Thank you. A quick sim suggests that TPC is a distinct possibility. As an added bonus you get more PM but less GM. You will most likely get a good THD improvement by using TPC.
Would avoid TMC for now as this adds extra loops that could result in more problems than advantages.
My Laptop has decided to mess about. So can't add any attachments tonight. But change to TPC as per Bob Cordells recommendations as to how to estimate the values and you should be pleasantly surprised.
It's getting late here, so will have a proper play tomorrow. And I'll have the works computer so can add attachments (images).
I don't find that V(x) is similar to V(Z)...
Paul
Would avoid TMC for now as this adds extra loops that could result in more problems than advantages.
My Laptop has decided to mess about. So can't add any attachments tonight. But change to TPC as per Bob Cordells recommendations as to how to estimate the values and you should be pleasantly surprised.
It's getting late here, so will have a proper play tomorrow.
And I'll have the works computer so can add attachments (images).I don't find that V(x) is similar to V(Z)...
Paul
... it looks remarkably similar to the outer loop, need to think about what that means
It means you have plotted the outer loop.
Best wishes
David
OK, I had a quick look at the inner loop.
Phase looks a bit odd. I suspect that is due to the feed-forward.
There is also a 180 phase flip, presumably from probe placement but gain should be almost independent of where the probe is placed in the inner loop.
I will need to think and experiment a bit.
I like to index my tests so, for the records, do you have a name?
Last edited:
...It means you have plotted the outer loop.
Didn't realize V(z) is outer and V(x) is inner - probably a consequence of playing with LTSpice until 3:00am (Was trying to use an inductor method from the web to plot the inner loop, without success, before trying your probe)
The name is Tim, it's nice to know that it goes into your index of tests
Hi Tim
No worries, so easy to make these simple slips. The IT industry and DIYAudio forum are very different but the technique is the same, let someone with fresh eyes have a look.
Practically useless for inner loops and more error prone than Tian probe even when it does work, no reason to ever use it I think.
Now I have a folder of your work called "Tim", that's much friendlier than "Level_shifted_folded_cascode_push-pull_VAS"
Thanks for the motivation to re-read Samuel's article.
An important observation is that his comparison with a standard amp is not really fair.
(One of the comments on the article also confirms this.)
So that undermines some of the rationale for the extra complexity.
Still very educational and fun.
Best wishes
David
The VAS cascodes could be Hawksford style, would this improve results?
No worries, so easy to make these simple slips. The IT industry and DIYAudio forum are very different but the technique is the same, let someone with fresh eyes have a look.
Practically useless for inner loops and more error prone than Tian probe even when it does work, no reason to ever use it I think.
The name is Tim, it's nice to know that it goes into your index of tests
Now I have a folder of your work called "Tim", that's much friendlier than "Level_shifted_folded_cascode_push-pull_VAS"
Thanks for the motivation to re-read Samuel's article.
An important observation is that his comparison with a standard amp is not really fair.
(One of the comments on the article also confirms this.)
So that undermines some of the rationale for the extra complexity.
Still very educational and fun.
Best wishes
David
The VAS cascodes could be Hawksford style, would this improve results?
Last edited:
I'm making changes to the schematic based on the various comments I've received so far, from you and others, starting with your comment about bias spreader. I went for the CFP spreader as a veteran member of this forum recommended it to someone in another thread, for it's lower impedance. It's probably not worth the risk with a EF3, but for my knowledge would it be actually beneficial in some other situations.
In fact I also thought about this before, one of the major selling point of this VAS is PSRR (and I haven't taken advantage of the slew rate benefit, limited by the IPS chosen), and I might achieve a good PSRR by adding a capacitance multiplier also, but I would like to have more learning than transferring a schematic from Self's book to a PCB ... and it turns out not bad - I still don't know how to interpret the inner loop plot, but the intellectual challenges continue.
Thanks and regards,
Tim
This question is beyond my knowledge limit, to start with I need to study what is a Hawksford style cascode first ....
The impedance compensation resistor discussed in Self and Cordell is simpler and probably almost as effective.
"Probably" because I've never actually done a detailed study, but looks reasonable.
My personal opinion is that it's a fair bit of extra complexity to solve PSSR and slew rate concerns when there are simpler ways to improve both.
But fun and education are valuable objectives in themselves.
This question is beyond my knowledge limit, to start with I need to study what is a Hawksford style cascode first ....
The classic reference is Malcolm Hawksford "Reduction of Transistor Slope Impedance Dependent Distortion..."
It's public domain on the Net, easy to find.
I'm not sure how his ideas would be applied to a cascode that is folded (like yours), as opposed to his examples, where none are folded.
Just wanted to have it noted, to remind me to maybe look later.
Similarly, I note that Samuel has only one Miller capacitor, specifically to avoid the mismatch problem I mentioned earlier, you have split it into two for implementation reasons. Maybe reconsider this and unify. (As an IT person you probably understand why I call this "normalization" and recommend it.)
Also, the EF transistors in the VAS run at low Vce, maybe worth a look later too.
Best wishes
David
Last edited:
Tim, I would add a very big caveat about your trying to build this amp.
I'm sure someone will correct me if I'm wrong but NO ONE HAS MADE A POWER AMP WITH THIS TOPOLOGY BEFORE.
Guru Groner's measurements were with an 'output stage' with bc550/560.
There is a huge difference between that and your 100W 8R o/p stage.
This is a very complex circuit. Toni describes the work in optimising a much simpler Blameless amp in 2stageef-high-performance-class-ab-power-amp-200w8r-400w4r.html.
This topology has been analysed and built by many but this thread shows there are important subtleties that neither Self or Cordell touch on .. especially in the area of overload behaviour.
Unless you have the experience and equipment that he's got, its unlikely that your build would work better than far simpler circuits.
Very small changes can result in big improvements.
I'm glad to see you've encountered some stability problems and been able to solve them but there's probably more yucky stuff you haven't seen.
You need to ask yourself, "What does the extra complexity buy?" "How can I check that it delivers?" "Do I know what to do if it doesn't?"
But don't let me put you off. As a pseudo guru, I'm as interested as the next person, in how this turns out but I'm not sure it is ethical to get you to foot the bill for this knowledge.
Can you measure THD20k down to <10 ppm?
Toni's thread is important to me as he replicates loadsa stuff I've done in Jurassic times. It shows where SPICE is telling the truth and where it might be lying .. and warns us there might be other things lurking.
Blameless isn't my favourite topology but it is the simplest if very low THD is the priority.
I'm sure someone will correct me if I'm wrong but NO ONE HAS MADE A POWER AMP WITH THIS TOPOLOGY BEFORE.
Guru Groner's measurements were with an 'output stage' with bc550/560.
There is a huge difference between that and your 100W 8R o/p stage.
This is a very complex circuit. Toni describes the work in optimising a much simpler Blameless amp in 2stageef-high-performance-class-ab-power-amp-200w8r-400w4r.html.
This topology has been analysed and built by many but this thread shows there are important subtleties that neither Self or Cordell touch on .. especially in the area of overload behaviour.
Unless you have the experience and equipment that he's got, its unlikely that your build would work better than far simpler circuits.
Very small changes can result in big improvements.
I'm glad to see you've encountered some stability problems and been able to solve them but there's probably more yucky stuff you haven't seen.
You need to ask yourself, "What does the extra complexity buy?" "How can I check that it delivers?" "Do I know what to do if it doesn't?"
But don't let me put you off. As a pseudo guru, I'm as interested as the next person, in how this turns out but I'm not sure it is ethical to get you to foot the bill for this knowledge.
Can you measure THD20k down to <10 ppm?
Toni's thread is important to me as he replicates loadsa stuff I've done in Jurassic times. It shows where SPICE is telling the truth and where it might be lying .. and warns us there might be other things lurking.
Blameless isn't my favourite topology but it is the simplest if very low THD is the priority.
Just to add, have you a scope / sig gen-- speaker loads etc to Test amplifier once its built.
Thanks again for the reminder, currently I'm using a single channel scope meter with built in signal generator, but it's more like a toy so a Rigol 2072A-S will be on it's way to where I live (late Christmas gift to myself but delivery takes 8 weeks). I plan to hack it to 200MHz when I receive it, it's not a high end scope but hopefully enough for my purpose.
Other equipment that I acquired since started this hobby:
- A bench top power supply that can do 2 x 0-60V/3A (or 2 x 0-30V/6A) with current limiting, that makes prototyping easier and less risky.
- A DIY curve tracer (PC based board made and sold by a forum member here), I do buy transistors from China, and need to measure their curves to be sure they're not fakes
- A LCR meter
- 2 DMMs, they can't do True RMS though
- A component tester to measure transistor parameters
- A QA400 to measure distortion
.
Last edited:
Thanks for your comments, especially the link to Toni's thread as it seems to be very useful. After all these kind warnings from yourself and others, it would be naive not to review if it's a bad choice to have selected that VAS, but as I have stated in the first post, the purpose of attempting to design this amp is to:
My main priority is actually the designing process, and not the product (the built amp) - I would be much better off with a low distortion amp by buying a Honey Badger PCB and solder the components onto it, if the amp is the target.
With the very useful comments I got so far, it has been fun and rewarding and I'll see where it takes me to, but if at some point this VAS is proven to be inherently unstable (either through simulation or prototyping), then I'll definitely change my course.
Thanks and regards.
Toni's thread is very educational. Some good tips in there and the end result is superb.
The proof is in the making. Maybe try to make a vero board version. Then you can simulate and experiment in real life simultaneously. This is when I learnt the most. Simulations only tell part of the story. Try to use cheap components and no expensive "fancy audiophile" rubbish. That way failure need not be expensive and there is one less excuse not to build.
I remember making my first CFA with a Hawksford Error Corrected Output Stage. Spent many many hours fighting to get it working. Did eventually succeed which was both satisfying and educational. I'm very stubborn so sometimes go my own way despite warnings.... all part of the fun.
Paul
P.S. Have you looked at the Keantoken H Amplifier? http://www.diyaudio.com/forums/solid-state/243901-keantoken-h-amplifiers.html
The proof is in the making. Maybe try to make a vero board version. Then you can simulate and experiment in real life simultaneously. This is when I learnt the most. Simulations only tell part of the story. Try to use cheap components and no expensive "fancy audiophile" rubbish. That way failure need not be expensive and there is one less excuse not to build.
I remember making my first CFA with a Hawksford Error Corrected Output Stage. Spent many many hours fighting to get it working. Did eventually succeed which was both satisfying and educational. I'm very stubborn so sometimes go my own way despite warnings.... all part of the fun.
Paul
P.S. Have you looked at the Keantoken H Amplifier? http://www.diyaudio.com/forums/solid-state/243901-keantoken-h-amplifiers.html
Last edited:
Tim, from the list of stuff you've acquired and also hints of your technical competence, it's obvious you are serious about becoming an amp design guru.
Well you've come to the right place and I'm certain you will have loadsa help ... and increase the font of human knowledge.
Just be aware that achieving a good working amp MAY be more convoluted than Toni's simpler (??!) design.
Indeed this is a great forum. Thank you for your kind words, I'm barely learning to walk right now, I hope in time I can walk properly.
How do you plan to do this?
I know the sample rate is adequate but I suspect there are other limitations.
Best wishes
David
Very true indeed. This is a year long project for me so I'm taking the time to get the basics right in theory and simulations, and the prototyping will come in Q2/3. Some transistors has arrived yesterday but they have to wait to be soldered, and quite possibly a good number of them will die oscillating in my experiments
With the comment of Struth I've removed the DC blocking caps, the THD1K numbers fall to sub-ppm level straight away, but the THD20K reductions are not as significant. It also reveal another issue - the optimum bias for the OPS seems different for different output power level, this is my attention right now ...
Attachments
The method can be found in eevblog.com.
If what I read is correct, the hardware for 70MHz, 100MHz, 200MHz and 300MHz versions are identical - this is to save R&D as well as manufacturing cost for Rigol. The only thing that differs is the license key, which put a software limit onto the scope.
So the method involved is to dig out an ID which is unique to each scope, and then use this ID to generate the keys for the more powerful versions (someone has written a software for that). Enter the key through the GUI of the scope and upgrade is done.
Many Rigol scopes are hackable and well described in that forum.
Last edited:
I know that sometimes manufacturers deliberately cripple a product so they can sell a "premium" version.
It usually occurs when one player has a dominant position. Intel was (is?) a classic case.
Did not expect this behaviour in the competitive world of oscilloscopes.
But it does explain why the sample rate looked excessive for a 70 MHz scope.
I planned to move up to a Tektronix from my current, slowish, dual trace CRO.
Maybe time to rethink
ThanksBest wishes
David
Indeed, at those price point I wouldn't take the risk of losing the maintenance support from Tektronix either
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.