Hi Damir
Nested feedback is important to the good sound of this amplifier in my opinion. My early versions didn't have it, and they sounded very good too, but I think it sounds even better with.
You are probably correct that Cdom can go down to 15p with the values specified. I had my unit down to 22p and the scope trace was clean, although I didn't notice a sonic improvement over 33p so I left it as is.
If I'm reading your schematics correctly that's a pretty dramatic improvement in THD on the TPC version. I've never tried TPC on a CFA and will have to give it a go sometime.
Nested feedback is important to the good sound of this amplifier in my opinion. My early versions didn't have it, and they sounded very good too, but I think it sounds even better with.
You are probably correct that Cdom can go down to 15p with the values specified. I had my unit down to 22p and the scope trace was clean, although I didn't notice a sonic improvement over 33p so I left it as is.
If I'm reading your schematics correctly that's a pretty dramatic improvement in THD on the TPC version. I've never tried TPC on a CFA and will have to give it a go sometime.
I meant "The reason for that, in my opinion, is two week drive of the output transistor(more visible at higher frequencies when output transistor input capacity came to play)."
Hi Damir,
Thank you for your marvellous examination of the quasi, it is very thorough and I think I agree that the 10R degen on the VAS is probably a waste of the some loop gain.......
However, from previous and recent work on a number of different designs the nested fb is very useful. It does increase the THD, no question, and also 'rounds' the sound more like a tube amp, but I think it sounds better although I agree it is non-conventional and goes against views about distortion and sound quality as per Doug Self.
I'm reminded of a comment from Charles Hansen, the designer of Ayre amps in Boulder, CO. He once remarked that once you have heard a zero global fb amp you will never listen to a full global fb amp ever again. I agree with him. The nested fb on the quasi restricts the global fb to around 31.5dB, much lower than the usual 50-70dB, so it clearly is a moderate, global fb amp. BTW, it was my idea to put in that nested fb.
The sound quality is a personal choice. Most who have heard this amp will probably say that it has some qualities which mark it unique compared to some very sophisticated and elegant designs - and yet it has a low part count.
I have been an admirer of your zero fb amp. I think it is a brilliant piece of design and I am ashamed to say I have not built and listened one! Generally I do not like to build amps these days with more than 10 transistors!
Again, thank you for your insightful charts!
Cheers,
Hugh
Thank you for your marvellous examination of the quasi, it is very thorough and I think I agree that the 10R degen on the VAS is probably a waste of the some loop gain.......
However, from previous and recent work on a number of different designs the nested fb is very useful. It does increase the THD, no question, and also 'rounds' the sound more like a tube amp, but I think it sounds better although I agree it is non-conventional and goes against views about distortion and sound quality as per Doug Self.
I'm reminded of a comment from Charles Hansen, the designer of Ayre amps in Boulder, CO. He once remarked that once you have heard a zero global fb amp you will never listen to a full global fb amp ever again. I agree with him. The nested fb on the quasi restricts the global fb to around 31.5dB, much lower than the usual 50-70dB, so it clearly is a moderate, global fb amp. BTW, it was my idea to put in that nested fb.
The sound quality is a personal choice. Most who have heard this amp will probably say that it has some qualities which mark it unique compared to some very sophisticated and elegant designs - and yet it has a low part count.
I have been an admirer of your zero fb amp. I think it is a brilliant piece of design and I am ashamed to say I have not built and listened one! Generally I do not like to build amps these days with more than 10 transistors!
Again, thank you for your insightful charts!
Cheers,
Hugh
Mooly:
I've looked at that cross connection between channels and can't figure out what it is for. Since it goes to the negative feedback node it must be related to that , but coming from a different channel ? Cross-feed as in ambience ? But the RC time constant indicates a Fc of 48.2Khz !
I have another question regarding the current circuit being tested.
1. The need to have a diode ( 1N4004) between the base and emitter of the input transistor. Protection of Q1 from hard clipping or a short ?
2. The resistor from the VAS stage to the emitter of the input stage ( 47 K at present ). This is an additional feedback point. Is that for stability which it otherwise didn't have ? Earlier circuits had a capacitor also. Usually it is just a cap . Then it was a cap with a resistor across it to be effective down to dc . Now it's just a resistor , additional wide band feedback . Have they been added after seeing some form of instability in the output ?
Are both these essential parts ?
Thanks.
Last edited:
Hi Hugh,
I am very happy that your health is good and allow you to help diyers to improve amps.
When I said I think that local(nested) NFB around VAS will not improve the sound, I came to conclusion only from FFT harmonic distribution not changed with it, just distortion increased. Of course I have not heard that amp and my argument is a weak one.
I know about Charles Hansen NGNFB preference, and I hoped to build my NGNFB amp, but you know, one in retirement has match less free time(very strange). So I never heard any NGNFB amp except some simple valve amp very long time ago.
Best wishes to you Hugh, Damir
Damir,
You are my senior, Sir, so in the notion of tempus fugit, I trust you are well too, and that you are robust in health and able to argue with your clever, motorcycling son, and play with your grandchildren! (Alas, none for me yet, but my eldest is an Amazon with a razor tongue and the men run away when she speaks... and the other lives in Brooklyn. Woe is me!)
I exhort you to analyse the FFT of the quasi with and without the NFB. You will notice that H2, H3, H4 and H5 all increase, particularly H2 and H4, but H7, H9, H11 and beyond are all four to five dB lower. This analysis of the FFT is predictable from the Taylor Series, and is the clearest indication to me that NFB does increase THD but reduces the H7 and beyond, which of course reduces the 'machine tones' which denigrate the musical presentation of most amplifiers.
Ciao,
Hugh
You are my senior, Sir, so in the notion of tempus fugit, I trust you are well too, and that you are robust in health and able to argue with your clever, motorcycling son, and play with your grandchildren! (Alas, none for me yet, but my eldest is an Amazon with a razor tongue and the men run away when she speaks... and the other lives in Brooklyn. Woe is me!)
I exhort you to analyse the FFT of the quasi with and without the NFB. You will notice that H2, H3, H4 and H5 all increase, particularly H2 and H4, but H7, H9, H11 and beyond are all four to five dB lower. This analysis of the FFT is predictable from the Taylor Series, and is the clearest indication to me that NFB does increase THD but reduces the H7 and beyond, which of course reduces the 'machine tones' which denigrate the musical presentation of most amplifiers.
Ciao,
Hugh
As this has sort of been a lesson in amp design theory for many of us, so for the benefit of the uninitiated, could someone highlight in the schematic where all the above technical discussion points are located at?
1. Local vs global feedback - where is that happening?
2. TPC?
3. Degeneration resistor?
4. Where would full global feedback go if implemented?
Thank you.
Btw, I spent some time comparing this amp with the JFET Circlotron (already mentioned), then also against my main reference amp, Vzaichenko's VHex+. I have to say, it has a unique sound that makes it very appealing. It sounds very different than the VHex+ but not in a bad way at all. I don't have an amp measurement rig yet so will have to resort words similar to how people describe wine tasting
"Sweet and full bodied, with a hint earthy organic hues and goes down the palette easily with just about any type of food." is perhaps the best description.
1. Local vs global feedback - where is that happening?
2. TPC?
3. Degeneration resistor?
4. Where would full global feedback go if implemented?
Thank you.
Btw, I spent some time comparing this amp with the JFET Circlotron (already mentioned), then also against my main reference amp, Vzaichenko's VHex+. I have to say, it has a unique sound that makes it very appealing. It sounds very different than the VHex+ but not in a bad way at all. I don't have an amp measurement rig yet so will have to resort words similar to how people describe wine tasting
"Sweet and full bodied, with a hint earthy organic hues and goes down the palette easily with just about any type of food." is perhaps the best description.
Thank you Hugh, I am 72 and in quite good heath. Yes, grandchildren are joy, I wish you a couple soon.
I will repete the simulation with the simulation with local NFB and look more carefully, thank for the hint.
Ciao
Damir
Hmm, never tried a Lambrusco. Australia has some splendid wines - last time I was in Sydney (1996) I brought a bottle back.
I was thinking a Siraz from mountains of Argentina.
AC Ideas
Hi X,
Big question....... here goes!
I have attached Prasi's excellent schematic dated 2 Jul 16. I have changed a few values, as follows:
1. C2 from 470pF down to 220pF (LF corner higher now, this is a current fb amp and is not vulnerable to RF injection from cellphones, TV, radio, etc).
2. C6 bootstrap back from 470uF to a smaller, adequate 100uF rated to 63VW.
3. R8 (shunt fb) changed from 47R to 68R to accord with 1k8 as R9. Sets the overall gain, with small reduction do to R10, the nested feedback resistor.
4. C3 from 470uF down to 22uF 16VW to reduce thumps at switch on.
That's it. To address your points:
1. Local vs global feedback - where is that happening?
Local feedback is gain taken from end to beginning of a SINGLE stage, usually a single transistor. We are dealing with transistors where input is at base, emitter is close to AC ground (a rail, for example) and output from collector. This is a common emitter transistor, the only way to set up a single transistor for voltage AND current gain. Local fb is also sometimes called degeneration, and in this amp the degeneration resistor, which reduces the gain hence the use degeneration, is R23, 10R.
2. TPC?
This is complex because it focusses on a huge, pivotal issue with most audio amplifiers - STABILITY. Now, take a long breath.... this is difficult. An amp signal takes a short time to pass from input to output, and IF we take a little feedback from the output node and pass it back to the input stage (the speaker output back to the emitter of Q1, in this case) it will arrive at the Q1 emitter just a little BEHIND the input signal - that is, it is delayed by nanoseconds! So the first stage, which compares input signal with output signal to create an ERROR SIGNAL, has to handle this fb signal and it is a bit late to the party. This is a problem, because if this delayed period is one half of the cycle of the input frequency, what we think is negative feedback now becomes POSITIVE feedback, and it has huge consequences.
Positive feedback means that the late period of the fb to the party looks like to the 'guests' to be a signal to foster the original input, that is, positive feedback, and the result is the electronic equivalent of a RIOT. In other words, a little enthusiasm at the input is met by a little more enthusiasm at the same input stage and things get completely out of hand!
This is DESTRUCTIVE OSCILLATON, and it usually destroys the amp, the speakers, and the audiophile's pocket. This must be avoided at any cost.
There are ways to fix this problem. We modify the amp's gain so that at the very high frequency where the negative fb turns positive, usually called the HIGH POLE, we adjust the gain of the amp at this frequency so it's gain is less than UNITY. Then, if the nfb becomes positive fb, the amp gain is less than one, and it cannot oscillate and destroy itself.
The usual way is to put a cap from the output of the voltage amplifier (the 'gain' active device) back to the input, the base. On this amp it is C5, and it's set as 33pF. This is a tiny capacity, but it is perhaps the most important cap in the amp; it is difficult to set it with simulation; it is so small that the parasitics on the pcb tracks and components force you to figure this out by listening tests and some checks with a CRO. This is called 'compensating the amp'. The method I've first described is lag compensation, the most common method, a single cap. But there is a more complex version named Two Pole Compensation, TPC, which uses two caps, and a resistor to create a HF filter to delay the reduction of the amp gain until a higher frequency but with a higher slope which pull the gain down below unity at that same HIGH POLE. This is the best way to do it, but it's more complicated (at least to me) and I have never heard an improvement in the sound quality. To me, I don't care too much about how we achieve stability; I'm most interested in the subjective sound, which I can tell in a good listen. Damir (Dadod) is a Telecoms Engineer and very au fait with this technology; you might ask him about his feelings.
3. Degeneration resistor?
As mentioned earlier, 10R R23 on the schematic. I am not convinced this is needed, actually. I think Damir is quite right here.
4. Where would full global feedback go if implemented?
Just about 98% of the commercial marketplace has full global feedback amplifiers. It works this way: An amplifier has an open loop gain, and a closed loop gain. If you set up the gain around 30dB, then this is the gain set by the feedback ratio; in this case close to (1k8+0.068k)/0.068 less the small margin due to the 68k nested fb. This is around 26.75, about 28.5dB, the usual THX standard. If you remove the fb network, the amp would operate at OPEN LOOP. This would be 5350, which corresponds to 74.5dB!! Now, if you subtract the close gain from the open gain we have (74.5 - 28.5) = 46dB, and this would be the normal global feedback for this amp if it used only global feedback.
In fact, this amp has nested feedback between first and second stages. This reduces the open gain available to this amp, and the global feedback then is somewhat less. It turns out the global fb is simple; it is 68k/1k8 = 37.8, or 31.5dB compared to the usual 46dB if we were using ONLY global feedback with no nested feedback. Even 46dB is a low feedback amplifier; 31.5dB is very low. Most tube amps have much less than this, 15dB down to none at all, that is, they operate in OPEN LOOP - another name for 'without feedback'.
This is thumbnail analysis of the AC considerations and mathematics behind the Ranchu QUASI. FB was discovered at Bell Labs in NYC in 1927 by Harold Black, and a good analysis is here: Feedback theory
Don't tell me you didn't ask..........
The history of fb is absolutely riveting, at least, to me!
Hugh
Hi X,
Big question....... here goes!
I have attached Prasi's excellent schematic dated 2 Jul 16. I have changed a few values, as follows:
1. C2 from 470pF down to 220pF (LF corner higher now, this is a current fb amp and is not vulnerable to RF injection from cellphones, TV, radio, etc).
2. C6 bootstrap back from 470uF to a smaller, adequate 100uF rated to 63VW.
3. R8 (shunt fb) changed from 47R to 68R to accord with 1k8 as R9. Sets the overall gain, with small reduction do to R10, the nested feedback resistor.
4. C3 from 470uF down to 22uF 16VW to reduce thumps at switch on.
That's it. To address your points:
1. Local vs global feedback - where is that happening?
Local feedback is gain taken from end to beginning of a SINGLE stage, usually a single transistor. We are dealing with transistors where input is at base, emitter is close to AC ground (a rail, for example) and output from collector. This is a common emitter transistor, the only way to set up a single transistor for voltage AND current gain. Local fb is also sometimes called degeneration, and in this amp the degeneration resistor, which reduces the gain hence the use degeneration, is R23, 10R.
2. TPC?
This is complex because it focusses on a huge, pivotal issue with most audio amplifiers - STABILITY. Now, take a long breath.... this is difficult. An amp signal takes a short time to pass from input to output, and IF we take a little feedback from the output node and pass it back to the input stage (the speaker output back to the emitter of Q1, in this case) it will arrive at the Q1 emitter just a little BEHIND the input signal - that is, it is delayed by nanoseconds! So the first stage, which compares input signal with output signal to create an ERROR SIGNAL, has to handle this fb signal and it is a bit late to the party. This is a problem, because if this delayed period is one half of the cycle of the input frequency, what we think is negative feedback now becomes POSITIVE feedback, and it has huge consequences.
Positive feedback means that the late period of the fb to the party looks like to the 'guests' to be a signal to foster the original input, that is, positive feedback, and the result is the electronic equivalent of a RIOT. In other words, a little enthusiasm at the input is met by a little more enthusiasm at the same input stage and things get completely out of hand!
This is DESTRUCTIVE OSCILLATON, and it usually destroys the amp, the speakers, and the audiophile's pocket. This must be avoided at any cost.
There are ways to fix this problem. We modify the amp's gain so that at the very high frequency where the negative fb turns positive, usually called the HIGH POLE, we adjust the gain of the amp at this frequency so it's gain is less than UNITY. Then, if the nfb becomes positive fb, the amp gain is less than one, and it cannot oscillate and destroy itself.
The usual way is to put a cap from the output of the voltage amplifier (the 'gain' active device) back to the input, the base. On this amp it is C5, and it's set as 33pF. This is a tiny capacity, but it is perhaps the most important cap in the amp; it is difficult to set it with simulation; it is so small that the parasitics on the pcb tracks and components force you to figure this out by listening tests and some checks with a CRO. This is called 'compensating the amp'. The method I've first described is lag compensation, the most common method, a single cap. But there is a more complex version named Two Pole Compensation, TPC, which uses two caps, and a resistor to create a HF filter to delay the reduction of the amp gain until a higher frequency but with a higher slope which pull the gain down below unity at that same HIGH POLE. This is the best way to do it, but it's more complicated (at least to me) and I have never heard an improvement in the sound quality. To me, I don't care too much about how we achieve stability; I'm most interested in the subjective sound, which I can tell in a good listen. Damir (Dadod) is a Telecoms Engineer and very au fait with this technology; you might ask him about his feelings.
3. Degeneration resistor?
As mentioned earlier, 10R R23 on the schematic. I am not convinced this is needed, actually. I think Damir is quite right here.
4. Where would full global feedback go if implemented?
Just about 98% of the commercial marketplace has full global feedback amplifiers. It works this way: An amplifier has an open loop gain, and a closed loop gain. If you set up the gain around 30dB, then this is the gain set by the feedback ratio; in this case close to (1k8+0.068k)/0.068 less the small margin due to the 68k nested fb. This is around 26.75, about 28.5dB, the usual THX standard. If you remove the fb network, the amp would operate at OPEN LOOP. This would be 5350, which corresponds to 74.5dB!! Now, if you subtract the close gain from the open gain we have (74.5 - 28.5) = 46dB, and this would be the normal global feedback for this amp if it used only global feedback.
In fact, this amp has nested feedback between first and second stages. This reduces the open gain available to this amp, and the global feedback then is somewhat less. It turns out the global fb is simple; it is 68k/1k8 = 37.8, or 31.5dB compared to the usual 46dB if we were using ONLY global feedback with no nested feedback. Even 46dB is a low feedback amplifier; 31.5dB is very low. Most tube amps have much less than this, 15dB down to none at all, that is, they operate in OPEN LOOP - another name for 'without feedback'.
This is thumbnail analysis of the AC considerations and mathematics behind the Ranchu QUASI. FB was discovered at Bell Labs in NYC in 1927 by Harold Black, and a good analysis is here: Feedback theory
Don't tell me you didn't ask..........
The history of fb is absolutely riveting, at least, to me!
Hugh
Attachments
Last edited:
there is an old saying here "he / she is between San Juan and Mendoza" when someone is awfully drunk
