High frequency noise with DIY LM3886 build - Help

[jan.didden]

The fact that the noise only happens when both are connected points to a ground loop.

In the schematic, I can't see a connection between SIGGND and GND nodes. How are they actually connected to each other?

Edit: Oh I see it now.

That is the wrong way to hook up an input and ground, You want the inputs to go directly to the PCB, not route the SGND through the power supply board.

Bingo! (Bold mine). Grounding 101.

Jan

 

[Njstroud]

Final update - Problem solved!

First I want to thank everyone who replied on this thread- it is my first time on this forum and you have all provided great guidance in a supportive way.

There were multiple problems-

1. Grounding: I did have a ground loop breaker, and the speaker inputs, as well as the signal inputs, were isolated. However, per the instructions I was following, I had all grounds go to the power supply PCB. Before I connected the signal ground to the lm3886 PCB, I was getting some odd behavior. I have since tied the signal ground and speaker ground to the ground of the lm3886 PCB. From there the LM3886 PCB ground gets tied to the other LM3886 as well as to my ground loop breaker.
2. Decoupling capacitors: Once I fixed the grounds, I was getting very reliable performance. Almost satisfactory. Yet I still had that constant 6-13kHz whisper. I shorted the input signal with various resistors, changed audio sources, used different power supplies (expensive benchtop power supply versus my power supply PCB), and I moved to different areas of the house while randomly shutting off fluorescent lights/ceiling fans/etc much to my wife's confusion. I could change the frequency and amplitude slightly by plugging and unplugging the speaker wires, but the whine did not relent. The only thing left to do was remove the 4.7uF and 22uF capacitors and solder them directly to the pins of the lm3886 as was suggested. I did that to just one of the boards, and it has solved the problem. Both boards are connected and running.

I'm going to go ahead and re-design the amplifier boards to fix both decoupling capacitors and grounding issues. Also the interconnects, they seemed like such a good idea at the time.

Thank you all so much!

 

[pitbul]

I would suggest adding 0.1uF (100nF) parallel to mentioned 22uF and 4.7uF caps as close as possible to the pins of the LM3886.
Also would like to recommend twisted pair from UTP cable for feeding the input of amp, not so thick (twisted) wires.

Enjoy in your sound...

 

[Vovk Z]

Attached is a picture of the schematic and PCB.

By the way, C12 have to be film type (PP or X2 PP will be better then PET) but not electrolyte.

 

[Vovk Z]

I would suggest adding 0.1uF (100nF) parallel to mentioned 22uF and 4.7uF caps as close as possible to the pins of the LM3886.

There isn't strong need in such amount of different decoupling caps. You may need only some small value near the IC pins (0.1-1.0 uF) and some large ones (1000 uf) a bit further.
If TS wants to leave them then 4.7 uF ones have to be placed very close to IC pins.

 

[nigelwright7557]

The critical components are the feedback components.
They must be as close to chip amp as possible.
For each 1mm of track you get 1nH of inductance causing phase shift.

 

[Mark Whitney]

Just use/copy An open source layout for LM3886?.

 

[tomchr]

First I want to thank everyone who replied on this thread- it is my first time on this forum and you have all provided great guidance in a supportive way.

You're most welcome.

Decoupling capacitors: Once I fixed the grounds, I was getting very reliable performance. Almost satisfactory. Yet I still had that constant 6-13kHz whisper. [...] The only thing left to do was remove the 4.7uF and 22uF capacitors and solder them directly to the pins of the lm3886 as was suggested. I did that to just one of the boards, and it has solved the problem.

Interesting with the whisper. That's a good observation. Good to hear that physics still works... 😉

I'm going to go ahead and re-design the amplifier boards to fix both decoupling capacitors and grounding issues.

I would suggest fixing your grounding scheme as part of that. If you handle the grounding on the PCB, all you need to do is to run one wire from it to the power supply. But whatever works. There are multiple right solutions here.

There isn't strong need in such amount of different decoupling caps. You may need only some small value near the IC pins (0.1-1.0 uF) and some large ones (1000 uf) a bit further.

OP's experience suggests otherwise. As does mine. Also, National Semiconductor (now TI) recommends a minimum of 470 uF where the power enters the PCB and 10 uF tantalum || 100 nF ceramic capacitor right at the chip. You can read their recommendations in the Application Section of the LM3886 data sheet.
But recall that recommendation was made in the mid 1990s when the LM3886 was made and we didn't have large X5R/X7R dielectric caps or low-ESR electrolytic caps (not many anyway). I had those things in mind when I wrote my article on LM3886 decoupling.

The 10 uF decoupling cap is definitely needed. Without it (or even if you go much below 10 uF) the LM3886 will oscillate at higher output currents (think ±32 or so volt supply rails and 4 Ω load). Ask me how I know... 😉

Tom

 

[Vovk Z]

I have observed two or three different boards with only 0.1 uF ceramic and 470-100 uF caps, and I haven't seen any oscillation. Possibly I was a bit luckier.