Have you run the amp with shorted inputs and speaker connected for a test? Just wondering if it could be picked from whatever is connected to the amp. Also does it get worse with your source device powered off and the amp on? I once had a condition where if you turned off the preamp and left the amp on, radio pickup was somewhat worse. Seems a powered off device can act as a better detector.
Many feedback amplifiers become unstable if the capacitance across the input becomes too high, as this moves the pole in the feedback to a lower frequency, creating a 180 degree phase shift in the overall loop response, or positive feedback, before the gain drops below 1. However in the case of the LM3886 it appears designed to accept this capacitance without causing instability hence appears a good idea to include near the input terminals, particularly if RF problems are encountered.
Indeed.
LM1875 especially does not like excessive capacitance as well.
Using typical filter on the + input will murder the phase margin too.
Probably good to check with the 3886. The models are more than accurate enough to check
LM1875 especially does not like excessive capacitance as well.
Using typical filter on the + input will murder the phase margin too.
Probably good to check with the 3886. The models are more than accurate enough to check
I live smack in the middle of Calgary. I have several commercial FM radio and TV broadcasters within 10-15 km of my house. I can 'see' 69 wifi devices (2.4 GHz, 5 GHz) from my listening chair. The FM broadcast is strong enough to be picked up by an oscilloscope probe ground lead. So I'd say there's plenty of RF to go around, but it's not like I have a broadcast tower right next door.
Both fascinating anecdotes. But did any of those systems use an LM3886?
I'm not denying that RF interference is a thing. After all, there's a reason I fit all my products with EMI/RFI filters on their inputs. I'm just saying that I don't find the LM3886 particularly susceptible to RF interference, even in a setup like this:
Or this:
Both were some experiments I played with in the very early stages of the Modulus-86 development.
Tom
... which is why there are ceramic capacitors in parallel with the electrolytic ones in the decoupling network.
Tom
Obviously not but I did encounter RF when I was developing this kit with the LM3886. It was downright annoying. It's only a few components yet I had this RF issue. I have more than 20yrs experience designing discrete amps and never had this RF once.
I did some research and chanced upon an interesting schematic of the Yamaha HS80M Active Speaker.
Apart the 220pF across the +- inputs, they inserted two inductors (L601/602).
My guess is to eliminate RF.
Don't get me wrong, the LM3886 is an excellent chip amp on it's own. You have brought it to a whole new level with your composite design. The DIY community should be grateful for sharing your work.
I'm just high lighting that one can encounter RF which is what the OP is having. Not everyone will encounter RF but some people do. I did.
Regards
Mike
AmpsLab
Interesting. Could it be because you're missing the bulk capacitance on the PCB and that the other decoupling capacitors are pretty far from the LM3886 on your PCB?
Also interesting with the inductors in the Yamaha design. I fail to see how they would keep RF out of the LM3886, though. On the contrary, they would open the feedback loop at RF, which would make any RF issues worse. I suppose the LC formed by those inductors and the capacitor across the inputs of the LM3886 could do something, but if they do, I bet they're intended to tame any buzz or rail sticking that occurs when the LM3886 exits clipping.
If the issue was RF getting into the LM3886, why wouldn't Yamaha have added an inductor in series with the non-inverting input? For example in series with R602?
Tom
Also interesting with the inductors in the Yamaha design. I fail to see how they would keep RF out of the LM3886, though. On the contrary, they would open the feedback loop at RF, which would make any RF issues worse. I suppose the LC formed by those inductors and the capacitor across the inputs of the LM3886 could do something, but if they do, I bet they're intended to tame any buzz or rail sticking that occurs when the LM3886 exits clipping.
If the issue was RF getting into the LM3886, why wouldn't Yamaha have added an inductor in series with the non-inverting input? For example in series with R602?
Tom
I'm sure Yamaha tried adding an inductor in series with R602 and found that it didn't help.
I think RF is picked up inside the chip amp, right at the input.
Because of that, no external filtering will help.
They solved it by tapping the RF from the output and injecting it into the -ve terminal.
That's where the inductors come in.
So basically, they are using the CMRR method to cancel out the RF.
Just an educated guess.
Regards
Mike
AmpsLab
I think RF is picked up inside the chip amp, right at the input.
Because of that, no external filtering will help.
They solved it by tapping the RF from the output and injecting it into the -ve terminal.
That's where the inductors come in.
So basically, they are using the CMRR method to cancel out the RF.
Just an educated guess.
Regards
Mike
AmpsLab
Interesting thought. I'm wondering what you're basing it on aside from that one bad experience that you haven't seem to have gotten quite to the bottom of.
If that was the case, they'd be marked as wire links in the schematic. No company of that size will pay for inductors that don't do anything.
And if you're now sure that Yamaha found the inductors ineffective, why did you mention them in support of your argument. I'm confused.
Look. I'm not denying that you've had an experience with RF pickup in an LM3886-based amp. But I am inviting you to look for alternate explanations instead of simply blaming the chip. If this was a widespread issue, we'd all know about it by now.
Tom
@ tomchr
Interesting thought. I'm wondering what you're basing it on aside from that one bad experience that you haven't seem to have gotten quite to the bottom of.
I did. I just solder a 220pF across the + and - inputs.
If I still get RF, I'll just abandon using the LM3886. It's not a big deal.
@ tomchr
If that was the case, they'd be marked as wire links in the schematic. No company of that size will pay for inductors that don't do anything.
Precisely the point. So, explain to me why the inductors are there?
@ tomchr
But I am inviting you to look for alternate explanations instead of simply blaming the chip. If this was a widespread issue, we'd all know about it by now.
This is not the first time I hear RF complains from the LM3886. Check it out yourself.
Tom
I understand you are making a living off this IC.
You give me the impression that I am trying to jeopardize your business by "blaming" the chip?
Rest assured, I have not intention of doing that.
The OP has a RF issue.
I offered a solution.
Try it. If it works, great.
If not, find another way.
End of story.
I really don't have time to argue with you.
I've got more important matters to attend to like settling my estate.
Mike
AmpsLab
Interesting thought. I'm wondering what you're basing it on aside from that one bad experience that you haven't seem to have gotten quite to the bottom of.
I did. I just solder a 220pF across the + and - inputs.
If I still get RF, I'll just abandon using the LM3886. It's not a big deal.
@ tomchr
If that was the case, they'd be marked as wire links in the schematic. No company of that size will pay for inductors that don't do anything.
Precisely the point. So, explain to me why the inductors are there?
@ tomchr
But I am inviting you to look for alternate explanations instead of simply blaming the chip. If this was a widespread issue, we'd all know about it by now.
This is not the first time I hear RF complains from the LM3886. Check it out yourself.
Tom
I understand you are making a living off this IC.
You give me the impression that I am trying to jeopardize your business by "blaming" the chip?
Rest assured, I have not intention of doing that.
The OP has a RF issue.
I offered a solution.
Try it. If it works, great.
If not, find another way.
End of story.
I really don't have time to argue with you.
I've got more important matters to attend to like settling my estate.
Mike
AmpsLab
Oh man. 30 years ago, a friend of mine bought a Patriot 10-meter ham radio modified to be able to talk on 11 meter (CB) at 100w.
He'd key up and you could hear him on every speaker in the house, and sometimes the TV would change channels.
Also, my dear departed extra-class uncle used to push 1kw of of 80-meter from his house. Before he assembled his antenna tower he acquired a big crate of assorted RFI filters and went door to door for a few blocks around letting people know what was up and promising them that if he interfered with their stuff, he'd fix it.
And if they got ornery about it he would refer them to FCC regulations specifying that their unlicensed equipment is REQUIRED to accept interference from his LICENSED equipment, but since he's a nice guy, he'd fix it for them.
The city of Downersgrove IL has an antenna tower ordinance (probably not strictly legal) because of his tower. But he moved there in part because he knew they didn't have one when he bought the house.
He'd key up and you could hear him on every speaker in the house, and sometimes the TV would change channels.
Also, my dear departed extra-class uncle used to push 1kw of of 80-meter from his house. Before he assembled his antenna tower he acquired a big crate of assorted RFI filters and went door to door for a few blocks around letting people know what was up and promising them that if he interfered with their stuff, he'd fix it.
And if they got ornery about it he would refer them to FCC regulations specifying that their unlicensed equipment is REQUIRED to accept interference from his LICENSED equipment, but since he's a nice guy, he'd fix it for them.
The city of Downersgrove IL has an antenna tower ordinance (probably not strictly legal) because of his tower. But he moved there in part because he knew they didn't have one when he bought the house.
This suggests that you didn't incorporate a 220pF cap across the input, nor if others complaining of RF had that capacitor installed. The inclusion of this capacitor can be considered as a critical inclusion for RF suppression, seeming being intended by the designers of the LM3886 specifically for that purpose. This is not to suggest that by its exclusion sonics can be improved.
Power amplifiers, not unlike the LM3886, can have low bandwidth output stages that require frequency compensation and/or control of the gain in totality of the loop response. Although the LM3886 shows high input impedance (appearing as a pseudo darlington input pair) it also shows degeneration resistors in the emitters. This limits the input stage gain as seeming intentional to support the inclusion of that 220pF capacitor across the input terminals without causing instabilities.
It seems the LM3886 should be considered in the whole of equivalency to contain an embedded 220pF before beginning to assess its value for better or worse.
I never felt personally attacked by you. All I'm saying is that I don't think you have enough data to blame the chip. At least I haven't seen any data from you that would indicate that the trouble is with the LM3886 rather than with its implementation on your PCB. I think you've jumped to a convenient conclusion. This is further exemplified by your mention of the inductors in the Yamaha which were first a clear sign of trouble in the chip (Post 27), then they probably didn't do much (Post 29), and now it's my job to explain what they do and to prove you wrong (Post 31). I don't have time or patience for that either. Similarly, I don't have patience for any argument that suggests that I have something to hide because I make a living selling amps that embody the LM3886. The performance of my products speaks for itself.
Great. Then don't.
All I asked for that you backed your claims up with facts rather than speculation.
Would you care to share a link? A google search for "LM3886 RF susceptibility" doesn't turn up much.
Tom
The main purpose of Cc across the inputs of the LM3886 is to tame the quasi-oscillations that occur when the amp exits clipping, especially with higher rail voltages and a 4 Ω load.
The trouble with Cc is that it distorts the transient response a bit. Rf2 and Cf are added to compensate for this. You can convince yourself of this pretty easily by running a simulation.
Tom
The inclusion of the 220pF can have multiple purposes as intended by the designer, being too large to incorporate into the die of the chip itself. That it can also tames quasi-oscillations isn't unreasonable, being of primary intent or otherwise.
Distortion often increases with reduction in loop bandwidth, hence this can affect transient behavior as you suggest. Yet simulations treat all lines as single nodes, hence this doesn't necessarily predict the manner of RF pickup or rectified behavior in the reality. Including Cc supports less than ideal circuit layouts as seeming a reasonable design criteria by the designer of the LM3886, even though the performance can be hindered to some extent.
Distortion often increases with reduction in loop bandwidth, hence this can affect transient behavior as you suggest. Yet simulations treat all lines as single nodes, hence this doesn't necessarily predict the manner of RF pickup or rectified behavior in the reality. Including Cc supports less than ideal circuit layouts as seeming a reasonable design criteria by the designer of the LM3886, even though the performance can be hindered to some extent.
I have mixed feelings about Cc, to be honest. I'm generally not a fan of capacitors across opamp inputs. That said, AC Test Circuit #2 in the LM3886 data sheet includes Cc, Rf2, and Cf. They're not in the Typical Application schematic (Fig. 1) but for some bizarre reason they show up in the Single Supply schematic (Fig. 3). So I'm not surprised that there's some confusion about their application.
I did at one point measure the clipping response of the LM3886 versus Cc, where Cc ranged from 0 pF to 330 pF. I found that the quasi-oscillations that occur when the amp exits clipping on the positive swing was reduced with Cc values around 180-220 pF. So that's why I started recommending them.
There's little doubt that the clipping response is the achilles heel of the LM3886. If I was to redesign it I would get rid of the quasi-oscillation and rail sticking that the LM3886 currently has. Improving the performance with 4 Ω load so it could run on ±35 V without running out of output current would be nice too. That said, it's an incredibly rugged chip and it's a good candidate for the output stage of a composite amp.
Tom
You could reference the inputs (left of Cin2) to GND either by a resistor per channel or by adding volume control. Change the filter to -3 dB point 100 kHz. A ceramic cap from the inputs RCA GND to casing may also work out OK. Another one to try out is to add tiny input ferrite beads around the input cabling. We can debate this to annoyance but without knowing the frequencies trial and error is a method just costing time while possibly offering a remedy. I always see it as a puzzle to solve with the reward being having solved it.
It is built into a metal casing (preferably connected to PE) I take? It definitely should be in a metal casing especially when being in a RF rich environment. Can be hard to solve! In tough cases RF may wander in via loudspeaker cabling and mains cables. Then ferrite rings and mains filters may be necessary.
RF/EMI has turned out to be audio's best enemy despite denial by many ("can't be heard").
Last edited:
I didn’t use any 220pF capacitors across the LM3886 input in my tests with various EC/Comp configurations. I also didn’t encounter any RF interference or stability issues.