SteveTheShadow wrote:
The rule of thumb with partial feedback is that the feedback resistor should be between 4.5 and 5 times the value of the driver stage load resistor
I've been wondering about this so-called rule of thumb I read on the interweb. As everyone knows by now, I have been struggling with the gain structure of this amplifier, since I changed the drivers to a pair of EF184 triode strapped pentodes per channel.
Going to Broskie's "Partial Feedback Amplifiers" where he talks in terms of V to I and I to V converters; basically the anode follower used as an output stage, there is mention of the voltage generated by a current operating against a resistance; simple Ohms Law, but easy to overlook in this situation.
Now my driver stage will swing 14mA pk-pk so this swing against the 53K feedback resistor will generate a 742Vpk signal across that resistor, overwhelming the feedback mechanism and exceeding the voltage capability of the power supply.
This condition results in bugger all feedback, sky high gain, sky high voltages across resistors not designed for them, high output impedance and gross distortion, or something. Anyway that's what happens as soon as the amp goes beyond 10 o'clock on the volume control, meaning that the feedback that is supposed to be there, must be missing in action somehow, so that'll do me for an explanation.
To prevent this happening according to Broskie, the output stage must have sufficient feedback from its plate to its grid so that the driver cannot overwhelm the mechanism, or generate crazy voltages. If you want that not to happen, you scale back the idling current in the driver stage, but if you do that with a triode, you have to flatten its loadline, that's if you don't want to end up in the cramped bottom of the curves on the negative swing. But wait a second, you can't flatten its loadline 'cos then you are trying to drive voltage into an output stage that looking at all this, makes you realise it really could do with a current source, which means that what you really want is a pentode driver with its high output impedance, plus a steep loadline that'll drive current without swinging big volts.
Now with a triode, it's much more difficult to raise its output impedance, unless you bias it with a very big un-bypassed cathode resistor, and with a 2 stage amp like mine, this is even more difficult to do as there is no possibility of direct coupling from the previous stage to bias the grid positive so that you can use a big resistor to give a + 2.5V difference from cathode to grid, so I had to find a workaroud.
So I have steepened the loadline of the EF184 and reduced the feedback resistor value to 27K. Applying 14mA against this resistor will generate a voltage of 378V which is about exactly what the HT across the output stage is, so no danger of exceeding the capability of the power supply, even at max volume.
The upshot of this is that now I have the right gain structure and can turn the amp up so that I give up before the amp does.
Anyway it works properly now with the triode driver, but what a bloody game to get it going properly.
My attempt to explain it probably has more holes than a colander but I've got there somehow, with my convoluted logic and am getting excellent results now. Sometimes I really wish I was a proper engineer.
It'll be soooo much easier to get it right with a pentode driver; so much easier, and you don't have to have a degree in astrophysics to get the thing going properly.
The sound of plate to grid feedback over one stage is well worth the hassle though, when got right.
Sgt. Baker started talkin’ with a Bullhorn in his hand.