Trying to linearise a significantly nonlinear amp by slapping on lots of feedback is exactly what valve fans often cricise SS for, yet this is what SE needs (for those who prefer low distortion).
Actually, I was speaking somewhat loosely: re-entrant distortion starts as soon as you put on any feedback so SE may be better without feedback. Perhaps I can rephrase: SE needs feedback but often can't cope with it, PP needs less feedback but can cope with lots.
Actually, I was speaking somewhat loosely: re-entrant distortion starts as soon as you put on any feedback so SE may be better without feedback. Perhaps I can rephrase: SE needs feedback but often can't cope with it, PP needs less feedback but can cope with lots.
That's what I don't understand - the phrase SE often can't cope with it. I really don't know what you mean by cope in that context. And what is re-entrant distortion??
Cheers
Ian
Re-entrant distortion is what happens when a somewhat distorted output from a circuit is fed back in again and interacts with itself and the original signal to generate higher order products. NFB does it, both global and local. Inadequate decoupling can do it. To a first approximation, the level of higher order re-entrant distortion varies as the square of the original distortion. This is why it is best to make an amp good before applying feedback.
An amp with high distortion will generate a lot of re-entrant distortion when you try to linearise it with feedback. Two solutions:
1. don't apply feedback, and instead learn to like the inherent distortion
2. apply lots of feedback, if loop gain and stability allow this i.e. the amp can cope with it
An amp with high distortion will generate a lot of re-entrant distortion when you try to linearise it with feedback. Two solutions:
1. don't apply feedback, and instead learn to like the inherent distortion
2. apply lots of feedback, if loop gain and stability allow this i.e. the amp can cope with it
I have tried to demonstrate this with actual circuit but not succeeded.
The re-entrant distortion must obviously be at very, very low level.
(<-100 dBc)
Rather a problem in theory than in practise.
Maybe I am confusing the original topic, as but we earlier discussed about distortion cancelling, I dare to come back to this subject.
I am presently nibbling with a simple SE amplifier with soviet made 6F5P (ECL85) triode-pentode.
The best operating condition for the pentode halve alone I found was a UL-connected with +Ub = 200 V, Ia = 40 mA and Ug1 = 22 V. The opt is 5k2.
I can get 2 W with 4.5 % THD. The clipping takes place at 2.5 W.
When I add the triode halve of 6F5P as voltage amplifier and bias it to optimum by only observing the THD at the output, I can get the same 2 W, but now with less than 1 % THD. No NFB is used.
The circuit is basically equal to what I presented at the post 20 earlier, but with different resistor values.
At the triode, Ra = 47 k, Rk = 2.03k (adjusted to optimum), Ua = 147 V.
And all these results are with real circuits, no LTspice simulations.
I have to say that the amount of distortion cancelling is amazing in this case.
I am presently nibbling with a simple SE amplifier with soviet made 6F5P (ECL85) triode-pentode.
The best operating condition for the pentode halve alone I found was a UL-connected with +Ub = 200 V, Ia = 40 mA and Ug1 = 22 V. The opt is 5k2.
I can get 2 W with 4.5 % THD. The clipping takes place at 2.5 W.
When I add the triode halve of 6F5P as voltage amplifier and bias it to optimum by only observing the THD at the output, I can get the same 2 W, but now with less than 1 % THD. No NFB is used.
The circuit is basically equal to what I presented at the post 20 earlier, but with different resistor values.
At the triode, Ra = 47 k, Rk = 2.03k (adjusted to optimum), Ua = 147 V.
And all these results are with real circuits, no LTspice simulations.
I have to say that the amount of distortion cancelling is amazing in this case.
Last edited:
Not a myth, but real. The issue is whether it is large enough when compared with the inherent higher order nonlinearity. In many cases not. Re-entrant distortion gets big as nonlinearity gets big.
Not a myth, just a poor way to put it. NFB does trade off gross non-linearities for low level, high order, harmonic distortion. Crowhurst demonstrated this both mathematically and empirically. This may explain why NFB trades off sound stage for the reduction of nasty, dissonant harmonics that quickly lead to listener fatigue, or that takes the "edge" off the sound.
This is just another way to say: "Fix your open loop design before applying the NFB".
Not in the least convincing. Specific cases have been chosen to illustrate what they want to show rather than illuminate the truth. They make no attempt to explain why the NFB suppresses ALL harmonics in the class A tube push pull case.
It is simply a comparison of different designs not an exploration of the properties of NFB in general.
I am with Bruno on this one:
http://www.edn.com/design/consumer/4420162/3/Negative-feedback-in-audio-amplifiers--Why-there-is-no-such-thing-as-too-much--Part-2-
Cheers
Ian
Last edited:
I've witnessed the positive results brought about by using plenty of feedback but I feel that the question whether feedback will ultimately make all amplifiers sound the same, and the implication that this might mean something doesn't take into account the damping characteristic of the amplifiers actual bandpass response (as with a slow roll-off woofer vs. an in band sharp one), in cases where large amounts of feedback have been required.
On the other hand unless large amounts of NFB is applied, it will be difficult to produce a linear feedback factor across the spectrum unless the open loop bandwidth is sufficient, creating an unwanted link between timbre and tone.
An SE triode based amp provides a relatively tuning free and stable way of producing a useable distortion character.. which is probably desirable for any amp before feedback is applied. IME applying a little feedback to these can bring a reduction in Zo, and not as much regarding noticeable distortion changes.
On the other hand unless large amounts of NFB is applied, it will be difficult to produce a linear feedback factor across the spectrum unless the open loop bandwidth is sufficient, creating an unwanted link between timbre and tone.
An SE triode based amp provides a relatively tuning free and stable way of producing a useable distortion character.. which is probably desirable for any amp before feedback is applied. IME applying a little feedback to these can bring a reduction in Zo, and not as much regarding noticeable distortion changes.
Fair enough. I've had different luck 😉.. ie. the last SE amp I built responded well to just -5dB feedback. The last PP amp dialled in when given at least -12dB.
The word "comparable" in my post 55 is crucial.
'Dialling' is something I used to do with my telephone.
'Dialling' is something I used to do with my telephone.
Just reviving this thread to see if there is any more data available.
One source recommends 8mA at 160v, and this has been put to me by users.
But the OP has different figures - his lowest distortion looks like these:
12mA, -8v at 278v = THD 0.1, diss. 3.34W
10mA, -8v at 265v = THD 0.11, diss. 2.62W
8mA, -10v at 295v = THD 0.1, diss. 2.36W
8mA, -8v at 248v = THD 0.11, diss. 1.98W
8mA, -6v at 204v = THD 0.13, diss. 1.63W
8mA, -4v at 160v = THD 0.18, diss. 1.28W
The 6SN7 data, however, states each section has a max dissipation of 2.5W.
So looking at the above I think I'd go for 8mA, -8v, 248v so not to go too hot.
One source recommends 8mA at 160v, and this has been put to me by users.
But the OP has different figures - his lowest distortion looks like these:
12mA, -8v at 278v = THD 0.1, diss. 3.34W
10mA, -8v at 265v = THD 0.11, diss. 2.62W
8mA, -10v at 295v = THD 0.1, diss. 2.36W
8mA, -8v at 248v = THD 0.11, diss. 1.98W
8mA, -6v at 204v = THD 0.13, diss. 1.63W
8mA, -4v at 160v = THD 0.18, diss. 1.28W
The 6SN7 data, however, states each section has a max dissipation of 2.5W.
So looking at the above I think I'd go for 8mA, -8v, 248v so not to go too hot.
Last edited:
Going off of the late GE 6SN7GTB datasheet 7.5 watt combined dissipation is acceptable-
http://www.radiostation.ru/tubes/6SN7.pdf
Personally the 8mA, -8v at 248v or 8mA, -10v at 295v operating point would seems like a good safe spot to run them even on earlier tubes. 8mA at 250 volts has been in my notes as my personal "sweet spot" for years. With a nice high supply voltage and a degenerated cathode it's a decent way to run them. Makes a bog standard 1K ohm cathode resistor an easy solution too.
http://www.radiostation.ru/tubes/6SN7.pdf
Personally the 8mA, -8v at 248v or 8mA, -10v at 295v operating point would seems like a good safe spot to run them even on earlier tubes. 8mA at 250 volts has been in my notes as my personal "sweet spot" for years. With a nice high supply voltage and a degenerated cathode it's a decent way to run them. Makes a bog standard 1K ohm cathode resistor an easy solution too.
You can spend many happy hours mucking about simulating, tweaking and testing design's when in reality all your efforts to reduce THD by 0.1 % are inaudible.
If you want an ultra low distortion amplifier, use transistors, loads of open loop gain and slap on loads of NFB; the classic straight wire with gain. Personally I build amps with valves because they are aesthetically pleasing to the eye and because a valve amp has a sonic signature, IE distortion and a certain frequency response probably caused mostly by the OPT.
If you want an ultra low distortion amplifier, use transistors, loads of open loop gain and slap on loads of NFB; the classic straight wire with gain. Personally I build amps with valves because they are aesthetically pleasing to the eye and because a valve amp has a sonic signature, IE distortion and a certain frequency response probably caused mostly by the OPT.