I have completely reworked the amplifier with positive results. (Work and real life intervenes and these things take a while.) Two things stand out:
1. The layout matters. Yes, it was said in this thread and yes, I've read Tom Christiansen's articles and many of his posts in this forum. This post, among others, was also interesting, LM3886 PCB vs Point-to-Point
I've tightened up the layout and in a very newbie fashion created wider power traces and ground plains using coper tape. There was no science to the layout of the ground plains (1 each for signal and power grounds) other than they fit the layout. I've attached a pic of the topside and underside of the board using this methodology. I also soldered component leads adjacent to the LM3886 directly to the pins rather than using the perf board.
I'm sure the layout is not ideal but stay tuned for results.
2. Ground. Ground. Ground. Again, thanks to Tom C's Taming the LM3886 Chip Amplifier
and others who have written in these forums on grounding. It also helps to read the datasheet. I should have started there. Changes to grounding made all the difference in the world in terms of both audible noise AND the wave form.
3. Finally, I thought this article from Ellliot Sound Products on capacitors
The attached pictures show the build to this point, top and bottom. (I used an acrylic insulator between layers of copper tape. It looks crude but it was a very effective way to swap out and test components. I've also attached:
- an oscope shot of the input and output 1 kHz sine and square waves
- an oscope shot of the detail on the output signal square wave showing a small amount of deviation from a true square wave. This is possibly due to the impact of the high pass filter. When I input very low frequencies and 10 kHz and higher, I get the classic shape of the square wave as it changes in relation to frequency input with a high pass filter
- an FFT analysis
The scope images use a 1kHz 500mv input. I'm using a 5 watt 8 ohm resistor for the load.
So how does it sound? In truth, I'm very pleased. I haven't hooked this up to my B&Ws, not sure it would drive a B&W very well, but after thorough testing, I'm using it on an old pair of KEF UniQ towers. No audible distortion, at least of the kind my ears can hear and very nice sound.
I know there are more improvements and I'm sure my layout and execution leave much to be desired but this whole exercise was to learn what makes a chip amp like this tick and I've learned a lot. I also made a lot of mistakes but the two most important lessons were layout and grounding. The difference in what I see on the scope and what I hear from these two elements were immediately discernable, even to a novice.
I'm going to keep experimenting and I want to create a higher power amp, possibly by running the LM3886 in a parallel configuration. I have a pair of Modulus 86 boards that I will eventually build as a final comparison and which I will use in the amp that I want to put into use but this has been a great exercise.