Why does my gainlcone oscillate?

[gmarsh]

Dear Carlos, I think there are quite many reports which tell us that a LM3875 and LM3886 actually _can_ drive a normal speaker but with that I didn't say every speaker.

Personal discovery - LM3886's *hate* my New Advents. I've tried two, and both oscillate.

Doesn't matter if there's a zobel or not, or an output inductor or not. The feedback path is short - 1/4W resistors directly across the leads. The amp's grounded correctly and there's tons of power supply capacitance. I'm using a non-inverting config.

Maybe my 3886's are bad or something, who knows.

edit: haha, wrong thread... i hit the reply button before the thread split.

 

[Peter Daniel]

What speaker cables are you using?

 

[gmarsh]

What speaker cables are you using?

plain 14 AWG stranded copper, 300V insulation, zip-tied together to form a pair. 10 feet per channel.

Nothing fancy with zobel networks integrated or anything.

 

[peranders]

My guess is that the grounding could be made better. Have you really grounded thing in the right order?. Irsft the signal ground, feedback network, then speaker, then power.

Can you get it stable with short cables? Why don't you test the amp with capacitors how much capacitance it can take. My estimation is that it should be able to handle at least 10 nF, preferable 100 nF.

 

[gmarsh]

My guess is that the grounding could be made better. Have you really grounded thing in the right order?. Irsft the signal ground, feedback network, then speaker, then power.

Same way I ground my solid state amps, and i've never had a problem.

The junction of the two electrolytics in my amp is "power ground". Hanging off this junction is about 1.5 inches of 12AWG solid wire, and I call the end of this wire "signal ground". Feedback and incoming signal grounds go to signal ground, speaker and power grounds go to power ground.

There's not much point to this thread anymore though - my LM3886 amplifier has been disassembled for almost 2 years now, and I've since focused on solid state amp design.

I still blame bad LM3886's.

 

[macboy]

I had a problem with LM1875's oscillating. I fixed it by placing a small (0.22 uF) ceramic cap across the two power supply pins. I also tried a 0.047 uF cap, which also did the trick. You may also try a small cap to bypass the feedback resistor. This reduces gain at very high frequencies and can help with stability. E.g. for a 100 k feedback resistor, use a 10 to 22 pF cap; with 10 k, use 100 to 220 pF.

 

[gmarsh]

I had a problem with LM1875's oscillating. I fixed it by placing a small (0.22 uF) ceramic cap across the two power supply pins. I also tried a 0.047 uF cap, which also did the trick. You may also try a small cap to bypass the feedback resistor. This reduces gain at very high frequencies and can help with stability. E.g. for a 100 k feedback resistor, use a 10 to 22 pF cap; with 10 k, use 100 to 220 pF.

Hmm.... I do this stuff instinctively when I design a solid state design, but don't think about it when I throw together a gaincloneclone using the minimum possible parts count.

I've still got a collection of LM parts, I might dig a few out of the bin and build a GC style amp for kicks. Maybe i'll even design a "proper" compensated feedback network for it.