Test of RFI circuit susceptibility to AM modulated signal with 1MHz carrier

[pma]

Test of RFI circuit susceptibility to AM modulated signal with 1MHz carrier

Well – this was tough. I wanted to test possible susceptibility if input analog circuits to AM modulated signals like AM transmitters etc. Sometimes we hear complaints about audio equipment capturing and demodulating local radio stations – so how to test. I admit I had similar troubles when I was a youngster and started to build my own preamplifiers and amplifiers, however it was about 50 years back from now and today I am even not able to replicate all the mistakes from those nice beginner years.

So I started thinking about the AM modulator and drawn this simple circuit with gain modulated by the D-S resistance of the JFET:

I built it and briefly checked the function – it is working though quite nonlinear, however it does not matter for the purpose. So I used my Audio Generator set at 1MHz as a carrier frequency generator and a soundcard as a modulation signal generator and connected them to the modulator inputs. The devices under test were a RIAA phono preamplifier and a linear 30dB amplifier. Outputs of those amplifiers were filtered by a 2xRC 30kHz passive LPF (to get rid of the HF) and then analyzed by a soundcard and monitored by a DSO – DSO directly without the LPF filter.

I decided to test the phono pre and the 30dB linear amp, with various opamps swapped. I started with the phono pre and a reasonable input level of some 10mV, but got no meaningful results, except for some method errors. Then I started to test the 30dB linear amplifier. Again, no results, until I decided to rise the input voltage level considerably, to some 212mVrms of the carrier frequency 1.006MHz, which is close to the AM transmitter in the vicinity. The modulating frequency was 3kHz. As the modulator is not very linear, we can see not only the Fo-Fm, Fo and Fo+Fm frequencies (1003, 1006 and 1009kHz), but also Fo-nFm and Fo+nFm frequencies, however as already mentioned, this does not matter for the susceptibility test.

Output of the AMP30dB with LM6171 driven by the AM modulated test signal

Output spectrum of the AMP30dB with LM6171 driven by the AM modulated test signal

So, what was tested and what is the result? NE5534, OPA627, OPA637, LM6171 and some other devices. To make the long story short, the only important parameters were GBW and slew rate. All the named opamps exhibited some degree of demodulation except for the LM6171, which was perfect. The demodulation started to occur when the output signal slew rate started to be comparable with the slew rate of the opamp used. Because the LM6171 is so fast, it did not have any problem to handle the AM signal. Though it is a BJT input opamp, there was no demodulation.

We can see some demodulation at 3kHz and multiples with the OPA637 (though it is a very fast opamp)

... and no demodulation with LM6171, extremely fast opamp.

The input level of the AM signal in the test was insane and nothing like this should occur under home conditions. With mV and 10mV signals, nothing suspicious was measured. However, I of course do not know the circuit diagram of the components that are reported to have troubles and also do not know the EMI fields there. So my conclusion only tells that under “normal” conditions with “normally” designed component there should be no troubles with AM transmitters demodulation at the amplifier input .

 

[pma]

After testing more solid state amplifiers, nothing much different was found and even the AIYIMA A07 performed quite well, with negligible demodulation effect.
The only "hit" was my SRPP tube preamplifier with ECC82 (12AU7) and 16dB gain. Under same conditions as in the post #1 the output spectrum looked like this

Just to mention, 0dBFS in those 20kHz spectral plots equals to 6V. The demodulation effect here seems to be a result of higher transfer function nonlinearity of the tube, as it is not slew rate limited.

 

[pma]

The same test on tube pre as above, but spectrum calibrated in dBV

Now another test, with the same 1MHz carrier modulated by 13+14kHz twin-tone IMD signal

This brings a lot of intermodulation components due to both modulator and tube non-linearities. And in audible level, however the input signal level was again higher than realistic.
At least we have some relative comparison of the parts more and less prone to AM demodulation effect. The best performers are the fastest opamps with very good transfer function linearity. Both low (and asymmetric) slew rate and nonlinear transfer function are killers of RFI demodulation immunity.

 

[pma]

So, this would be "the golden nail". The test of the LME49720 (LM4562), my least favourite opamp concerning RFI immunity. I already wrote a thread
https://www.audiosciencereview.com/...pecially-of-the-lm4562-lme497x0-family.10687/

and now it was put under this AM modulation test. The tested circuit is the TI LME49720 evaluation board as described in the linked thread. It has inverting voltage gain of -1x, two 10 kohm resistors in the simplest inverting circuit.

Yes, the IC does not like RF. The input voltage was high, 2.74Vrms of 1MHz carrier. This makes dv/dt (aka slew rate) = 24.3 V/us, just near the opamp specs. This was modulated by 2.5kHz and quite mildly, about 10-15%. And here is the result

Yes the input level was high, very high. But remember it was only 1MHz carrier. With 100MHz carrier (like FM radio transmitters), 100x lower level, i.e. 27mV would make the same dv/dt thus the same issue. So the designers should make a proper input RF filtering with respect to HF design rules.

The parts have different RFI immunity and it is good to know. Interestingly enough even MicroCap simulation shows the differences.

 

[chris719]

At work, we blast things with +33 dBm GSM source in the 850 and 900 MHz bands, and +30 dBm in 1800 and 1900 MHz bands. The 217 Hz TDM really gets into everything and is an easy way to find immunity issues.

I cant believe the amount of trash that came your way in your previous thread on this topic. It's pretty clear that LME497x0 is an outlier. I see almost no reason to use it now when you consider EM susceptibility, popcorn noise, and the availability of OPA1612 and OPA1656.

 

[pma]

At work, we blast things with +33 dBm GSM source in the 850 and 900 MHz bands, and +30 dBm in 1800 and 1900 MHz bands. The 217 Hz TDM really gets into everything and is an easy way to find immunity issues.

I cant believe the amount of trash that came your way in your previous thread on this topic. It's pretty clear that LME497x0 is an outlier.

Yes it is an outlier. I think that the reactions are based on audio industry interests. This family of opamps has become quite a standard in mid-priced audio equipment. The producers and designers argue with low distortion specs and nothing else. It is a way of "blinded" approach. Too much interests in it. And they count on numbers, on useless inaudible numbers and do not want to hear about possible issues. However, time after time we can read about problems the users have.

AM carrier modulated by DIM, one of the worst scenarios. It is interesting that it remains bad even if the opamp SR is not exceeded.

the test signal spectrum

 

[chris719]

Yes it is an outlier. I think that the reactions are based on audio industry interests. This family of opamps has become quite a standard in mid-priced audio equipment. The producers and designers argue with low distortion specs and nothing else. It is a way of "blinded" approach. Too much interests in it. And they count on numbers, on useless inaudible numbers and do not want to hear about possible issues. However, time after time we can read about problems the users have.

I doubt it’s an audible issue for 99% of situations, but I agree there is too much focus on the spec war. I prefer a product that is robust, ESD protected, and has undergone some testing in a representative setup - not sitting alone on a turntable in an EMC chamber.

I’m sure you can design a device with these op-amps that doesn’t have a problem even in harsh environments, but I doubt most audio gear has put the required mitigations in place.

 

[KSTR]

Excellent as always, Pavel!
Thanks for putting this up.

 

[restorer-john]

@pma Your laptop battery appears to be very swollen. I would be removing and disposing of it safely before something bad happens (fire).

 

[pma]

This shows how incredibly good is circuit simulation even in a case like this - MC11 simulation:

 

[AnalogSteph]

Now the interesting question is why would the LM4562 be that much more sensitive to EMI than its specs would suggest? I don't think very much is known about its architecture, except that it seems to use a substantially degenerated input stage (which would make the thing less prone to RF if anything) and nested feedback much like a 5532. Something to do with input protection or bias current cancellation perhaps? Most audio opamps don't even employ the latter, given that it generally has a negative impact on current noise.

It might be interesting to test some other parts with bias current cancellation then. OPA(2)227, OP27, OPA(2)211, LT1028, LT1115, that sort of stuff. Also cross-check with a few other common parts just in case (NE5532, NJM4580, OP275).

 

[pma]

Now the interesting question is why would the LM4562 be that much more sensitive to EMI than its specs would suggest? I don't think very much is known about its architecture, except that it seems to use a substantially degenerated input stage (which would make the thing less prone to RF if anything) and nested feedback much like a 5532. Something to do with input protection or bias current cancellation perhaps? Most audio opamps don't even employ the latter, given that it generally has a negative impact on current noise.

It might be interesting to test some other parts with bias current cancellation then. OPA(2)227, OP27, OPA(2)211, LT1028, LT1115, that sort of stuff. Also cross-check with a few other common parts just in case (NE5532, NJM4580, OP275).

I did huge set of various EMI tests on all these and other opamps about 12 years ago and I have the results archived. Some of the data I might disclose, some I would like to keep.

In general, when talking about opamps and AM demodulation, the output voltage waveform must not approach to the slew rate limit. It is a disaster then. It may be fixed by a proper input RC filter, which would not allow for such dv/dt. For obvious reason, not to limit the BW too much, the slew rate should be at least 20V/us and the input filter must be used with cut off frequency no higher than some 200 kHz.

Another than this, JFET input stages are more resistant than BJT input stages and the higher the input stage idle current, the better. Another disaster are very low power opamps.

I have been curious and measured today my discrete headphone amp for AM demodulation
https://www.audiosciencereview.com/forum/index.php?threads/pm-h2-headphone-amplifier.21014/
Though it has not much higher SR than the 4562, it passed the test without signs of demodulation, probably also because of the input RC and higher input stage idle current and of course the input JFETs.
On the other hand, I have to admit that even the OPA2134 (JFET input) was prone to AM demodulation when the dv/dt approached SR specs.

 

[chris719]

You may start to get demodulation from ESD protection devices at some point, even on JFET input op-amps.

 

[Maki]

This is really neat. I remember a few years ago I built a headphone amp and screwed up grounding on the whole thing. By touching the input jacks and using my body as an antenna I could pick up local radio stations. It was a battery powered nuhybrid in a wooden case so that may explain some of the issues.

 

[pma]

@pma Your laptop battery appears to be very swollen. I would be removing and disposing of it safely before something bad happens (fire).

Just in time, John!! . The best thing is that tis notebook works even without battery!

 

[Speedskater]

Audio Engineering Society Fellow Jim Brown, an EMI/RFI did several old school interference studies, some time ago.
http://www.audiosystemsgroup.com/publish.htm

 

[Skeeter]

At work, we blast things with +33 dBm GSM source in the 850 and 900 MHz bands, and +30 dBm in 1800 and 1900 MHz bands. The 217 Hz TDM really gets into everything and is an easy way to find immunity issues.

I cant believe the amount of trash that came your way in your previous thread on this topic. It's pretty clear that LME497x0 is an outlier. I see almost no reason to use it now when you consider EM susceptibility, popcorn noise, and the availability of OPA1612 and OPA1656.

I think the OPA164x devices are generally superior to the LM4562, but the patter’s PSRR is superb so I use it as a reg. As a GP audio amp stage, it has been shown to have many flaws. Quite why TI haven’t sorted it. Might be they know it’s days are numbered and the OPA devices are better.