1970s Design Indulgence

Dear Graham,

Let us now consider the role of ground wires in amplifier stability.

One inch of 32/0.2 1mm^2 wire, which is capable of handling the current, has an inductance of around 20nH. We could increase the thickness considerably, but that would not reduce the inductance by much.

To get from the central big smoother capacitor negative terminal to the amplifier grounds, which are either end of the case, requires wire lengths 7 inches long. The inductance of those wires is approximately 140nH (nano Henries).

Having pushed the amplifier to the edge of stability, to obtain more high-frequency negative feedback, to reduce mid-frequency distortion (harshness), the inductance in the ground wires could push it into the risky region of fewer than 45 degrees phase-margin.

The two grounds: one to the output stage, and one to the voltage stage, are obviously separated by this 140nH inductance. The grounds become ineffective above some high frequency.

One solution is quite apparent that the compensation capacitors need to be made larger in value. That is to say, just that little bit larger to force the phase margin a decent amount above 45 degrees.

Another method which I'm not too fond of is to resistively join the two grounds at the circuit board. This method was used by the Dinsdale and Quad 303 power amplifiers, which were not "star earthed." The resistive tie of only a few ohms made for reduced stereo distortion but increased central image distortion. A value of 10 ohms was suggested by E M Cherry even with separated grounds.

I made the decision to slightly increase compensation, losing 2-3dB of 20kHz NFB. A simulation showed a return to a larger phase-margin, if only in the '50s of degrees.

Whether or not I should use the tie resistor will be something I will need to consider if upper-mid distortion arises again.

At present, I am using the switched-mode power supplies, with individual dummy loads that have brought their resonating noise down to almost inaudible. The S/N isn't as good due to the prototype placement of components, but I feel this will be remedied with proper separation.

Edited by Graham Slee - 07 Dec 2020 at 1:48pm

For the benefit of any skim-readers whose eyes might have photographically harvested "Sibilant harmonics at 70MHz", I do hope you take it in the context you should have done.

Anybody spouting-off that Graham Slee claims to have heard 70MHz will be sent packing!

Harmonics of sibilance are a significant problem in audio work of any kind and can ruin the concentration and enjoyment of a piece of music.

Very often, the lack of vocal control of a singer, even a temporary slip, can result in a clipped signal in the recording, along with its shower of harmonics. The recording engineer will always do his/her best to minimise the impact.

The amplifier and all other components in the "signal path" must also do their bit. It is here where power amplifiers seem to have their most significant problem.

It is no good working on the assumption that reducing all forms of distortion below a particular threshold will result in the banishment of the evils of sibilance. The sibilant shower of harmonic distortion lasts longer than the vocalists slip and represents a form of chaos.

We are not dealing with mono. Pieces of harmonics are contained in each channel, perhaps more to one than the other. Current is being drawn through inductive conductors from a source and is returning to that source through the similarly inductive ground conductors. The designer has to see all this in slow motion through the mind's eye and compare it with his knowledge, and sometimes that knowledge is incomplete!

Knowledge is gained through dissemination from those who have traveled the road before. If they skimmed over the subject or ignored it completely, it is up to us to either glean knowledge or "chalk chuff under it."