Hi Fred,
There are many judgment calls in picking the IPS emitter degeneration and the optimums are quite broad. I usually set emitter degeneration at a factor of about 10. This means that the degeneration resistor is about 10X the value of re' of the transistor being degenerated. Put another way, approximately, 10X degeneration reduces the CE transconductance of a stage by a factor of about 10. Introducing degeneration does indeed increase input voltage noise, but the relative amount by which it increases it depends on the degree of contribution from other sources, such as the resistances in the base circuits. These noise issues are covered in the noise chapter in my book. For a given size of Cm, increased degeneration allows for increased LTP current, which increases slew rate. However, with a BJT input stage, there is a point at which the resulting input base current shot noise will increase overall noise if the impedance in the base circuits is high enough.
I often like to keep Cm in the range of 30 pF when using a 2T VAS, but this is again a broad optimum. I have at times gone lower and higher. Going significantly higher forces the VAS to supply more current at high frequencies to charge it, which is not necessarily desirable.
I also usually degenerate current mirrors by a factor of about 10, as discussed by Mark below. A current mirror with little or no degeneration will increase noise. Degenerating to the point where there is 200-300 mV across the degeneration resistor is usually a good choice, and usually corresponds to about 10X degeneration.
Cheers,
Bob
Thanks, Bob. I think I'm finally getting this thing tuned in, though it appears that there may be some strange artifacts as a result of going out to a breadboard, rather than a perf board in the unit. I may make the changes on the installed board and see what happens.
I DO have the bias spreader tuned in quite well. The drivers are running at a high enough current that the speed up capacitor can be eliminated. This means that the drivers will heat up initially but will then remain at a relatively constant temperature. So, it is necessary to have 2 diode connected transistors, one sandwiched on a driver, the other clamped to a TO-3 output.
Temperature was stabilized, bias set to 24 mV. Unit then allowed to get cold overnight. On cold startup, bias was close to 24 mV within 10 seconds and tracked properly until temperatures stabilized. Next, the output was driven until the transistors were too hot to touch. Returned to idle and the bias returned to close to 24 mV within 5 seconds. I don't think we can get any better than that.
Cheers,
Fred
I DO have the bias spreader tuned in quite well. The drivers are running at a high enough current that the speed up capacitor can be eliminated. This means that the drivers will heat up initially but will then remain at a relatively constant temperature. So, it is necessary to have 2 diode connected transistors, one sandwiched on a driver, the other clamped to a TO-3 output.
Temperature was stabilized, bias set to 24 mV. Unit then allowed to get cold overnight. On cold startup, bias was close to 24 mV within 10 seconds and tracked properly until temperatures stabilized. Next, the output was driven until the transistors were too hot to touch. Returned to idle and the bias returned to close to 24 mV within 5 seconds. I don't think we can get any better than that.
Cheers,
Fred
There was an article by Taylor in Aug/Sept Wireless World 1977 that discussed the non-linear
input impedance of a diff pair that we discussed back in 2005. andy_c offers some insight in
this post:
https://www.diyaudio.com/forums/sol...ordell-otala-gilbert-papers-3.html#post548556
This is the reference, there was a part 2 in Sept:
"Distortion in low-noise amplifiers" by Eric F. Taylor, Wireless World August, 1977
I seem to recall that we had this scanned, not sure if it is posted anywhere. I have hard
copy somewhere around here from when I read it back in the 1970s.
Found it online: http://www.keith-snook.info/wireles...d-1977/Distortion in low-noise amplifiers.pdf
More Wireless World articles: ~ Scanned and cleaned up Wireless World Articles ~
Last edited:
An unexpected and generous financial gift from my sister has at last allowed me to order from Amazon the 2nd Edition. The site claimed only two copies were left, will a second print run become available?
I'm also requesting the King County Library System order at least one copy for their circulation. They've got a first addition already, but there's so much new material I think it's justifiable. My book collection doesn't cover Class D in any detail, so I'm looking forward to get caught up on that subject.
I'm also requesting the King County Library System order at least one copy for their circulation. They've got a first addition already, but there's so much new material I think it's justifiable. My book collection doesn't cover Class D in any detail, so I'm looking forward to get caught up on that subject.
Last edited:
Hi Damon,
Thanks for your kind words about my second edition and hope you enjoy the read. Not to worry about availability. Amazon occasionally runs out of stock, but that is not an indication of overall availability. I do know that sales of the book usually increase in the holiday season. Books can also be obtained directly from the publisher.
Cheers,
Bob
FYI, I have the amplifier, called "BC-1" in chapter 5 built & operational.
I approached Bob earlier asking if there would be something to build in the new book. The answer was yes and no, there is a complete design but no pcb's available for easy assembly. So, I took it upon myself with Bob's help and encouragement to design a set of pcb's for BC-1. This was done in anticipation that their maybe some interest in book owners wanting to buildup the BC-1 amplifier.
I have sent the pcb's to Bob Cordell for evaluation. Bob graciously offered to review, and now evaluate my design implementation and the build.
Once Bob has given them his approval I will be able to release them for others to use.
I will attach a pic of the set of pcb's in operation. There is a Mouser BOM in a shopping cart in which you will be able to order the parts against. It will be a complete amp with out the chassis and mechanical bits. A 3 board set, a simple & cost effective, rectifier/filter, BC-1 (AFE), BC-1(OPS).
Since I was on a roll I also did up the DH-200C design as shown in chapters 9 & 14. Bob is also evaluating that one too, like Bob needs more work, especially from me
Hi All, Is there any news on these PCB's ? Does anybody know of there a BOM for the BC-1 and protection boards on Digikey or Mouse?
Hi Stuart,
The PCBs and BOM are coming, but things have gone slower than hoped. As you may know, Rick and I have also been working on the DH-220C Hafler amplifier upgrade, and have decided to get that one completed first, so that has delayed some of the work on the BC-1. We will definitely keep you posted on progress, and deeply appreciate your interest.
Cheers,
Bob
Hi Shadders,
You are absolutely right, and it is my bad that I have not yet gotten the errata up on my website. I have been hammered with so much stuff to do. I will resolve to get those errata up on the site soon, and deeply appreciate both the reminder and your bringing another possible error to my attention. Today I have to go out and shop for Christmas presents for my four grandchildren before Christmas gets here
.Cheers,
Bob
Hi Bob,Hi Shadders,
You are absolutely right, and it is my bad that I have not yet gotten the errata up on my website. I have been hammered with so much stuff to do. I will resolve to get those errata up on the site soon, and deeply appreciate both the reminder and your bringing another possible error to my attention. Today I have to go out and shop for Christmas presents for my four grandchildren before Christmas gets here
Cheers,
Bob
Thanks - no problem.
I had to do Jury Service and a lot of waiting around meant i had spare time to read. Merry Christmas too.
Regards,
Shadders.
Hi,
I have not read all the book, so this may be already covered.
There is a lot of useful information in the book about the benefits of using specific circuit configurations. One area that i am woefully deficient in knowledge, is the theory of the circuit operation for dynamic conditions and the reduction of THD. As an example, a cascode implemented with the VAS output reduces the THD in the audio band (mostly) due to the feedback it applies. How is this analysed ?. and if so, which equivalent transistor circuit is most appropriate to use to illustrate how the feedback will be effective at which frequencies and reduce in effectiveness at higher frequencies etc ?.
For poles - i have B2Spice simulation software, which produces as an output of all the poles in the system, but what it does not indicate is the nodes which generate the poles. For any simulation, and amplifier built, how do you determine where the pole resides ?. From my initial testing of my amplifier circuit i have built, it seems to be the VAS that is the issue - with one output transistor pair, it was ok, but when i added the final pair oscillation occurred at approx 70kHz. I can add a capacitor to the VAS output as per the book for lag compensation which resolves, but i can also change the VAS resistor values which has an effect too. So, which is the optimal approach - i assume changing resistors and add capacitors as a last resort ? These aspects would be very helpful in the next edition ?
Thanks and regards,
Shadders.
I have not read all the book, so this may be already covered.
There is a lot of useful information in the book about the benefits of using specific circuit configurations. One area that i am woefully deficient in knowledge, is the theory of the circuit operation for dynamic conditions and the reduction of THD. As an example, a cascode implemented with the VAS output reduces the THD in the audio band (mostly) due to the feedback it applies. How is this analysed ?. and if so, which equivalent transistor circuit is most appropriate to use to illustrate how the feedback will be effective at which frequencies and reduce in effectiveness at higher frequencies etc ?.
For poles - i have B2Spice simulation software, which produces as an output of all the poles in the system, but what it does not indicate is the nodes which generate the poles. For any simulation, and amplifier built, how do you determine where the pole resides ?. From my initial testing of my amplifier circuit i have built, it seems to be the VAS that is the issue - with one output transistor pair, it was ok, but when i added the final pair oscillation occurred at approx 70kHz. I can add a capacitor to the VAS output as per the book for lag compensation which resolves, but i can also change the VAS resistor values which has an effect too. So, which is the optimal approach - i assume changing resistors and add capacitors as a last resort ? These aspects would be very helpful in the next edition ?
Thanks and regards,
Shadders.
Shadders, what you want is "The Method Of Open Circuit Time Constants". I'm sure you can find articles and tutorials about it, on the web, for free. If not I am happy to recommend a textbook used in EE graduate courses at Stanford and other universities, which includes a dozen pages about TMOOCTC. However, as you might imagine, it's a textbook so it is very expensive.
TMOOCTC was developed in the pre-computer era, when all they had was theory and slide rules, so it's heavy on theory. Maybe you'll like that, maybe you won't.
TMOOCTC was developed in the pre-computer era, when all they had was theory and slide rules, so it's heavy on theory. Maybe you'll like that, maybe you won't.
Hi Mark,
Thanks. I did a quick search - was not aware of this method. I assume i will have to examine the model parameters and use the hybrid-pi or the ebers-moll model. The hybrid-pi seems to take into account the capacitances, so i assume will be a more accurate model for pole location ?
Although the method may be harder - it will let you know where the pole resides - simulation just gives you the results.
Is there a book you can recommend ?.
Thanks and regards,
Shadders.
Thanks. I did a quick search - was not aware of this method. I assume i will have to examine the model parameters and use the hybrid-pi or the ebers-moll model. The hybrid-pi seems to take into account the capacitances, so i assume will be a more accurate model for pole location ?
Although the method may be harder - it will let you know where the pole resides - simulation just gives you the results.
Is there a book you can recommend ?.
Thanks and regards,
Shadders.
I recommend this book
https://www.amazon.com/Design-CMOS-...9220/ref=mt_paperback?_encoding=UTF8&me=&qid=
but be prepared: it is top heavy with theory. It is LADEN with theory. I hope you remember your complex variables, differential equations, and Laplace transforms.
https://www.amazon.com/Design-CMOS-...9220/ref=mt_paperback?_encoding=UTF8&me=&qid=
but be prepared: it is top heavy with theory. It is LADEN with theory. I hope you remember your complex variables, differential equations, and Laplace transforms.
Pay no attention to Amazon's stock levels. In the last year+ Amazon has worked to reduce their costs on the Prime free 2-day shipping program, by stashing small quantities of books in a dozen+ warehouses, minimizing delivery distance. Since this could get out of hand they also have algorithms to make-up numbers of required stock level at each stash.
High sellers like Where the Crawdads Sing or Dog Man: For Whom the Ball Rolls, Amazon can predict large sales and reasonable stock levels. Cordell's book is no $8 comic, sales are (much?) less, with quantum uncertainty. Amazon's algorithms have been pretty goofy at this sales-level. Self-publishers say Amazon will order an initial box of 17 books (which could fit 24 to a box) then even before they arrive order another 5 books (doubling the shipping cost to get books TO Amazon), because of algorithm burp.
At this moment, Amazon is not showing ME any stock-level on Cordell's book. Just "In Stock". Maybe they re-ordered since yesterday. Maybe it has not sold as well in outer Boston as it has in outer Tacoma, you ran low first. And it has been shown that Amazon may have books in-hand for 2 weeks yet not list them as orderable. And one self-publisher told me that an innocent click on his side caused his book to become unavailable forever (two algorithms disagreed and got in a snit, took a week of human intervention to find and patch the glitch).
The cure to all this of course is submitting your book for print-on-demand. No need to stock anything, and they never run out of copies to ship.
Bob's book may be too large for that, although I have seen 500+ pages books in print-on-demand. And the quality has gone up so much that it takes a trained printer to see the difference.
Jan
Bob's book may be too large for that, although I have seen 500+ pages books in print-on-demand. And the quality has gone up so much that it takes a trained printer to see the difference.
Jan
I think the reduction in THD is more due to two effects:
- Elimination of early effect in VAS transistor
- Keeping VAS transistor base-collector voltage more-or-less constant eliminates effects of its non-linear base-collector capacitance
In this specific case, clearly you need a model that incorporates early effect and non-linear base-collector capacitance.
Bruce Hofer (of Audio Precision) wrote an article "Designing for Ultra-Low THD+N" that talks briefly about using Taylor Series expansions to predict harmonic distortions in transistor circuits. It's unclear to me how useful this would be in a power amplifier employing many discrete transistors (it's unclear to me if the enormous effort of manual analysis like this would yield any significant insight that would enable the designer to optimise the circuit beyond what is possible from other approaches, such as spice simulation* and application of engineering intuition).
* Although this also requires significant effort to ensure all transistor models are accurate; sadly manufacturer-supplied ones are usually atrocious.
Sounds like you definitely need to learn how to use Tian/Middlebrook probes in spice. I think this is a reasonable introduction. You need a simulator that can combine results from two different simulation runs, such as LTSpice (never heard of B2Spice so I don't know if it can do that).
You will then be able to analyse feedback loops in the amplifier (i.e. the "global" loop, as well as "local" loops such as the one around the VAS) and determine their stability margins.