But "we are fully aware", as you are saying.

Happy New Year to everyone!

View attachment 44580

So why are there so many successful implementations of these parts? The 5532/4 is everywhere in audio.

Im afraid its an inescapable conclusion.

 

[March Audio]

Experienced suggestions?

My suggestion if you see something like this with a very standard extremely widely and successfully implemented component such as the 5532, is to question what you have done wrong instead of saying the component is problematic.

 

[pma]

It is an indisputable fact that JFET input opamps are much less sensitive to EMI than bipolar input opamps. Bryant, Jung and Kester state that BJT input opamps are 1500x more sensitive to HF interference than JFET input opamps.

https://www.analog.com/media/en/training-seminars/design-handbooks/Op-Amp-Applications/Section7.pdf

As we have seen, there are differences in BJT input opamps as well. IME, the worst behavior may be expected from low power, low supply current parts. The issues that I have shown are related not only to the parts itself, but also to the products where they are used, if they are not designed strictly to meet HF design and shielding rules, and audio products are not. We can see the same issue in some of the soundcards and measurements performed with them. The problem is then sometimes incorrectly addressed to the DUT and its PSU problem. As an example, please see measurements of noise output of the M-Audio Fasttrack USB card. It is standing freely on my test bench. I have measured it in one placement freely standing, then put a hand on the top cover, and then moved the soundcard about 20cm on the table. Please see the 100Hz multiples exactly as in 497X0 measurements. The bipolar input opamp of this soundcard is also prone to EM field in its vicinity. It was measured with XLR input shorted and nothing else than USB cable connected to the card.

 

[SIY]

So why are there so many successful implementations of these parts?

Because they're not using the test circuit shown in Pavel's original post which has 86dB of noise gain. In a more conventional feedback circuit, one sees dramatically better noise performance and this stuff is basically a non-issue.

 

[KSTR]

Because they're not using the test circuit shown in Pavel's original post which has 86dB of noise gain. In a more conventional feedback circuit, one sees dramatically better noise performance and this stuff is basically a non-issue.

Nope, see eg here: https://e2e.ti.com/support/audio/f/6/t/282765?LME49720-Noise-and-RFI
Note the incompetent answer from the TI app guy....
I've seen the issue on my bench as well. This, plus the large percentage of LM4562/LME49720 that have extreme levels of random popcorn noise made me avoid those part as much as possible (after an expensive learning curve). It could be that some RFI burst demodulation actually were mistaken as popcorn noise, that is, in hindsight.

 

[March Audio]

I. The issues that I have shown are related not only to the parts itself, but also to the products where they are used, if they are not designed strictly to meet HF design and shielding rules, and audio products are not.

View attachment 44582

View attachment 44583

View attachment 44584

No.

These are not problems that exist with competently designed products. It doesn't require specialist design measures.

Your circuit or test design has created a problem that does not exist for competently designed 5532 circuits. As such it's impossible to consider any conclusions as correct.

This appears to be another example of your penchant for creating pathological conditions which are not representative of real world designs or situations.

Btw I hope you don't use that sound card for measurement. Nothing I have ever seen on any decent one.

 

[March Audio]

Nope, see eg here: https://e2e.ti.com/support/audio/f/6/t/282765?LME49720-Noise-and-RFI
Note the incompetent answer from the TI app guy....
I've seen the issue on my bench as well. This, plus the large percentage of LM4562/LME49720 that have extreme levels of random popcorn noise made me avoid those part as much as possible (after an expensive learning curve). I could be that some RFI burst demodulation actually were mistaken as popcorn noise, that is, in hindsight.

Look, we are not seeing sensible performance from a 5532. Whilst different op amps have different sensitivities to the issue this is not a competent examination of the subject.

 

[SIY]

Nope, see eg here: https://e2e.ti.com/support/audio/f/6/t/282765?LME49720-Noise-and-RFI
Note the incompetent answer from the TI app guy....
I've seen the issue on my bench as well. This, plus the large percentage of LM4562/LME49720 that have extreme levels of random popcorn noise made me avoid those part as much as possible (after an expensive learning curve). I could be that some RFI burst demodulation actually were mistaken as popcorn noise, that is, in hindsight.

And yet, here's the spectrum of an 18dB flat gain stage using the 4562, with nothing unusual done in my test setup to prevent noise. Implementation and layout are key- like I mentioned earlier, the trick with that particular part is reputed to be a small stopper inductance right at the input pin.

edit: In case I was unclear, this is NOT my circuit design or build, I was just testing it.

 

[mansr]

What about differential opamps like OPA1632 or LME49724?

 

[KSTR]

And yet, here's the spectrum of an 18dB flat gain stage using the 4562, with nothing unusual done in my test setup to prevent noise. Implementation and layout are key- like I mentioned earlier, the trick with that particular part is reputed to be a small stopper inductance right at the input pin.

Yes of course, good design practise would certainly circumvent the issue. The RFI levels we see today maybe weren't fully anticipated for these chips at the time they were developped... there is a reason why most of the current opamp chips now have EMIRR specs given in the datasheets.

 

[SIY]

Yes of course, good design practise would certainly circumvent the issue.

100% agreed, and the corollary is that poor design practice for the intended application will cause issues. And indeed, different op-amps often have differing requirements for optimal design practice (and thus my scorn for the "op=amp rolling" craze).

 

[ajawamnet]

So I designed, ordered and paid the double-sided PCB with metal through-holes and top shielding ground plane.

View attachment 44452
Fig.16. New test circuit that comprises 86dB noise gain/-1 signal gain circuit and also 40dB signal gain circuit, for dual opamps.

View attachment 44453
Fig.17. And this is the photo of the new test circuit

You might want to consider stitching those top/bottom copper pours together with vias...
Typical RF design with lots of vias:

The tool I use (Altium) has a via stitching function that will create an array of vias. It can also generate via walls for GBCPW (ground-backed coplanar waveguide [ which is what you fabbed minus the vias)

Kicad has a cool via fence tool for co-planar waveguides

the "oh so cool" music thing kinda sucks... I hate vids with that.


It might also be that using SMT components will allow a lower profile and reduce the EMI profile of the more critical parts of your test setup.

Interesting paper on RF isolation and EMI :
https://core.ac.uk/download/pdf/84312564.pdf
Where :
"The highest isolation seen in this study was achieved with stripline structures. ... Stripline provides approximately 20dB better isolation over GBCPW under the same layout conditions."

I do a lot of commercial analog designs for ICRI and mil applications and have very little issues with susceptibility; unless I do stoopid (which I'm really good at). Emissions... conducted as well as radiated ... well, see my other posts on that. That can be nightmare, esp for MIL-STD-461 where you need to be below 24dBuV...

For silly sensitive applications, we use guard rings around opamps for applications like medical gear and things like strain gauges, where the output of the device feeding the opamp is low amplitude and a higher impedance.

Another issue is Seebeck voltages that can be caused by thermoelectric voltages

See the AD App Notes:
https://www.analog.com/en/technical-articles/layout-for-precision-op-amps.html#
www.analog.com/media/en/technical-documentation/data-sheets/AD8551_8552_8554.pdf

I've only had a few instances where I had to deal with thermoelectric issues and these were silly sensitive topologies that had little room for output error... nothing that an audio product would ever deal with.


Another appnote from TI on using opamps for lower EMI susceptibility:
www.ti.com/lit/slyt660



There's also the section in Chapter 11 of the AD book Basic Linear Design Guide - Section 11.3:
ftp://ftp.analog.com/pub/cftl/ADI%20Classics/Basic%20Linear%20Design%20(Linear%20Circuit%20Design%20Handbook),%202007/Chapter_11_Overvoltage_Effects_on_Analog_Integrated_Circuits.pdf

 

[ajawamnet]

Obviously I am not expecting you to be listening inside the plane in the picture under
Since you wrote air coupled EM fields in your title I am just curious what it can be nothing more.
With a probe in the air connected to scope or better a spectrum analyzer I can catch lots of external/internal RF...

View attachment 44514

Me too... tho it depends on what's on in my lab.

As to your photo - here's where I spent my summer, grumbling about EMI issues :
http://thehowlandcompany.com/ISTF/ASIL.htm

 

[Arpiben]

Me too... tho it depends on what's on in my lab.

As to your photo - here's where I spent my summer, grumbling about EMI issues :
http://thehowlandcompany.com/ISTF/ASIL.htm

View attachment 44605

Thanks.
At that time (90s), I had a job proposal there. But, I ended using microwaves for only carrying data from a point A to a point B rather than destroying electronic targets at point B.

 

[ajawamnet]

Thanks.
At that time (90s), I had a job proposal there. But, I ended using microwaves for only carrying data from a point A to a point B rather than destroying electronic targets at point B.

They do more that EWS or ECM... any device that goes on any platform has to go thru things like 461 at a facility like that. Pax just happens to be close by and is silly huge...

 

[ajawamnet]

Here's an interesting article from ICD that uses quantum mechanics and string theory to explain CPW simulation in field solvers:
http://www.icd.com.au/articles/CPW_PCBD-Mar2017.pdf

 

[ajawamnet]

I do a lot of commercial analog designs for ICRI and mil applications and have very little issues with susceptibility; unless I do stoopid (which I'm really good at). Emissions... conducted as well as radiated ... well, see my other posts on that. That can be nightmare, esp for MIL-STD-461 where you need to be below 24dBuV...

I should clarify that most of our designs are minimum 4 layers... using a plane for power distribution (with good bypassing) really helps, as does having the ground/power plane closer to the routing layers...

 

[pma]

Alan @March Audio , these are time records that belong to the plots you did not like. It was the original unshielded universal PCB, as already described in post #1. The double-sided PCB with groundplane was produced just after these measurements and it significantly reduced the spikes, though not removed completely in case of LM4562. However, nothing can neglect the fact that different opamps behaved very differently in the same setup and that sensitivity of LM4562 to EM field was and is highest.
All plots for 86dB noise gain and -1 signal gain, input shorted.

Alan, may I ask you not to make cherry picking, not to pick up my measurements from the context of the post??

Universal unshielded PCB

LM4562 universal PCB

NE5532 universal PCB

TL072 universal PCB

Double-sided test PCB with ground plane

LM4562 double-sided test PCB with ground plane. The level of spikes is still high enough to be easily detected by spectral analysis.

LM4562 double-sided PCB, output amplified 30x and filtered by 20kHz 2RC LPF. Measured by oscilloscope only. USB not connected, PC is OFF.

 

[JimB]

... the large percentage of LM4562/LME49720 that have extreme levels of random popcorn noise ...

I have this experience, too - low level "rustling" in the background. It seemed (going back 12 years) that some parts were bad, others not. But later, some previously good parts became bad. Most annoying. Ten years ago, I meant to put in some OPA1612s (samples then provided by a fellow audiophile how worked for BB/TI), which required getting package adapters, but never got to it. Maybe this year?