Troubleshooting moving_electron's inverted LM3886 amp

[moving_electron]

I put together one channel using Digi01's regulated 3886 boards. When I fired it up I had very large AC at the output .5-1A. I got these large values no matter if I had no input or if I had a load or no load. Zobel or no Zobel.

The heatsink got very hot etc. The LM3886 probably shut down.

I checked the board carefully and could find no bad solder joints, solder bridges etc. I substituting in another Digi01 3886 board and had the same problem although with slightly different AC values.

My meter has a frequency measurement capability and it showed 100kHz-104kHz at the output.

I finally put a 22pF cap in parallel with the 220k feedback resistor and the problem went away. It is what I had on hand, looks like a ceramic disk. Probably would be good to change it to a monolithic/multilayer (better at very high frequency?). Or perhaps not since things are working fine now. AC at output is minimal and DC offset is now set to near 0vDC.

I live near numerous transmission towers and typically have to put 220pF caps across 10k feedback resistors on op amps or I have similar problems.

The amp channel is now playing through a cheap speaker. Sounds like clean sound through a cheap speaker.

I will probably try it through my main speakers later today.

[moving_electron]

Originally posted by moving_electron
I put together one channel using Digi01's regulated 3886 boards. When I fired it up I had very large AC at the output .5-1A. I got these large values no matter if I had no input or if I had a load or no load. Zobel or no Zobel.

Correction: I meant to say .5-1V AC at the output.

[digi01]

Originally posted by moving_electron


Correction: I meant to say .5-1V AC at the output.

have you set up a trim pot on the board?

[moving_electron]

Originally posted by digi01


have you set up a trim pot on the board?

Yes, I have a trimpot on the board. I started with it set to 50k. After I got the oscillation to go away I was able to trim the DC offset to about 0V.

I am kind of surprised to get such a severe oscillation when I normally don't see people report it as an issue. But having multiple radio towers just up the hill certainly may give me much stronger RF.

The 22pF across the 220k feedback resistor eliminated the oscillation. I expect to build the IGC with this always.

[Upupa Epops]

Try lower value - with 22 pF is probably too narrow frequency characteristic.

[peranders]

The LM3886 won't oscillate if you take care of the supply voltage and decoupling _and_ make the grounding right. Remember this is a high current device.

Can you show us exactly how you have done, both schematic and a picture of the pcb?

[moving_electron]

Originally posted by Upupa Epops
Try lower value - with 22 pF is probably too narrow frequency characteristic.

I don't have anything lower on hand it appears. What lower capacitance value are you suggesting to try?

Jim

[moving_electron]

Originally posted by peranders
The LM3886 won't oscillate if you take care of the supply voltage and decoupling _and_ make the grounding right. Remember this is a high current device.

Can you show us exactly how you have done, both schematic and a picture of the pcb?

I should probably take the built PCBs and draw the schematic from that. I will likely not get that done for a few days.

In the mean time let me describe a bit.

I was sort of throwing a lot at it so it has tsome he CarlosFM part values for the RIGC as well as the unregulated "snubber" adds.

On the power supply for the first channel I did not have the 2.2k resistor on hand yet so the power supply looks like this:
12,000 uF
100nF 200v poly
1 Ohm + 100nf monolythic ceramic
Regulator parts targeting 30.5v ( I need to check what I put in)
After reg on PS PCB :
two 100nF mono ceramics in series for 50nf
That finishes out the PS board description.

I measured at V+ and V- on the 3886 board and got no AC to speak of (0-.1mv?) and no frequency measurement showed on the meter. So I was thinking the power loooked pretty good. I do not have a scope.

On the 3886 PCB there is a 47nF bypassed with a 100nf mono ceramic in parallel.

The input capacitor was a 3.3uF polypropelyne. It is large so it is off board. Probably 4cm from the cap to the board and about about 8cm from cap to input jack. None shielded yet. However, the oscillation occured with or without the cap hooked up.

Grounds are temporarily hooked up by but all soldered together.

The two grounds from the PSU boards are connected to the earth ground first. Then the ground from the 3886 board. Then the signal ground is attached to that.

I was not using the .1ohm output resistor. I had that jumpered instead. I have an on board 2.7ohm+100nf zobel drom output to ground. The oscillation occured both with a load hooked up (4 ohm wirewound) and when no load was hooked up. tried a .5 ohm in palce of the .1ohm spot but no effect on the osciallation.

[moving_electron]

Originally posted by Upupa Epops
Try lower value - with 22 pF is probably too narrow frequency characteristic.

To me the highs (cymbals etc.) were a bit reserved compared to the BrianGT amp. We tend to like the highs fairly bright. Changing the input cap to Black Gate 4.7uF instead of the 3.3uf polypropylene 630v made the highs more prominent and crisp. They have only run for 1/2 hr though.

The Blackgate low value non polar caps as input caps always seem fairly bright to me and quite detailed. Still it is possible that it is rolled off in the very upper range. Nothing I would put my finger on though.

I will try the lower value in parallel wih the 220k resistor to see what happens. I have to get some though.

Jim

[moving_electron]

Putting 100pF from Vin+ to Vin- allowed me to remove the 22pF across the feedback resistor. Would 220pF instead have much effect or improve the safety margin.

I thought I had tried that but I had actually put it between Vin+ and ground. Perhaps not good to work on it late at night. :-)

All seem well now. Since it was late at night when I was trying things I was only able to play it fairly softly. Able to turn it up now. Sounds very nice so far.

By the way, I am not using a pot so I have a 100k resistor from input to ground. If that was not there I understand that would have been a source of instability.

What is the correct way to calculate the effect of the 22pF cap across the 220k feedback resistor? Searching through some other threads it looks like the feedback was being filtered low pass at ~40kHz?