NFB net gain or loss
- Thread starter rick57
- Start date
rick57 said:
Feedback helps to lower amp non-linearities/ distortion, and therefore it is often used. If you use a lot (strong feedback), you have to be careful not to get instabilities like ringing or oscillations.
But, you can only use feedback if the open loop (no feedback) gain of the amp is higher than the gain you want to have. The difference, called sometimes the excess gain, can be turned into feedback. Very roughly, the decrease in non-linearity is about equal to the excess gain available for feedback.
Solid state amps generally have a very high open loop gain, thus a lot of excess gain that can be turned into feedback, and the measured non-linearities are often very low in SS compared to tube amps. Tube amps have generally less excess gain, therefore less curative feedback and higher measured non-linearities.
Despite the higher measured non-linearity, many people prefer tube amp sound. Most probably one factor involved is that the specific type of non-linear distortion generated by tube amps gives a tube amp a special 'warm' sound. SS amps don't have this effect or at least much less so people often say they sound 'cold' or 'analytical', but it is actually a more transparent rendition of the music from the source. Preferences here are all about taste of course.
The higher the excess gain used for feedback, the higher the risk for instabilities. So, this risk is lower in tube amps, although the phaseshift resulting from an output transformer aggain increases the risk. But, generally, instability is mainly an issue in SS amps.
So, to your question: It doesn't really make things that much harder, it is all well established technology. The cost in components is trivial. That people don't do it surely is not because it is hard, but maybe because they like the non-feedback sound, or, there is so little excess gain that it doesn't make a heck of a difference anyway.
Jan Didden
EDit: Colt45's post came in while I was typing, but I see his point and I think it illustrates what I was trying to say also.
Probably the best advice I've heard on this subject is, "Get your amp performing well without negative feedback first, and then to apply as little NFB as you can get away with."
Probably the most important criterion for most people is damping for the speaker. For example, with a pentode output stage, you will need quite a lot of NFB to get a reasonable damping factor. With a distributed load arrangement (e.g. ultralinear), you will still need NFB but rather less than with straight pentode connection. With a triode OP stage, it's debatable whether you need any NFB at all. However, I find triode stages produce rather boomy bass with the speakers I have, so I apply a modest amount of NFB (say 6 to 10dB).
Other things that NFB can do (including making the amp's gain less device dependent, flattening the frequency response and reducing distortion) have to be considered too. They may or may not make a significant difference, depending on the design and devices used in an individual amp. H.J. Leak claimed, in the early 1950s, that the open loop gain of an amp should be very high, so that a large amount of NFB (at least 26dB) could be used and THD could be reduced to a maximum of 0.1%. His amps typically had an input sensitivity of 150mv for full output, and that was after feedback was added!
Not many people knew enough to argue with Harry Leak in those days but the popular view today is different. NFB is seen as something to be used in moderation and with caution. Excessive use of NFB can lead to instability and some claim that it can make an amp sound "stifled". The type of feedback is also deemed important, local feedback being generally considered "benign" and "safe", while global NFB (such as Leak and most of his contemporaries advocated) is considered "risky".
Probably the most important criterion for most people is damping for the speaker. For example, with a pentode output stage, you will need quite a lot of NFB to get a reasonable damping factor. With a distributed load arrangement (e.g. ultralinear), you will still need NFB but rather less than with straight pentode connection. With a triode OP stage, it's debatable whether you need any NFB at all. However, I find triode stages produce rather boomy bass with the speakers I have, so I apply a modest amount of NFB (say 6 to 10dB).
Other things that NFB can do (including making the amp's gain less device dependent, flattening the frequency response and reducing distortion) have to be considered too. They may or may not make a significant difference, depending on the design and devices used in an individual amp. H.J. Leak claimed, in the early 1950s, that the open loop gain of an amp should be very high, so that a large amount of NFB (at least 26dB) could be used and THD could be reduced to a maximum of 0.1%. His amps typically had an input sensitivity of 150mv for full output, and that was after feedback was added!
Not many people knew enough to argue with Harry Leak in those days but the popular view today is different. NFB is seen as something to be used in moderation and with caution. Excessive use of NFB can lead to instability and some claim that it can make an amp sound "stifled". The type of feedback is also deemed important, local feedback being generally considered "benign" and "safe", while global NFB (such as Leak and most of his contemporaries advocated) is considered "risky".
You might not want it...
As other posters have said, the main need for feedback is to lower the output resistance of the amplifier sufficiently to allow the loudspeaker to achieve the bass response that the designer intended (and which assumes zero output resistance). In practice, because the DC resistance of the voice coil is in series with the amplifier, the difference in the loudspeaker's bass response between having an amplifier with zero output resistance and 1 Ohm output resistance is quite small.
As it happens, yesterday, I applied 19dB of feedback to a PP 2A3 amplifier and at the onset of output stage grid current it reduced distortion from 0.9% to 0.4%. I didn't think that was very good value for money! And yes, the amplifier was carefully compensated for best 10kHz square wave response.
As other posters have said, the main need for feedback is to lower the output resistance of the amplifier sufficiently to allow the loudspeaker to achieve the bass response that the designer intended (and which assumes zero output resistance). In practice, because the DC resistance of the voice coil is in series with the amplifier, the difference in the loudspeaker's bass response between having an amplifier with zero output resistance and 1 Ohm output resistance is quite small.
As it happens, yesterday, I applied 19dB of feedback to a PP 2A3 amplifier and at the onset of output stage grid current it reduced distortion from 0.9% to 0.4%. I didn't think that was very good value for money! And yes, the amplifier was carefully compensated for best 10kHz square wave response.
All that being said, there are many ways to apply feedback.
With tube amps, since direct coupling tends to be much more difficult than with SS, and you (may) be wnclosing an output transformer inside a feedback loop, you are more likely to run into LF instability (motorboating) than HF, i.e. you are unlikely to get outright oscillation at HF, but likely to get ringing on square waves.
Fortunately, there are ways to avoid this problem by applying 'shorter' than overal NFB - for instance, partial FB from output tubes to driver before transformer. Also, there are many versions of local feedback (cathode degeneration, cathode coupling on outputs, and many more) and/or nested feedback techniques which reduce the need for overal NFB in the first place, perhaps to the point where you may not need it at all. So, the answer is not simple. If your amp is made linear before the application of overal feedback, the net gain in applying it will likely be an exercise in diminishing returns.
With tube amps, since direct coupling tends to be much more difficult than with SS, and you (may) be wnclosing an output transformer inside a feedback loop, you are more likely to run into LF instability (motorboating) than HF, i.e. you are unlikely to get outright oscillation at HF, but likely to get ringing on square waves.
Fortunately, there are ways to avoid this problem by applying 'shorter' than overal NFB - for instance, partial FB from output tubes to driver before transformer. Also, there are many versions of local feedback (cathode degeneration, cathode coupling on outputs, and many more) and/or nested feedback techniques which reduce the need for overal NFB in the first place, perhaps to the point where you may not need it at all. So, the answer is not simple. If your amp is made linear before the application of overal feedback, the net gain in applying it will likely be an exercise in diminishing returns.
Well, I think everything was said about NFB, except just a little thing about pulse response and intermodulation distortion. A guy from Fairchild indroced the relationship between NFB amount and the sound from SS amplifier. He was Hans Palouda and presented an application for a five transistor amplifier with a very very nice sound. He stated that excessive nfb in SS amplifiers where the cause of a kind of distortion that occurred when a high intensity bass bulky pulse where to be processed. I tried out that SS amp and I found it was really thrue. As everyone stated here tube amps uses no or fairly low NFB, this is another explanation of why they operate better than most SS amplifiers. As it is said, fairly low NFB helps to reduce gross distortion, flattens response band, lowers output impedence and do not influence transient distorsion. What it will work in a PP amp with UL local NFB I will say after the amp I'm building get finished.
Cheers
Larry.
Cheers
Larry.
Thanks for the range of excellent perspectives! I will experiment in local moderation.
The only thing that doesn’t click with me - EC8010 – when
“I applied 19dB of feedback to a PP 2A3 amplifier and at the onset of output stage grid current it reduced distortion from 0.9% to 0.4% - I didn't think that was very good value for money!”
It’s not a massive reduction, but if you say not good value, was it expensive??
Larry or anyone, can you add to EC8010’s experience with local NFB in a PP amp?
Where/ how specifically did you do it/ did it sound better?
The only thing that doesn’t click with me - EC8010 – when
“I applied 19dB of feedback to a PP 2A3 amplifier and at the onset of output stage grid current it reduced distortion from 0.9% to 0.4% - I didn't think that was very good value for money!”
It’s not a massive reduction, but if you say not good value, was it expensive??
Larry or anyone, can you add to EC8010’s experience with local NFB in a PP amp?
Where/ how specifically did you do it/ did it sound better?
If I apply 19dB of feedback, I expect to see a reduction in distortion of 19dB, so 7dB reduction was poor value for money.
Did the feedback actually cost money? Depends on how you look at it. One resistor and two capacitors were required, so that hardly broke the bank. On the other hand, the amplifier was deliberately built to have sufficient gain in hand to allow feedback, and that certainly did cost money because an extra stage was required.
Did the feedback actually cost money? Depends on how you look at it. One resistor and two capacitors were required, so that hardly broke the bank. On the other hand, the amplifier was deliberately built to have sufficient gain in hand to allow feedback, and that certainly did cost money because an extra stage was required.
Modern speakers and pentodes want NFB.
Speakers are tuned for high damping. Nowadays, nominally zero source impedance, or DF>>40. If they are tuned flat with high DF, and you give zero DF, the bass resonance will rise 10dB or 20dB.
Even guitar amps, where the speaker can be selected for best spectral balance on a hi-Z amplifier, often use about 6db NFB for DF near 1, instead of DF near 0.2 as a naked pentode gives.
Triodes give DF like 2 to 5 without any extra NFB. Rise at bass resonance is only like 3dB. While not "perfect", it usually vanishes in real-room bass nodes.
Ultralinear splits the difference.
Yes, you can enjoy a speaker without good damping, especially if you design the speaker that way. Some of the great consoles of the past ran very low damping with speakers and cabinets adjusted for pleasant bass response.
The problem with NFB and tubes is that: in many practical situations you don't have much excess gain, and you have so many coupling caps and the output transformer that you can't stablize more than 10dB-20dB NFB without crippling the amp or your wallet. That's about enough to damp a speaker, but small amounts of NFB increase IM distortion and high-order distortion.
> Solid state amps generally have a very high open loop gain, thus a lot of excess gain that can be turned into feedback, and the measured non-linearities are often very low in SS compared to tube amps.
The nonlinearity of any simple BJT stage without NFB is huge compared to any tube. But so is the gain. And you can often reduce coupling rolloffs enough to run NFB with stability. They may hide it, but it is very unusual to find a truly no-NFB BJT amp. (FETs are a lot like tubes for gain and linearity, but can often use BJT-like NFB.)
> I applied 19dB of feedback to a PP 2A3 amplifier and at the onset of output stage grid current it reduced distortion from 0.9% to 0.4%.
NFB will actually increase THD above the clipping point.
What was the change in THD at say 3dB below clipping? Just guessing: 0.5% no-NFB, <0.1% with 19dB NFB.
Of course this raises a different question: do we want THD to stay low up to a point, and then rise like a brick wall? Or a more gentle rise through and past the gross-distortion point? And of course that depends on our needs: if we never need to touch clipping, NFB will give lowest distortion in our operating zone; if we like to crank it up until it comes out bent, the gentle rise and low slope of no-NFB may be less obnoxious. (In college, with low efficiency speakers, I cranked the heck out of a 25 watt amp to get 50 watt sound. Decades later at home, I built a similar low-NF amp of 40 watts but with a high efficiency speaker, and just never felt inspired to crank it past the limit.)
Speakers are tuned for high damping. Nowadays, nominally zero source impedance, or DF>>40. If they are tuned flat with high DF, and you give zero DF, the bass resonance will rise 10dB or 20dB.
Even guitar amps, where the speaker can be selected for best spectral balance on a hi-Z amplifier, often use about 6db NFB for DF near 1, instead of DF near 0.2 as a naked pentode gives.
Triodes give DF like 2 to 5 without any extra NFB. Rise at bass resonance is only like 3dB. While not "perfect", it usually vanishes in real-room bass nodes.
Ultralinear splits the difference.
Yes, you can enjoy a speaker without good damping, especially if you design the speaker that way. Some of the great consoles of the past ran very low damping with speakers and cabinets adjusted for pleasant bass response.
The problem with NFB and tubes is that: in many practical situations you don't have much excess gain, and you have so many coupling caps and the output transformer that you can't stablize more than 10dB-20dB NFB without crippling the amp or your wallet. That's about enough to damp a speaker, but small amounts of NFB increase IM distortion and high-order distortion.
> Solid state amps generally have a very high open loop gain, thus a lot of excess gain that can be turned into feedback, and the measured non-linearities are often very low in SS compared to tube amps.
The nonlinearity of any simple BJT stage without NFB is huge compared to any tube. But so is the gain. And you can often reduce coupling rolloffs enough to run NFB with stability. They may hide it, but it is very unusual to find a truly no-NFB BJT amp. (FETs are a lot like tubes for gain and linearity, but can often use BJT-like NFB.)
> I applied 19dB of feedback to a PP 2A3 amplifier and at the onset of output stage grid current it reduced distortion from 0.9% to 0.4%.
NFB will actually increase THD above the clipping point.
What was the change in THD at say 3dB below clipping? Just guessing: 0.5% no-NFB, <0.1% with 19dB NFB.
Of course this raises a different question: do we want THD to stay low up to a point, and then rise like a brick wall? Or a more gentle rise through and past the gross-distortion point? And of course that depends on our needs: if we never need to touch clipping, NFB will give lowest distortion in our operating zone; if we like to crank it up until it comes out bent, the gentle rise and low slope of no-NFB may be less obnoxious. (In college, with low efficiency speakers, I cranked the heck out of a 25 watt amp to get 50 watt sound. Decades later at home, I built a similar low-NF amp of 40 watts but with a high efficiency speaker, and just never felt inspired to crank it past the limit.)
ray_moth said:
EC8010 said:
PRR said:
It's nice to be reminded that a music system must be considered as a whole. If you make your own amps and speakers, you have more degrees of freedom. You can adjust system Q with the amp or speakers. My main amps have an output impedence of 6ohms, which would be considered poor by most current standards. But on my speakers the bass is tight and solid, not underdamped in the least. If you bi or tri amp, you have even more ways to tune.
Sheldon
No discussion of damping factor should be without this analysis by Dick Pierce:
http://www.diyspeakers.net/Articles/Richard Pierce DAMPING FACTOR.pdf
For those unfamiliar with Mr. Pierce he's the antithesis and often the Usenet bane of audio tweakdom and spent many years as an industry consultant. He considers DF above 10 as gravy. I believe that's still out of reach of most non-NFB tube designs, at least SE.
http://www.diyspeakers.net/Articles/Richard Pierce DAMPING FACTOR.pdf
For those unfamiliar with Mr. Pierce he's the antithesis and often the Usenet bane of audio tweakdom and spent many years as an industry consultant. He considers DF above 10 as gravy. I believe that's still out of reach of most non-NFB tube designs, at least SE.
An iconoclast after my own heart! I'm sure the only reason there is a market for expensive speaker cables and connectors is to preserve the very low output impedance of some amplifiers. After all, there's no point in having an amp with OP impedance < 0.1 ohm if your cables & connectors add a much higher impedance, is there?
The Impedance of your speaker typically rises at low frequencies with a peak at the resonant frequency. It also rises at high frequencies as the inductance of the speaker (tweeter) voice coil becomes more significant.
With a very low damping factor amp the power delivered to the speaker tends to follow the impedance curve of the speaker and the lower end and upper end of the frequency response tends to be exagerated as more power is delivered into the higher impedances.
I have found a damping factor of 4 is more than adequate to balance the power vs frequency issues. In fact I have several amps with damping factors as low as 2 which sound quite glorious. This is speaker dependent. If this damping factor can be achieved without recourse to global NFB then I always choose to do it that way (eg local FB loops, triode connection, ultralinear connection, cathode feedback etc). If I need some global feedback to achieve that balanced frequency response then I have no problems with adding it. I try to keep global NFB to around 6dB maximum BUT have run as high as 15dB with some loss of stereo imaging.
Cheers,
Ian
With a very low damping factor amp the power delivered to the speaker tends to follow the impedance curve of the speaker and the lower end and upper end of the frequency response tends to be exagerated as more power is delivered into the higher impedances.
I have found a damping factor of 4 is more than adequate to balance the power vs frequency issues. In fact I have several amps with damping factors as low as 2 which sound quite glorious. This is speaker dependent. If this damping factor can be achieved without recourse to global NFB then I always choose to do it that way (eg local FB loops, triode connection, ultralinear connection, cathode feedback etc). If I need some global feedback to achieve that balanced frequency response then I have no problems with adding it. I try to keep global NFB to around 6dB maximum BUT have run as high as 15dB with some loss of stereo imaging.
Cheers,
Ian
Old thread, sorry, but still topical.
I'm replying to a post from 2005, but I see that PRR is still here and I hope for a reply.
Suppose we have a standard (let's say Blackface/Silverface Fender style) 2*6L6 guitar amp with the usual NFB loop from the secondary back to the driver.
Are you saying that 6db of NFB can give something like a 5-fold increase in speaker damping compared with the same amp without the NFB?
I need a rough rule of thumb or a single concrete example of the difference in DF between such an amp with and without the feedback loop. For example, see the attached schematic.
Could anyone estimate the DF of that circuit with/without the feedback loop?
Thank you!
I'm replying to a post from 2005, but I see that PRR is still here and I hope for a reply.
Suppose we have a standard (let's say Blackface/Silverface Fender style) 2*6L6 guitar amp with the usual NFB loop from the secondary back to the driver.
Are you saying that 6db of NFB can give something like a 5-fold increase in speaker damping compared with the same amp without the NFB?
I need a rough rule of thumb or a single concrete example of the difference in DF between such an amp with and without the feedback loop. For example, see the attached schematic.
Could anyone estimate the DF of that circuit with/without the feedback loop?
Thank you!
