I'm sticking my neck out here as a beginner in these things.
If slew rate is related to the speed of the amplifier. The faster an amp the higher the ULGF. The higher the ULGF the more available feedback at high frequencies. Therefore there is more THD reduction due to feedback. This leaves the amps original linearity as a factor. This assumes the compensation schemes in each amp is of the same order.
As CFB amps and VFB amps can share the same VAS / Output topologies other than the front end. Can we not boil this down to CFB front ends vs LTP front ends?
Please forgive me if I am completely wrong.
If slew rate is related to the speed of the amplifier. The faster an amp the higher the ULGF. The higher the ULGF the more available feedback at high frequencies. Therefore there is more THD reduction due to feedback. This leaves the amps original linearity as a factor. This assumes the compensation schemes in each amp is of the same order.
As CFB amps and VFB amps can share the same VAS / Output topologies other than the front end. Can we not boil this down to CFB front ends vs LTP front ends?
Please forgive me if I am completely wrong.
I didn't quote you? You weren't the one in a sudden mentioning listening tests?I thought I gave technical reasons why your assumptions were not correct.
Slew rate and ULGF can be varied independently in VFA - you can have very high SR's and still have a ULGF of 1 MHz for example. Typically, you would set ULGF to between 1 and about 1.5 MHz in a VAF. Much higher than this and phase shifts in the output stage make it difficult to make the amp stable. Similarly, the -3dB open loop corner frequency in a VAF can be set to a very low frequncy (few Hz) and still produce a very fast, high SR amplifier. In power amplifiers, the slew rate can be independently set by adjusting the LTP tail current in MC designs; MIC alows even more freedom because the feedback capacitor (which sets the front end ULGF) is driven from the TIS, so the tail current does not set the slew rate.
You can experiment with LTspice to see these relationships and how they work.
Did-you mean this http://www.diyaudio.com/forums/solid-state/193923-simple-symetrical-amplifier-52.html#post3163129 ?
About distortion of LTP in VFB, let see what happens in a simple way.
Let say the amp is open loop gain of 100, a closed loop gain of 1, and the LTP with a gain of 1.
Signal = 'S' and distortion of each side of the LTP= 'd'.
After amplification by the positive input of the LTP, the signal will be S+d.
After the 100X amplification of the following stages (considered as perfect) and feedback bridge attenuation, the feedback signal applied to the - input will be:
0.99 (S+d).
Once subtracted: S+d -0.99(S-d)= 0.01S+0.01d witch is the final signal. Amplified by the gain of the amp. result at the output in closed loop will be: S+d.
This is for the positive input and a perfect feedback path.
Now, let see what happens with the distortion added by the inverting input.
With a signal 'T'. similar to S.
The inverting input invert the previous numbers:
-(d- 0.99d)= -0.01d added at the collector.
So the final result is: S+d-(0.99(S+d))=0.01S+0.01d-0.01d = S
DS was right on this point, and i was wrong. i have to apologize. Shame on me.
I recall a short discussion in one of my threads where much higher ULGFs were discussed. 40Mhz was seen as a reasonable upper limit. Self's book was used as a reference. Is this only true for CFAs? I'm thinking 1Mhz ULGF is a little low.
I understand the idea of "excess phase" from the output stage. But would this not be consistent for both VFA and CFA amps?
My understanding is that CFB front ends can drive much higher currents than LTPs and this would account for the higher slew rates for a given miller compensation scheme.
I will experiment in spice as you suggest to see the relationships you mention above. They challenge my belief system so need to be investigated.
I understand the idea of "excess phase" from the output stage. But would this not be consistent for both VFA and CFA amps?
My understanding is that CFB front ends can drive much higher currents than LTPs and this would account for the higher slew rates for a given miller compensation scheme.
I will experiment in spice as you suggest to see the relationships you mention above. They challenge my belief system so need to be investigated.
Please excuse my comment for being incomplete. I was ridiculed at diyaudioproject.eu, not on diyaudio.rs.
You have the same right to your opinion as I have to mine.
I wonder then why some of the very old members, such as Sipi, have left the forum, not to even mention several very good people who know of it, but are not interested in joining it.
Why the management allows some people to behave in a most uncivilized way is a mystery to me.
Ultimately, this is to the detrement of quite a few members of the silent majority, who are probably quite nice people, but it's the bad boys who are always the loudest. And the best organized.
And this is all I have to say about it, I won't post anything else on this subject.
Yes, the phase shift in the output stage applies to both CFA and VFA designs - no way to get around that. Clever compensation techniques in VFA like TPC and TMC allow the designer to place more feedback around the amplifier at HF (TPC) or to transition the output stage out of the feedback loop at HF (TMC) in order to reduce HF distortion. In CFA's, the LG tends to be flatter so that although at LF its lower, you often find that at HF is much higher and the result can be lower distortion.
However, in both cases, once you get beyond a few MHz, besides the phase shift issues, you start to see layout parasitics become an issue - 10nH per cm is a good rule of thumb. So, if you want to use mosfets and close your loop at much more than 2 or 3 MHz, you better be prepared to simulate and mitigate all those parastitics, because they become the limiting factor very qucikly.
thank you, Bonsai.
oh &*%$. (At least I've done parasistic simulation.)Just to confirm you get the ULGF from the closed loop response?
But, at least in simulation you can use TMC with some shunt compensation on VAS/TIS on CFB amps.
The only person I ever ridicule is myself . It is plenty enough work I find .
Mad Max Pearson it is said .
Seriously . Every opinion carries the truth with it . That is either it is or isn't true . It is for the person reading it to decide . The difficult part is believing something simple experiment would prove untrue , that is doctrine .
Someone wrote of under biasing being better ( Ivan ? ) . Logically it must be . Notice how well a zero bias amp measures below 1 kHz . To make a tough sub-woofer amp zero bias is often used ( a mistake I feel ) . To take an amp just below critical bias up to 30 kHz should be OK . Equally high bias AB is OK . If it can go to 5 watts power levels in class A it might outperform true class A amps ( more dynamic power , less hum modulation ) . FET's hide their problems well , they do not work with the theory just given . They have their own rules . I often set FET's by finger . That's too hot now I guess it's about right . Saxophone solos also , bass power ( e.g Duke Ellington Black Beauty ) .
We had a comedy program called Little Britain . Very bad taste . The unhelpful assistant in the shop always said . Computer says NO . With FET's computer doesn't rule the day .
So what can be the advantage many people recognize to CFB sound reproduction ?
First, the previous analyze of error cancellation in LTP applies in a perfect world, where input devices are identical and add no delays in the servo loop (phase turns at HF).
I never noticed any real audible differences in harmonic distortion between the two topologies.
To try to understand and explore, i had modified, some times ago, the *same amp* in both topologies, VFB and CFB and compare them.
Of course, no way to optimize the two with the same VAS currents (while you can keep them identical if you want). So, i tried a CFA optimized in front of a VFA optimized. But, even in my tests with identical values compromised for the two, CFA was measuring and sounding much better.
It is here : http://www.esperado.fr/creations_audio/le-crescendo-revisite.html
(sorry it is in French)
This mod has been done by several people with the same results. (CFA in,green, VFA in red)
Well. my conclusions are:
CFB need care in the input stage to cancel power supply ripple (and distortion ;-). Good use of FAST CCS.
Once this done, the reduced delays in feedback path before addition to the original signal (due to the speed of the common base topology), the increased bandwidth of the feedback path due to the combination of the low impedance of the feedback with the reduced parasitic capacitance at the emitter of the input stage (compared to the base in VFB) means the feedback servo is a lot faster/better in CFA.
That we can see in this image that i had already published in other threads.
Those images shows signal bandwidths at the input of the VAS. It is the resulting signal of ['input' - 'feedback'] resulting signal in closed loop operation.
And shows the X5 increased flat bandwidth in CFB. (5Khz VS 1Khz)
It shows too that we are far away with those two VERY fast amps, to be fast enough for a constant feedback ratio in all the audio Bandwidth.
About distortion, the optimized CFB was better than the original VFB amp, but who care about those low Harmonic numbers ? Intermodulation is the issue, where CFB showed its amazing superiority.
On a listening point of view, the CFB version (it is my amp now) is more 'liquid', less harsh, better separation between instruments. Basses are more solid and 'fast', kick drums and basses nicely separated, treble seems less present, but more natural and less agressive: you can listen to the metal of cymbals and hit of the stick where it was just 'tshhh' in the VFB amp..
The VFB amp was impressive, but brought listening fatigue: I can listen to the CFA endless, and forget-it.
First, the previous analyze of error cancellation in LTP applies in a perfect world, where input devices are identical and add no delays in the servo loop (phase turns at HF).
I never noticed any real audible differences in harmonic distortion between the two topologies.
To try to understand and explore, i had modified, some times ago, the *same amp* in both topologies, VFB and CFB and compare them.
Of course, no way to optimize the two with the same VAS currents (while you can keep them identical if you want). So, i tried a CFA optimized in front of a VFA optimized. But, even in my tests with identical values compromised for the two, CFA was measuring and sounding much better.
It is here : http://www.esperado.fr/creations_audio/le-crescendo-revisite.html
(sorry it is in French)
This mod has been done by several people with the same results. (CFA in,green, VFA in red)
Well. my conclusions are:
CFB need care in the input stage to cancel power supply ripple (and distortion ;-). Good use of FAST CCS.
Once this done, the reduced delays in feedback path before addition to the original signal (due to the speed of the common base topology), the increased bandwidth of the feedback path due to the combination of the low impedance of the feedback with the reduced parasitic capacitance at the emitter of the input stage (compared to the base in VFB) means the feedback servo is a lot faster/better in CFA.
That we can see in this image that i had already published in other threads.
Those images shows signal bandwidths at the input of the VAS. It is the resulting signal of ['input' - 'feedback'] resulting signal in closed loop operation.
And shows the X5 increased flat bandwidth in CFB. (5Khz VS 1Khz)
It shows too that we are far away with those two VERY fast amps, to be fast enough for a constant feedback ratio in all the audio Bandwidth.
About distortion, the optimized CFB was better than the original VFB amp, but who care about those low Harmonic numbers ? Intermodulation is the issue, where CFB showed its amazing superiority.
On a listening point of view, the CFB version (it is my amp now) is more 'liquid', less harsh, better separation between instruments. Basses are more solid and 'fast', kick drums and basses nicely separated, treble seems less present, but more natural and less agressive: you can listen to the metal of cymbals and hit of the stick where it was just 'tshhh' in the VFB amp..
The VFB amp was impressive, but brought listening fatigue: I can listen to the CFA endless, and forget-it.
Last edited:
thank you, Bonsai.
Just to confirm you get the ULGF from the closed loop response?
But, at least in simulation you can use TMC with some shunt compensation on VAS/TIS on CFB amps.
No - closed loop response and loop response are not the same.
The closed loop response might cross the UGF much higher than loop response.
It's the loop response that determines the amplifiers stability. To measure loop response, , ground the input and insert a voltage source between the output and the feedback resistor. Label the output Vo and the feedback resistor side Vx.
Run a small signal analysis and plot Vo/Vx.
This is a basic technique and not as good as Middlebrook, but it gets you in the ballpark for conventional compensation design.
Forget for one moment CfbP outputs . Look at them at any point of an amplifier . Whenever I have used them it has been with better results than a Darlington . A word of warning that they should be optimized from the point of view of Early effect . That is one transistor doing most of the work . I used one in a capacitance multiplier with good results . VAS also . Why CfbP devices are not sold as a single package I will never know .
For an amp they are superb in class A ( or over bias AB ) . I can see they would be troublesome in B . Would a supersonic triangle-wave help ( sort of dither ) ?
The old Gogny was an amp I admired . It was 50 watts 1 ohm for a ribbon speaker . 1967 LTP + VAS as today . It summed the feedback at the amp input .
For an amp they are superb in class A ( or over bias AB ) . I can see they would be troublesome in B . Would a supersonic triangle-wave help ( sort of dither ) ?
The old Gogny was an amp I admired . It was 50 watts 1 ohm for a ribbon speaker . 1967 LTP + VAS as today . It summed the feedback at the amp input .
GEGO (Gerard Gogny) amp designed for its 'Orthophase' speakers ?
Yes, an OPA works better in inverting configuration, because no added delay in the feedback path before it is mixed with the original signal.
But even here, i found CFA to be better, because the needed high impedance of the input (so the one of the feedback) and the parasitic capacitance of the base making a low pass filter in the feedback operation.
Last edited:
DVV
I expected at least your Public Apology to DIY members from DIY Audio RS site !
BTW ,
Personally I was read many of your`s post here on Big DIY Audio site with special attention ! , but from now Not Anymore !
Same as you I will not post further anything else related to this off topic !
By
Understood. I know the frequency measurement you are talking about.
Was getting mixed up...
Thank you
PS. Still in some for some interesting times but less extreme.
So what can be the advantage many people recognize to CFB sound reproduction ?
First, the previous analyze of error cancellation in LTP applies in a perfect world, where input devices are identical and add no delays in the servo loop (phase turns at HF).
I never noticed any real audible differences in harmonic distortion between the two topologies.
To try to understand and explore, i had modified, some times ago, the *same amp* in both topologies, VFB and CFB and compare them.
Of course, no way to optimize the two with the same VAS currents (while you can keep them identical if you want). So, i tried a CFA optimized in front of a VFA optimized. But, even in my tests with identical values compromised for the two, CFA was measuring and sounding much better.
It is here : http://www.esperado.fr/creations_audio/le-crescendo-revisite.html
(sorry it is in French)
This mod has been done by several people with the same results. (CFA in,green, VFA in red)
Well. my conclusions are:
CFB need care in the input stage to cancel power supply ripple (and distortion ;-). Good use of FAST CCS.
Once this done, the reduced delays in feedback path before addition to the original signal (due to the speed of the common base topology), the increased bandwidth of the feedeback path due to the combination of the low impedance of the feedback with the reduced parasitic capacitance at the emitter of the input stage (compared to the base in VFB) means the feedback servo is a lot faster/better.
That we can see in this image that i had already published in other threads.
Those images shows signal bandwidths at the input of the VAS. It is the resulting signal of 'input' - 'feedback' resulting signal in closed loop operation.
And shows the X5 increased flat bandwidth in CFB. (5Khz VS 1Khz)
It shows too that we are far away with those two VERY fast amps, to be fast enough for a constant feedback ratio in all the audio Bandwidth.
About distortion, the optimized CFB was better than the original VFB amp, but who care about those low Harmonic numbers ? Intermodulation is the issue, where CFB showed its amazing superiority.
On a listening point of view, the CFB version (it is my amp now) is more 'liquid', less harsh, better separation between instruments. Basses are more solid and 'fast', kick drums and basses nicely separated, treble seems less present, but more natural and less agressive: you can listen to the metal of cymbals and hit of the stick where it was just 'tshhh' in the VFB amp..
The VFB amp was impressive, but brought listening fatigue: I can listen to the CFA endless, and forget-it.
Thanks for sharing your experiences and lab results. That's exactly what was needed. Yesterday I listened Modern Jazz Quartet "Last Concert" on my PeeCeeBee PMI version of VSSA and I do not remember that it was ever possible to listen to this disc several times without pause on any VFB amp. (I can imagine how good it will be when I build Bonsai's Ovation nx and full CCS version of VSSA or Fet hex Explendit). I have a torture test that uses Jimmy Smith's "Root Down" CD. First track Sagg Shooting' His Arrow has hi-hat and cymbals mixed so loud and the track time is almost 12 minutes. It is impossible to listen to this track for more than 3-4 minutes on most VFB amps. With CFB you can listen whole track without any fatigue. Yes, the VFB has impressive dynamics and massive attack, but also weird combination of a bit overblown, usually ill-defined bass, unmusical mids and distorted aggressive highs. Astrud Gilberto's voice was never really satisfying with VFB. And while with popular music VFB can have certain charm because of it impressive massive sound, with classical music there is no comparison: CFB is clear winner! With VFB you can not follow contrapunctal lines in Bach's Branderburgisches Konzert Nr.5. It's a mess. With this kind of material CFB rules supremely.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.