Ok, I will make it easier for you!
Product description
Most audio amplifiers are designed and measured using well established practices that have hardly changed during last thirty or more years. Despite their widespread use, these practices are not necessarily the best available. Harmonic distortion levels are in many cases below a realistically audible limit. Still, many people claim to hear differences between amplifiers with similar THD performance. Therefore, it is clear that harmonic distortion is not the metric that should be optimized for best subjective quality. That indicates that there are other errors to be considered in amplifier design, which do not receive enough attention. In this study, Dr. Kolinummi analyses many sources of distortion and measurement methods. His focus is specifically on dynamic and transient errors, which are not normally considered in measurements, and what Kolinummi calls ‘memory errors’1. Optimization of internal linearity is thought to be the best way to optimize dynamic and transient performance and to maintain good linearity and predictable characteristics over a wide frequency and dynamic signal range. Various existing circuit topologies are analyzed and compared using both simulation and prototyping in open loop configuration. Error correction methods including negative feedback are studied to better understand their advantages and drawbacks. Several new circuit variations are proposed, specifically optimized for very high performance in open loop. In addition to almost two hundred prototype units, three complete audio power amplifiers were built and analysed during the course of this research and were used in different actual audio systems to verify the findings. This study demonstrates that there are error sources having a greater impact on amplifier performance than harmonic and intermodulation distortion, and those errors do not become visible in common measurements. Most amplifiers on the market are poorly optimized to minimize dynamic errors. It was found that the common amplifier topologies produce a large amount of both static and dynamic distortion in open loop, leading to strongly rising distortions with frequency even when the feedback loop is closed. The proposed circuit structures reduce static and dynamic open loop distortion to negligible levels and improve overall performance. Although these circuit structures can be used within a global negative feedback loop, they are eminently suited for very high performance open loop topology power amplifiers. This book does not give you a cookbook design that you can copy into your own unit, but rather aims to help you develop an insight in sources of non-linearities in amplifier circuits; methods and ways to correct or cancel these non-linearities; and methods and ways to interconnect amplifier stages into a very high performance linear system. Those who are knowledgeable in high performance analog design will have no problem to apply these methods and ways into their own designs. I am indebted to Dr. Kolinummi for allowing the use of his study material for this book. I am also thankful for critical and constructive comments during the development of this book from Edmond Verreijn-Stuart, Dejan Veselinovic and Stuart Yaniger. Jan Didden, Publisher/Editor, Linear Audio
About the Author
Arto Kolinummi is a audio enthusiast who previously worked as a university reseacher focusing on audio and medical instrumentation and teaching audio, analog electronics and reliability. PhD work about audio was finished during those ten years at the university. Currently Kolinummi works as a CEO at Icraft who develop high-tech measuring instruments for various challenging industrial applications.
Product description
Most audio amplifiers are designed and measured using well established practices that have hardly changed during last thirty or more years. Despite their widespread use, these practices are not necessarily the best available. Harmonic distortion levels are in many cases below a realistically audible limit. Still, many people claim to hear differences between amplifiers with similar THD performance. Therefore, it is clear that harmonic distortion is not the metric that should be optimized for best subjective quality. That indicates that there are other errors to be considered in amplifier design, which do not receive enough attention. In this study, Dr. Kolinummi analyses many sources of distortion and measurement methods. His focus is specifically on dynamic and transient errors, which are not normally considered in measurements, and what Kolinummi calls ‘memory errors’1. Optimization of internal linearity is thought to be the best way to optimize dynamic and transient performance and to maintain good linearity and predictable characteristics over a wide frequency and dynamic signal range. Various existing circuit topologies are analyzed and compared using both simulation and prototyping in open loop configuration. Error correction methods including negative feedback are studied to better understand their advantages and drawbacks. Several new circuit variations are proposed, specifically optimized for very high performance in open loop. In addition to almost two hundred prototype units, three complete audio power amplifiers were built and analysed during the course of this research and were used in different actual audio systems to verify the findings. This study demonstrates that there are error sources having a greater impact on amplifier performance than harmonic and intermodulation distortion, and those errors do not become visible in common measurements. Most amplifiers on the market are poorly optimized to minimize dynamic errors. It was found that the common amplifier topologies produce a large amount of both static and dynamic distortion in open loop, leading to strongly rising distortions with frequency even when the feedback loop is closed. The proposed circuit structures reduce static and dynamic open loop distortion to negligible levels and improve overall performance. Although these circuit structures can be used within a global negative feedback loop, they are eminently suited for very high performance open loop topology power amplifiers. This book does not give you a cookbook design that you can copy into your own unit, but rather aims to help you develop an insight in sources of non-linearities in amplifier circuits; methods and ways to correct or cancel these non-linearities; and methods and ways to interconnect amplifier stages into a very high performance linear system. Those who are knowledgeable in high performance analog design will have no problem to apply these methods and ways into their own designs. I am indebted to Dr. Kolinummi for allowing the use of his study material for this book. I am also thankful for critical and constructive comments during the development of this book from Edmond Verreijn-Stuart, Dejan Veselinovic and Stuart Yaniger. Jan Didden, Publisher/Editor, Linear Audio
About the Author
Arto Kolinummi is a audio enthusiast who previously worked as a university reseacher focusing on audio and medical instrumentation and teaching audio, analog electronics and reliability. PhD work about audio was finished during those ten years at the university. Currently Kolinummi works as a CEO at Icraft who develop high-tech measuring instruments for various challenging industrial applications.
Yes. it's a rewritten version of the author's PhD thesis.
Error sources and their reduction techniques in audio power amplifiers - Tampere University of Technology
Error sources and their reduction techniques in audio power amplifiers - Tampere University of Technology
Last edited:
I don't understand the controversies about GNFB.
Open loop bandwidth, open loop gain (and the possible amount of feedback it brings), phase turns and phase margins (poles, stability), distortions.. al this create a multidimensional map where everybody is free to put his own sweet spot.
And each choice of a designer regarding the balance of those elements, each one with its own benefits and drawbacks, in absence of any indisputable audibility thresholds about harmonic and IM/FM distortions, phase linearity, overshoots, dumping factor etc., are just points of views and a matter of personal preferences.
As I read JC, for a given "C" closed loop bandwidth, with close levels of harmonic distortions, he prefers B to C as open loop. (see attachment)
So do I, for various reasons, including listening experiences.
If you agree with this (you are free to do not) it seems logical to prefer C ... as open loop bandwidth ;-)
No one deny the benefits of GNFB and Bruno P. tell the same, saying "if you use GNFB, use a lot". I would add ... "as long as it does not introduce too much drawbacks on other aspects of the amp behavior".
It is obvious that the first thing to try, when we begin to design an amplifier with GNFB, is to improve as much as possible the open loop performance.
And the second, obvious for me, that we have to improve as much as possible the way the feedback signal is subtracted to the original signal in the input stage (VFA VS CFA VS passive = inverting configuration).
Not to forget that we are feeding speakers, IE very imperfect devices, with a dynamic signal constantly changing : Some aspects of the amp behaviors will influence both the way the speakers will react to a given signal and the way the amp will react to the e.m.f they will produce (when gnfb, the output is an input too).
Open loop bandwidth, open loop gain (and the possible amount of feedback it brings), phase turns and phase margins (poles, stability), distortions.. al this create a multidimensional map where everybody is free to put his own sweet spot.
And each choice of a designer regarding the balance of those elements, each one with its own benefits and drawbacks, in absence of any indisputable audibility thresholds about harmonic and IM/FM distortions, phase linearity, overshoots, dumping factor etc., are just points of views and a matter of personal preferences.
As I read JC, for a given "C" closed loop bandwidth, with close levels of harmonic distortions, he prefers B to C as open loop. (see attachment)
So do I, for various reasons, including listening experiences.
If you agree with this (you are free to do not) it seems logical to prefer C ... as open loop bandwidth ;-)
No one deny the benefits of GNFB and Bruno P. tell the same, saying "if you use GNFB, use a lot". I would add ... "as long as it does not introduce too much drawbacks on other aspects of the amp behavior".
It is obvious that the first thing to try, when we begin to design an amplifier with GNFB, is to improve as much as possible the open loop performance.
And the second, obvious for me, that we have to improve as much as possible the way the feedback signal is subtracted to the original signal in the input stage (VFA VS CFA VS passive = inverting configuration).
Not to forget that we are feeding speakers, IE very imperfect devices, with a dynamic signal constantly changing : Some aspects of the amp behaviors will influence both the way the speakers will react to a given signal and the way the amp will react to the e.m.f they will produce (when gnfb, the output is an input too).
Attachments
Last edited:
Since its not difficult why aren't people just hanging a current probe on the output of an amp connected to a nasty speaker and collect some real data? I once configured an analog multiplier on the output of an amp to plot power and load impedance. It was fun but no exceptional breakthroughs came from it.
Actually, I think a lot of folks here bought it read through it.
I did.
It has lots of interesting ideas.
Would be nice if some were discussed in this thread or elsewhere on the site.
But, for me, one key thing missing from the book is construction photos.
Some of his measurements are quite impressive and I would like to see how the prototypes were built (PCB layout, wiring, etc.) were done to facilitate such results.
That would be some REALLY useful info!
Unfortunately, the author has moved on to other things, so it's not likely such photos would ever be seen ...
mlloyd1
I did.
It has lots of interesting ideas.
Would be nice if some were discussed in this thread or elsewhere on the site.
But, for me, one key thing missing from the book is construction photos.
Some of his measurements are quite impressive and I would like to see how the prototypes were built (PCB layout, wiring, etc.) were done to facilitate such results.
That would be some REALLY useful info!
Unfortunately, the author has moved on to other things, so it's not likely such photos would ever be seen ...
mlloyd1
Have you received the amps?
By the way, where is possible to buy that nice stepped volume control?
BR Damir
That stepped control is way, way expensive. But, I can ask who makes it.
I dont have the amps in California yet.
-Richard
I dont have the amps in California yet.
-Richard
The Chinese eBay DACT doesn’t have lots of steps but is likely better than that. At least it beats another RN one greatly (but it may be a question of the switch itself)
I’ve got two of her songbook albums (plus a bunch of stuff on CD) - both very good recordings. Never tire of them and the recording quality (especially the Irving Berlin one) is superb. Can’t beat brass off an LP IMV.
- Status
- Not open for further replies.