Michael,
the trouble is that these guys probably never built an amp with resonably low distortion components before applying NFB. As usually, one must get through on his own to be able to assess.
Seems your are not that busy today? 😀
Pavel
the trouble is that these guys probably never built an amp with resonably low distortion components before applying NFB. As usually, one must get through on his own to be able to assess.
Seems your are not that busy today? 😀
Pavel
On the original premise.......
Best sound - either very little feedback, or heaps of it >60dB.
More fb factor means more lag compensation to bring the OLG to below unity by the pole frequency. This can be bad for sonics.
Less fb factor means less lag comp, and has the further advantage of linearising the amp in OL. However, bandwidth can suffer.
There is a balance between more and less fb factor, and their corresponding lag compensation schemas.
The active components must also be fast, and more fb generally means they must be faster still to permit clean correction without formation of higher order IM products. You can't have too fast a VAS; though more correctly perhaps, the interelectrode capacitances must be kept very low because the depletion capacitance in common emitter is highly non linear. If depletion capacitance is kept low, it is swamped by the very linear lag comp, and the sonics are better.
VAS and front end have more influence on sonics than most realise. Component choice is vital.....
Incidentally, Pavel, a one man race is the only valid kind. In this case you prove something only to yourself - the only legitimate judge. Progress depends on unreasonable men......
Cheers,
Hugh
Best sound - either very little feedback, or heaps of it >60dB.
More fb factor means more lag compensation to bring the OLG to below unity by the pole frequency. This can be bad for sonics.
Less fb factor means less lag comp, and has the further advantage of linearising the amp in OL. However, bandwidth can suffer.
There is a balance between more and less fb factor, and their corresponding lag compensation schemas.
The active components must also be fast, and more fb generally means they must be faster still to permit clean correction without formation of higher order IM products. You can't have too fast a VAS; though more correctly perhaps, the interelectrode capacitances must be kept very low because the depletion capacitance in common emitter is highly non linear. If depletion capacitance is kept low, it is swamped by the very linear lag comp, and the sonics are better.
VAS and front end have more influence on sonics than most realise. Component choice is vital.....
Incidentally, Pavel, a one man race is the only valid kind. In this case you prove something only to yourself - the only legitimate judge. Progress depends on unreasonable men......
Cheers,
Hugh
Hugh,
your post might be valid for a particular circuit design topology. IMO it is not valid in general.
your post might be valid for a particular circuit design topology. IMO it is not valid in general.
Hi, PMA,
Who do you mean ?
the trouble is that these guys probably never built an amp with resonably low distortion components before applying NFB.
Who do you mean ?
PMA said:Michael,
the trouble is that these guys probably never built an amp with resonably low distortion components before applying NFB.
Pavel
Would you mind telling us how?
jcx said:
I can think of only 3 ways to linearize a gain stage (which have already been mentioned here and Ed Cherry has spelled out several times)
Increase the bias to signal ratio - use less of the nonlinear gain curve
Cancellation – really only useful with diff pair and push-pull emitter followers, and even order distortions
Negative Feedback – local feedback as in degeneration, emiiter/source followers or larger feedback loops including more gain stages
Negative feedback, applied locally is the most common method of building a low distortion amp – before adding global feedback
So then the question is how much gain and where to apply the negative feedback
There is a very good case for global feedback giving the largest reduction in distortion with a given total available device gain
In a unilateral gain linear system it can be shown that any loop characteristic from local feedback loops can be matched by some global feedback – several of these conditions don’t apply in real amplifiers so the “optimum” assignment of local and global feedback in a real amplifier depend on amplifying device inherent nonlinearities, circuit topology, signal and load properties
One-liners about linearizing a amp “before applying feedback” aren’t very useful
engineering design is about balancing tradeoffs and requires understanding what the options and consquences are - the main choice is about where to apply feedback to meet some desired amplifier peformance within some power, cost, complexity, ect
Hugh Dean wrote...
What transistor did you choose for your VAS?VAS and front end have more influence on sonics than most realise. Component choice is vital.....
Hi Bam,
Toshiba 2SC3423...... 150V 50mA 200MHz Cob 1.8pF 5W .
Very good device, inexpensive too.
Cheers,
Hugh
Toshiba 2SC3423...... 150V 50mA 200MHz Cob 1.8pF 5W .
Very good device, inexpensive too.
Cheers,
Hugh
AKSA said:
Toshiba 2SC3423...... 150V 50mA 200MHz Cob 1.8pF 5W .
Note.
Thanks, Mike.
Crossover distortion: - use feedback, bias out or use class A ?
Hi,
Crossover seems like a real pain! (It sounds horrible and is worse at HF)
Does biasing out crossover distortion reduction require a scope ?
Would it be possible to pulse the output transistors of a class A amplifier at a high frequency so to reduce heat dissipated ?
MikeB said:Yes, the harmonics created by crossover distortion are way too high frequency to be really compensated/reduced by global feedback. (especially for trebles)
Mike
Hi,
Crossover seems like a real pain! (It sounds horrible and is worse at HF)
Does biasing out crossover distortion reduction require a scope ?
Would it be possible to pulse the output transistors of a class A amplifier at a high frequency so to reduce heat dissipated ?
Hi, Ash Dac,
Nice idea 😀 But how to get a cct that only operates at classA at only high frequency?
The open loop gain is descending like a downhill towards higher frequency. This maybe simpler than your request. If we can get more OL gain at higher frequency, flat, not descending downhill, (but stable offcourse), it will be nice. The feedback will be working as effective good in LF and HF. The amp will sound nice 😀
Nice idea 😀 But how to get a cct that only operates at classA at only high frequency?
The open loop gain is descending like a downhill towards higher frequency. This maybe simpler than your request. If we can get more OL gain at higher frequency, flat, not descending downhill, (but stable offcourse), it will be nice. The feedback will be working as effective good in LF and HF. The amp will sound nice 😀
ash_dac said:
Pre-amplifier stage - Power amplifier stage:
-Input stage -Driver stage -Power stage
From what I understand the input stage will be bandwidth limited
with feedback often applied from power to the inverted input stage.
The feedback is bandwidth limited to prevent oscillation of the power amplifier stage.
The power stage should be linear in operation.
Does this mean that feedback in amplifier is ultimately the driver of sound quality ?
.
jcx said:1. global negative feedback does change important parameters of even “perfectly linear” amplifiers in a desirable direction
2. there is a feedback “error” signal even without nonlinear distortion in the amplifier gain
3. negative feedback trades excess open loop gain of a amplifier
for reduced variation of closed loop parameters such as gain, frequency response and output impedance
at the price of a lower closed loop gain
defined by the much more linear and stable feedback network
I can think of only 3 ways to linearize a gain stage
A 🙂 Increase the bias to signal ratio
- use less of the nonlinear gain curve
B 🙂 Cancellation
– really only useful with diff pair
and push-pull emitter followers,
and even order distortions
C 🙂 Negative Feedback
– local feedback as in degeneration, emiiter/source followers
or larger feedback loops including more gain stages
😎 The differential pair input with degeneration can be linear enough for any practical application
– a subtlety of global negative feedback is that forward loop gain directly reduces the signal at the diff pair input,
increasing the bias to signal ratio by decreasing the signal
and further “linearizing” the diff pair as a consequence of large negative feedback
Cheers!
hi
ash_dac interesting and challenging question:
Feedback in amplifier is ultimately the driver of sound quality?
Not in my lineup book it is,
the ultimate way to get good sound.
Even if reduce distortion a bit, using NFB, this can make your sound less jolly nice.
And so destroy and take away some positive things,
from your amplifier Output Signal..
🙂 But also in my book,
there is no MORE EFFECTIVE way,
than Global Feedback,
to Give an amplifier Hi Fi Quality.
If we by Such Quality mean:
... OUTPUT signal is as close as possible a PERFECT CLONE
... of the INPUT signal
jcx replied, giving his view on this thing called 'Negative Feedback':
I can think of only 3 ways to linearize a gain stage
jcx 😎
there has to be more ways than 3 !!!!
to linearize signal handling in an audio circuit and amplifiers ..
Anybody?
regards, lineup
jcx said:You don’t need credentials to show that Rod’s wrong on this – think of a simulation as a type of mathematical proof available for all to use, view, critique:
An externally hosted image should be here but it was not working when we last tested it.
Green trace is the output of a “pure square law” device B1, properly biased to get a single ended output
(yellow is on top of green at 1KHz in the fft, B1 circuit is inverting in the wave plot, B2 is inside a +1 unity gain feedback loop, input signals adjusted for identical output level)
The second harmonic of the no feedback design is entirely predictable:
(sin(wt))^2 = (1-cos(2wt))/2
If you increase the forward gain in the feedback circuit you'll see the distortion decrease. Never do any new higher order products appear, or increase as you increase the gain.
🙄 If you closely study feedback, you will realize that feedback will produce harmonics in addition to the pure 2nd harmonic that was created by the square law device.
This is simply because the amount of feedback is modulated by the distortion as this distortion changes openloop gain...
This is nothing mystique, just mathematic.
If feedback is high enough and/or openloop distortion low enough, these new harmonics are very low in amplitude.
Mike
This is simply because the amount of feedback is modulated by the distortion as this distortion changes openloop gain...
This is nothing mystique, just mathematic.
If feedback is high enough and/or openloop distortion low enough, these new harmonics are very low in amplitude.
Mike
MikeB said:🙄 If you closely study feedback, you will realize that feedback will produce harmonics in addition to the pure 2nd harmonic that was created by the square law device.
This is simply because the amount of feedback is modulated by the distortion as this distortion changes openloop gain...
This is nothing mystique, just mathematic.
If feedback is high enough and/or openloop distortion low enough, these new harmonics are very low in amplitude.
Mike
Once you pay the price and get yourself the feedback, increasing the feedback always reduces the distortion, provided that you don't turn your square law device into some 14th law device by cascading too many stages.
Yes, it will always reduce the original harmonic, but introduce new ones.
Assuming you have an amp generating 2nd harmonic only open loop, applying feedback will reduce the 2nd harmonic and at the same time generate new harmonics. You see exactly this behavior in jcx's sim.
If feedback is high enough and open loop distortion low enough, these new harmonics are well below noise floor and are no threat.
This is because of the nature of feedback.
The formula for closed loop gain is:
Cg = K / (1 + K/Og)
where: K = feedback net factor, Og = Open loop gain (Which also contains the distortion)
You see, the higher Og, the more Cg approaches K, which would be linear.
But, for low Og, you transfer the original transfer function into a new one, with very low feedback you can even increase thd.
Mike
Assuming you have an amp generating 2nd harmonic only open loop, applying feedback will reduce the 2nd harmonic and at the same time generate new harmonics. You see exactly this behavior in jcx's sim.
If feedback is high enough and open loop distortion low enough, these new harmonics are well below noise floor and are no threat.
This is because of the nature of feedback.
The formula for closed loop gain is:
Cg = K / (1 + K/Og)
where: K = feedback net factor, Og = Open loop gain (Which also contains the distortion)
You see, the higher Og, the more Cg approaches K, which would be linear.
But, for low Og, you transfer the original transfer function into a new one, with very low feedback you can even increase thd.
Mike
traderbam said:Yes.
Blue is spectrum of 20kHz + 23kHz signals through a perfect cubed function, y = X^3.
Red is spectrum when a modest amount of NFB is applied.
The worse your non-linearity the more gain you need to fix it, but enough gain will fix it. It's when that gain is poorly implemented (as in y = x^3, that feedback gets a bad rap.
Unfortunately in the world of discrete electronics we don't have access to super-beta's like an op-amp designer has.
It's not that simple, feedback can't fix, just attenuate. You can't increase feedback as you need, your amp would get unstable.
To keep a feedback amp from oscillating, you have to decrease feedback with increasing frequency. But, this decrease with frequency gives you the problem that the higher order harmonics generated by the feedback itself will not get attenuated to acceptable levels.
If you look at traderbams fft, feedback will not be able to reduce the harmonics generated above 100khz.
You see, whether you use feedback or not, the first step in amp design is to reduce open loop distortion.
Mike
To keep a feedback amp from oscillating, you have to decrease feedback with increasing frequency. But, this decrease with frequency gives you the problem that the higher order harmonics generated by the feedback itself will not get attenuated to acceptable levels.
If you look at traderbams fft, feedback will not be able to reduce the harmonics generated above 100khz.
You see, whether you use feedback or not, the first step in amp design is to reduce open loop distortion.
Mike
fizzard wrote:
This may seem bizarre but you have to consider both the mathematics of a feedback system and the way our brains hear things. NFB reduces the total distortion but is the total distortion the only factor in good sound...or even the most important factor?
I think the original debate was about Rod Elliot's claim that NFB does not introduce new frequencies. This is incorrect EXCEPT in a linear system. As MikeB implies NFB will reduce its own splatter as it is increased but only up to a point where it becomes unstable.
I have spent a lot of time listening to circuits. For me, the "bad rap" is because NFB generally makes a circuit sound worse. By worse I mean it sucks the life out of the music. Which is counter-intuitive, especially if you have had years of education about the benefits of NFB in linear systems. In a practical circuit, which is not linear, special conditions are required to get the benefits without sacrificing the sonic performance.The worse your non-linearity the more gain you need to fix it, but enough gain will fix it. It's when that gain is poorly implemented (as in y = x^3, that feedback gets a bad rap.
Unfortunately in the world of discrete electronics we don't have access to super-beta's like an op-amp designer has.
This may seem bizarre but you have to consider both the mathematics of a feedback system and the way our brains hear things. NFB reduces the total distortion but is the total distortion the only factor in good sound...or even the most important factor?
I think the original debate was about Rod Elliot's claim that NFB does not introduce new frequencies. This is incorrect EXCEPT in a linear system. As MikeB implies NFB will reduce its own splatter as it is increased but only up to a point where it becomes unstable.
- Status
- Not open for further replies.
- Home
- Amplifiers
- Solid State
- Feedback in amplifier is ultimately the driver of sound quality ?