Ah, you are looking at the original (20% feedback) schematic at the beginning of the thread. I linked to the post showing the later version where I used 30% feedback with much better results.
20% requires less drive voltage and can indeed tolerate a lower negative rail for the fet as you showed. Personally, I think that the 30% is worth it but that's a personal preference. Lots of ways to skin a cat or make an amp.
20% requires less drive voltage and can indeed tolerate a lower negative rail for the fet as you showed. Personally, I think that the 30% is worth it but that's a personal preference. Lots of ways to skin a cat or make an amp.
Correct.
Yes, corrected non-linearities of an output stage are reflected on it's input resistance that loads the driver stage. Free cheese is only in a mousetrap.
Just different ways to skin the same cat.
Attachments
Take it to an extreme. Driver drives power tube grid up to zero. Power tube plate pulls down to zero. Driver tube gets NO current.
Yes, the plate can't pull all the way to zero. But for LARGE swing (your sim is medium swing) you get WILD swing of driver tube current, including down to low-low current and low gain.
True, this is the same place the power tube gain maxes-out. So there is some distortion cancellation. However in most practical cases, using ONLY the feedback resistor to feed DC to the driver is "too much", you want the resistor to a steady voltage to keep good current flowing all through the cycle.
Earlier mention was made of stability issues with nested loops which got me to thinking (always a risky business). Obviously long loops and or loops which include complicated reactances such as transformers engender stability issues but does the simple nesting of short loops necessarily bring stability in question?
For example with the single stage plate to grid (A.K.A plate to plate) being discussed here, would the addition of a loop including both the driver and output tube via output plate to driver cathode necessarily call stability into question such that measures would need to be taken or stability ensured by o-scope probing?
I also wonder if combining these two types isn't more than just belt and suspenders but actually methods that aim for slightly different targets leading to a sum greater than the parts. Also, is it possible to successfully combine these two design elements without specialized test equipment?
For example with the single stage plate to grid (A.K.A plate to plate) being discussed here, would the addition of a loop including both the driver and output tube via output plate to driver cathode necessarily call stability into question such that measures would need to be taken or stability ensured by o-scope probing?
I also wonder if combining these two types isn't more than just belt and suspenders but actually methods that aim for slightly different targets leading to a sum greater than the parts. Also, is it possible to successfully combine these two design elements without specialized test equipment?
It is what I do in my amps routinely. My latest Edelweiss-3 amps have 3 loops.
Here are some results:
Wavebourn - Testing high-end headphone version of... | Facebook
And some video:
Structure of magic: Facebook
Here are some results:
Wavebourn - Testing high-end headphone version of... | Facebook
And some video:
Structure of magic: Facebook
IIRC your amps also have a global loop which guarantees the need for stability compensation. Would compensation be needed if only the inner loops were used? Are the relative amounts of FB in each loop arrived at by trial and error?
Well, my simulation was for the case 1, from the OP.
"1- use a resistor from the output tube anode to the driver's anode."
You are refering to:
"2- use the feedback resistor as the only load for the driver, so we get a I-V converter (like Browskie showed on one of his articles)."
http://www.tubecad.com/march2001/Partial-Feedback%20Amplifier%20Topology.png
The extensive article at TCJ presents this topology as a valid and nice one.
Some simulations I've been performing on this topology reveal a dramatic improvement in Zout, which is very nice, but much less power output too, due to the driver's anode "eating" some B+ from the output stage, as shown here:
http://www.tubecad.com/march2001/Relative Voltages.png
No need to, since thanks to a local loop OL gain by the global loop is not enough for instability. As you can see from pictures, ringing is less than 1 period on 100 kHz frequency. However, if needed I would apply it.
No. Compensdsation would be needed only if on some frequency gain through the loop is more than 1 and phase shift is 180 degrees. How a single stage add such a phase shift?
Yes, trials and errors, but not for stability, but for "inaudible saturation", while still maintaining very low THD and IMD on normal listening levels. I still did not get that work finalized, need to organize some "audibility of errors" listening sessions and build my own waterfall.
I suspect, Nelson Pass has some, since he often shows dependence of distortions on output power for his amps.
- Status
- Not open for further replies.