PGP (Pretty Good Poweramp)

[matze]

Concerns about electrolytics in feedback loop

I would like to support the concerns in terms of microphonic effects in polarized media from jcx and KSTR.

Are there any serious investigations about that (we are talking about micro-volts certainly not visible when hitting the capacitor housing and looking at an osciloscope)?
THD measurements, at least when without acoustic impact onto the capacitor, might be inappropriate here.

My simple - hopefully needless - concern is that electorlytics are quite prone to microphonic effects already due to their sheer size and therefore lower mechanical compactness/stability.
[ This might even apply to power supply blocking capacitors in small-signal stages, if the PSRR is not large enough ]

Matze

 

[matze]

Dear Ovidiu,
Dear Edmond,

my (late) congratulations and respect for your project. It is certainly one matter to design and simulate such a complex and demanding piece of hardware, but it must be a completely different effort to build and evaluate it both in terms of objective and subjective performance.

Chapeau!
Matze

 

[Bob Cordell]

Dear Edmond,
Dear Bob,

I am wondering about the following. The unknown load capacity and the output inductance form a series resonance. Using e.g. a 2 uH or larger inductance, one can easily image circumstances, where the resonance frequency falls into the region of the unity gain crossover of the global feedback loop. Although the resistor parallel to the inductance dampens this resonance, it will influence the phase margin of the amplifier.
As the listening results to the PGP (and my own experience) suggest, a phase margin change can influence the amplifier's sound, even if the unity gain crossover frequency is really high. (I would therefore also opt for a inductance as low as possible.)

[[ The engineers of Bryston probably found a nice trick. They use a quite large inductor, but also load the output with a rather large capacity of 100nF (perhaps a capacitor with high dielectric losses, providing good intrinsic damping). The unity gain crossover of their main feedback loop lies *above* this resonance. Therefore, each additional load rather improves the behaviour of the overall circuit. This also could explain the result of certain reviews that Bryston amplifiers are that well suited e.g. for planar speakers. ]]

Finally the question: did you investigate the influence of the output resonance? It puzzles me a bit, that none of the well-known audio power amplifier design books seems to further elaborate on this point.

Best regards,
Matze

Hi Matze,

This is an interesting question you bring up. It should be answerable by simulation. To some extent, one might argue that any change in the load, whether isolated from the output stage or not, will alter the phase margin to some degree. Interestingly, the R-L corner frequency of 1 uH and 2.2 ohms (as an example L-R network) lies around 350 kHz, a frequency below most amplifier gain crossover frequencies. This means that the network looks largely resistive at the higher frequencies near the gain crossover frequency. With the feedback tapped off before the series R-L network, it tends to get isolated by that resistance from whatever is going on on the other side.

As you point out, any series resonance with load capacitance will be of a rather low Q for any reasonable value of capacitance - and this assumes a load capacitance with no series resistance of its own.

Cheers,
Bob

 

[matze]

Hi Bob,

thank you for the fast reply. Even if it is perhaps pure speculation or hair-cutting, I would like to add some points.

Including the resistive and capacitive amplifier load into the simulation is one point. This will bring us information about the loop behaviour under different conditions. The way how these technical results translate to perceived sound quality is (sadly, or the other way around for our hobby - happily) still unclear.

1. Audiphiles make the observation that "linearizing" the speaker impedance, i.e. deliberately reducing its value at high frequencies, often may improve the sound quality. If there were a technical problem with the output resonance, this would fit together. The reduced (resistive) impedance would further dampen it.

2. Most of us will aggree that it is easier to make an implifier "sound right" if the output bias current is high. One argument is of course the lower crossover distortion. Another (speculative ...) argument might be as follows. Independently of the "VAS" output impedance, the lowest possible open-loop EF-OPS output impedance is of the order U_T/I_E. If we have an amplifier with low bias current, then, at the point of current take-over between the two output halfes, the open-loop output impedance will remarkably increase. This effectivley means that the NFB phase margin will be modulated with signal excursion if we have a heavy reactive load. Again, this fits together with my uneasy feeling about neglecting the issue of a possible output resonance in the vicinity of the unity gain crossover.

Cheers,
Matze