Yes he did. I didn't want to assume your answer was his answer.
In that case, are you saying that you have found the complementary diff stage to sound better, and therefore THD is a secondary consideration? It seemed to me you started by saying a comp diff is better because the THD is twice as good.
Now, this begs the question of why the comp diff sounds better. Have you a theory to share about this?
Brian
In that case, are you saying that you have found the complementary diff stage to sound better, and therefore THD is a secondary consideration? It seemed to me you started by saying a comp diff is better because the THD is twice as good.
Now, this begs the question of why the comp diff sounds better. Have you a theory to share about this?
Brian
What do you mean with cheaper ?
Cheaper on the background of using Bybee purifiers ?
Or cheaper on the background of a $17000 Blowtorch ?
Are you kidding ?!!!
So what is the discussion about ?
In order to get "cheap" and reliable, I would take an LME49XXX or AD797 (Btw.: thank you Scott!!) for a line stage, instead of messing around with esoteric dual FETs (select them manually) and hoping that you still get them in the next few months.
Where have you been in the last 10 years ?
Did you know, that SMT is around us everywhere and many of your beloved through-hole components didn't survive the RoHS change ?
OK, here goes:
I was given a SPECIAL project that had to be VERY LOW DISTORTION, but the signal passing through was only a single sine wave between 10 and 100,000Hz.
This was for an audio THD oscillator driver, NOT audio in general. Global negative feedback can work wonders in this application.
SINCE, I had to work to 100,000Hz with less than .001% distortion or so, I chose to use 2 stage (12dB/octave) compensation.
It was easier to use essentially the SAME fet pair that Bob Cordell uses to minimize the non-linear input capacitance and reduce its effect at 100,000Hz, because P channel devices have MORE non-linear input capacitance than N channel devices.
SO, it was easier to make, (and compensate) a more conventional topology, that I have used since 1973, than to use the more difficult to adjust complementary differential fet input topology that I have used for even longer.
I was given a SPECIAL project that had to be VERY LOW DISTORTION, but the signal passing through was only a single sine wave between 10 and 100,000Hz.
This was for an audio THD oscillator driver, NOT audio in general. Global negative feedback can work wonders in this application.
SINCE, I had to work to 100,000Hz with less than .001% distortion or so, I chose to use 2 stage (12dB/octave) compensation.
It was easier to use essentially the SAME fet pair that Bob Cordell uses to minimize the non-linear input capacitance and reduce its effect at 100,000Hz, because P channel devices have MORE non-linear input capacitance than N channel devices.
SO, it was easier to make, (and compensate) a more conventional topology, that I have used since 1973, than to use the more difficult to adjust complementary differential fet input topology that I have used for even longer.
Thanks. I guess my underlying question is whether you have a quantitive measure of sound quality other than THD? So that you don't have to rely on your ears alone to make a choice of topology.
When you used the term "brute force" to describe NFB, I infer you consider NFB brutalizes the sound quality despite its favourable effect on THD.
When you used the term "brute force" to describe NFB, I infer you consider NFB brutalizes the sound quality despite its favourable effect on THD.
This is a semantic or use of English, confusion of word definition. Brute force can be just a bigger hammer, not a rougher one, or the use of explosives.
OK good sounding topology is more important than brute force THD ? Aha !
What is a topology worth if it was judged by 60+ year old ears or by a 30 year old memory ?
And why using your differential complementary open loop approach, if other non complementary open loop approaches sound better than yours, but have slightly higher THD, if "brutal" low THD is bad ?
Are there two kinds of low THD out there, one stands for good sound and the other bad one for the use NFB ?
I really come more and more to the conclusion that JC's approach is a big bag full of hot air ...
What is a topology worth if it was judged by 60+ year old ears or by a 30 year old memory ?
And why using your differential complementary open loop approach, if other non complementary open loop approaches sound better than yours, but have slightly higher THD, if "brutal" low THD is bad ?
Are there two kinds of low THD out there, one stands for good sound and the other bad one for the use NFB ?
I really come more and more to the conclusion that JC's approach is a big bag full of hot air ...
PMA said:
I don't believe this to be true, if each is designed with equal care and expertise. I beleive this applies whether or not negative feedback is used.
I guess we're back to the subject of the Unipolar vs Complementary thread.
Could you please elaborate briefly on the particular reason why you assert that the complemntary has lower distortion?
BTW, it is most certainly the case that the complementary pair approach can incur GREATER distortion if the usual pair of Miller capacitors is returned back to the input of each of the upper and lower VAS stages. This is due to VAS fighting. The individual Miller feedback capacitors provide shunt feedback, which lowers output impedance. We thus end up with the upper and lower VAS stages as a pair of low-outpt-impedance stages connected in parallel - bad design practice - and they will fight, creating distortion.
They will fight to the extent that the output voltage they wanted to produce is not equal. This inequality will occur if the miller feedback capacitors are mismatched by as little as 5%. It will also occur if the transconductance of the P input pair and the N input pair do not match. This is a special problem for N/P channel JFET complementary pairs.
It is notable that John's JC-1 DOES NOT use the usual pair of Miller capacitors back to the VAS inputs, and so appears not to be subject to this problem.
BTW, I'll be up at FSI in Montreal tomorrow through Sunday. Anybody else going to be there?
Cheers,
Bob
john curl said:
verbal diarrhea:
A serious disease which, once it has control of a person, causes them to spew forth incoherent babble from the bowels of the voicebox. Often extremely frustrating for the victim and extremely hilarious for the observer
And about when were you cured from this disease?
Mr. Cordell , would that "vas fighting effect" also hold true
with unipolar/comp. "hybrids" like the APT or symasym ??
OS
Bingo !
Thanks god, that someone with more "prerequisites" and as famous as Mr. Curl attends again this discussion !
Hi John,
Here! You're skipping there! 😀
On your complimentary differential input preference...
In a perfect world, yes, it may have lower distortion, and it will have lower DC offset for a given beta and tail current.
Now for the depressing part .....
In every production amplifier I have seen designed with complimentary differential input, the best I have seen was each polarity with a reasonable match. Of course we now have some DC offset on the order of a single differential. This because complimentary differential input pairs tend to be run at higher tail currents because the offsets will mostly cancel. Measured distortion is not remarkable, being about the same (both sets close) or worse than a single matched pair (normally).
The average case ...
Normally I find none of the input transistors match. This doesn't sound very good for one, and it sure measures worse. Completely expected, no surprises here. After matching these parts by hand (a very time consuming process), both the distortion and DC offset are reduced considerably. DC offsets normally run less than 5 mV after this. The audible improvement is noticeable. Of course, no way are production lines going to halt and wait for weeks before enough matched sets are gathered to begin the run. No, they will stick in what they have received from the distributors. The 2SA798 and 2SC1583 dual transistors were an attempt to make this work reasonably in production environments. I find that over time, these even drift apart in beta.
In closing ....
As far as I'm concerned, complimentary differential inputs only work in hand built products. In a normal mass market product, they are a no-flier and everyone is better off with single matched pairs. Servo or no. They are more expensive to service if done correctly due to the extra time required to match 2 pairs rather than 1 pair.
I'm not disagreeing with you John. Just illustrating the real world after some years have passed, or mass production examples.
-Chris
Here! You're skipping there! 😀
On your complimentary differential input preference...
In a perfect world, yes, it may have lower distortion, and it will have lower DC offset for a given beta and tail current.
Now for the depressing part .....
In every production amplifier I have seen designed with complimentary differential input, the best I have seen was each polarity with a reasonable match. Of course we now have some DC offset on the order of a single differential. This because complimentary differential input pairs tend to be run at higher tail currents because the offsets will mostly cancel. Measured distortion is not remarkable, being about the same (both sets close) or worse than a single matched pair (normally).
The average case ...
Normally I find none of the input transistors match. This doesn't sound very good for one, and it sure measures worse. Completely expected, no surprises here. After matching these parts by hand (a very time consuming process), both the distortion and DC offset are reduced considerably. DC offsets normally run less than 5 mV after this. The audible improvement is noticeable. Of course, no way are production lines going to halt and wait for weeks before enough matched sets are gathered to begin the run. No, they will stick in what they have received from the distributors. The 2SA798 and 2SC1583 dual transistors were an attempt to make this work reasonably in production environments. I find that over time, these even drift apart in beta.
In closing ....
As far as I'm concerned, complimentary differential inputs only work in hand built products. In a normal mass market product, they are a no-flier and everyone is better off with single matched pairs. Servo or no. They are more expensive to service if done correctly due to the extra time required to match 2 pairs rather than 1 pair.
I'm not disagreeing with you John. Just illustrating the real world after some years have passed, or mass production examples.
-Chris
- Status
- Not open for further replies.