John,
Is this the Halo P2(JC2) you are referring to, or is it an entirely different CTC version?
I'm confused, I stole the design from Chas Hansen! What is the schematic that goes with your picture?
This is an old design. I did it with first with bipolars as a power amp driver at Ampex Research in 1969, and then improved the feedback connection and used it as a line driver for the Grateful Dead in 1970. The Blowtorch is not this topology, even though it is superficially similar.
psu again
Dear John,
You have given us a lot of hints about the power supply of your Blowtorch.
In the particular case of this circuit which draw a constant current from the psu, you seems to care much more about noise (self induced and coming thru the mains) and constant, purely resistive output impedance of the psu and this over a wide frequency range than an absolute low Zout.
I think that I understand why considering that this output impedance is in series with the resistance (200 ohm on the proposed schematics) that feeds the folded cascode. I mean 200 or 201 ohm will not make a big change.
Will you have the same priorities with something that doesn’t have this topology? Are the priorities the same for a follower for instance?
Best regards.
Philippe (who hopes that neither is English nor his question are too “convoluted” and who is fooling around with different psu on a simple follower with more or less satisfactory results)
Dear John,
You have given us a lot of hints about the power supply of your Blowtorch.
In the particular case of this circuit which draw a constant current from the psu, you seems to care much more about noise (self induced and coming thru the mains) and constant, purely resistive output impedance of the psu and this over a wide frequency range than an absolute low Zout.
I think that I understand why considering that this output impedance is in series with the resistance (200 ohm on the proposed schematics) that feeds the folded cascode. I mean 200 or 201 ohm will not make a big change.
Will you have the same priorities with something that doesn’t have this topology? Are the priorities the same for a follower for instance?
Best regards.
Philippe (who hopes that neither is English nor his question are too “convoluted” and who is fooling around with different psu on a simple follower with more or less satisfactory results)
Plep, it is not the ABSOLUTE VALUE of a specific resistor or resistance of the power supply, but it is the stability of that value when passing signal through it, that would potentially cause distortion. Therefore, if you had a power supply buffer with 10 ohms output impedance, for example, it had better not change with changes in current or voltage across it to any measurable degree, or it will effect the sound quality.
The need for very low impedance is important when several channels share the same power supply output. If the impedance is not very low, then the various channels will talk to each other through the power supply. I made this mistake in the JC-2 preamp phono module in 1973 and didn't catch it until I specified a similar buffer for a multichannel mixer in 1975 for a Swiss company. It was embarrassing to see my oversight, but I now always use separate simple buffers for each channel or gain module, so there is no problem.
For the record, followers have inherently more power supply rejection than would complementary cascode stages fed with a 200 ohm resistor.
The need for very low impedance is important when several channels share the same power supply output. If the impedance is not very low, then the various channels will talk to each other through the power supply. I made this mistake in the JC-2 preamp phono module in 1973 and didn't catch it until I specified a similar buffer for a multichannel mixer in 1975 for a Swiss company. It was embarrassing to see my oversight, but I now always use separate simple buffers for each channel or gain module, so there is no problem.
For the record, followers have inherently more power supply rejection than would complementary cascode stages fed with a 200 ohm resistor.
john curl said:
This is incorrect. If the power supply Zout is non-zero, any varying signal load through that Zout will cause a signal-dependent ripple on the supply line. Simple ohms law.
That signal dependent ripple will cause distortion, depending on how good the amp power supply rejection ratio is. Basically, with an output series resistor in the supply line you rely on the psrr and the local decoupling to contain the damage done by that resistor.
That resistor partly undoes the advantages of the low supply impedance.
Even if you use multiple stages fed by the same supply, a low impedance supply will always have less ripple at the common supply line than when using the resistor where each stage will have self-generated ripple.
Jan Didden
Hi,
Surely, what John was trying to get over here is that it is the "stability" of ANY 'preceeding' resistance which is more important here, than any actual value of resistance.
Regards,
Surely, what John was trying to get over here is that it is the "stability" of ANY 'preceeding' resistance which is more important here, than any actual value of resistance.
Regards,
They are both right, each in their own way............
If you have 0 ohm output impedance, and feed your modules through a 10 ohm resistor, a 10 mA variation in current draw will of course cause 100 mV of ripple, being fed into the other module/channel or whatever....
if the 10 ohm is nonlinear vs. current draw, the situation is even worse.
Johns previous musings about separate regulators just makes a lot of sense, and is quite common on other types of low noise equipment.
If you have 0 ohm output impedance, and feed your modules through a 10 ohm resistor, a 10 mA variation in current draw will of course cause 100 mV of ripple, being fed into the other module/channel or whatever....
if the 10 ohm is nonlinear vs. current draw, the situation is even worse.
Johns previous musings about separate regulators just makes a lot of sense, and is quite common on other types of low noise equipment.
Bobken said:
... which is nonsense, of course. Surely the value is not up for grabs?? Would 10 ohms +/- 1 ohm be better or worse than 100 ohms +/- 0.5 ohms? Why/why not? These are just anecdotes without any factual basis.
Jan Didden
Hi Jan,
You seem determined to misunderstand or ignore that any current passing through an *unstable* resistive 'element' (however this is achieved) will cause further non-linear distortion in the circuits being fed.
As far as I am aware, no-one suggested that the impedance per se is not important, but that (in normally low impedance situations) it is important to ensure that this impedance does not vary in accordance with the signal.
Regards,
You seem determined to misunderstand or ignore that any current passing through an *unstable* resistive 'element' (however this is achieved) will cause further non-linear distortion in the circuits being fed.
As far as I am aware, no-one suggested that the impedance per se is not important, but that (in normally low impedance situations) it is important to ensure that this impedance does not vary in accordance with the signal.
Regards,
Bobken said:
Hi Bob,
My point is (possibly poorly expressed) that any resistance added will cause non-linear distortions swamping any varying resistance.
As an example, take a series resistor of 10 ohms, and a varying signal load current of say 10mA pk to peak. That will cause at the supply line 100mV signal-dependent ripple. Now let is assume that series resistance is unstable and varies 10%. The ripple will, all other things remaining equal, also vary 10%. So what? I say, do away with that 10 ohms. A simple supply regulator easily gets down to 0.1 ohms output Z. That same load current now only gives 1mV ripple. Is that not MUCH better than having 100mV?
The folklore about the need for stability of the resistance over its minimization comes and goes, and nobody seems to question it.
Jan Didden
Hi Jan,
I understand what you are saying, and it is clear that for some kinds of distortion, doubtlessly reducing PS impedances is the way to go.
However, it is not simply a question of ripple which you mention, and there are other objectional distortions to take care of here for the very best sonic results. This thread has (mostly!) been about the highest quality sounds from a pre-amp, and most of the recent queries seem to have been about improving the sonic results, whether all of these can be properly measured or not.
Making comparisons where a single PS is used to feed several amplifying stages, changing the associated resistors which decouple these stages from the rails is a very instructive exercise, and I have spent a lot of time on such trials.
Increasing or reducing these resistors clearly has some effect by altering the amount of 'crosstalk' between the stages, but as they are generally amplifying the same basic signals in the same channel, the differences are not overwhelming (within normal reason).
However, changing the type and/or the power ratings of decoupling resistors, in my experience has a quite profound effect, by comparison.
As one might expect, using very low tempco highly spec'd resistors, which seem to be more linear in holding their values under dynamic conditions, do generally 'sound' better than some others. Also, frequently using a higher rated resistor of the same value, will give sonic benefits over the same make and type of resistor, and this is most likely (I cannot say for sure) because their temps are less affected by the varying currents passing through them.
Regrettably, we disagree on the importance of how circuits actually sound in practice, but to suggest that no-one ever questions this, is not so, as my above account explains.
Regards,
I understand what you are saying, and it is clear that for some kinds of distortion, doubtlessly reducing PS impedances is the way to go.
However, it is not simply a question of ripple which you mention, and there are other objectional distortions to take care of here for the very best sonic results. This thread has (mostly!) been about the highest quality sounds from a pre-amp, and most of the recent queries seem to have been about improving the sonic results, whether all of these can be properly measured or not.
Making comparisons where a single PS is used to feed several amplifying stages, changing the associated resistors which decouple these stages from the rails is a very instructive exercise, and I have spent a lot of time on such trials.
Increasing or reducing these resistors clearly has some effect by altering the amount of 'crosstalk' between the stages, but as they are generally amplifying the same basic signals in the same channel, the differences are not overwhelming (within normal reason).
However, changing the type and/or the power ratings of decoupling resistors, in my experience has a quite profound effect, by comparison.
As one might expect, using very low tempco highly spec'd resistors, which seem to be more linear in holding their values under dynamic conditions, do generally 'sound' better than some others. Also, frequently using a higher rated resistor of the same value, will give sonic benefits over the same make and type of resistor, and this is most likely (I cannot say for sure) because their temps are less affected by the varying currents passing through them.
Regrettably, we disagree on the importance of how circuits actually sound in practice, but to suggest that no-one ever questions this, is not so, as my above account explains.
Regards,
Bobken said:
No problem Bob, I tried to give factual arguments on my views, and anybody can do with that what they wish. I'll live anyway 😉
It's of course useless to argue based on how it sounds to each of us as these opinions are not comparable.
I have always worked on the principle that if my amp changes sound if I change the type or brand of a resistor in that amp I have designed a truly spectacularly incompetent design and I really wouldn't want to advertise that!
Sorry to interrupt this cozy thread.... I'll go back to my scopes and meters 😎
Jan Didden
Jan, you make a point, but a confusing one, I think. All else being equal, a very low impedance power supply is best, and perhaps necessary if: (1) The supply is shared by many loads. (2)If the loads are varying their current draw significantly, such as class B operation.
However, very low impedance usually means high negative feedback, which then implies a tendency to transient overshoot and ringing, and this is the problem that I consider really important, even more important than a finite value final output resistance of a power supply buffer.
However if I use a simple fet buffer, for example, I have found that I can ONLY use it for an individual circuit stage, and that the stage must run in class A operation, so that the change in current across the buffer is minimum. If the resistance change of the buffer itself is minimal, then I don't see much harm done by the finite resistance of the buffer itself, so long that it is not too great a value, and 100 ohms might be a problem.
I will have to think about this further. Perhaps I can learn something here.
However, very low impedance usually means high negative feedback, which then implies a tendency to transient overshoot and ringing, and this is the problem that I consider really important, even more important than a finite value final output resistance of a power supply buffer.
However if I use a simple fet buffer, for example, I have found that I can ONLY use it for an individual circuit stage, and that the stage must run in class A operation, so that the change in current across the buffer is minimum. If the resistance change of the buffer itself is minimal, then I don't see much harm done by the finite resistance of the buffer itself, so long that it is not too great a value, and 100 ohms might be a problem.
I will have to think about this further. Perhaps I can learn something here.
- Status
- Not open for further replies.