1500 Watt

[wg_ski]

Originally posted by ostripper
BTW - WG ,what amp is that?

OS

A home brew PA amplifier. An example of something on the order of "the spoof from the ESP site" that works perfectly. It does generate a lot of heat, and makes the electric meter spin really fast.

[wg_ski]

Originally posted by AndrewT

can you give us more detail on this NFB/Signal Ground link and how not to do it.

The natural tendency would have been to run the NFB trace and signal ground up through the middle or around the edge of the PCB. Right next to a bunch of buses and connections that have very large half-sine currents in them. That will induce distortion into the feedback line - and this is indistinguishable from desired signal. The other thing designers often do is to just tap the speaker out off one side, and the NFB off the other. That keeps the trace short and fixes induction problems, but with a large bus with a bunch of taps for the output transistors the voltage at one end of the bus is NOT the same as the other. The load and sampling need to occur at the same node.

So the first attempt was to run the NFB and ground up and over everything, keeping them at right angles to the extent possible. Everything was fine except when another amp's power trafo was sitting right above it. Hummed like crazy. Then I tried coax (shielded cable) to run the NFB and signal ground back to the other end of the board where speaker and speaker return were. Now anything that tries to get induced gets cancelled. The input side is balanced so it doesn't need ground return current.

[MJL21193]

Originally posted by wg_ski
But if the speaker output is the yellow wire, it's at the wrong end. The speaker out and the feedback tap need to come off the same end of the emitter resistor bus. Preferably, Bus-->Spk-->FB. Otherwise, a distorted signal can end up being fed back.

I see my thingy has sparked some comment.
That is the speaker out wire but the NFB pick off point is close to it - there is a trace on the bottom that runs to it: