Lumanauw
My experience says that the best (stability, linearity) output tripple darlington stage looks a bit different:
-the pre-drivers are biased by means of base-to-emitter resistor of drivers (150R or so)
-drivers have their emitters connected through a resistor in parallel with cap (I use 40R||33n).
-no base resistors
-output bjt's biasing a little bit higher than optimum, say, 100mA for 0R3 emitter resistors.
My experience says that the best (stability, linearity) output tripple darlington stage looks a bit different:
-the pre-drivers are biased by means of base-to-emitter resistor of drivers (150R or so)
-drivers have their emitters connected through a resistor in parallel with cap (I use 40R||33n).
-no base resistors
-output bjt's biasing a little bit higher than optimum, say, 100mA for 0R3 emitter resistors.
As far as Bob's article about slew rate and feedback factor:
I have great respect to mr. Cordell, but here he seems to have missed the point.
Higher feedback DOES generate higher TIM, but this happened at MENTAL stage. The engineers were so delighted with feedback and its possibilities, especially in solid state OTL designs, that they missed some more important issues. Thanks godnees, Matti came and the "maximum feedback from minimum number of devices, who cares about anything as feedback will correct" thinking is gone.
Or not entirely gone...
I have great respect to mr. Cordell, but here he seems to have missed the point.
Higher feedback DOES generate higher TIM, but this happened at MENTAL stage. The engineers were so delighted with feedback and its possibilities, especially in solid state OTL designs, that they missed some more important issues. Thanks godnees, Matti came and the "maximum feedback from minimum number of devices, who cares about anything as feedback will correct" thinking is gone.
Or not entirely gone...
darkfenriz said:
NO, Cordell is absolutely right.
Applying feedback is the process of 1. having lots and lots of open loop gain and then 2. giving away most of it
to get other nice things: higher input impedance, lower output impedance, flatter gain, less distortion.
But the requirement is: You've got to have a lot of gain to give away in the first place.
If you have an incompetently designed differential+VAS stage that cannot support the current to charge its Cdom,
then you get maybe a 10V step at 10 KHz for a 1 V input step. This is an effective gain of 10 and that even without closing
the global feedback loop.
A factor of 10 does not qualify as "lots of excess gain" if you need that much already to reach the required output level.
Therefore, even if you draw a global feedback voltage divider: You don't get global feedback at all.
Otalas point of view is usually shorted to "global feedback is baaad"
while in reality, a slewrate-limited amplifier is punished for not having enough open loop gain dynamically
to support any global feedback.
regards, Gerhard
If you read carefully, you'd notice I am not talking about technical side of Cordell's article, which is of course more than correct.
I was talking aboutM E N T A L (sorry, but only capital letters last time weren't enough) mechanisms.
There are many ways to get more 'dumb' gain, for example skip diff. pair emitter restistor, which lead to problems with slew rate. Robert's way of limiting gain does not change slew.
Look at schematics from 60's and early 70's and use a bit of imagination to see how someone had to say: "stop, feedback's not a cure-all".
Historically speaking, the desire of more and more feedback and carelessness of anything else led to slew problems.
Of course one can imagine high feedback amp of similar of even better slew rate than low feedback amp, but that's not really the point.
That's my view.
I was talking aboutM E N T A L (sorry, but only capital letters last time weren't enough) mechanisms.
There are many ways to get more 'dumb' gain, for example skip diff. pair emitter restistor, which lead to problems with slew rate. Robert's way of limiting gain does not change slew.
Look at schematics from 60's and early 70's and use a bit of imagination to see how someone had to say: "stop, feedback's not a cure-all".
Historically speaking, the desire of more and more feedback and carelessness of anything else led to slew problems.
Of course one can imagine high feedback amp of similar of even better slew rate than low feedback amp, but that's not really the point.
That's my view.
AKSA said:
Don't know.
But when I see that 47 labs, makers of the original uncloned Gainclone
brag about their ultrashort feedback path (definitely not global!) and
measure the feedback path as 9 mm only, then I guess the reason
is not qualification.
They say it's a good thing that they need only 1000uF in the power supply for > 50 W,
and that this makes it sonically superior.
I suppose they want to make positively sure that their customers
are terminally braindead, so nobody wants his money back.
Gerhard
darkfenriz said:
Well, for me TIM is still the same as slew rate limiting causing IM, which was known, ohh, probably before Otala was born.
But it hapens always. Someone starts to study, stumbles on something, isn't aware that it is old hat and presto! a new 'discovery' is born!
Jan Didden
Hi, AKSA,
I think about this recently. I think there is some aspect of global feedback power amp that is overlooked. It is the situation when global feedback poweramp is in use. Power amp is used by putting speaker cables and then connected to speakers, via passive xover.
From Walt Jung's article he mentioned that output node is actually another input to the feedback system. He is right, because output node is feeded-back to the differential via resistive voltage divider. So what happens at the output node is actually goes back in to the feedback process.
Another clue that I can link is one of the reason why Charles Hansen don't use global feedback. I forgot his exact words, but he said that the environment after output binding post (speaker cable+passive x-over+speaker) is acting like a huge antenna.
He makes sense, because I found out even a 10cm of PCB track can generate voltage between the ends when exposed to electromagnetic field, looked in the osciloscope. Speaker cables usually have more length than FM/AM portable antenna, so why speaker cables should not receive signals like antenna?
People put LPF in the non-inverting input of the power amp. This filters signal about 100khz-300khz. But none of the people put this kind of LPF in the inverting input, where the cables are actually receiving voltages (small?) generated by the speaker cables acting as an antenna. Small or not, this HF is entering the feedback system without any LPF blocking it, since it enters via inverting input. None of people put LPF in the inverting input/feedback input. If this really gets into feedback system, this, with the actual music signal, will be both amplified with the HF "junk" intermodulate the music signal. The feedback system (differential + VAS+ output stage) will be processing this "junk" (over-riding the music signal), because the transistors are into 10's or 100's of MHZ operation capability.
One coincidental way out (that John Curl tries to advoid along with the usage of capacitors) is to use output filter (output inductor=L). But nowdays, many amps don't use output filter anymore. I don't know if I'm thinking right about this, though.
Jan,
Norman Crowhurst wrote of such things back in the '50s....is this what you were refering to?
On a more general note I have often found that high feedback designs do tend towards all the characteristics which are usually blamed on solid state devices; hardness, glassyness, edgy etc
even when they have no (or not many) solid state devices in them. I have heard valve amps which use a lot of NFB by valve amp standards sound this way. I once designed a hybrid which was valve apart from the output pair of MOSFETS. It had loads of feedback, 0.005% THD, was fast and gave great 20Khz square waves but, unfortuneatly, it did not sound good....it sounded just like you would expect a high feedback amp to sound
As to WHY high NFB amps sound like this.....I wish I knew
Norman Crowhurst wrote of such things back in the '50s....is this what you were refering to?
On a more general note I have often found that high feedback designs do tend towards all the characteristics which are usually blamed on solid state devices; hardness, glassyness, edgy etc
even when they have no (or not many) solid state devices in them. I have heard valve amps which use a lot of NFB by valve amp standards sound this way. I once designed a hybrid which was valve apart from the output pair of MOSFETS. It had loads of feedback, 0.005% THD, was fast and gave great 20Khz square waves but, unfortuneatly, it did not sound good....it sounded just like you would expect a high feedback amp to sound
As to WHY high NFB amps sound like this.....I wish I knew
jez said:Jan,
Norman Crowhurst wrote of such things back in the '50s....is this what you were refering to?
On a more general note I have often found that high feedback designs do tend towards all the characteristics which are usually blamed on solid state devices; hardness, glassyness, edgy etc
even when they have no (or not many) solid state devices in them. I have heard valve amps which use a lot of NFB by valve amp standards sound this way. I once designed a hybrid which was valve apart from the output pair of MOSFETS. It had loads of feedback, 0.005% THD, was fast and gave great 20Khz square waves but, unfortuneatly, it did not sound good....it sounded just like you would expect a high feedback amp to sound
As to WHY high NFB amps sound like this.....I wish I knew
It's quite complicated. This weekend I again listened to some LP's recorded on CD. The CD sounded 'just like LP'. No hardness, no glare, 'round' sound. What does that tell you? Possibly we blamed the CD for the wrong things. It proved that CD's can sound quite realistic: the LP recording sounded like an LP. The conclusion should be that the LP process itself is not glaring, hard, etc: it apparently is the recording.
To play the devils advocate: that LP recording obviously missed a lot of hf energy that you would normally have on the CD because of it's superior hf capability. If you then switch from listening to LP to listening to CD, what you don't like apparently is the more realistic sound, i.e. the hf tones are there but were missing before on the LP.
What is 'better' ? Don't know. But possibly the difference between tube amps and low feedback amps vs high feedback amps could be in part to the fact that the high feedback amps are more transparently and therefore show everything that is on the source, including strong hf signals. That may not sound 'nice' to peoples taste.
I'm not saying this is the TRVTH (tm), just another possible explanation in addition to the many others that are not very satisfactory.
Jan Didden
lumanauw said:
David,
Yes, the output is also an input, just like the supply connections are inputs in a sense. But that doesn't mean it HAS to have an (audible) effect. The signal coming in through the speaker cables is heavily attenuated by the series output coil (but now seen from the other side) and the amp output impedance. I agree that at hf that output Z probably is a lot higher than for audio. So, your suggestion to put an hf filter in the feedback loop at the output side looks sensible.
So, try it out, and see if there is an audible difference. Or even a measurable difference
Jan Didden
Hi, Janneman,
I tried this sometimes ago (output using or not using output inductor).
Using output inductor the sound become more "calm", the harshness become less. Then I begin to think, is it the output of the amp that is LPF'ed by the output inductor (output inductor L + speaker load R forming LPF) or is it the "filtering" of antenna effect (by speaker cables, passive xover, etc outside the amp) entering back to the differential pair, also by the same output inductor, that is making the effect that I hear.
I tend to pick the 2nd possibility. The more "calm" sound is not because the output of the amp is LPF'ed by output inductor, but it is more that the output inductor is "preventing" any HF junk picked by antenna effect entering back to the differential system (since the feedback input has no HF blocking LPF like non-inverting input).
I will try putting LPF in the feedback voltage divider for an amp without output inductor (although I'm afraid it will alter the transfer function since it will be in the closed loop of the differential system, while output inductor is indeed outside the closed loop of the differential system)
The main idea is this : HF signal (>20khz up to Mhz) will change the audible effect by intermodulation in a global feedback amp. So preventing any HF junk, especially entering global closed loop, will make the sound better.
Since amp's input is not only from non-inverting input (also from inverting input and supply rails like you said), while we usually only put LPF on non-inverting input only, trying to isolate the possible cause one by one, maybe we can find out is it the feeback itself that is contributing to the "not nice" sound, or other factors involved , like the HF junk entering the global feedback loop that is causing this
I tried this sometimes ago (output using or not using output inductor).
Using output inductor the sound become more "calm", the harshness become less. Then I begin to think, is it the output of the amp that is LPF'ed by the output inductor (output inductor L + speaker load R forming LPF) or is it the "filtering" of antenna effect (by speaker cables, passive xover, etc outside the amp) entering back to the differential pair, also by the same output inductor, that is making the effect that I hear.
I tend to pick the 2nd possibility. The more "calm" sound is not because the output of the amp is LPF'ed by output inductor, but it is more that the output inductor is "preventing" any HF junk picked by antenna effect entering back to the differential system (since the feedback input has no HF blocking LPF like non-inverting input).
I will try putting LPF in the feedback voltage divider for an amp without output inductor (although I'm afraid it will alter the transfer function since it will be in the closed loop of the differential system, while output inductor is indeed outside the closed loop of the differential system)
The main idea is this : HF signal (>20khz up to Mhz) will change the audible effect by intermodulation in a global feedback amp. So preventing any HF junk, especially entering global closed loop, will make the sound better.
Since amp's input is not only from non-inverting input (also from inverting input and supply rails like you said), while we usually only put LPF on non-inverting input only, trying to isolate the possible cause one by one, maybe we can find out is it the feeback itself that is contributing to the "not nice" sound, or other factors involved , like the HF junk entering the global feedback loop that is causing this