Douglas did say it would be published on the first of July.
Actually, amazon.com states that the publication date is the 24th of june.
Last edited:
5mA to 10mA. It's a trade between the stage gain (which is, before any VAS degeneration, Au=gm*(Ri||R0)=40*Ic*(Ri||R0) where Ri is the output stage impedance and yes, a triple may allow a lower VAS current, for the same gain) and the VAS output impedance R0, which appears in parallel with Ri and is approximatively VA/Ic where VA is the Early voltage. So a large Ic may increase the transconductance, but decreases the stage load. If the output stage input impedance would be very large, then the VAS gain would be independent of Ic, unfortunately this approximation never holds.
Cascoding the VAS stage with a high VA device largely removes this limitation, then the maximum VAS current is limited mostly by the power dissipation and the SOA of the active device, and some stability criteria (some cascodes may be subject to potential local stability issues).
Cascoding the VAS stage with a high VA device largely removes this limitation, then the maximum VAS current is limited mostly by the power dissipation and the SOA of the active device, and some stability criteria (some cascodes may be subject to potential local stability issues).
Symmetrical slew rates can be achieved by other means. The complementary push-pull TIS is simply not a good or efficient way of doing this, with the possible exception of Shinichi-Kamijo's approach:
http://www.ne.jp/asahi/evo/amp/index.htm
Last edited:
You mean .. you mean you only have ONE good circuit and all your others are cr*p? 😱
If this is the case, I'm sure Jan Didden will make an exception and allow us a sneak preview of your masterpiece. It will whet our appetite to buy the next issue of Linear Audio and may even persuade indigent beach bums to get a job. 🙂
Err.rh! When you modify someone's circuit, it seems to be invariably worse. I haven't seen ALL your 'improvements' but for those you recommended for my circuits, THD is sadly at least 20dB worse and stability too. 😡 Is there some esoteric improvement I've missed that compensates for these 'trivial' points?
Err.rrh! I've posted at least 3 of my own circuits where this isn't the case in SPICE world and more on astx & dadod's circuits .. and more importantly, mine are stable in the 'real world'. None of them require evil shunt caps at the Holy HiZ VAS output.
You asked explicitly for examples and I provided you with a long list on at least 2 occasions.
I totally accept that YOU have the ability to take one of my circuits and make it unstable in SPICE world and probably in 'real life' too. You are the MASTER at this. But then, you seem to put little credence to 'real life' compared to your sims. and (perhaps wonky?) theories.
You seem to be a master at another useful technique : When you ask for evidence and it is presented in detail, you conveniently forget it happened. 😱
I now count at least 4 people who have posted SPICE simulations of this useful property of VAS degeneration. Perhaps we are all cerebrally challenged and only you know how to use SPICE properly. You certainly seem to be the one that is most reliant on SPICE to the exclusion of 'real world' considerations.
I did better than that. I used YOUR circuits & models that you posted in #3170 bob-cordells-power-amplifier-book-317.Have you actually tried to replicate or otherwise perform my simulations with the "correct" circuits and models of your choice? Hell No!
I promised you and Bob the results but I didn't post them cos the example (like all of yours I've seen) had poor choices of currents & operating points such that stability & THD were marginal. Mea maxima culpa 😱
This is the Inner Loop.
You'll note without R10, the VAS emitter resistor, the stability is seriously wonky. Is this the case with ALL your designs? Adding the emitter resistor and my lucky decoupling cap C3, improves stability.
Should point out that the context was suggesting to Bob that decoupling the VAS emitter resistor was another useful string to the stability bow. Some of us are always on the lookout for new & better dodges to improve stuff ... rather than degrade stuff. BTW, the effect of VAS emitter resistor & judicious decoupling is apparently a 'Bode Step' to da pedants. I prefer to think of it as a kink to Nyquist.
I didn't do more work on this as your design (Unobtainium devices, no Zobels or output Inductor etc) would at best, be marginally stable with real loads and have poor distortion. It's performance is easily beaten by a much simpler, more stable, 'real life' circuit using my favourite 'pure Cherry' or even lesser compensation methods. 🙂
Not so. If eg Iq is inadequate, there are often stability effects. This is sometimes seen as bursts of oscillation on the THD residual or just higher THD on very specific loads. You may like to investigate this in the 'real world'.
________________
If you don't mind, I won't waste time looking at any more of your circuits ... unless you tell us that they illustrate good performance & practice. Looking at an endless line of cr*p circuits is usually NOT productive. 🙂
Of course if you post a good circuit with good performance, we will be eager to learn how you do it ... especially if this is backed up with 'real life' 😀
Attachments
Last edited:
Michael, can you tell us which of Shinichi-san's circuits you are referring to? Can I assume its not any of the valve circuits?
Keep this away from personal remarks and jibes, or there will be points.Stick to the subject and technical arguments.
Well, a Bode Step is a rather specific response function, not just any arbitrary kink.
Nyquist is just a particular presentation of the response.
Experts like Lurie and Horowitz find Nichols plots more convenient for this kind of work and I have started to think so too.
You may want to try them.
Best wishes
David
I just ran your sim. and found that Michael does have a point.
If R10 is simply increased from 0.1 to 22 ohms then the stability does deteriorate.
Of course the increased value of R10 does allow freedom to increase C3.
But it does appear that the simple increase in R10 does destabilize that circuit with those models.
I wish Michael's models were better, now I have to decide whether to spend the time to determine if it is someone's error, a quirk, or more fundamental.
Best wishes
David
Ha! Guru Zan is a pedant 🙂
OK, a Bode Step is an arbitrary kink in Nyquist that improves stability 😀
Please lobby LTspice to include dis fancy stuff. Den please Guru Zan, prepare da Idiot's guide to Nichol for us unwashed masses. 😕 I think some of the big $$$ SPICEs do dem.
Damn. Caught out by the Head Prefect. Dis was Michael's attempt to dis one of Bob Cordell's circuits.
Yes. I knew but was hoping no one would notice 😉 I didn't keep my other sims of Michael's stuff cos they all gave awful THD & stability.
As I said, the original intention was to suggest yet another stability dodge to Bob.
I think all Michael's models need A LOT of work to come up with anything worth keeping. Try a 8R load and doing THD20k with this one ... or indeed any of his others. 😀
BTW, Cherry predicts this.
Yes, I know, that's partly why I checked.
I still don't have a feel for how it varies with circuit or transistor parameters.
As to Nichols plots. I will try to do a template in LTSpice.
It is not quite as simple as I had hoped.
There seems to be no way to equate a real variable with the real part of a complex variable, a Type CAST in other words.
Best wishes
David
For the avoidance of doubt, I will say it just ounce more: I was, in fact, referring to small signal transistors and only in the context of minor and major loop gain.
So, you can actually hear small signal transistors in an amplifier with large quantities of feedback, can you?
Er,... no, my soul has been stolen by Matti Otala forever since 1974, when his first text on TIM appeared in IEEE. Others, like Richard Miller of H/K and John Curl, did the mop up work, and I am a dedicated follower of the high local, low global NFB school.
This approach seems logical to me, and in real life, it generally tends to sound better to me than other approaches. However, I am no exclusivist either, zero global NFB amps sound a bit loose, not quite focused to me, at least those I have heard.
So, I believe an amp SHOULD have global NFB, but in moderate terms, ideally 20-26 dB, and that only assuming its open loop full power bandwidth hits at least 40 kHz/8 Ohms, preferably more. I also make it a point to have run at least over 350 kHz at rated power, after which I do what reVox/Studer and Sony typically do, I install a 200 kHz low pass filter at the input.
That said, I must come clean and admit that in my time, I have heard several high global NFB amps which I would love to have, first and foremost from a now gone German company, called Linear Audio Systems (LAS). I loved it so much than I plan to actually make it, don't care if it came out in 1977, don't care if I don't really need it.
If there's interest, I can dig up the schematic and post it here. Minimalists would love it, those with hard-to-drive speakers would adore it.
Last edited:
LTspice Waveform Arithmetic help:
so I guess the number's type still stays complex - but separable
don't see the reverse conversion - how to put real into the imaginary part - could ask for the feature, also ability to select vertical axis in AC, Nyquist Plot
so I guess the number's type still stays complex - but separable
don't see the reverse conversion - how to put real into the imaginary part - could ask for the feature, also ability to select vertical axis in AC, Nyquist Plot
Last edited:
After a few minutes to think about this I am not so sure.
Cherry talks about the outer loop.
The inner loop is a different problem.
OK, this is worth the effort to check, even if Mike's models are suspect.
Best wishes
David
Reading a test review by the late Geoffrey Horn of a NAD amplifier he said inspired by H C Lin . Geoff in my opinion was a bit lazy when reviewing . He would say a speaker showed a touch of frequency doubling at 60 Hz or whatever . What he was saying was not Quad ESL .
I took the first ever copy of the book on Audio to Texas with me . I went to please my then wife staying with her friend . It came from Oxford library , if lost I would have bought them a new copy . It was high summer and I went reading by a small river having been coached in the problems of snakes . Whilst there at 90 F + I penned a 1964 amplifier using the first silicon transistors commonly seen . It was just a bit of fun and was putting the best ideas in the book together with what would been available . I had to respect the fact that in 1964 the cost of materials was about 90% that of building with valves . I used one extra transistor in the LTP . The VAS CCS a bootstrap .
Many years later my friend asked me to evaluate a Chinese amplifier , 95% was dreams come true . Telephone exchange volume control with relays ( talking audacity more than sound , was good on that also ) . Shunt regulated PSU up to the VAS . Raw DC to outputs . No loop feedback class B .
The latter feature was it's downfall . Exceptionally punchy and often sounding like a kazoo . Having now built similar I realize it was too low driver current rather than bias problems .
My friend said could I redesign the output , he strongly warned me not to give away too much ? I put the 1964 design in using their superb parts bin . The sound was so different . Very slightly dainty for want of a better word ( VAS degeneration of 47 R I suspect ) . 50 were made and my sample never got to me . As far as I know it continues in China ?
Now the crunch . I sent the circuit . They asked me how much of the original I wanted to keep . Nearly all I said . The PSU especially . Then the killer question . How do we set 0V ? Like an op amp I said . H C Lin although graduating from a Chinese university ( ? ) had not entered their design culture .
The one thing I religiously did was the bias as per book .
Wally , thanks for VAS current answer .
I took the first ever copy of the book on Audio to Texas with me . I went to please my then wife staying with her friend . It came from Oxford library , if lost I would have bought them a new copy . It was high summer and I went reading by a small river having been coached in the problems of snakes . Whilst there at 90 F + I penned a 1964 amplifier using the first silicon transistors commonly seen . It was just a bit of fun and was putting the best ideas in the book together with what would been available . I had to respect the fact that in 1964 the cost of materials was about 90% that of building with valves . I used one extra transistor in the LTP . The VAS CCS a bootstrap .
Many years later my friend asked me to evaluate a Chinese amplifier , 95% was dreams come true . Telephone exchange volume control with relays ( talking audacity more than sound , was good on that also ) . Shunt regulated PSU up to the VAS . Raw DC to outputs . No loop feedback class B .
The latter feature was it's downfall . Exceptionally punchy and often sounding like a kazoo . Having now built similar I realize it was too low driver current rather than bias problems .
My friend said could I redesign the output , he strongly warned me not to give away too much ? I put the 1964 design in using their superb parts bin . The sound was so different . Very slightly dainty for want of a better word ( VAS degeneration of 47 R I suspect ) . 50 were made and my sample never got to me . As far as I know it continues in China ?
Now the crunch . I sent the circuit . They asked me how much of the original I wanted to keep . Nearly all I said . The PSU especially . Then the killer question . How do we set 0V ? Like an op amp I said . H C Lin although graduating from a Chinese university ( ? ) had not entered their design culture .
The one thing I religiously did was the bias as per book .
Wally , thanks for VAS current answer .
I'm not totally sure i can reach the ~1000V/µs and the 6Mhz of my power amp with some kind of 2N3055. 😕
In real world, even the layout of the board affect stability and sound, as well the same device from different manufacturers. It is obvious for caps, it is true for active devices too.
Sims are nice to explore a circuit and get a deeper knowledge of its behaviors. It save time in calculations and paper. But i do not believe in it: The real work begin with ... a soldering iron. And, on an industrial point of view, you can only be confident after the first 100 samples of the pre-serial had been tested, measured and tortured.
My son is pretty good in flight simulator. I would not travel in an Airbus he would fly 🙂
Last edited:
No need to try convince anyone, if after 30 years of solid evidence that he was wrong they still dont get it youre just wasting your time. 😀