One advantage of ripple current and not having pure dc ! Although one could assume that any audio signal ends in a magnitude that decays to zero, and hence silence does end at the origin.
Thanks @smoking-amp , is it the same reason why, on the japanese 12AT7 circuit I proposed, you suggested to set it to just nullify the primary resistance and no more than that? Will it oblige to add negative voltage feedback from the secondary as well to avoid instability, is it correct?
May I ask you how do you dimension the PSU for it? I'm slowly designing a SE amp with GU50 that will swing from 450 down to 50 V.
Considering heaters needs 12V, I need to have at least 15V to take care of the CCS. Will the PSU swing down as the plates of the tube?
Will I then need to have 135V for the heaters? So will it mean that the CCS will dissipate the rest of the power?
Excuse my dumb questions.
The secondary does not carry any magnetizing current. So there is no need to over do the correction. The sampled cathode current contains magnetizing current in the primary, so over compensating the tube drive to also reduce secondary resistance would put in inverted magnetizing current which would not be "consumed" in the secondary. To cancel secondary resistance, one has to use a different scheme that measures secondary current going to the speaker.
The primary R cancelling is useful since the only other way to monitor actual xfmr output (after magnetizing current does it's dirt) would be to put an unloaded secondary on the OT for Voltage Fdbk. Some classic amplifiers did do this. Two secondaries.
One CAN put in local Negative Voltage Fdbk from the primary/plate to null out the output tube Ri, and would be a good idea (since tube Ri is variable with current, unlike xfmr resistance). Nulling out everything (ie., secondary R too) can get you close to an oscillator easily. Global Neg V Fdbk can null the secondary resistance safely anyway.
Last edited:
Thanks again. Is it what you are doing in your amp with feedback from anode to cathode of the ccs-loaded pentode driver?
That would conceptually be the most straight forward, and practical. I prefer the idea of taking voltage Fdbks from the UL taps, if possible, since they are usually more closely coupled to the secondary (for UL purposes). But then there is more chance of phase shift and instability under load with the higher loop gain (driver in loop and part of the primary as well). Depends on the OT, just have to try it, with resistor adjustments for the different tap voltages. Ultimately, one adjusts the various loops for best FFT and lowest output Z before any global N Fdbk is applied. And then check for sound quality. I would set up a set of local N Fdbk resistors for each case, adjusted for equal local Fdbk, and then try one or the other with load attached. (and then a modest cap across the load too, to check stability.)
As to the amount of local N Fdbk, 20% seems to be typical for single tube Schade like setups, but with the driver in the loop you can obviously go for a lot more. One ends up with a trade-off between stability of the local loops versus stability of (any) global loop. I would go for the best balance (ie., not having one type loop near instabilty and the type other not) Both loop types are enclosing the output tubes with plenty of gain, so no need to maximize one excessively.
You probably want to do the positive current Fdbk (neg. resistance) thing last, after optimizing the Local/Global balance to keep complexity and interactions down. Magnetizing current effects are fairly minor in P-P OTs anyway, just a small tweak.
As to the amount of local N Fdbk, 20% seems to be typical for single tube Schade like setups, but with the driver in the loop you can obviously go for a lot more. One ends up with a trade-off between stability of the local loops versus stability of (any) global loop. I would go for the best balance (ie., not having one type loop near instabilty and the type other not) Both loop types are enclosing the output tubes with plenty of gain, so no need to maximize one excessively.
You probably want to do the positive current Fdbk (neg. resistance) thing last, after optimizing the Local/Global balance to keep complexity and interactions down. Magnetizing current effects are fairly minor in P-P OTs anyway, just a small tweak.
Last edited:
Late edit:
You can check relative instability of the Local versus Global N Fdbks. By checking Amp output (versus input) phase shift at audio HF on an O' Scope, between both loops not connected and one loop connected conditions, with an output load connected. Ultimately you want the least phase shift with both loops connected. So you might want to check to see if changing one loop gain has more phase effect than the other, and calc. where the least delta phase sum would be.
For tiny phase amounts, there is the HP3575A Gain-Phase Meter (EBAY) for perfectionists, if you plan on doing a lot of designs. Hmmm, they were cheap, I guess not any more.
You can check relative instability of the Local versus Global N Fdbks. By checking Amp output (versus input) phase shift at audio HF on an O' Scope, between both loops not connected and one loop connected conditions, with an output load connected. Ultimately you want the least phase shift with both loops connected. So you might want to check to see if changing one loop gain has more phase effect than the other, and calc. where the least delta phase sum would be.
For tiny phase amounts, there is the HP3575A Gain-Phase Meter (EBAY) for perfectionists, if you plan on doing a lot of designs. Hmmm, they were cheap, I guess not any more.
Last edited: