Feedback resistor and capacitor calculation, anyone good at maths?

[Gabe001]

Just want to pick your brains on a feedback calculation with a different OPT impedance.

I am cloning an Armstrong 220, which has 29db feedback
The circuit diagram shows a 75k resistor and a 47pf resistor in parallel, using the 16 ohm OPT tab. Circuit is shown. The feedback goes to an ecc83 with a 2.2k cathode resistor.

I don't have a 16ohm OPT output, I'll use the 4 ohm impedance output instead

I've calculated the adjusted feedback resistor and capacitor values for the 4ohm OPT tab, as follows:

Resistor value= 37.5k ohm
Capacitor value = 94pf (91pf if you factor in the 2.2k)

The thought process behind this is as follows:
The speaker rms voltage for 16ohm is twice that for the 4ohm impedance output for the same power (eg 1w at 4ohms is 2v, 1w at 16ohms is 4v), so the resistor size ratio is 2:1

The capacitor value was calculated using the low pass filter formula. I used an online calculator. The cutoff frequency is 44/45khz. For the same cutoff frequency with a resistor of half the size, the cap size needs to increase accordingly. There are a number of constants involved, including the valve capacitance but these should largely cancel out being the same on both sides of the equation.

Can anyone please check my results?

Thanks
Gabriel

[wd40addict]

Mathematically you're correct, nearest preferred values would be 39k and 100pf. But, bear in mind that as you're not using the original transformer the cap value may need tweaking if there are signs of instability.

[Craig Sawyers]

Even with Audio Research mega-priced valved power amps (N x £10k where N can be anything up to 20), they just take feedback from the 16 ohm tap. Just like with the Armstrong 200.

Craig

[Gabe001]

Thanks guys. My budget only stretches to Chinese 8k ultralinears with 4 and 8 ohm options. I've used them before. They're ok, but tend to pick up magnetic coupling from mains transformer a bit more than others I've used. Correct placement some distance away from mains transformer is crucial.

[wd40addict]

Radford did it properly of course, selecting a different output tap also selects a different feedback network

[Craig Sawyers]

Depends on the amp. And yes they did on the original 15W, 25W and 100W amps. But the Renaissance only had a single output winding, no switching and no gain selection.

There are also those out there who rip out all the gain selection gubbins, connect the transformer tap to match the speakers they are using, and hard wire in the relevant feedback network.

Craig

[trobbins]

The capacitor value may be related to different frequency characteristics of the OPT depending on where phase margin is being encroached on the most. As such, it may be significantly different when selecting a different secondary winding configuration.

[Diabolical Artificer]

I use a 20k pot then whilst I IP a 1khz sine wave I monitor the OP on an audio analyser and scope, all the while reducing R, when the amp starts going unstable I back off a bit. Then comes some frequency response testing, hum & N testing etc. Pot off, measure replace with nearest fixed R value. Once I'm happy with that I use capacitance decade box across a now fixed R whilst I IP a square wave of 1khz & 10khz selecting C to get the best square wave at 1k & 10k, I then drop down to 100hz to see how that looks.

Before any of that though I do an open loop frequency sweep from 10hz to 100khz with an 8r dummy load looking for peaks and frequency's where distortion rises dramatically, then do the same with a capacitive load of 100n, then 1u if needed looking for the same things. If I hit a peak at 35khz say where THD = 21% I'll fit an RC filter - HPF from anode to ground of the OP valve/s to reduce high frequency gain. this goes on untill I've got an unconditionally stable amp with all loads, then comes fitting the RC FB network.

All this is a lot of work but needed if your going to sell an amp, if it's just for your use you might get away with bunging in a 39k etc, but every OPT is different, I've heard some right horror story's of month's of work to try and get an OPT to work in a 5-10 for instance.

Lastly NFB seems to work better when you split Rk into two, a 2k2 might be a 1k8 & 470r say, Ck is also semi bypassed, why this is I've no idea but feeding back straight to the cathode on low power amps doesn't seem to have as much affect.

All this not as 100% correct as doing 5 pages of equations but works for me, Andy.

[Gabe001]

Thanks for the replies. Yes I do know that it can be tricky business with a different opt and a different feedback loop, but I've got to start somewhere I suppose. I expect to have to do some "snagging" anyway for a few days afterwards until I'm satisfied, so if this involves messing around with the feedback loop, so be it.

I had no trouble with the 5-10 luckily - actually I did with the first OPT that gave me 1v of parasitic inaudible oscillation only at zero volume (thanks GJ for your thread on instability which made me realise this) - but I swapped it for another from the spares box and it's completely stable - so fingers crossed with the Armstrong.

Andy, I have used the pot approach before (I think you may have suggested it elsewhere), but I'm using an established design here and I'm building it stock, except for the solid state rectifier, so for now I'll stick with the official recommend nfb and I'll let you know if I run into trouble.