Audio Power Amplifier Design book- Douglas Self wants your opinions

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
www.hifisonix.com
Joined 2003
Paid Member
There is no such thing as MIC ("Miller Input compensation"). As I pointed out somewhere, it is really phase lead compensation with the output stage excluded; certainly, pole splitting does not occur with this arrangement, which explains why other supplimentary compensation is invariably required.

Simple SPICE simulation demonstrates these facts, so both D. Self and R. Cordell were wrong in calling it "input stage inclusive compensation". Moreover, it has nothing to do with Mr J. M. Miller, so it should never have been called "MIC".

While technicalities you describe may right, it Is MIC and that is what it will remain, as is the case with CFA.

You cannot go renaming things because you disagree with the their commonly understood terms.
 
Last edited:
this means the "extra" loop gain is not available to improve the LTP characteristics, nor the VAS/TIS supply reference PSRR – except through a slight change in the available global outer loop single pole gain from the TMC bootstrapping increasing VAS/TIS loop gain

With TMC the VAS/TIS + OS local loop gain increase this stage
input impedance , hence reducing the amplitude of the TIS/VAS
input AC current and thus decreasing the loading of the input stage ,
as with a current mirror load the IPS gain is mainly modeled by
the TIS/VAS input impedance.
 
While technicalities you describe may right, it Is MIC and that is what it will remain, as is the case with CFA.

You cannot go renaming things because you disagree with the their commonly understood terms.

Only Bob calls it MIC, and, no, if, as you concede, the "technicalities you describe may be right" why continue calling it MIC thereby misleading an entire host of budding engineers?

As for renaming "CFAs", these were always known as VFAs until the marketing men at the major IC manufacturers renamed them. :nod:
 
Last edited:
Only Bob calls it MIC, and, no, if, as you concede, the "technicalities you describe may be right" why continue calling it MIC thereby misleading an entire host of budding engineers?

As for renaming "CFAs", these were always known as VFAs until the marketing men at the major IC manufacturers renamed them. :nod:

Hi Mike,

Yes, I am guilty as charged for naming it Miller Input Compensation, or MIC. I gave it a name simply because I had not seen the technique before nor seen a name for it. I used MIC in my MOSFET power amplifier back in 1983.

As explained in my book, MIC is just a form of input compensation. Passive lag input compensation has been used in the past to also achieve high slew rate. It has its own disadvantages. So my thinking in 1983 was to use input compensation to get the good slew rate, but to achieve it in a Miller feedback way so as to eliminate the shortcomings of passive lag input compensation.

As for the workings of MIC, we have been down this road before, and have to agree to disagree. It is certainly not lead compensation. Moreover, it is indeed a form of Miller compensation and does indeed posess the properties of pole splitting at the input and output of the integrating entity, in this case the combined input stage and VAS. If you look at the output impedance of the VAS with MIC, you can easily see that it is made low and has been pole-splitted.

Most importantly, Miller compensation, or perhaps more correctly a Miller integrator, is certainly not limited to a forward path gain element with only one stage, as is a conventional Miller-compensated VAS. The fact that some compensation is needed within that loop is completely irrelevant, and does not disqualify the arrangement as a Miller integrator. One need look no further than the multitude of Miller integrators built with an op amp encircled by a Miller integrating capacitor. The op amp so enclosed necessarily has compensation included within it.

Cheers,
Bob
 
Bob Cordells 1982 or '84 design was purely theoretical and had no PCB layout, full schematic/components etc and as far as I am concerned you might as well put it in the bin. Complete waste of space. There is absolutely no evidence nor measurements it could slew at 300V/us or have THD20K @ 0.001%.

As Douglas Self himself said in a riposte to JLH in his remark, that he does not believe THD measurements below 0.001% had any credibility in his experience, was that the only reliable way of testing sub 0.001 THD distortion is with a A.P. system 1 (itself involving 50+ patents I believe!). I think the AP system 1 was mid 90's equipment....not '82 or '84....

Kevin

Hi Kevin,

The MOSFET Power Amplifier with Error Correction circa 1983 was most certainly not just a theoretical design and I am disappointed that you do not recognize this and think it was a complete waste of space. The amplifier was presented in a paper in the JAES. That paper is available on my website at CordellAudio.com - Home. It shows component values and many measurements.

Just because it did not involve a printed wiring board and was not presented as a DIY product is no eveidence that it was not built and measured. You should really do at least some research before you make such completely irresponsible assertions.

That 50-watt amplifier did indeed achieve 300V/us and <0.001% THD at 20kHz (with a 200kHz measurement bandwidth, I might add, not the oft-used 80kHz bandwidth).

As far as I know, that amplifier was the first to crack the THD-20 0.001% distortion barrier. If there was another one that preceded it I am unaware of it and would certainly be interested in finding out what it was.

Doug was wrong if he asserted that THD measurements below 0.001% could only be made with an AP. The THD analyzer I built as a construction project for Audio magazine appeared in the same time frame as the AP1 and achieves performance equal to that of the AP1 in terms of the THD floor.

If what you attribute to Doug saying is true, it is surprizing and disappointing that he was unaware of my THD analyzer and its capabilities back then. Over 200 sets of printed wiring boards for that analyzer construction project were sold.

BTW, I believe the Tek AA501 (designed by AP founders Rich Cabot and Bruce Hofer) was also capable of measuring 0.001% THD at 20kHz.

In order to measure the very small THD of my MOSFET amplifier I did build and use a version of my Distortion Magnifier to aid in testing the amplifier. That is also touched on in the JAES article.

Kevin, please get your facts straight in the future.

Cheers,
Bob
 
Member
Joined 2011
Paid Member
Doug,

I'm confident you've read Linear Audio Volume 5, as it includes a very nice article written by you. It also includes an article by Morgan Jones, whose single paragraph Conclusion conflicts with parts of APADH 5th edition. If you find Mr. Jones's article convincing, you may want to consider modifying Fig 9.2 (p.272) and the section "RF Emissions from Bridge Rectifiers" (pp. 284-5). Or, if you don't, this section might be a convenient place to insert your own opinions, measured results, and real-world experiences with commercial products.

Best regards,
MJ
 
Doug,

I'm confident you've read Linear Audio Volume 5, as it includes a very nice article written by you. It also includes an article by Morgan Jones, whose single paragraph Conclusion conflicts with parts of APADH 5th edition. If you find Mr. Jones's article convincing, you may want to consider modifying Fig 9.2 (p.272) and the section "RF Emissions from Bridge Rectifiers" (pp. 284-5). Or, if you don't, this section might be a convenient place to insert your own opinions, measured results, and real-world experiences with commercial products.

Best regards,
MJ

I agree that D. Self's new line preamp with variable frequency tone controls is very nice. I hope a PCB will be made available.
I think it could be very useful to bass-boost closed boxes with low Qtc which can be obtained by various ways : initial very low Qts of the driver, velocity servo, amplifiers using negative resistance, or parametric equalisation which then makes it as an approximate transform. However in these cases, a maximal boost of 15 dB would be more adequate.

Morgan Jones's article seems to confirm Eva's experiments which I use since I know them :
http://www.diyaudio.com/forums/power-supplies/66542-fast-recovery-rectifier-diodes-4.html#post753891
I very much dislike anything which emits RF or acts as an unwanted RF receiver.
 
Hi Mike,

Yes, I am guilty as charged for naming it Miller Input Compensation, or MIC. I gave it a name simply because I had not seen the technique before nor seen a name for it. I used MIC in my MOSFET power amplifier back in 1983.

As explained in my book, MIC is just a form of input compensation. Passive lag input compensation has been used in the past to also achieve high slew rate. It has its own disadvantages. So my thinking in 1983 was to use input compensation to get the good slew rate, but to achieve it in a Miller feedback way so as to eliminate the shortcomings of passive lag input compensation.

As for the workings of MIC, we have been down this road before, and have to agree to disagree. It is certainly not lead compensation. Moreover, it is indeed a form of Miller compensation and does indeed posess the properties of pole splitting at the input and output of the integrating entity, in this case the combined input stage and VAS. If you look at the output impedance of the VAS with MIC, you can easily see that it is made low and has been pole-splitted.

Most importantly, Miller compensation, or perhaps more correctly a Miller integrator, is certainly not limited to a forward path gain element with only one stage, as is a conventional Miller-compensated VAS. The fact that some compensation is needed within that loop is completely irrelevant, and does not disqualify the arrangement as a Miller integrator. One need look no further than the multitude of Miller integrators built with an op amp encircled by a Miller integrating capacitor. The op amp so enclosed necessarily has compensation included within it.

Cheers,
Bob


This arrangement was also recommended by J.L. Hood incidentally.

I have run major loop gain simulations of an ordinary Miller compensated amp., and then merely connected the end of the compensation capacitor connected to the input of the second stage to the inverting input of the amp and run the simulation again.

The results demonstrate that pole splitting does not occur with your "MIC". Period. In fact, the results with "MIC" are virtually identical to those of obtained with the compensation capacitor disconnected entirely from the second stage and connected instead to the inverting input and the output, viz. phase lead compensation.

As you clearly cannot be bothered to run the simulations above to enlighten yourself, I suppose we'll have to agree to disagree.
 
Last edited:
michaelkiwanuka,
How about you actually build one of these circuits instead of your simulations and get real measurements? Why should we trust your simulations, you could be leaving things out that in reality give results that don't track your sims. I suspect that Bob has done real circuits and real testing with accurate test results. Sims are only as good as the program and the input that you use. As they say garbage in gets you garbage out. I would put much more faith in what Bob Cordell says than your insistence that you can do a proper simulation of the real circuit.
 
The purpose of simulation is to obtain insight into circuit behaviour that cannot be obtained from real circuits, or is difficult to obtain with real circuits.

I have gathered from Bob's responses that he hasn't simulated or obtained from real circuitry the loop gain responses for comparison of "MIC" compensated amplifiers, Miller compensated amplifiers or lead compensated amplifiers.

My simulations are valid and they confirm the truth about "MIC" that no one here has disproved with real circuitry.
 
Michael,
It seems that simulations really only show general trends and not necessarily real results. I have read in many threads where the simulations show no problems until a real circuit is built and then things show up that will not be evident in the simulation. The missing parasitic information that is often missing in sims cause many problems that have to be corrected in actual circuits to stabilize the circuits that can look so good in a simulation. From what I gather most if not all of the designers that reside on these threads do actual circuit analysis to prove their designs and to test for real stability and look for oscillations that aren't apparent with simple simulation. I am not going to say that your simulations are not relevant or correct, but you could be missing something by relying only on untested actual circuitry. I think I will wait to see if Bob comments on how he has come to his conclusions. I also have his text and Doug's books and will look and see what is in those books as a follow up to this exchange.
 
paper (or CAS math software) analysis is often much less "real" than sim

no one I've seen anywhere I've worked as an engineer even gets up to Cherry's JAES audio amplifier modeling articles' admittance matrix level of analysis

almost all detail is thrown out to limit complexity to what little humans can handle

it can be useful for "centering" design values, tradeoffs if the required sufficiently simplifying assumptions don't destroy applicability to the real circuit's operation


Lundberg is not very happy with the "pole splitting" rubric: http://seit.unsw.adfa.edu.au/staff/sites/hrp/teaching/ct2/ACC04_opcomp.pdf
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.