I am at a loss as to figuring out how to track down this short.

[Hextejas]

I have built 2 of these amplifiers(Right and Left) and R works and L doesn't.
For the faulty one, I have built it twice so the next time will make 3. I would like some guidance to try and figure out what I am doing wrong.
I built it (I thought) according to the schematic but both times when I put it in the chassis and power it up, I get a .22 shot sound, smoke, and in one instance, a fire. Both times it seemed to be on resistor 4, 20K.
The resistor looks like it has 2 chips taken out of it but it still tests 20K.
Resistor 4 is soldered to the back side of the PCB. The PCB is mounted on 1/2 inch plastic risers.
I completely desoldered board L1 and tested all the parts and they all tested ok, including resistor 4. The only thing that I couldn't or didn't know how to test was the LM3886 IC. So, I bought some new ones and put it in.
So, L2 was with a new LM3886 and the salvaged parts from board L1. Before I put it into the chassis, I hooked it up to my bench power supply and ran it up to +30DC, and -30DC and it seemed to work ok.
I put L2 into the chassis along with the R1. applied power and L2 .22 shot sound, and smoked again.
I took L2 out and have been trying to test it outside of the chassis and it seems to work ok.
However when I put it back into the chassis, at least I dont get any shots nor smoke but I think that it is not performing correctly.
I took some voltage readings and tried to compare both boards.
Something strange that I noticed was this. V- was measuring -5 Vdc with no connection to the power supply. With it normally connected, it should read -26Vdc

I dont know where to start. I was going to build another amplifier board using the scavenged parts from L2 but I am reluctant to plug it into the chassis and apply power to it.
I think that I will build it and swap it with the working amplifier, R1. If R1 shorts then that will tell me something,,,,, that I will have to think about.
Could it be the wiring into the chassis or the power supply ?
I dunno

[Andy Watson]

. Before I put it into the chassis, ...

Have you checked that the tab of the LM3886 is isolated from the chassis? The datasheet suggests that the tab is connected to the "V-" pin.

[Hextejas]

. Before I put it into the chassis, ...

Have you checked that the tab of the LM3886 is isolated from the chassis? The datasheet suggests that the tab is connected to the "V-" pin.

Why no I haven't ! I actually have not been paying any attention to it. It is connected to the heat sink with some compound in between and I am pretty sure that the heat sink may be connected to the chassis.
Do I need to pay attention to that ?

[james_s]

It depends on the design, but usually yes. The tab will almost always be connected to some element of the semiconductor die.

[cdev]

Here's what I do, I have some LCD containing material that changes color depending on temperature. So, this is a cheap form of thermography. For example, one changes color over the range 20 to 25 degrees C. Thats usually the most appropriate one. I would connect up the board to a current limited supply at a fraction of the voltage and current it normally draws and I would use something like my temperature sensitive sheeting to see where it gets warmest. I'd start there. That lets you see tiny variations in temperature visually. You can get this material at science museums, etc. Its not at all expensive.

Since you also have a working unit, I would start looking wherever they differ. The working unit at the same (lower) voltage + current wont work either but any major difference in the two boards warming pattern should lead you right to the problem. Remember to start at a very low voltage, and limited current, just high enough to overcome any voltage drops and get a small current flowing, enough to warm something up just a degree or two, not to damage anything.

[donmr]

Clean up the extra solder and wire ends, wash off the flux, then look at it closely.  There are a couple of places that look like they might be shorts but you can't tell from the pictures and with all of that flux.

[SiliconWizard]

. Before I put it into the chassis, ...

Have you checked that the tab of the LM3886 is isolated from the chassis? The datasheet suggests that the tab is connected to the "V-" pin.

Why no I haven't ! I actually have not been paying any attention to it. It is connected to the heat sink with some compound in between and I am pretty sure that the heat sink may be connected to the chassis.
Do I need to pay attention to that ?

Well, as Andy Watson said, the LM3886's tab is internally connected to V-. If it's not isolated from the heatsink, the heatsink is connected to the chassis and the chassis to ground (which is likely), there you have your short between V- and ground.

See page 2 of the datasheet.

[Hextejas]

. Before I put it into the chassis, ...

Have you checked that the tab of the LM3886 is isolated from the chassis? The datasheet suggests that the tab is connected to the "V-" pin.

Why no I haven't ! I actually have not been paying any attention to it. It is connected to the heat sink with some compound in between and I am pretty sure that the heat sink may be connected to the chassis.
Do I need to pay attention to that ?

Well, as Andy Watson said, the LM3886's tab is internally connected to V-. If it's not isolated from the heatsink, the heatsink is connected to the chassis and the chassis to ground (which is likely), there you have your short between V- and ground.

See page 2 of the datasheet.

Thank you Silicon and Andy. I wondered about that as I was installing the heat sink. It seemed to me that the heat sink and consequently the tab of the LM3886 was connected to the case.
I took a long look at the attached picture and it seems to show the heat sink attached to the chassis.
Would you please take a look at the photo and see what you think.
I am about to make version 3 of the board so I would like to not smoke it up again.

[TrickyNekro]

Wwwwoooww wwowow wow woow....

V- together with ground together with torroidal ground, it´s a bit how ya doing :-P. Even the single screw that mounts the torroidal can act as an antenna under load.

Having a Star Ground connection is not a bad idea also, but how you implement it... The output wires should be as close to the amp as possible.
It doesn´t matter if it´s ground or anything else as your out-. Star ground is meant for powering stuff, but no drive wires.

You subject like that your whole circuit to both rail fluctuations and output fluctuations. Double the noise for nothing. Plus their phase beat noise.
Output wires should always come out of the power stage as close to the power stage as possible! Still try to isolate the tabs from the heatsinks, maybe with some mica pads and test again.

Edit: And you also do that for the "input" ground no... star ground is meant to be around your main IC or your main rail points for copper pours, like make a star path for power stage and another for pre-amps and buffers whatever that share the same plate, and even there you got to be careful where on the plate you connect it. By plate I mean ground pour btw.

[cdev]

You likely need a substantial number of holes for ventilation under and over the heat sinks, and feet on the bottom to allow air to get into the bottom ones. If I missed them just ignore my post.

. Before I put it into the chassis, ...

Have you checked that the tab of the LM3886 is isolated from the chassis? The datasheet suggests that the tab is connected to the "V-" pin.

Why no I haven't ! I actually have not been paying any attention to it. It is connected to the heat sink with some compound in between and I am pretty sure that the heat sink may be connected to the chassis.
Do I need to pay attention to that ?

Well, as Andy Watson said, the LM3886's tab is internally connected to V-. If it's not isolated from the heatsink, the heatsink is connected to the chassis and the chassis to ground (which is likely), there you have your short between V- and ground.

See page 2 of the datasheet.

Thank you Silicon and Andy. I wondered about that as I was installing the heat sink. It seemed to me that the heat sink and consequently the tab of the LM3886 was connected to the case.
I took a long look at the attached picture and it seems to show the heat sink attached to the chassis.
Would you please take a look at the photo and see what you think.
I am about to make version 3 of the board so I would like to not smoke it up again.

[Hextejas]

Re the LM3886 to heat sink to chassis connection, is that correct or not ?
The picture seems to say yes, yet I am thinking that you folks don't think so.
I am totally lost.

[Doctorandus_P]

You've bolted the heat sinks with angle irons to the Chassis.

I would replace the angle irons with some plastic blocks, Acrylly, Nylon, or  3D print something.
Also make sure the sides of the heatsinks can not contact the (side off the) chassis.

It may not be absolutely nessacarry to do this for correct operation, but it lessens the probability that horrible things happen when you are poking around with measurement equipment.
Alo: A label on the heatsink to remind you that is has a live voltage can be usefull.

[kripton2035]

you have the chance to have one working unit: use it !
use a multimeter in diode mode, put the red plus probe on the ground of one board, and note the value in different parts of the board
then do the same on the bad board, and you will have a better idea of where the problem is, where the values read are different.

[David_AVD]

Why not just install a mica or silicone insulator pad (with insulating bush) and be done with it?  Leaving the heatsink isolated from chassis but "live" is asking for trouble.

[Hextejas]

Why not just install a mica or silicone insulator pad with insulating bush) and be done with it?  Leaving the heatsink isolated from chassis but "live" is asking for trouble.

I will do that David and thanks for the suggestion. This electronic stuff is very new to me. Could you or anybody  tell me how one amplifier board works fine yet the other one produces a short ? They both have the heatsink connected to the chassis.

[james_s]

Maybe there's a something insulating one of them? Anodized heatsink with thermal paste on the part might provide just enough insulation to prevent a short, I certainly wouldn't rely on it though. It's also possible that you do have a solder bridge or defective component on the shorted board.

[cdev]

Whenever I have one working unit and one non working, a problem is usually findable quickly by comparing the two electrically. Even if one doesnt know electronics beyond the basics, simply compare the two.

First with power removed and capacitors discharged, check for shorts and opens.

[ciccio]

LM3886T's tab is connectedt to V-, and if it is conected to ground via a grounded heatsink the IC will blow, together with other components.
You need an insulator, but it is difficult to source  because it is larger than  standard TO220 or plastic TO3 insulators.
There is a simple solution:
LM3886TF has an insulated tab, and it will work in your circuit.
I've used many hundreds of them, but now I do not find them on TI site anymore, even if Mouser has them still available.

Best regards

[Hextejas]

LM3886T's tab is connectedt to V-, and if it is conected to ground via a grounded heatsink the IC will blow, together with other components.
You need an insulator, but it is difficult to source  because it is larger than  standard TO220 or plastic TO3 insulators.
There is a simple soluti3886TF has an insulated tab, and it will work in your circuit.
I've used many hundreds of them, but now I do not find them on TI site anymore, even if Mouser has them still available.

Best regards

I should shoot myself at all the confusion, $$$$$, and troubles I have been causing myself. The original parts list called out the TF version but I, in my ignorance, was not careful when ordering. A oainful lesson.
Sheesh !!

[james_s]

Well it's not that big of a deal, you can get thermal pad material that can be cut to size.

[David_AVD]

The correct size mica / silicon pads with the hole are not exactly rare, but a newbie may be better off swapping the output devices with the TF version to avoid dealing with that and the insulating bushes. 

[cdev]

What is the best two sided thermal interface pad material for heat sinks? (insulator or non-insulator)

I need something that stays attached and doesn't (let say, a heat sink placed on top of a chip) migrate downward (or fall off) when it gets hot!

[SiliconWizard]

The insulator pad is necessary, but not sufficient. Beware! You also have to insulate the screw from the IC. It's easy to forget about this. The screw may/will touch the thermal pad of the IC if you don't add an insulator. Guaranteed! (It will never lie exactly centered and never touching the edges of the hole... real life is not disney , and a simple nylon washer is thus not enough).

So add a "nylon screw insulator".

[james_s]

The term I'm familiar with is "shoulder washer".