Some really constructive ideas here. Time to do some serious cut and paste into notepad so I don't loose them. The Hafler will be my test unit as it is easy to work on, has many comments published, and fails the listening test. It for sure has physical layout issues. Mine seemed to be kit built so the wiring was not even up to the manual. It will be a slow slog as I need to do a lot of background reading. My old FA lab days were mostly related to component failures due to miss use so, quite clear to those here who design amps for a living, I only know enough to ask questions. Every day is for learning.
As I use the term, propagation delay, measured in the single unit of time, is the difference between an input signal at some chosen threshold, and the equivalent threshold on the output. It is not usually considered frequency dependent. My background is more digital so there may be a more common use in linear.
Slew is a rate of change measured in change-time, as in volts per microsecond. So, it clearly defines the bandwidth upper limit.
Every circuit has propagation delay as electrons only move about 7 inches a second through a wire. Not an AF problem, but fundamental for any who look for literal exactness. Of course, all circuits have a maximum bandwidth, the limit of such in the universe is yet to be fully described but is assumed to be the Planck frequency of 2 times 10 to the 43rd Hz. Today's trivia.
Slew is a rate of change measured in change-time, as in volts per microsecond. So, it clearly defines the bandwidth upper limit.
Every circuit has propagation delay as electrons only move about 7 inches a second through a wire. Not an AF problem, but fundamental for any who look for literal exactness. Of course, all circuits have a maximum bandwidth, the limit of such in the universe is yet to be fully described but is assumed to be the Planck frequency of 2 times 10 to the 43rd Hz. Today's trivia.
Sy,
It is not the definition of propagation delay, it is how to derive an approximation of the value for a given circuit.
Propagation delay is literally the time by which a signal is delayed by the circuit it passes through, clearly bandwidth is normally the main determinant (there are exceptions), especially for feedback amplifiers.
Ciao T
It is not the definition of propagation delay, it is how to derive an approximation of the value for a given circuit.
Propagation delay is literally the time by which a signal is delayed by the circuit it passes through, clearly bandwidth is normally the main determinant (there are exceptions), especially for feedback amplifiers.
Ciao T
Hi,
This arguably is very true, but not in the way you wish to represent it.
Degeneration lacks the distortion multiplication effect of looped feedback (e.g. shown by Baxandall, Olson et al). According to extant research (Boyk at all) degeneration does in fact lower harmonic distortion for nearly all distortion products (dependent somewhat on circit and device) by the approximately the amount of the degeneration applied, something which looped feedback cannot accomplish.
Further, if there is less distortion fed to the following stage distortion multiplication effects are reduced and whatever feedback we choose to apply is more effective dealing with the low order harmonics that remain, rather than with the multiplied products.
So degeneration is not the same as global looped feedback, because it is, if smartly applied considerably superior in reducing all distortions than to just apply more global looped feedback.
What you point out is in direct opposition to widely published observations. So if you wish to have your contentions to be considered, you would have to provide proof that widely published observations are in fact wrong and your claims that are in opposition to them are in fact true.
On the contrary, I did not downplay the effects, but I included them and mentioned them explicitly.
The Miller compensation forces the undegenerated input stage to swing more signal for a given output voltage, so it's distortion is increased. However the VAS stage's distortion after the input stage is decreased. Overall it leads usually to an increase in the distortion compared to no miller feedback.
You are of course welcome to present your own reasoning and math in detail if you wish.
Ciao T
This arguably is very true, but not in the way you wish to represent it.
Degeneration lacks the distortion multiplication effect of looped feedback (e.g. shown by Baxandall, Olson et al). According to extant research (Boyk at all) degeneration does in fact lower harmonic distortion for nearly all distortion products (dependent somewhat on circit and device) by the approximately the amount of the degeneration applied, something which looped feedback cannot accomplish.
Further, if there is less distortion fed to the following stage distortion multiplication effects are reduced and whatever feedback we choose to apply is more effective dealing with the low order harmonics that remain, rather than with the multiplied products.
So degeneration is not the same as global looped feedback, because it is, if smartly applied considerably superior in reducing all distortions than to just apply more global looped feedback.
What you point out is in direct opposition to widely published observations. So if you wish to have your contentions to be considered, you would have to provide proof that widely published observations are in fact wrong and your claims that are in opposition to them are in fact true.
On the contrary, I did not downplay the effects, but I included them and mentioned them explicitly.
The Miller compensation forces the undegenerated input stage to swing more signal for a given output voltage, so it's distortion is increased. However the VAS stage's distortion after the input stage is decreased. Overall it leads usually to an increase in the distortion compared to no miller feedback.
You are of course welcome to present your own reasoning and math in detail if you wish.
Ciao T
I was thinking back to our problems in industry with jfet input op amps. Just picking one up off the static pad with tweeter, strap, ion flow, was enough to get esd caused damage to the inputs enough to cause input offset current. We discovered this due to a very bad sensor design. It seems they were only safe to about 20v, and you could exceed that just moving it through the dry Colorado air.
Hi Thorsten,
I'm being lazy (in not doing the circuit analysis) but intuitively I would expect to get the 10dB reduction in distortion only if the stage's gains are increased back the same pre-degenerated value.
Thanks
-Antonio
You have a problem Tvr, and no real experience. I have had only 1 Blowtorch input failure in my experience, AND it was most probably due to an input transient from another piece of equipment (tube), as the SAME customer blew up 2 Parasound power amps as well, in the same way.
I lost maybe 2 Vendetta phono stages over 20 years, due to lightning, and 1 Vendetta jfet follower xover that was tracked to a house wiring problem. In the meantime, as far as fet input op amps, I have servoed thousands of power amps and preamps, using LF411, AD711, and OPA134, with no significant failures.
However, if Colorado is a real pain, because of the lightning storms, etc, I would highly recommend a lightning protection box to be used with any equipment that I design. We find that making a design 'idiot proof' by putting in input zeners, for example, takes away a certain percentage of the sound quality. I have had to add them, however for problem customers, or customers with problems, depending.
I lost maybe 2 Vendetta phono stages over 20 years, due to lightning, and 1 Vendetta jfet follower xover that was tracked to a house wiring problem. In the meantime, as far as fet input op amps, I have servoed thousands of power amps and preamps, using LF411, AD711, and OPA134, with no significant failures.
However, if Colorado is a real pain, because of the lightning storms, etc, I would highly recommend a lightning protection box to be used with any equipment that I design. We find that making a design 'idiot proof' by putting in input zeners, for example, takes away a certain percentage of the sound quality. I have had to add them, however for problem customers, or customers with problems, depending.
Last edited:
This is what we discovered in our failure analysis lab. It was a long time ago, but I think they were TLO82's. We discovered the problem in-circuit, and then discovered how fragile the parts were so much so they did not survive the automated assembly. We switched to a different chip, from LT if I remember, and the problem was cured. The circuit was not an input exposed to the world, but an optical sensor for a bellows vacuum gauge within the system. A better designed circuit would not have been so sensitive to input offset.
As Anderson states, a device subject to ESD is a pregnant part, one that will give birth to a failure in the future. ( inventor of pink poly among other things) Analysis by Lockheed on failed avionics showed the myth of components being "safe" once in the circuit and in the box, so recommendations for well designed input protection and power supply protections can't be stressed enough.
Another esd story was that we shipped a unit containing a wire-wrap mother board on a pallet with a poly bag over it. I did a statistical analysis and discovered a majority of failures of components all attached to a mother board pin. Pulling the bag off the unit allowed all those ww pins which looked the world like perfect HV antenna, to get zapped. A piece of carbon foam and pink poly bag solved the problem.
I have no audio amp consumer experience, ( obviously) but I did failure analysis for ten years before running corrective action. Our systems I worked on as an engineering tech contained linear power amps for servo, read/write amps, and big linear power supplies. I then did field support for a few years before a bit of systems administration and then left industry for systems architecture. Audio is my hobby now I sold my TVR.
My guess is the LP filter on an amp input to do a petty good protection job on esd. You can't make it fool prof as 1: those fools are so damn clever, and 2: they keep making better fools.
As Anderson states, a device subject to ESD is a pregnant part, one that will give birth to a failure in the future. ( inventor of pink poly among other things) Analysis by Lockheed on failed avionics showed the myth of components being "safe" once in the circuit and in the box, so recommendations for well designed input protection and power supply protections can't be stressed enough.
Another esd story was that we shipped a unit containing a wire-wrap mother board on a pallet with a poly bag over it. I did a statistical analysis and discovered a majority of failures of components all attached to a mother board pin. Pulling the bag off the unit allowed all those ww pins which looked the world like perfect HV antenna, to get zapped. A piece of carbon foam and pink poly bag solved the problem.
I have no audio amp consumer experience, ( obviously) but I did failure analysis for ten years before running corrective action. Our systems I worked on as an engineering tech contained linear power amps for servo, read/write amps, and big linear power supplies. I then did field support for a few years before a bit of systems administration and then left industry for systems architecture. Audio is my hobby now I sold my TVR.
My guess is the LP filter on an amp input to do a petty good protection job on esd. You can't make it fool prof as 1: those fools are so damn clever, and 2: they keep making better fools.
Good choice: both examine similar, but different enough topologies in details.
So global feedback while lowering thd increases propagation delay ...?
LOL.....
....
- Status
- Not open for further replies.