Bootstrapping capacitor to Q5/Q6 emitter sound like a good plan, as Q5/Q6 emitter are closer to unity gain with respect to Q3/Q4 emitter than the final output node. This will require a bit more current drive from Q5/Q6, but likely this is not particularly significant. Didn't look into detail though, so verify if everything's fine at least in simulation.
I think I'd split R3 in two equal halfs to form a common bootstrapping point.
Samuel
I think I'd split R3 in two equal halfs to form a common bootstrapping point.
Samuel
Thank you for replay Samuel...As the matter of fact, I have already simulated that particular configuration (discrete J-fet input AD797 with triple EF output stage headphone amp as per AES paper, I have a working prototype), but as I'm still in learning curve and simulators are not to be trusted, so I asked ;-) and I can confirm that simulator gives lowest distortion from bootstrapped configuration, it's even better then simple LED based CCS.
That's exactly what I have done...Since my bootstrapped resistors are low in value (2 x 1k2, ~6mA current and +/-15VDC regulated supply), what would be good value for bootstrapped capacitor (have tried "real" ones in sim from 100-1000uF, but have not been able to see any improvements)?
Basically I'd make sure that the cutoff-frequency (i.e. the frequency below which bootstrapping becomes ineffective) defined by the bootstrapping capacitor and resistor (1/(2*pi*R*C)) is at least 3x lower than the lowest signal frequency of interest, say 10 Hz for audio. So 100 uF is enough.
Samuel
Samuel thanks for the paper and the hard work you put into it, some very good info you have brought to light. All of it is not new however, there arent many books available on amplifier design but it doesnt mean that all designers are not aware of other distortion mechanisms beyond Dr Self book or how to deal with them. Your paper is a good step forward.
I can provide two examples of where designers made use of the ideas you put forward. There is a patent or was, I dont know if it is still valid, for using a mosfet as the follower before the vas , this patent dates to 1986 if I remember correctly, I might still have a copy of this and if I remember correctly Van Alstine was the author. In it the author correctly calls the vas a transimpedance stage. I have tried amps with this combination and I can say that in 90 percent of the cases it did bring distortion down, most Self type amps would benefit from this modification if better current mirror is used. I also know at least one commercial designer that has transgressed this patent in commercial designs but I dont think the parties involved are aware of this fact as some manufacturers tend to hide their schematics, so much for having patents ......
The triple output stage you show has been used 25 years ago by pioneer in some of their exclusive and top end amps so its not new in any way and is the earliest example of such a outputstage design that I have seen. The model was the exclusive M6 if you want to try to see the schematics.
I can provide two examples of where designers made use of the ideas you put forward. There is a patent or was, I dont know if it is still valid, for using a mosfet as the follower before the vas , this patent dates to 1986 if I remember correctly, I might still have a copy of this and if I remember correctly Van Alstine was the author. In it the author correctly calls the vas a transimpedance stage. I have tried amps with this combination and I can say that in 90 percent of the cases it did bring distortion down, most Self type amps would benefit from this modification if better current mirror is used. I also know at least one commercial designer that has transgressed this patent in commercial designs but I dont think the parties involved are aware of this fact as some manufacturers tend to hide their schematics, so much for having patents ......
The triple output stage you show has been used 25 years ago by pioneer in some of their exclusive and top end amps so its not new in any way and is the earliest example of such a outputstage design that I have seen. The model was the exclusive M6 if you want to try to see the schematics.
Samuel, if you will give me your E-mail address, I will send you a copy of my 1978 IEEE paper: 'Omitted Factors in Audio Design'. You will find direct measurements of cross-coupling tantalum caps the way that you found with aluminum caps, except that tantalum caps distort even worse, on average. You might find it interesting.
I am going through all 50+ pages of your work. Good stuff!
I am going through all 50+ pages of your work. Good stuff!
Again thank you for the interest!
I'd be very glad to have the patent number; of course the question is if the designer was actually aware of all/the advantages of the use of a MOSFET in this position. Note that in the "Self Blameless" configuration the VAS input resistance is often the least significant contribution to what I call lumped resistance RL. Quite a substantial reduction in VAS output impedance is possible just with the improved current mirror and input stage cascode.
There's also the possibility to add a second emitter follower (running at low Ic) instead of the MOSFET. Improvement is similar, although at higher parts count.
Thanks, I do have your excellent paper. You may have missed that I gave reference to it ([19]).
Of course. I'm not claiming any innovation here--last but not least I give two references and even say that it has "found widespread use".
Samuel
I'd be very glad to have the patent number; of course the question is if the designer was actually aware of all/the advantages of the use of a MOSFET in this position. Note that in the "Self Blameless" configuration the VAS input resistance is often the least significant contribution to what I call lumped resistance RL. Quite a substantial reduction in VAS output impedance is possible just with the improved current mirror and input stage cascode.
There's also the possibility to add a second emitter follower (running at low Ic) instead of the MOSFET. Improvement is similar, although at higher parts count.
Thanks, I do have your excellent paper. You may have missed that I gave reference to it ([19]).
Of course. I'm not claiming any innovation here--last but not least I give two references and even say that it has "found widespread use".
Samuel
The patent by Aado Perandi is no. 4801893, Forward transimpedance amplifier.
I haven't read the full text yet but so far it seems like he didn't claim reduced VAS output Z/sensitivity to VAS loading. Also his preferred embodiment does show a standard mirror with low output resistance, so the FET follower can't be very effective to reduce VAS output Z/sensitivity to VAS loading anyway.
But still thanks for pointing me to the patent, it is good reading!
Samuel
I haven't read the full text yet but so far it seems like he didn't claim reduced VAS output Z/sensitivity to VAS loading. Also his preferred embodiment does show a standard mirror with low output resistance, so the FET follower can't be very effective to reduce VAS output Z/sensitivity to VAS loading anyway.
But still thanks for pointing me to the patent, it is good reading!
Samuel
that triple (fig 59). Thanks, excellent work.. (it shall be built).
Thank you Samuel, for this 'commentary' . Walt Jung turned me on to it, or I might have missed it. It is just what we need to update ourselves. I'm glad that you have my paper, it also shows why IC's need to behave properly above and beyond 20KHz, as you must have noted.
Tomorrow, I hope to go to a gathering of analog designers, you know: The IC designers. Tell you more, if, and when it happens.
Tomorrow, I hope to go to a gathering of analog designers, you know: The IC designers. Tell you more, if, and when it happens.
Great paper Mr. Groner. It seems to me that if you replace Q5 and Q6 in Figure 59 with Sziklai pairs, then at a first approximation, there is no temperature dependence issue with the power devices Q7 and Q8. The trade off then is being able to reliably and accurately set the quiescent output bias current. So the question is, is it worth it?
Bela
Bela
Last edited:
The patent doesnt go into any detail but merely claims better performance, I experimented with it a couple of years back using high gm small signal zetex mosfets and it brought the distortion down in the Self designs 98 percent of the time, despite using standard mirrors, my mirrors had larger degeneration though.
On the subject of the perfect vas transistor, how well do you suppose a very high beta cascoded transistor would work regarding low frequency distortion. I have seen designs using very high beta darlington and cascoded for vas duty. Would a high VA transistor still outperform or how would it compare with a high VA transistor which is cascoded.
Excellent work Samuel, I might actually buy Self's book now so that I can get a better perspective. This is an interesting circuit that uses a MOSFET buffer and common base VAS:
http://www.diyaudio.com/forums/solid-state/122378-cambridge-audio-a3i-repairs-mods.html
http://www.diyaudio.com/forums/solid-state/122378-cambridge-audio-a3i-repairs-mods.html
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.