Load Ground ?

[petemoore]

  Perhaps someone can straighten me out on the 3886amp Grounding.
  http://www.generalguitargadgets.com/pdf/LM3886.pdf
  Instead of star grounding most of it, understand which is the:
  Load Ground
  Output compensation ground
  And the Input / Feedback Ground.
  ...Thing is I count 7 ground points,
  Input pot ground
  C3 feedback loop tail' ground
   the four capacitors used to bypass power supply _/Gnd/+.
  Pin 8 has a cap to ground
  ...the data sheet lists those three^ and doesn't even use plurals, 3+0=3, 7-3=4...I count 4 capacitor grounds from the power supply.
  I'll be going off GGG's PCB layout to try sorting it out, that might help me with definition of terms, putting the puzzle together.

[R.G.]

This all gets less confusing if you think of where what currents are going.

There are three kinds of ground: reference ground, a high quality source of 0V; sewer ground, which returns the "used" electricity to the power supply, just to complete the loop, and safety ground.

Safety ground is outside this discussion.

Sewer ground is where you dump the electricity that has already been used. It is a path for that electricity to flow back to the power supply for recycling. What is important about sewer ground is that it carries the output signal in current form, perhaps mixed with whatever else is dumped into that sewer. Since no wire is perfect, there is a resistance in the wire; that makes a sewer signal. If you connect something that should be connected to a high quality reference ground to someplace on the sewer, you will pick up the sewer current signal and mix that in with your signal. This is bad. You can cause oscillations, as well as it just sounding bad.

Let's go through the grounds you quoted:

Load Ground

Load ground is the recycling sewer from the speaker. It comes off that 8 ohm load, and is the highest current sewer ground in the circuit. This ground MUST run on its own wire directly back to the power supply center tap. Otherwise it generates junk that gets mixe into whatever you connect it to.

Output compensation ground

There is no output compensation network on the National Semi datasheet circuit. If there were, it would need either its own ground or to be connected into the speaker return sewer.

And the Input / Feedback Ground.

These are the reference grounds.

Input pot ground

The input signal is impressed across this pot. That has to be a high quality reference ground. This one is the most critical in the circuit.

C3 feedback loop tail' ground

Another reference ground.

  the four capacitors used to bypass power supply _/Gnd/+.

These are sewer grounds. Any junk on the power supply is shunted back to the main centertap ground. Interestingly, the upper half cycle of the power signal is pulled out of the positive side one, the lower half out of the negative side one. Ideally these would be connected separately back to the centertap on their own sewer. The ground for the two can be connected together since only one polarity is working at a time. Keep these away from reference ground.

Pin 8 has a cap to ground

This one is not critical. Reading the data sheet, a DC current is pulled from the pin. DC might put an offset voltage on the reference ground wires, so in a critical application for DC accuracy, you would not want to do that. But here, it only carries 500uA and it's relatively constant. It may not need its own ground wire.

For grounding think of
- what currents flow into that "ground" symbol?
- will these upset high quality reference ground if the flow in the same wire?

[petemoore]

  Not to go OT but,
  I guess it's just the wattage blowing my little test speakers..
  I get clicks, but then the speaker's dead, didn't really have time to source input, touch the wire..
  I wouldn't think idle [no input] would blow a [guess] 2w speaker, I must be doing something else wrong.
  I'll go look at the data sheet to sort what kind of output coupling I have wired up.
  When looking at the schematic of the GGG project and the data sheet typical application diagram, I get the idea I don't need an output transformer or capacitor.
  However when looking at the 'inside' diagram of the chip, I get the idea I do need output coupling...
  One thing is fer sher, cone coils are gettin' damaged..maybe I need a bigger one, I'm prefer the idea of not using my bigger speaker coils for testing.
  Anyway I just completely star grounded the circuit, somewhere I saw a cheet sheet, musta been for a TDA or something, they had "S", and two other letters assigned to all the grounds, "S" would mean 'Signal grounds'..all the grounds with "S" get a wire to the star ground, I forget the other two letters, but 3 ground wires, to star ground.
  Any advice on how to test without having any more coils fry would be helpful, I have more test speakers but would like to try changing something first.

[brett]

Hi Pete
the output from a 386 sits at a DC voltage of half the supply.
Hence the need for a DC-blocking output cap.
If your speaker is fried, there's a good chance that it's due to a faulty output cap.

Caps are cheap and speakers aren't, so go for a new cap.
cheers

[petemoore]

  Thanks Brett !
  I put big 'ol caps BTB on the speakers [I figure it's gonna be swinging +/- STSpeak], cant' miss with a bipolar cap build. I don't know what was blowing the Speaks for sure, haven't blown one since though..
  Amp sounds about the same with short across these caps too.
  Data sheet is still a touch cryptic for my vocabulary, I didn't see 'center tap' mentioned, it's all probably there for those with vocabulary DOT [definition of terms] and better comprehension.
  I got a chance to have focus and read your post RG, I need to go back through and...rewire here and there, speaker sewer and input pot for sure.
  I did get the noise way down from where it was, replacing the solid core wires with stranded thicker wires, also I had about 40% shielding, I upped that to about 90%...only the small sides pointing <l circuits l> are 1/2 open, this helped too.