Charles Hansen said:
Charles, I'll ping you privately I don't want to give away anyone's discovery without their permission.
Re: Re: Re: current driven?
Hmmm... I see nothing common between current control and flat, constant, well defined beta, whatsoever.
Well, if you can't answer this question, can you answer the other one? Where do you get voltages to control devices, if all of them are current sources?
Edmond Stuart said:
Hmmm... I see nothing common between current control and flat, constant, well defined beta, whatsoever.
Well, if you can't answer this question, can you answer the other one? Where do you get voltages to control devices, if all of them are current sources?
Re: Re: Re: Re: current driven?
For obvious reasons, I'm not willing to answer that question, that is, giving examples of 'very wrong thinking'.
Is the output of a pre-amp a current source? NO.
Is the output of a Miller compensated VAS a current source? NO.
Is the output of an emitter follower a current source? NO.
Is the output of a LTP a current source? YES. But most of the time, this one is tied to a next stage with low impedance, a VAS for example, which transfer function is mainly determined by gm and the compensating cap and far less by beta. Moreover, the transfer function of the LTP itself is mainly determined by gm, which in turn is defined by the exponential relationship of Ic/Vbe and emitter degeneration resistors.
As for beta, in all cases, this wildly variable variable is supposed to be just high enough and should not have a major impact on the performance. If it does have a major impact, I regard such a design in question as flawed.
Cheers,
Edmond.
Wavebourn said:
For obvious reasons, I'm not willing to answer that question, that is, giving examples of 'very wrong thinking'.
Is the output of a pre-amp a current source? NO.
Is the output of a Miller compensated VAS a current source? NO.
Is the output of an emitter follower a current source? NO.
Is the output of a LTP a current source? YES. But most of the time, this one is tied to a next stage with low impedance, a VAS for example, which transfer function is mainly determined by gm and the compensating cap and far less by beta. Moreover, the transfer function of the LTP itself is mainly determined by gm, which in turn is defined by the exponential relationship of Ic/Vbe and emitter degeneration resistors.
As for beta, in all cases, this wildly variable variable is supposed to be just high enough and should not have a major impact on the performance. If it does have a major impact, I regard such a design in question as flawed.
Cheers,
Edmond.
I hope that it has become more clear that beta, especially non-linear beta is an important factor in many design approaches. Especially when you are trying to do the most with the least number of active devices. Many quality audio designs try to do the most with with least number of series junctions. This is but one approach, but it is my approach and it has been successful for decades. We call it 'stage count' and I did not invent the concept. Unfortunately, our definition of 'stage count' differs from a more common definition counting the number of voltage amplifying stages. Our definition adds up each device using 1 for a gain device and 1/2 for a cascode or follower. Minimum 'stage count' is often the best, all else being equal.
minimum stage count
John, that's my philosophy as well.
Some people laugh at me when I'm promoting a common mode control loop (CMCL) in order to tightly define the standing currents of the a complementary VAS topology, which, at the same time, also boosts the gain of VASes with over 20dB without adding more stages to the signal path.
Cheers,
Edmond.
John, that's my philosophy as well.
Some people laugh at me when I'm promoting a common mode control loop (CMCL) in order to tightly define the standing currents of the a complementary VAS topology, which, at the same time, also boosts the gain of VASes with over 20dB without adding more stages to the signal path.
Cheers,
Edmond.
Lumba Ogir said:
I would suggest you the reading of the book from P.E. Gray and C L Searle Electronic Principles, your view of odd teacher will change
This is the course at MIT in the late sixties and seventies.
It is the best book IMHO on transistor operation linking devices physics and electrical models in an easy, accessible but in depth way.
A nice view of transistor operation is that it acts at the collector as a controlled current source.
There are 4 points of view concerning the control of this current source: emiter-base voltage, excess charge in base, base current, emitter current.
The book explains well why the emitter -base voltage as a control element is the most usefull view and why the current control is not.
The charge control is another usefull view introducing the non linear capacitance of a dynamic model.
The base current is a recombination and injection current essentially a defect current coming from second order effects.
A parameter (delta) the base defect quantifies this and is the original parameter: Ib=delta Ic
alpha or beta are defined using this delta:
alpha = ( 1/1+delta) = Ic/Ie
beta = 1/delta= Ic/Ib.
Between collector current and emitter base potential difference, there is an accurate exponential relation based on the the statistics of carrier distribution. these are first order physical laws that can be measured accutately.
delta is a parameter difficult to control, coming from second order effects and showing large variation in the same familly. It is wrong to design based on this parameter that is ill quantified.
A complement to Scott's explanation.
JPV
I am not sure who is saying what, anymore, but let's get down to basics.
In the early days, the mid-'60's, especially with germanium transistors, BETA was everything, 15 years earlier, ALPHA was everything.
About 1965, linear IC design was introduced and it utilized Vbe tracking and current mirrors, almost exclusively, yet discrete designs did NOT follow this guideline, because, 1. It did not have to, to get good performance, and transistors cost real money, in those days. 2. The circuits that we developed for audio were NOT used to pass DC through them, in fact this would be a bad thing, so our circuits relied much more on beta and its linearity, than almost anything else.
Now Scott mentioned something that on the face of it, would make me laugh, but I am sure that it is a misunderstanding.
That is Gm of semiconductors. You don't have to MEASURE Gm in a bipolar transistor, it is always the same for a GIVEN CURRENT. YES, Gm is always 40,000 at 1ma, changing up or down, depending on standing current. IF you know the standing current, you know the Gm, that is all there is to it, I am pretty sure. Please prove me wrong, if you can.
However, both tubes and fets always have somewhat less Gm at 1 ma. This depends on the construction of the device and its operating point as to how close it can track a transistor. That is why we measure Gm in tube testers. For fets, we normally don't have meters that directly measure Gm, so we usually look it up for a given current from the data sheet. The fets are so consistent within their own group, even with time, we don't need to measure them over time to see if they are still doing what they are supposed to do. This is not the case with tubes, and why there are Gm tube testers.
In the early days, the mid-'60's, especially with germanium transistors, BETA was everything, 15 years earlier, ALPHA was everything.
About 1965, linear IC design was introduced and it utilized Vbe tracking and current mirrors, almost exclusively, yet discrete designs did NOT follow this guideline, because, 1. It did not have to, to get good performance, and transistors cost real money, in those days. 2. The circuits that we developed for audio were NOT used to pass DC through them, in fact this would be a bad thing, so our circuits relied much more on beta and its linearity, than almost anything else.
Now Scott mentioned something that on the face of it, would make me laugh, but I am sure that it is a misunderstanding.
That is Gm of semiconductors. You don't have to MEASURE Gm in a bipolar transistor, it is always the same for a GIVEN CURRENT. YES, Gm is always 40,000 at 1ma, changing up or down, depending on standing current. IF you know the standing current, you know the Gm, that is all there is to it, I am pretty sure. Please prove me wrong, if you can.
However, both tubes and fets always have somewhat less Gm at 1 ma. This depends on the construction of the device and its operating point as to how close it can track a transistor. That is why we measure Gm in tube testers. For fets, we normally don't have meters that directly measure Gm, so we usually look it up for a given current from the data sheet. The fets are so consistent within their own group, even with time, we don't need to measure them over time to see if they are still doing what they are supposed to do. This is not the case with tubes, and why there are Gm tube testers.
john curl said:
Jeeezzz John... Given the bias, don't you know how to calculate a JFET or tube gm?
It is virtually impossible to know what a fet Gm is without measuring it. That is NOT true with bipolar transistors, either Germanium or Silicon, and probably most else.
JPV said:
It is not wrong; it is difficult to use comprehensive equations. But it is rewarding, if to do properly. Simplified approach is valid when topology and technology forgives such simplifications, but such a way you can't design something new and exciting. The result will be always the same: nice looking devices, good measurements, but sound worse than what Edmond Stewart calls "flawed designs".
The only way to escape from the cage of topologies that allows you to omit variables hard to use in calculations, is the way when you have to face the reality as it is, instead of through rose glasses of cheap and easy models.
Far the more, I am getting better results from breadboarding and measurements than from calculations. Such a way I don't need to say calculate using dependencies of capacitances on voltages driven by sources with non-linear resistances; I can breadboard, measure, extrapolate, optimize. I can use calculations that involve each time parameters that play significant role in the particular topology, instead of using topologies that are easy to model.
"Where you have lost your keys, Sir?" "Over there, Officer, on the parking lot!" "But why are you searching here?" "It is lighter here, under the street light!"
What is good for design of artificial devices for a virtual SPICE world, not always is the best way to design High-End audio that still is the art of design and engineering, kind of magic for the most of designers who can't understand why end users prefer "flawed" designs over their "perfect" ones. And the more designers rely on simplified models, the blacker is the magic of natural sounding audio for them.
Re: Re: Re: Re: Re: current driven?
Come on... Sometimes what believers call "very wrong thinking" is the key for heretics to get better results. However, in such cases believers have the standard belief: "Heretics prefer wrong results!" 😀
Yes, it is source of current that depends on load impedance.
Yes, as soon as it drives any load.
Yes, and current it supplies depends on input impedance of the source that drives the emitter follower.
Most of the time because it is better to do so? No, because it is easier to make ICs with predictable output such a way.
All your cases involve wildly variable impedances. It does impact on performances. How linear is open loop opamp, and how it's transfer function depends on frequency? How it depends on temperature variations caused by the signal itself? And so on, and so forth...
...what is " stable and linear" for plastic boombox is "wildly variable" for high-end gear.
Edmond Stuart said:
Come on... Sometimes what believers call "very wrong thinking" is the key for heretics to get better results. However, in such cases believers have the standard belief: "Heretics prefer wrong results!" 😀
Yes, it is source of current that depends on load impedance.
Yes, as soon as it drives any load.
Yes, and current it supplies depends on input impedance of the source that drives the emitter follower.
Most of the time because it is better to do so? No, because it is easier to make ICs with predictable output such a way.
All your cases involve wildly variable impedances. It does impact on performances. How linear is open loop opamp, and how it's transfer function depends on frequency? How it depends on temperature variations caused by the signal itself? And so on, and so forth...
...what is " stable and linear" for plastic boombox is "wildly variable" for high-end gear.
Re: minimum stage count
Hi Edmond,
Some people are interested by CMCL :
http://www.diyaudio.com/forums/showthread.php?postid=1213768#post1213768
and following posts.
There were also some schematics of CMLC published in EWWW (Jager's hybrid amp, Marcel Van de Gevel's mosfet amp and one from you, I think).
Edmond Stuart said:
Hi Edmond,
Some people are interested by CMCL :
http://www.diyaudio.com/forums/showthread.php?postid=1213768#post1213768
and following posts.
There were also some schematics of CMLC published in EWWW (Jager's hybrid amp, Marcel Van de Gevel's mosfet amp and one from you, I think).
JPV said:
Paul and Campbell were both teachers, the book with covers worn off still sits near my desk.
john curl said:
So did the Japanese and they almost exclusively employed qualified engineers to design their gear and they have many patents to prove it.
I don't know anyone who owns your gear but I know plenty of people who own Japanese equipment from the 70's, 80's and 90's so what are you trying to say ??
Ray, we run in different circles. If you have never been near any of my designs over the last 35 years, so be it.
Just this weekend, I received a request for an upgrade to a Levinson design that I made 34 years ago. That individual surprised me in even knowing about the upgrade, because I only did it once, and to my own personal unit.
For everyone: It might be time to reset, and determine what we want to discuss. This is getting pretty tiresome, and it is circular in argument.
Scott Wurcer makes pretty good IC's, I use some of them myself.
Wavebourn, apparently makes some interesting stuff, and he has a unique point of view, but I don't know how the rest of us can use his input.
PMA has seen the light! Good for him.
Dimitri is a good friend of mine, and a good engineer.
I will leave it at that.
Just this weekend, I received a request for an upgrade to a Levinson design that I made 34 years ago. That individual surprised me in even knowing about the upgrade, because I only did it once, and to my own personal unit.
For everyone: It might be time to reset, and determine what we want to discuss. This is getting pretty tiresome, and it is circular in argument.
Scott Wurcer makes pretty good IC's, I use some of them myself.
Wavebourn, apparently makes some interesting stuff, and he has a unique point of view, but I don't know how the rest of us can use his input.
PMA has seen the light! Good for him.
Dimitri is a good friend of mine, and a good engineer.
I will leave it at that.
- Status
- Not open for further replies.