HELP! Hearing radio interference in simplest unity gain buffers

[xst]

Hello everyone,

There seems to be a radio station very close to my apartment (according to google maps). I'm not sure the inference is really coming from there but I do hear (probably AM) radio from some of my audio equipment (one is JOYO American Sound, and the other one is a vintage cassette player if you place a wire on top of the chassis).

I was never able to figure out a solution, but now I seem to pinpoint where the problem is.

I was trying to build a very simple unity gain buffer using the TL072 op-amp on a breadboard. The schematic is as follows



Just a pretty standard voltage follower with some DC blocking filter. However, even this simple circuit is picking up radio signals! No wonder why I can hear radio from other equipment.

The 470pF capacitor and 1k resistor before the input DC blocker was intended to be a low-pass RF filter.



Seem to be a standard practice to filter out RF signal, but it is not working. Even if I bump the capacitor to 100uF, the radio signal is still there and it kills most of the highs from audio signal. Even if I ground the input signal the radio is still there, so I'm pretty sure the source of RF interference is not from input.

Here is everything else I tried:
1. Add capacitors around power sources and virtual ground. not working
2. Use 9V battery as voltage source. not working
3. Use other input voltages. not working, but there are some observations
   a) the output is pure RF interference under 5V, until output audio is gone somewhere below 3V.
   b) increasing from 5V-9V gradually attenuates radio interference, but it's not gone completely.
   c) increasing voltage any higher does not attenuates radio further
4. Add capacitor in feedback path. not working
5. Add capacitor post-op-amp output as RF low-pass. not working
6. Maybe it's the problem of the long breadboard wiring, so I tried the same circuit on stripboard. not working
7. Changing input output cable. not working
8. Shielding with metal case. not working. probably leaking from audio jacks. Actually I tried the same thing with JOYO American Sound (seems to be using TL062), with enhanced shielding using copper tape and aluminum foil and it was not working either.

I also tried to build a similar circuit using the MXR Microamp schematic here, just a non-inverting op-amp, but it's still picking up radio signals just like the buffer. It's very faint, but it's still there.

I suspect it might be the complex transistor stages in TL072 that picks up the RF interference, so I tried something simpler, just a BJT emitter follower using a single 2N5088, roughly the same structure as tube screamer:



But it still picks up radio! The input RF filtering is not working, and I can hear radio even when the input is grounded.

Though I have some new observations, the RF seems to be coming from Vbias. When I disconnect R4 and left input connected, the radio is gone, but without the bias the buffer is not working correctly.

Does anyone has any experience handling radio interference like this, and what else should I try?

There has to be a solution in circuit design since the some other devices I have do not have this radio pickup issue, such as RC-30 looper, Focusrite 2i2, etc. I even opened up the metal case of RC-30 looper so it runs without shielding and there is no radio pickup.

[antonis]

I tried something simpler, just a BJT emitter follower using a single 2N5088, roughly the same structure as tube screamer:



But it still picks up radio! The input RF filtering is not working, and I can hear radio even when the input is grounded.

Hi & Welcome..

I'd move R3 right on Q1 Base (between bias resistor and Base) and, of course, C3 wired across B-E junction.. 

[GibsonGM]

Would it help to also place a smaller cap in parallel with C2 at Vbias? 470p, or 1n?

How about a small value cap across the input jack, from tip to shield, of 100p etc?

Next you may have to try getting some ferrite beads   

[Rob Strand]

Try raising the 1k input resistor to 10k to see if it helps at all.

You should also have 470 ohm to 1k in series with the opamp output, to make sure it doesn't oscillate with capacitive loads like the output cable.

[anotherjim]

^ +1
But when you say you added extra power capacitors, did you try adding any ceramic discs? 100nF is my favourite. These disc caps work better for RF due to lower inductance but you still need the electro caps to handle low-frequency noise.

[PRR]

If you are in sight of a high power AM transmitter, you can get reception even on passive devices. In my youth we had only passive telephones, yet the phone company had to add suppressors on the line and on each earpiece; and almost every year (I dunno why they wore out). I could get the ball game with a finger, a foot of wire, and an earphone.

[xst]

Would it help to also place a smaller cap in parallel with C2 at Vbias? 470p, or 1n?

How about a small value cap across the input jack, from tip to shield, of 100p etc?

Next you may have to try getting some ferrite beads 

Holy cow! I think I have some breakthrough discovery. The RF noise doesn't seem to come from either from Vbias or the input, but the TL072 op-amp itself!

I don't have any ferrite beads in hand, so I tried using inductors instead. Turns out placing a 1mH inductor RIGHT IN FRONT OF the non-inverting input of op-amp kills the radio interference, but the inductor is generating a very high noise floor. I tried different values of inductors. Lower values seem to reduce the noise floor a bit, but the minimum is 330uF to filter out radio signal:



Here is before and after comparison. Notice that I am recording with highest input gain from my audio interface

https://on.soundcloud.com/qPPpZ

Notice something interesting that even if I disconnect the input to op-amp you can still hear RF noise, but adding the inductor kills it.

I tried to put the inductor elsewhere to see if I can fix the noise floor problem. What I observe is that
1. inductor right after input or between C4 and R4 doesn't filter out RF noise, so the RF noise is not coming from input
2. inductor after power supply, between V+ and V- power rail of op amp to power supply, or between Vbias and R4 doesn't filter out RF noise, so RF noise is not from supply.
3. inductor after op-amp output does not fix the problem, so RF pickup is not after op-amp
4. any inductor values lower than 330uF gives both noise floor and radio noise

Basically, the conclusion is that the RF pickup is from op-amp itself, and the inductor does work. I discovered that there is another arrangement that kills the RF noise: put the 330uF inductor in the feedback path of op-amp, but it still has the same noise floor problem.



Maybe I should get some ferrite beads to see if they can improve the noise floor.

I noticed that ferrite beads are used at the input and output of big muff tone wicker, which I happen to got one today, and I can tell you it's not working. The RF pickup problem is definitely not from input. Extra ferrite beads need to be placed somewhere else in the circuit.

I'll have to figure out a same placement strategy for BJT.

[xst]

Holy cow! I think I have some breakthrough discovery. The RF noise doesn't seem to come from either from Vbias or the input, but the TL072 op-amp itself!

Actually, I was reading the datasheet of TL072 more carefully. I noticed there is a new version called TL072H with enhanced RF filter. Maybe using that will also fix the problem? Too bad it only has SMD.

The problem is definitely not specific to Tl072, I tried LM358 with the same circuit and it still have the same problem.

There seems to be a special variant of op-amp with hardened EMI rejection, but the one used in video also only available in SMD

https://www.youtube.com/watch?v=CQZUXRbpAxU

Has anyone used one of these that has a through hole variant?

[Rob Strand]

op-amp with hardened EMI rejection,

No amount of hardening can protect it from some users.

[xst]

Another breakthrough discovery!

I was reading this guide on EMI rejection from STMicroelectronics in op-amp design and found something that works!

https://www.st.com/resource/en/application_note/an5798-how-to-improve-the-robustness-of-operational-amplifiers-against-electromagnetic-interferences-emi-stmicroelectronics.pdf

Instead of using inductor, adding a capacitor between the non-inverting and inverting input of the op-amp kills the radio, and it also doesn't increase the noise floor! 100pF is the minimum for me to filter out radio entirely. Anything below that will give some residual. This seem to be the best solution in terms of op-amp buffer:



Here is an audio comparison:

https://on.soundcloud.com/hWNKE

I'm not sure why it works, it seems to form a lowpass filter according to AC analysis from SPICE. The C3 capacitor at the input doesn't make a difference. Only C1 is needed and necessary to filter out the radio. However, I notice that C1 introduces a resonance peak in RF range if used without C3 in my SPICE simulation



Adding back C3 tames the resonance peak



Since C3 does not contribute much to RF filtering, I tuned its value to 100pF. Now there is almost no loss in audible frequency range.

I also found similar designs in the schematics of Yahama HS5/7/8 speakers, which happens to be one of the audio gears that don't have the RF pickup issue.



It also uses ferrite beads (the inductors) at the input, but I think it's the capacitor that makes the difference.

Seriously, every pedal manufactures using op-amp should add this capacitor! It makes a HUGE difference in RF rejection.

Now need to figure out the BJT buffer design.

[bluebunny]

Your schematic only shows a single opamp.  There are two in the TL072.  If the other one is unconnected, then it may be oscillating.  You'll hear this.  And a naked circuit on a breadboard is a fancy antenna and will pick up all sorts of crud (not just radio stations).  It will be quieter when it gets into a metal enclosure.

[FSFX]

I was reading this guide on EMI rejection from STMicroelectronics in op-amp design and found something that works! ...
...... adding a capacitor between the non-inverting and inverting input.

The reason it works is that by putting that capacitor between the non-inverting and inverting input you provide a low impedance path to both inputs for any RF signal picked up on the input wiring. This RF signal is then a common-mode signal rather than a differential signal and so is not amplified by the op amp. 

[xst]

The noise on the BJT buffer seems to be more complicated to fix.

The best I can get is to put a 100pF capacitor between the BE junction.



This fix is used in the Germanium FF and mentioned above by antonis, but there is still some residual radio noise, and strangely ONLY 100pF works, lower values and higher capacitor values can't fix the noise problem. Using higher capacitor values not only reduces the highs of audio but also let radio interference creep back in.

The frequency response is also very weird. Higher values of capacitor acts as a shelving filter some how. It's not as good as the high order filter formed by the capacitor connecting non-inverting and inverting input of op-amp.



Some other observations:
1. move input resistor R3 right before base of BJT doesn't fix the problem. Increasing its value attenuates the frequency response of RF in SPICE, but in reality it does nothing for me
2. shunt a 100pF capacitor between R5 to ground also reduces the noise, but its effect is less than the BE junction capacitor.
3. adding an inductor in front of base also fixes the problem, but again it has very high noise floor.

There are some related videos on youtube on this topic found from this post:

At least the solution in the first video doesn't help me. The only thing left to check is ferrite beads. I need to order some of them to see if they can help with the noise floor using inductors.

[FSFX]

The noise on the BJT buffer seems to be more complicated to fix.

The best I can get is to put a 100pF capacitor between the BE junction. .... but there is still some residual radio noise, and strangely ONLY 100pF works, lower values and higher capacitor values can't fix the noise problem.

It may be that the 100p is resonating with some inductance in your circuit wiring to provide a tuned circuit around the frequency of the offending radio station. Using ferrite beads on their own is unlikely to eliminate the problem completely as they provide a very small inductance that may not be sufficient to filter out the offending station unless they are combined with the correct value of capacitance to act as an 'RF trap'. I am sure any radio amateur (ham radio) members here will be well familiiar with all of this.

Here are some measurements of the effective inductance of a few ferrite beads with various turns of the wire threaded through them . As you can see, the inductance is in the microhenry range and so doesn't provide much reactance except at very high frequencies. Adding a capacitor in parallel to form a tuned circuit at the desired frequency and placing that in series with the input close to the BJT base may help. Similarly, a series resonant circuit from the input connection to ground would also help.     

[antonis]

How an innocent Emitter follower (a) can be turned into a wicked oscillator (b)..



Usually, a "base stopper" resistor (100Ω to 2k7) and a series output resistor of 100Ω prevent oscillations while a cap (10nF or so) across Base-Emitter junction prevents RF rectification..

[Rob Strand]

Input and output resistors can help suppress the oscillation.
Output resistors can help RF and EMC as well on both opamp and BJT circuits.
Noise/RF can get in via the output lead.  (For the opamp case the input is connected to the output!)

This thread looks at the BE cap on the BJT buffer, with appropriate safe-guards for preventing oscillation,
https://www.diystompboxes.com/smfforum/index.php?topic=124166.msg1224204#msg1224204

This is very quick demonstration of buffer oscillation and the counter measures,
https://www.diystompboxes.com/smfforum/index.php?topic=119501.msg1118183#msg1118183

The input wiring inductance is a factor.  You don't need much, so oscillation is a very real problem.

[amz-fx]

But it still picks up radio! The input RF filtering is not working, and I can hear radio even when the input is grounded.

The V+ power rail is also an input and needs to be decoupled with an RC filter. I suspect that is how the RF is getting into your simple BJT buffer.

best regards, Jack

ps: I apologize if someone has already mentioned this.

[xst]

This is the design that finally fixes the problem for the BJT buffer. Somehow it needs a ferrite bead at the emitter.



Putting the ferrite bead at the base, collector, or power supply are not as effective. The ferrite bead also needs to have quite a large impedance at 100MHz. The one I use is Fair-Rite 2944666631, which is listed as 690 Ohm @100MHz

https://www.mouser.ca/ProductDetail/623-2944666631

This ferrite bead is quite large in size though, so I kind of prefer the 100pF capacitor at BE junction solution despite it leaving some residual noise though. Smaller 60 Ohm @100MHz ferrites don't work that well.

[Rob Strand]

This is the design that finally fixes the problem for the BJT buffer. Somehow it needs a ferrite bead at the emitter.

A resistor (470R to 2k2 or so) in series with the output should achieve a similar effect.  You should be able to move the bead to the output lead as well.  (You could put the resistor in the emitter but it causes unnecessary signal attenuation to the audio.  There's no advantage to it there.)

[xst]

This is the design that finally fixes the problem for the BJT buffer. Somehow it needs a ferrite bead at the emitter.

A resistor (470R to 2k2 or so) in series with the output should achieve a similar effect.  You should be able to move the bead to the output lead as well.  (You could put the resistor in the emitter but it causes unnecessary signal attenuation to the audio.  There's no advantage to it there.)

I just realized this indeed is another solution. The ferrite bead just make the resistor only apply to RF frequency. I tried to use an output resistor before, but the value was too small. It needs to be around 1k for me to attenuate the radio signal enough to be barely audible. Adding back the BE junction capacitor kills the radio completely. This one should be the simplest RF-hardened BJT buffer design.