I'm not sure why this should be the case. Can you elaborate?
How exactly did it improve distortion performance (e.g. was the frequency-independent low-frequency floor reduced)? I realise that my previous statement that VAS input resistance is often the least significant contribution to lumped resistance at the VAS input is misleading. If the VAS is built from low-hFE parts and run at rather high current it might easily dominant lumped resistance, so a MOSFET for the follower would improve things even without too much attention to current mirror etc.
There is surely some prior art to this patent which shows the same or very similar arrangements. I doubt the patent was ever really valid. In any case I believe that true progress is only given if the exact mechanism, which causes an observed distortion improvement, is fully understud. Everything else is just blind repetition of guesswork.
The cascode connection is likely superior in most cases as Early effect is reduced by at least one, more like two orders of magnitude.
Surely a good investment!
How do you conclude that the output stage from figure 6 (and any other model amplifier in my paper) is not class A? It's not easy to calculate the exact output stage quiescent current. In fact it measures at about 6 mA, so ample class A current to drive a medium-impedance feedback network. As you can see from figure 47 ("no load" plot) there is surely no significant class B crossover distortion going on.
Samuel
Great Job!
Hi Samuel,
I finally got to read your paper in depth on a plane ride to San Francisco last night. This is really a great piece of work. Anyone who buys Doug’s book should surely have this at their side as well. Doug does a good job of covering IPS/VAS distortion sources in most cases, but your paper really adds much more insight to what is going on and why some choices are better than others. Here are a couple of questions and comments. I don’t have Doug’s book in front of me, so that could be the origin of a couple of things I missed.
On page 2 you mention the EFA mirror in figure 1, but I could not find it in your figures. Was that a Self figure? By EFA mirror I assume you mean a “helpered” Widlar mirror where base current to the two mirror transistors is supplied through an emitter follower. I like this current mirror and often use it because it does permit better control of LTP current balance when used in conjunction with a VAS transistor preceded by an emitter follower. Choice of the bias current in the current mirror EF transistor gives us that opportunity.
Your analysis of input-referred noise contribution is very insightful and a contribution that is easy for many of us to neglect or under-estimate, including myself, especially when we are using heavily-degenerated input stages. As you point out, making it quite a bit smaller is an easy kill as long as DC operating margins are not compromised, My only comment there is that noise levels of any contributors in a power amplifier need to be put into context with objectives and other sources.
For example, the feedback network shunt resistor is often the biggest contributor if the feedback network resistance values are not kept very low (500 ohms or less). Also, any power amplifier whose input-referred noise is less than about 10 nV/rt Hz is already quite exceptional in the noise department given the line level signals present at the input. In my amplifier of my book Figure 3.8, input-referred noise using a 200 ohm NFB network is 6.5 nV/rt Hz using 470 ohm current mirror degeneration resistors and drops to 5.7 nV/rt Hz using 1k current mirror degeneration resistors. Nevertheless, if one can pick up a couple of dB by just a modest increase in the current mirror degeneration resistors that is a good thing.
Your mention of VAS distortion as a function of VAS emitter degeneration is great food for thought. I must admit I just about always just degenerate the VAS by a factor of 10, and have not looked closely at VAS distortion reductions that can be had by upping the VAS transconductance by reducing or eliminating VAS emitter degeneration. I’ll look into this with some simulations.
That’s all for now, but I may get back with some additional comments.
Once again, this is a fantastic piece of work that is well-written and a very enlightening read. Thanks for all the hard work you put into this.
Cheers,
Bob
Hi Samuel,
I finally got to read your paper in depth on a plane ride to San Francisco last night. This is really a great piece of work. Anyone who buys Doug’s book should surely have this at their side as well. Doug does a good job of covering IPS/VAS distortion sources in most cases, but your paper really adds much more insight to what is going on and why some choices are better than others. Here are a couple of questions and comments. I don’t have Doug’s book in front of me, so that could be the origin of a couple of things I missed.
On page 2 you mention the EFA mirror in figure 1, but I could not find it in your figures. Was that a Self figure? By EFA mirror I assume you mean a “helpered” Widlar mirror where base current to the two mirror transistors is supplied through an emitter follower. I like this current mirror and often use it because it does permit better control of LTP current balance when used in conjunction with a VAS transistor preceded by an emitter follower. Choice of the bias current in the current mirror EF transistor gives us that opportunity.
Your analysis of input-referred noise contribution is very insightful and a contribution that is easy for many of us to neglect or under-estimate, including myself, especially when we are using heavily-degenerated input stages. As you point out, making it quite a bit smaller is an easy kill as long as DC operating margins are not compromised, My only comment there is that noise levels of any contributors in a power amplifier need to be put into context with objectives and other sources.
For example, the feedback network shunt resistor is often the biggest contributor if the feedback network resistance values are not kept very low (500 ohms or less). Also, any power amplifier whose input-referred noise is less than about 10 nV/rt Hz is already quite exceptional in the noise department given the line level signals present at the input. In my amplifier of my book Figure 3.8, input-referred noise using a 200 ohm NFB network is 6.5 nV/rt Hz using 470 ohm current mirror degeneration resistors and drops to 5.7 nV/rt Hz using 1k current mirror degeneration resistors. Nevertheless, if one can pick up a couple of dB by just a modest increase in the current mirror degeneration resistors that is a good thing.
Your mention of VAS distortion as a function of VAS emitter degeneration is great food for thought. I must admit I just about always just degenerate the VAS by a factor of 10, and have not looked closely at VAS distortion reductions that can be had by upping the VAS transconductance by reducing or eliminating VAS emitter degeneration. I’ll look into this with some simulations.
That’s all for now, but I may get back with some additional comments.
Once again, this is a fantastic piece of work that is well-written and a very enlightening read. Thanks for all the hard work you put into this.
Cheers,
Bob
Thank you ver much for your interest and kind words! It is my pleasure that the efforts put into this paper are of help.
It's on page 3 of my paper.
The current mirror noise stuff is surely not of dramatic practical relevance. However I just don't feel comfortable as designer if I'm not aware of any significant tradeoff involved...
Samuel
It's on page 3 of my paper.
The current mirror noise stuff is surely not of dramatic practical relevance. However I just don't feel comfortable as designer if I'm not aware of any significant tradeoff involved...
Samuel
a very profound statement.
Could it be what defines the difference between designer and assembler?
Maybe!
Just noticed a subtle, but possibly evil drawing error in figure 57. The output of U2 should connect to Q13 base, not Q14 base. Q14 base should be grounded. Otherwise the polarity of the common-mode control loop is wrong and very large VAS currents will be enforced.
My apologies for this, I hope it did not so far, and will not in the future, cause any accidents. I uploaded a new PDF with a comment added; corrected schematic maybe later.
Samuel
My apologies. I was considering a slightly different architecture.
Attachments
Hello all
Just to let you know that I am actively discussing with Samuel the various issues raised by him- and doing a few experiments to hopefully shed further light.
The results will be uploaded to my website The Douglas Self Site and no doubt to Samuel's site also.
Just to let you know that I am actively discussing with Samuel the various issues raised by him- and doing a few experiments to hopefully shed further light.
The results will be uploaded to my website The Douglas Self Site and no doubt to Samuel's site also.
I think for some the answer is "because the challenge is there..."
An observation of mine is that judging by the number of folk who are still "searching" for a "better" sound, and having built amps of super low distortion, that low distortion in itself is not the full answer to achieving a satisfying listening experience.
That's another story though, and I have the greatest admiration for the designers and designs they produce. They are coming ever nearer to the old "wire with gain" ideal.
I see Mooly, i do admire them a lot too.... i do also respect them a lot
But it seems to me this is not to obtain better sound, as i suppose we cannot listen that when speakers goes messing the whole stuff...just an exercise.... the fight with numbers, to break some barrier...the challenge.
I understood the point... has anything with audio quality for humans to listen.... so, as you said... "because the challenge is there"
Thank you,
regards,
Carlos
But it seems to me this is not to obtain better sound, as i suppose we cannot listen that when speakers goes messing the whole stuff...just an exercise.... the fight with numbers, to break some barrier...the challenge.
I understood the point... has anything with audio quality for humans to listen.... so, as you said... "because the challenge is there"
Thank you,
regards,
Carlos
Carlos, I don't know how to explain it quickly, but the distortion of a speaker is of a diferent nature than an electronic component. We can actually hear very well <0.1% amplifier distortion 'through' a speaker wirth 3% distortion.
If we couldn't, all amps (yours too) would sound the same!
jan didden
Oh!....this is interesting Janneman...and that kind of thing was what i wanted
to listen....my trouble is to understand why so good guys have concerns about to reduce " non listenable" distortions.....so....they are not really non listenable....We can perceive something even having speakers messing with sonics...at least in such levels you pointed.
Yep...sometimes we do not perceive in our conscience, but brain is stimulated with something we perceive in a such way we detect that there's "something present"....alike frequencies above 16000 cicles...we do not listen (old buggers alike me) in the conscience...but we feel...one equipment with larger bandwidth is perceived as better than other having smaller bandwidth....things we capture...goes encoded into the brain but we do not decode in the conscience....so, not listened the way we understand as "to listen" something.
Was not my intention to disrespect these wonderfull brains we have in this thread....all i want is to "get their point"....because for sure, none of these guys are stupid or mad...so..there's something in there i could not yet fully understand.
Thank you Janneman....really thank you very much.
regards,
Carlos
to listen....my trouble is to understand why so good guys have concerns about to reduce " non listenable" distortions.....so....they are not really non listenable....We can perceive something even having speakers messing with sonics...at least in such levels you pointed.
Yep...sometimes we do not perceive in our conscience, but brain is stimulated with something we perceive in a such way we detect that there's "something present"....alike frequencies above 16000 cicles...we do not listen (old buggers alike me) in the conscience...but we feel...one equipment with larger bandwidth is perceived as better than other having smaller bandwidth....things we capture...goes encoded into the brain but we do not decode in the conscience....so, not listened the way we understand as "to listen" something.
Was not my intention to disrespect these wonderfull brains we have in this thread....all i want is to "get their point"....because for sure, none of these guys are stupid or mad...so..there's something in there i could not yet fully understand.
Thank you Janneman....really thank you very much.
regards,
Carlos
Hi Samuel,
For those interested, I have answered Carlos' question over on the thread about my book, since I addressed some of those kinds of issues in my book.
Cheers,
Bob
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.