The driver part of the amp is more or less a classic Blameless voltage feedback circuit. It worked absolutely fine apart from the clipping issues, giving low THD across the whole audio band. The clipping was fixed with a single Baker clamp diode.
I parted company with Self on the output stage as he prefers the CFP, but I believe the crossover distortion from an emitter follower output stage is subjectively less bothersome.
Or another option...?
Make the feedback network out of many resistors of the same value (David Zan advocates this method).
Apart from increasing the chances of component failure and size, I can't see any disadvantages to this. If you have to get from A to B then it's either copper or component.
A single diode across the Cdom did the job. Not the best place to put a diode, as the nonlinear junction capacitance causes distortion, but it seemed to work well enough. I used the smallest diode I could find with an adequate voltage rating, the BAV20. I also reduced R21 to help turn off Q1 quicker.
The clipping problems only appeared on the negative half. The current source on the positive side didn't seem to need any clamping.
I think this is a bad idea. A big long chain of resistors will have distributed capacitance to ground, adding a whole clump of extra high frequency poles to the feedback loop.
Last edited:
I sometimes use two 2-watt metal film feedback resistors in series for the feedback path, but that is mainly to keep the dissipation very low when the impedance of the feedback network is kept low (for example, a 215 ohm feedback shunt and a pair of 2150 ohm feedback resistors in series for a gain of 21).
Cheers,
Bob
Hi Bob,
How about 215R for the shunt and 20 x in series 215R for the feedback? Apparently, all sorts of distortions cancel by doing this. What's your opinion? Is this overkill?
Paul
This would certainly work, and keep the working conditions of all of the resistors the same, but it would take up a lot of board space. Not sure how bad the distributed lumped R-C line would be with excess phase added.
Maybe we could add a 22pF capacitor (or pick your value) across each of the 21 resistors for good measure 🙂.
Cheers,
Bob
There is a tradeoff here. According to B.Hofer from API low ttc resistors have low voltage non linearity . This feedback resistor sees the full voltage variation and THD is not attenuated by feedback. It should then be low ttc. It is one of the most critical for low THD. But these résistors are expensive and low power; a serie of them would be perfect but at what cost ( in my active crossover system with subs I have 11 power amplifiers). Have you measured the THD impact of these résistors in this position. What would be an optimum from a diminushing return pov ?
Thanks
JPV
Appears more than one person has concerns over the parasitic C using this method.
We are talking about a large number of components to achieve this. If we go series parallel to increase reliability and small (22pF) caps across these then we are looking at 84 components. When this was discussed before the idea of a daughter board populated with SMD parts for the feedback resistor was suggested.
Then there is this parasitic C. Using 22pF to "cancel" this suggests that the HF poles are likely to influence amplifier stability. Would be good to model this in LTSpice. What sort of values would you use for parasitic C per resistor?
Are the parasitics of a chain of components that much worse than a long copper track?
Is all this worth it or do we go for z-foil resistors? Or are these snake oil?
It's all a trade off. At present I'm of the mode of going with a slower amp with more features. Slowing things down means these HF poles are less troublesome.
Sorry for the ramble... 🙂
Paul
Last edited:
I am bringing this up based on an RF project I did, a signal strength meter for an antenna tuner. I used a series chain of carbon film resistors to attenuate the RF voltage. I was surprised to see that the resistors near the hot end dissipated most of the power, to the point that the first one went black and crispy while the last one was almost cold.
To me this proves that the resistor chain must act as a string of Rs and Cs.
To me this proves that the resistor chain must act as a string of Rs and Cs.
The parasitics are similar. But you can arrange it so that the track is driven by a low impedance (the output impedance of the amp) whereas each resistor in the chain is driven by the impedance of the resistors before it.
Last edited:
But so would using a single feedback resistor of a sensible size. 🙂 We are always taught to keep the wiring associated with the feedback node of an opamp as short and low capacitance as possible. Why should an audio power amp be any different?
I habitually use 600mW resistors for my builds.
For the feedback route I ensure that the feedback resistor never has to dissipate more than 60mW (10% of Pmax) when signal is at maximum peak value.
This ensures the resistor/s stay cool to cold and that the dynamic variation in Pdiss is kept very low. Equals low resistance variation = low resistor distortion.
For the feedback route I ensure that the feedback resistor never has to dissipate more than 60mW (10% of Pmax) when signal is at maximum peak value.
This ensures the resistor/s stay cool to cold and that the dynamic variation in Pdiss is kept very low. Equals low resistance variation = low resistor distortion.
Agreed, this would suggest a leakage path to ground. And would suggest that a long chain of resistors is sub-optimal. This is unless Bob's suggestion of paralleling small caps with the resistors counters this.
A single resistor can be used when you are dealing with "VFA" amps but with "CFA" amps the power dissipation could be too high.
Agree with keeping the feedback path as short as possible. I see amplifiers as a specialised form of control system.
Would this not cause parasitic C problems as well?
Paul