Has anyone seen this front-end before?

[AndrewT]

I'm still looking for the stitching. Did you use green or blue thread?
Nice. No, not Nice, Cote d....
Foreign languages, I'll never get to understand all these different pronunciations for the same word.

 

[Edmond Stuart]

PLIL

Hi Edmond
[..]
Do you have the article by John Ellis, Ive been working on a certain design where I need to curtail ULGF by a fair margin and there might be useful info I could use.

Hi Alex,

I finally found the article by John Ellis. Please look here.
I've also found some 'letters to the editor' about his article and my own one about MIC. If you are interested, I could scan them as well and put them too on my website.

Cheers,
E.

 

[Edmond Stuart]

/OT, stitching

I'm still looking for the stitching. Did you use green or blue thread?
Nice. No, not Nice, Cote d....
Foreign languages, I'll never get to understand all these different pronunciations for the same word.

Hi Andrew,

I can't see the stitches, because I did it a Nice way.
(As a matter of fact, I used PTGui to glue the separate pictures together)

Cheers,
E.

 

[Edmond Stuart]

typo

Yet another typo, aaargh!
I can't see the stitches = You can't see the stitches

 

[Dave Zan]

I've also found some 'letters to the editor' about his article and my own one about MIC. If you are interested, I could scan them as well and put them too on my website.

Cheers,
E.

I would be interested in all those letters so thank you for the offer, it would be appreciated, especially since it was an effort to track down the Ellis article when it was first referenced.
"So nice in Nice" as the tune says. Winter down here and ice on the leaves makes it look even better, if that is possible.

Best wishes
David

 

[Dave Zan]

Welcome back Edmond!

Just ignore the Waly's!!!!

I know there is some history of personal clashes that I don't wish to be involved in, but I found Walter's comments made me think and clarify my ideas.
I dismiss any posts that contain personal insults rather than by author.

Best wishes
David

 

[Edmond Stuart]

I would be interested in all those letters so thank you for the offer, it would be appreciated, especially since it was an effort to track down the Ellis article when it was first referenced.
"So nice in Nice" as the tune says. Winter down here and ice on the leaves makes it look even better, if that is possible.

Best wishes
David

Hi David,

I've merged all letters about PLIL in one pdf file which you can find here, as well as my comment on the compensation of the so called 'Hitachi MOSFET amp' (also known as the 'Maplin MOSFET amp').

OT/
>"So nice in Nice"
That's what my GF says too. She lived there for awhile and speaks French fluently, which makes holidays over there much easier.

Cheers,
E.

 

[homemodder]

Thanks Edmond, much appreciated, I had googled it to death and couldnt find a copy anywhere.

 

[Dave Zan]

Hi David,

I've merged all letters about PLIL in one pdf file which you can find here, as well as my comment on the compensation of the so called 'Hitachi MOSFET amp' (also known as the 'Maplin MOSFET amp').

Thanks for that. The letters reassured to see I wasn't the only one with reservations about that particular application of PLIL. I prefer your comments on the 'Hitachi' amp compensation. It would be even better if you could include the Kessler letter so one could see the circuits that you discuss. If it doesn't violate copy laws, of course.

Best wishes
David

 

[Bonsai]

I took a look at the Ellis article and frankly don't think it adds to much to the compensation debate, although I felt some of his critics went over the top. He notes that in designs that take the Miller cap back to the inverting input (MIC), there tend to be stability problems in some designs and this scheme is not always stable. Well, I think this is so because if you use this type of comp scheme, you enclose the front end LTP and the VAS, and that loop often requires comp - in fact Cordell discusses this in his book. In these designs, you see RC networks across the LTP load, or sometimes the VAS is shunt loaded to ground. Ellis also discusses the problem with MC and couches it in terms of a delay between the input and the feedback signals - well its just a phase difference and needs to be seen in that light. I think MC is very effective for its simplicty, but of course if you are looking for sub 1ppm performance at 20KHz, then its not going to be adequate, and you need to go for more advanced schemes.

Anyway, thanks for sharing Edmond.

 

[Edmond Stuart]

PLIL

I took a look at the Ellis article and frankly don't think it adds to much to the compensation debate, although I felt some of his critics went over the top. He notes that in designs that take the Miller cap back to the inverting input (MIC), there tend to be stability problems in some designs and this scheme is not always stable. Well, I think this is so because if you use this type of comp scheme, you enclose the front end LTP and the VAS, and that loop often requires comp

Hi Andrew,

That's precisely the reason why I reduced the gain of the TIS to unity. Nevertheless, the SuperTis still needs some lag compensation at the inputs of the IPS, though with a pole (and a zero) at a higher frequency.

- in fact Cordell discusses this in his book. In these designs, you see RC networks across the LTP load, or sometimes the VAS is shunt loaded to ground. Ellis also discusses the problem with MC and couches it in terms of a delay between the input and the feedback signals - well its just a phase difference and needs to be seen in that light.

You're quite right on this point. Thinking in terms of delay is rather inappropriate, as it is virtually absent. Indeed, it's the phase difference that's matters.

I think MC is very effective for its simplicity, but of course if you are looking for sub 1ppm performance at 20KHz, then its not going to be adequate, and you need to go for more advanced schemes.

MIC or PLIL alone is not sufficient for sub 1ppm performance, but it does a great job in reducing the load on the IPS. As a result, less distortion from this stage and a higher slew rate. In case of a complementary push-pull TIS, it also circumvents the fighting TIS (or VAS) issue. That makes this scheme so attractive.

Anyway, thanks for sharing Edmond.

You're welcome.

Cheers,
E.

 

[Bonsai]

BTW Edmond, while you were away, I finished my new amp and did a write-up. It's on the first page of my website. It can be jumper selected for MC, TMC and also 'wide banded'. EF3 output stage.

 

[Edmond Stuart]

The Ovation 250

Hi Andrew,

Very nicely built and excellently documented. Chapeau!
It would be interesting if you also publish THD figures using the MC and WB option (or did I overlooked it?)
The coming days I will have a closer look at it (at this moment I'm in the middle of a simulation of a new AM2-OPS) and I'm sure I'll learn a lot of your experience with this amp.

BTW, do you also sell the PCB's?

Cheers,
E.

 

[Dave Zan]

He notes that in designs that take the Miller cap back to the inverting input (MIC), there tend to be stability problems in some designs and this scheme is not always stable. Well, I think this is so because if you use this type of comp scheme, you enclose the front end LTP and the VAS, and that loop often requires comp - in fact Cordell discusses this in his book. In these designs, you see RC networks across the LTP load, or sometimes the VAS is shunt loaded to ground.

That's precisely the reason why I reduced the gain of the TIS to unity. Nevertheless, the SuperTis still needs some lag compensation at the inputs of the IPS, though with a pole (and a zero) at a higher frequency.

Bob Cordell shows MIC with a shunt loaded VAS (p.181) but also discusses the possibility to return the VAS load to the input rather than to earth (p.177)
I asked him about this and he said he hadn't tried it in his JAES amp simply because there was already a lot of new stuff to think about.
Have you simulated this Edmond? Since Miller compensation usually works better than shunt then it seems it should be a little better even for a minor loop. I expect the difference is small so maybe it's academic but it looks attractive.

Best wishes
David

 

[Bonsai]

Thanks Edmond.

I sim'd distortion. See the specifications page. About 8ppm on TMC and c. 40ppm on MC.

No AP measurements I am afraid. In Tokyo I a SYS272 in the lab, but I am in Taipei now and the lab is focused on digital computing an power. Sound card measurements to follow in the next month or so though.

 

[Edmond Stuart]

Bob Cordell shows MIC with a shunt loaded VAS (p.181) but also discusses the possibility to return the VAS load to the input rather than to earth (p.177)
I asked him about this and he said he hadn't tried it in his JAES amp simply because there was already a lot of new stuff to think about.
Have you simulated this Edmond? Since Miller compensation usually works better than shunt then it seems it should be a little better even for a minor loop. I expect the difference is small so maybe it's academic but it looks attractive.

Best wishes
David

Hi David,

Yes, I've simmed this alternative compensation scheme and it turned out that the high frequency loop gain gets too high. So PM and GM get worse. IOW, not recommended.

Cheers,
E.

 

[Dave Zan]

Yes, I've simmed this alternative compensation scheme and it turned out that the high frequency loop gain gets too high.
E.

Thanks. I think Return Ratio is like money, more should be better. Seems there should be a productive way to "spend" the extra. Like Miller compensation but smarter. I will re-read Ed Cherry on nested feedback loops and hope for a better idea.

Best Wishes
David.

 

[Edmond Stuart]

Hi David,

As for 'better ideas', at the moment I'm investigating a 'TMCed' frequency compensation of the AB2-OPS. THD-20k is more than halved, but, as usual, it has a price: stability is also less.

Cheers,
E.

 

[Dave Zan]

As for 'better ideas'... THD-20k is more than halved, but, as usual, it has a price: stability is also less.

That trade-off should be called "Bode's iron law"
I like conservation laws, so you know what is just a trade-off and what quantity must actually increase for a real improvement.
My next idea is obvious so I expect you must have simulated it. Probably subject to the iron law too but it will help me learn so -
What happens when the TMC resistor-capacitor node is loaded with a resistor to earth? Or, put another way, the TPC resistor-capacitor node is bootstrapped?

Best wishes
David

BTW. Bode was of Dutch ancestry. What is it with the Netherlands and feedback?

 

[Bonsai]

You can never escape the fact that as frequency goes up, phase shift accumulates with the 'conventional' topologies we talk about in audio. The thing that stops us easily getting ultra low distortion without having to do circuit gymnastics is the output stage bandwidth, which limits the amount of feedback we can apply to correct system errors. For the small signal stages, a 10-20MHz or even higher full power bandwidth is relatively easy to acheive

Maybe we have to re-think this whole story

1. A ultrawide band output stage - so we need mosfets.
2. Enclose this output stage in a fast local feedback loop. Use a WB Video op-amp (to hell with HEC) for error correction (or feedback configuration). Lets target 100ppm at 1MHz worst case load condition
3. Front end: Ultra wide bandwidth with c. 90db loop gain out to at least 100KHz and then 20dB/decade after that
4. Closed loop system gain 25-35dB
5. TMC+ [ultra low distortion, WB output stage + ultralow distortion WB SS stage] = ultralow distortion amplifer

Challenge: 2 output stage mosfet devices + 1 video op-amp; 10 devices for voltage amp - total 12 devices+video op-amp. No extra 3 pin or greater devices for anti-saturation; single +- supply rails only; use zener references/bootstrapping to supply the video op-amp

Target specs:-

Class AB
100W into 8 Ohms
THD sub 100ppb at 20KHz at any power level up to full power into 8 Ohms
5Hz - 300KHz -3dB
SR =>300VuS
DF > 100 @ 1KHz
Supply rails +-45V, but can go to +-50V

;-)