joey wrote:I would raise the 100n blocking cap to 10uf-22uf, and move it between 33K and the and the Pi input, I would then also move the presence circuit to the other side of this cap. (where it meets the 33K rf resistor)
Or leave the 100n aka 10uf-22uf entirely out. The Depth pot will be a bit scratchy after this, since now DC is across it, but this would enhance the stability of the power stage. In my British Purist I don't use a cap in the feedback loop, the scratchiness is negligible, but the power stage is so stable, that I can use a 12K/4.7K NFB w/o getting motorboating even with the Depth pot on zero.
joey wrote:... anytime you put a cap in your signal chain you introduce a phase shift and a frequency dependent time constant. In this case we don't want the blocking cap to act like a fixed resonance control even if the resonance pot is shunted out of the circuit. making it's value large will pass all frequencies of interest through unaltered. It may not be a huge deal since you have a -3db knee at 25hz-27hz with the .022uf PI output coupling caps, and the 220K+output impedence load, but the stock slo has half of that so.....
flemingmras wrote:Just out of curiosity Joey, can you explain this in layout form?
Ok, then I'll go and try to explain it with my German's English, what contains the risk, that only a few will understand it. But if I'd explain in German (much easier for me), then nobody here will understand it...
A NFB loop means, that a signal voltage is taken at the output and fed back to a previous stage of the amp, but the NFB signal is inverted - otherwise the power stage would become an oscillator. The reason for the NFB loop is to tame some uncontrolled reactions of the power stage in its interaction with the speakers, which do have an intense inductive interaction with the OT
As I've said above, the NFB signal must be inverted, to be able, to tame the power stage and not to activate oscillation, what would be the case with non inverted (positive) feedback. But the desired NFB signal to be "exactly" inverted - or in other words, to be 180° out of phase to the signal phase of the stage, where it's fed in unfortunately is only theoretical like so much in a "perfect world", only existing in the heads and/or on the paper. The reality is different!
The theory is quite simple: In a Marshall (and many more similar) power stage, there are 3 points, where the signal voltage becomes inverted (180°) between the point, where the NFB is taken from (the OT's secondary) and where its fed in (the 2-nd system of the PI). The 3 points are the PI tube, the power tubes and finally the OT itself. Let's calculate theoretical: 3 * 180° phase shift = 720° phase shift = one complete cycle + 180° = exactly inverted
But
that's as already said only theoretical and only valid for a certain frequency range. Due to the usual RC coupling in the amp everywhere and due to the output impedances and the input capacitances of the tubes itself, some additional phase shift occurs. For example in a low-pass-filter (high-cut-filter) at the -3dB point an additional phase shift of 45° occurs, the same at the -3dB point of a high-pass-filter (low-cut-filter).
With an OT it's still worse, because it's an LC unit (inductivity & capacity), hence on its both frequency band ends at the -3dB points the phase shift even is 90°
So let's check a power stage's schem now and let's follow the NFB loop:
1-st we have there, where we take the NFB off of the OT a phase shift of 90° off of our desired 180° at the -3dB points - and any OT does have a -3dB high-cut-point (highs roll off point) as well as a -3dB low-cut-point (lows roll off point).
2-nd we have there, where (maybe) the 100n is sitting in the NFB loop just before the Depth pot a phase shift of 45° off of our desired 180° - This caps forms together with the (variable) NFB resistance (impedance) and together with the 4.7K to ground on bottom of the PI in parallel with the input impedance of the tube a high-pass-filter with a certain -3dB point - the source's impedance (the OT) in series with the loop's resistance forms a low-pass-filter together with the presence cap, again with a certain -3dB point.
3-th we have a comparable situation with the 100n input cap to the 2-nd PI system, so again certain -3dB points to the highs as well as to the lows.
4-th we have the PI's coupling cap, which forms together with the bias R in parallel to the tube's input impedance a high-pass-filter and the PI tube's output impedance in series with the grid stopper forms a low-pass-filter together with the output tube's input capacitance.
So -3dB points everywhere on both ends of the audible (and inaudible) ends of the frequency range with each an additional phase shift of 45°, in case of the OT even 90° - hence we come way off of our desired, ideal and unfortunately only theoretical NFB phase of 180°
At about a total phase shift of 270°, consisting of our designed 180° and some unwanted others, the power stage becomes instable at this frequency of shift additions. As soon as the total phase shift, consisting of our designed 180° and some unwanted others reaches the 360° point, we do have the "perfect" oscillator. But oscillation starts already way below the possible max. of 360° - and we have to differ in:
a) 'high frequency oscillation', what's inaudible but power and sound robbing and/or is squealing like a pig, together with some nasty noises, which are the result of intermodulations of the signal with the RF oscillation
b) 'low frequency oscillation', what is known as "motorboating" and might go down to very low frequencies of 1-3 Hz, what's then more kind of a breathing of the amp. You can see the breathing on the scope, when the complete sinus or square wave is steadily jumping slightly up & down
The above described occurs, when all our low -3dB points and/or all our high -3dB points met on top of each other at the same or close together located (low cut or high cut) frequencies. So the designer's mission now is:
1-st to have as less -3dB points as possible in the entire NFB loop, what might be achieved by skipping i.e. the 100n in the NFB loop just before the Depth pot
2-nd to design all these different -3dB points in the way, that their frequencies are as much apart of each other as possible. As a rule of thumb the
absolute minimum is a distance of at least one octave between all of the -3dB points, but way better would be a distance of two octaves.
An example of a good NFB loop design would be:
Low cut (-3dB) points at i.e.: 3Hz; 12Hz; 48Hz
High cut (-3dB) points at i.e.: 6kHz; 24kHz; 96kHz
That's an ideal and especially to calculate the high & low roll off points of an OT only by the taken measurements of the primary inductivity and the stray inductivity is nearly impossible, hence some final R&D with a good scope is inevitable
But please don't have me now, to calculate your power stages!
I gave you the 'tools' above, so now go and DIY !
Believe it or not, but to write this all I've been sitting altogether nearly two hours
The same, but in German would have been probably just 30 Minutes.
Larry
