So? that doesn't stop it being the future. For a long time Class-D was rubbish but now there exists the combination of suitable devices and suitable understanding to make very not-rubbish class-D amplifiers.
With more and more high performance BJTs becoming EOL (i.e. most of the Japanese ones) I feel it's time to embrace the inevitable: class-D with high-order control loops can deliver amplifiers with higher fidelity over a wider range of operating conditions than any realistic (i.e., not one that doubles as a room heater and/or costs thousands of $ in parts and labour) linear amplifier. Having said that, for a little while there may be room for hybrid (linear + switch-mode) amplifiers.
"Definitely more attention to laterals. These days, the SS manufacturers are putting much more emphasis on switching type applications. Hopefully, more attention will make more devices useful for linear amplification more available. It's not good that this has been so thoroughly short changed lately. There are already more than enough sub standard SS amps on the market, and SS doesn't have to sound as horrible as it does all too often. However, sonic excellence still needs devices that can perform well in that application."
Very few people are laterals for switching applications nowadays. Most this has shifted to Trench technology, which has very low Rdson/mm2. These a great for class D but other than for a very few cases, the linear operating SOA is weak compared to conventional lats and verticals. The advantage of trench is that you can get very high cell densities, and the perfect place to make them is in a DRAM fab - preferably one that fully depreciated.
Very few people are laterals for switching applications nowadays. Most this has shifted to Trench technology, which has very low Rdson/mm2. These a great for class D but other than for a very few cases, the linear operating SOA is weak compared to conventional lats and verticals. The advantage of trench is that you can get very high cell densities, and the perfect place to make them is in a DRAM fab - preferably one that fully depreciated.
Certainly more on nested and multiple loops and their stability! I already asked here for references because it's very hard to find useful information on this topic. Still on the lookout.
And more on LTP to VAS interconnections! Not much contributed to my current thread. Not sure if this is lack of interest or of information.
Class D will indubitably increase, it's just that I think it would be better to have really comprehensive core material and leave the class D to a separate book. A really excellent book and a possible second beats one book somewhat compromised by space restrictions. Kind of an aesthetic issue too. The class D is so clearly a distinct unit that it sticks out for a separate and probably more detailed examination.
Best wishes
David
Last edited:
The problem with in depth design references is that they address mainly IC design. The objective of power amplifier topologies is reduction of distortion and stability. Ic design is also concerned by low power design, rail to rail, enough gain in submicron geometries, faisability of components, different matching performances, use of low voltage power supply
Therefore it is ( at least to me ) not obvious if the different advanced topologies found in IC design would have an advantage in amp design. At the very low level of distortion components, it is difficult to measure, simulate accurately and then rank these topologies.
The books of Cordel and Self are very interesting in this aspect but a comprehensive approach of the usefulness of different topologies in power amplifier design is still to be completed.
JPV
Multi-loop modulators look even more difficult to analyse than multi-loop amplifiers so I don't see how they help my comprehension
But thanks for the advice, I'll have a try, so where do you recommend I start?
Do you have any references?
Yes, that is my point. I want Bob to include advanced topics like multiple loops that are relevant but hard to find elsewhere without a lot of IC specific clutter.
And thank you for your comments in post #2573 that relate to my question about differential current mirrors.
Best wishes
David
Multi-loop modulators look even more difficult to analyse than multi-loop amplifiers so I don't see how they help my comprehension
But thanks for the advice, I'll have a try, so where do you recommend I start?
Do you have any references?
Yes, that is my point. I want Bob to include advanced topics like multiple loops that are relevant but hard to find elsewhere without a lot of IC specific clutter.
And thank you for your comments in post #2573 that relate to my question about differential current mirrors.
Best wishes
David
The topology we are discussing is a typical example of an IC design where active elements are used to create gain because high resistors are not available as in discrete design.
IIRC, when M. Leach was asked why he didn't use current mirror load in the first stage of his amplifier, he said that he could generate enough gain with simply resistors. Self (again IIRC) measured that the use of a current mirror forces balance in the current of both transistors reducing even order distortion components. But if we use enough degeneration in th differential amplifier and if we gain back the loss by using high resistors and why not Fet's inthe Vas to improve loading, we could have the same performance without use of current mirrors and their own non linearities. Spice is not simulating very well Early distortion so to answer these questions, it is necessary to build and measure at this low level of distortion.
As a side remark, I would call the topology in Cordell's book: high impedance self biasing active load. Samuel's input is certainly as valuable.
JPV
Elevated impedance levels have their own problems with stray capacitance, non-linear device capacitance becomes more of a problem, noise pickup and so on.
I plan to use a compensation scheme that can usefully exploit more feedback. Hence my interest.
More noise from increased emitter resistors of course. FETs in the VAS was looked at by Samuel and he had some concerns that I need to re-read. FETs in the OPS driver looks to have promise I think.
Yes, MANY trade-offs to be explored! I just want to understand them better.
Best wishes
David
High gain stage requires high impedance node.
You could generate overall high gain by cascading but then you have stability problems.
It seems that there are many trade offs like noise and distortion
JPV
I think local error correction and positive feedback are key. Think Baxandall CCS and Tringlotron. Edmond Stuart's TIS implements the former, but does not eliminate the intrinsic logarithmic transfer curve distortion. Although in this case the output stage generates most of the distortion anyways. If we correct transfer curve distortion, then we don't need to make the compromise of degeneration, or bother with "swamping out" nonlinearities. I think the field of BJT analog computing may hold great revelations in this regard. After all they are useful for this purpose because of the reliability with which they conform to their logarithmic transfer curve.
I have studied the AD797 that uses the elevated impedance approach. Nice on an IC where stray capacitance can be minimized but I worry for discrete circuitry.
Rather that raise the collector load impedance I think to lower the emitter impedance - it is the ratio that matters of course. Potential for lower noise and better stability at the same time. Need to increase the LTP current, so the trade-offs become with thermal problems. Maybe paralleled LTP transistors to spread the heat and lower Rbb at the same time. Lower the feedback network too. Burns a little power but is not too complex or expensive, uses what discrete circuitry does best. Sound reasonable?
Best wishes
David
Multi-tanh input could be another addition to Bob's second edition (or second volume!) but there may be more productive topics to cover.
Has anyone ever used this in an audio amplifier?
Best wishes
David
Last edited:
This book good for beginners? Can someone suggest a book for someone just starting? I have some experience with a Bottlehead Preamp, but nothing with high voltages or dealing with amplification. I understand parts of concepts, and am planning an F5 Turbo but am not near the level of comprehension that I could build it with just a schematic and no hand holding. I might be able to... But I'd never try. I am planning to try maybe an LM4780 Kit or PCB and BOM to throw one of those together a little cheaper and use that one to learn before ripping into an F5. I just feel I need to up my chops, and rad more.
Is this more of an advanced book? Does the level of comprehension really make or break this book, or can a beginner pick it up, tough through it and learn something properly, or should one be taking baby steps?
Is this more of an advanced book? Does the level of comprehension really make or break this book, or can a beginner pick it up, tough through it and learn something properly, or should one be taking baby steps?
I would not teach electronics with this book. It's more about theory and design, and it really helps to have basic electronic understanding from the beginning. I don't think this book will help with what you're trying to do, but it might once you start with more advanced designs. If you want to learn electronics, you might try the free online course here:
https://www.edx.org/
https://www.edx.org/
Oh yeah, the multi-tanh! It does not correct Vbe distortion though, it just combines and you get a weird wavy transfer curve.
This seems like such a great idea but if you talk to audio guys about it their eyes glaze over and they start repeating "no... no... no...". I suppose I should have ignored them at the time. It would introduce it's own special kind of distortion, I know that much. But the total distortion would be less. Sounds fun.
This seems like such a great idea but if you talk to audio guys about it their eyes glaze over and they start repeating "no... no... no...". I suppose I should have ignored them at the time. It would introduce it's own special kind of distortion, I know that much. But the total distortion would be less. Sounds fun.
Hi mortron,
I would not recommend it for a true beginner, although it depends a bit on the definition of a beginner. I tried very hard to start off easy and then go deep in the book, but of course there is a big compromise in that when one is limited to about 600 pages. There are a lot of books out there that could precursor reading this one. If you know about and have build some transistor circuits and understand schematics and devices like transistors, capacitors diodes and resistors you have a fighting chance. You definitely do not have to have an EE degree to get a lot out of this book and enjoy amplifier-building as a great hobby. I deliberately tried to keep the math to a reasonable level. You certainly don't need to know the Laplace transform and the like for this. I do in the early chapters try to give a person an easily digestible idea of things like poles and zeros and their relationship to frequency response. Similarly, I try to explain in straightforward terms negative feedback and stability.
You can probably go onto Amazon and read pieces of it to get a flavor for the level of it and how well it matches your current technical level.
Hope this helps.
Cheers,
Bob
there is a "maximally flat" version - no "waveyness"
but it is easy to have enough global loop gain at audio that the the input error V at audio frequencies is so low that ordinary diff pair tanh distortion is sub ppm
a question is what is the practical import of RF/EMI infiltration, possible rectification, mixing in the diff pair - that is when maximally flat multi-tanh may offer improvement - from input diff gm linearity improvement for frequencies way above negative feedback "working band"
another case is for those who insist on "low feedback" - then even the audio frequency (linear) error V input at the diff pair can become large
Last edited:
This is the "Bob Cordell's Power amplifier book" thread.
Of course you should buy Bob Cordell's Power amplifier book.
That's what I meant when I wrote that "there may be more productive topics."
But if there are multiple input transistors to reduce noise then it is an attractive idea to also lower distortion as a bonus. Do you have any ideas or references?
So far I have the Gilbert article from IEEE JSSCircuits.
And does anyone know of an audio amp example?
Best wishes
David
Hi Magna,
I'm sorry I took so long to get back to you on this. For a Vbe multiplier bias circuit, I often mount a TO126 device on the heat sink between the P-channel and N-channel devices. The MOSFETs are more tolerant of thermal drift than BJTs, so mounting on the H/S in close proximity to the P and N devices id fine. Note the different kinds of Vbe multiplier bias spreaders shown in the book that allow one to adjust the amount of compensation. In some cases one junction will be put on the heat sink and another may be at ambient to achieve less compensation. While vertical MOSFETs need bias spreader temperature compensation, designs with lateral MOSFETs often don't have it, since the temperature coefficient of current with temperature for fixed Vgs goes through zero at about 200mA.
Cheers,
Bob
Hi Dave,
I'll definitely have to expand the five chapters currently devoted to class D because that really is the wave of the future and its quality is getting so much better every year.
One thing that I found at the time of the writing of the book was that there was a great deal of information out there on class-D, but it was spread out all over the place. The other thing that is different about class D is that it is not quite as DIY-friendly as conventional amplification in regard to the ability to tinker with and refine the performance of the design. Even for the various incarnations of PWM, the art is more specialized. The very high speed switching needed to achieve good performance can be quite layout-sensitive, for example. Things get even more difficult for DIY design in the realm of more advanced modulation techniques like sigma delta and DSP-enhanced designs. However, class D is still DIY user-friendly if one is talking about building with modules to create complete amplifiers.
Cheers,
Bob