1970s Design Indulgence

The musical differences of capacitor types is nicely highlighted on this website, all be it as a key component of loudspeaker crossovers. A bit of light reading perhaps! 

After removing the "imaginary" emitter resistor and combining its value in the NFB divider resistor, the "life" in the music departed.

The "imaginary" resistor is actually real, but its contribution has not been properly explained by anybody on earth yet, as far as I know, and I've dug deep.


The actual values changed to R2: 0 ohms; R3: 620 ohms; R4 560 ohms; R8: 12k ohms; C5: 100uF.

This has now changed to R2: 180 ohms; R4: 390 ohms; R8: 8.2k ohms (C5 remains 100uF as the turnover frequency here is much below 1Hz).

R2 is therefore the "imaginary" resistor. Although its action can be explained, the differences cannot be measured by any conventional means, and therefore its explanation can be called into question.

Regardless, it's for keeps (I think...) because the sonic qualities reflect what the explanation explains.

The junction R7/R4 (via C3) is the negative feedback node. With R2 = 0, the signal at that point is not discontinuous with the input signal, and that can be mathematically proven. However, should distortion be present in the fed back signal it becomes discontinuous and an error voltage will be generated to eliminate it.

Now, if distortion is present, the discontinuity will mean R4 is not purely an extension of the intrinsic emitter resistance (re), and therefore it must have an effect on the value of gm. And as the value of gm works with the value of Cdom to set the stability values of the amplifier, these will change for the duration and magnitude of the distortion to some extent.

The duration and magnitude of the distortion will vary with signal amplitude and frequency and so it is always changing. It is therefore always generating an error voltage to some degree, and therefore the stable point is always changing. And because it is generating an error voltage the gm must increase rather than decrease.

The increase in gm should be countered by an increase in Cdom, but that is fixed, and so any error voltage leads to a narrowing of phase margin. Even the slightest narrowing will take it more toward ringing.

Therefore gm needs to be made more linear, but for reasons that the explanation against it being absolutely linear would take some digesting, I will simply state it can only be made more linear.

So by taking a proportion of Re from R4 and placing it in series with the emitter outside the NFB node (R2), we have made gm less prone to changes influenced by the need for distortion-eliminating error voltages.

We still have the same open-loop gain and so the error voltages should not be any larger. And so measurable distortion should remain the same.

So, the action of the "imaginary" resistor has now been explained, and I only hope it will stand scrutiny, because the audible improvement (yes, improvement) is undeniable.

But, as I will continue to state, improvements can be fleeting because of other factors so this will now be subjectively tested for a further lengthy period.

(erratum: C3 is now shown where it should belong)

Edited by Graham Slee - 05 Feb 2019 at 6:36am