Graham Slee Power Amplifier

Thanks, but I am bi amping so I need 4 conductors, the Canare cable could be worth a play with sometime though.

So how did it go?

Well, attempt number one isn't going to make it into production. Based on JLH's class A amp of 1969 (circuit diagram below), it simply doesn't perform, period!



It probably would perform as shown but then again I believe the design to be flawed as when modeled it has little phase or gain margin, and as JLH mentioned in his article, it readily oscillates with a small capacitance across R3, the NFB resistor. Other designers may think that obvious, but with sufficient margin it should not - although the capacitance does reduce the margin, and therefore the original circuit had little leeway in this respect.

My simulation model based on this circuit was absolutely stable, and it was in practice too. However, what I was trying to do is back bias it so it wouldn't run so darn hot. The result being that it added crossover distortion which in my opinion was beyond the acceptable. I could have increased bias to the point this distortion became within acceptable limits, but the temperature rise would necessitate a much larger heatsink, and hence larger case leading to a much more costly product - and would it be salable?

What I do like about this configuration is its sound. Camouflaged extremely well, it appears in a number of great sounding op-amp output stages. Indeed it also features in the much revered "gain-clone" amplifier, but as yet I have to obtain a simulation that I feel has the right stability margins: the LM3875 data sheet tends to labour the point regarding additional stability components, suggesting it doesn't have much margin which is borne out by the open-loop frequency response plot (see data sheet here: http://www.national.com/ds/LM/LM3875.pdf).

Another point about the LM3875 is its non-existent flat portion of its open loop response - it falls all the way from a few hertz at 20dB/decade and all the good points about the output stage will be negatively dominated by this - it is not "ultra-linear".

For now it's back to the drawing board. I am determined to find a way of biasing this superb sounding output stage into class AB and the LM3875 proves that can be done.

There are often little nuggets of such information in these discussions.

A good reason to follow along even when sometimes half of it is over my head.

Pity it would not play as hoped for.