Parallel LM3886, input ground, power ground, safety resistor and oscillations.

[markiemrboo]

Hi all,

I have just finished soldering a 3 x lm3886 in parallel amp and begun testing.

I have kept the input / low level grounds separate from the power ground, but have soldered a 10R (and later 0.47R) resistor connecting the input to power ground in case the input ground wire ever became loose. I read this was a good idea (and it sounds like a good idea), as it would stop the output from sticking to one of the rails.

The first time I tested, I left the input ground unconnected from the power ground in order to test out this safety resistor. I had no load and no input connected to the input +ve. I saw oscillation

I then tried shorting the input +ve to the input ground, but leaving the input ground only connected to the power ground by the resistor. The oscillation had disappeared.

It was at this point that I decided maybe 10R was too big, and bodged a 0.47R in parallel with the 10R to test my theory. Indeed, the oscillation seemed to have disappeared even when the input +ve was not shorted. All seemed well, I even put a 4.7R resistor dummy load and it still wasn't oscillating, and had a measured 14mV DC offset. I was very happy!

I gave it a rest for a few hours, as I don't like to mess with this stuff for too long in one sitting.

Not long ago I returned to it to test it with some input signal, still using the 0.47R resistor to connect the input and power ground together. I turned the volume right down on the preamp, played some music and switched on. The 4.7R test resistor (only 2W) made a small fizzling noise and I instantly switched it back off again!

I tried without the test resistor, but the wave seemed to shoot off the scope display really quickly even at 500mV/div, which would be 5v/div with a X10 probe? I disconnected the input music and it was still happening.

I finally managed to start seeing something on the display, which again looked like oscillation, but in a jagged form. Like what you might see on the rails when the caps are charging under load, /|/|/| kinda like that but elongated a bit. Very fuzzy though, not sharp.

I measured the 4.7R dummy load and it seems to have become open circuit instantly with the fizzle. I have now checked the 10R and 0.47R resistor and these too seem to have become open circuit! This seems to explain what I am seeing, to a degree. I guess because the input ground is now no longer connected to the power ground, the output is latching to one of the rails? I think this would also cause a lot of current to flow, so I would see the pattern I am seeing like the caps are recharging constantly. I am not sure about the oscillation, mind. Should this cause oscillation?

I'm also wondering why it is that these resistors have all managed to fail open circuit! There were no bangs or anything. Just a rather quiet fizzle. Physically they all even still look OK.

If anyone's got any suggestions as to whats going on here, what to look out for, things that I am doing which are incredibly stupid (testing with the resistor connecting input and power ground?!) etc I would love to hear from you please!

Sorry for the long post.

Thanks,


EDIT: actually, thinking about it... wouldn't having the input ground disconnected not lead to rail sticking but oscillation because the gain would be at unity, and these chips are unstable below a gain of 10?

I'm still stumped as to why the resistors blew in the first place though!

[BWRX]

First things first, can you post a schematic of what you built?

[markiemrboo]

Originally posted by BWRX
First things first, can you post a schematic of what you built?

Hi BWRX,

Sorry. Absolutely!




Just to clarify what has happened, the 0.47R resistor connecting the input stuff to the power ground has gone open circuit, which I think would make the gain of each chip 1, which would no doubt result in oscillation. I can't figure out why the 0.47R resistor would go open circuit as soon as I connected an input though!

[markiemrboo]

I fear this may have been somewhat of a false alarm everyone.

I've been looking at the schematic and PCB all morning. It seems that I made a very, very stupid mistake. Whilst looking it dawned on me that I had the oscilloscope probe ground connected to the speaker +ve output, and the oscilloscope +ve part connected to ground.

I switched these around, replaced the safety resistor with another 0.47R.......... and I finally saw what I was expecting, an amplified version of the input signal!

What a terrible mistake that was. I hope I haven't damaged anything by doing that

[AndrewT]

Hi,
yes, the safe way to use an oscilloscope is to earth link the probe. Causes all sorts of problems, not least trying to measure low voltages that the probe ground is intent on contaminating.

The 0r47 between NFB/Zin and power ground is leaving the input ground referenced only through the main audio ground.

I would try connecting the the input ground/Zin ground/NFB ground together. Then run a wire from the input RCA ground to the main audio ground. You can try this wire with or without a series resistor in it.

Does C12 have a second series cap in the input feed somewhere?
If your source is DC coupled then the RC high pass filter on the input is set a little low. You have actually set it lower than the NFB filter. The input filter should define the passband of the amplifier, not the internal circuits.
You have no low pass filter in the feeds to the power amps. Fit something. Any LP with RC~0.5uS (470pF) to 1uS (1nF) will do. If this cuts the apparent treble then move it down to 0.3uS (330pF).

[markiemrboo]

Originally posted by AndrewT
Hi,
yes, the safe way to use an oscilloscope is to earth link the probe. Causes all sorts of problems, not least trying to measure low voltages that the probe ground is intent on contaminating.

The 0r47 between NFB/Zin and power ground is leaving the input ground referenced only through the main audio ground.

I would try connecting the the input ground/Zin ground/NFB ground together. Then run a wire from the input RCA ground to the main audio ground. You can try this wire with or without a series resistor in it.

I will connect them together with wire to a single point as I usually do. The idea behind the resistor was supposed to be so that if the input ground wire did ever some how become disconnected, the input would still be connected to a ground through the resistor. I was just testing the behavior It actually seems to work very well indeed now I have everything sorted.

Does C12 have a second series cap in the input feed somewhere?
If your source is DC coupled then the RC high pass filter on the input is set a little low. You have actually set it lower than the NFB filter. The input filter should define the passband of the amplifier, not the internal circuits.

Would it be OK if they were set to the same frequency? A 2.2uF there, which I happen to have, would result in -3dB at 1.5Hz with the NFB filter at 1.6Hz. It's either that or change the NFB caps from 100uF to about 220uF, which would make the NFB frequency 0.7Hz.

Even as it is, it seems to be working OK. I don't have any idea what kind of negative effects having the input frequency lower than the NFB frequency has though. Whoops!

You have no low pass filter in the feeds to the power amps. Fit something. Any LP with RC~0.5uS (470pF) to 1uS (1nF) will do. If this cuts the apparent treble then move it down to 0.3uS (330pF).

I will do this. I have some 470pF's somewhere I think