No-no, nothing complicated.
http://www.diyaudio.com/forums/class-d/188887-china-fp10000q-service-manual-2.html#post5371127
Hi Pavel
I think that some sort of variable rail is probably the ultimate solution.
Drive them from a class-D type switcher for efficiency and have the OPS always with low volts across the transistors to minimize power dissipation in the linear section.
Efficient and very low distortion.
I believe DIY member Harry Dymond did a paper on this, or similar.
Just a bit complicated at the moment
Maybe when I have time to study switchers/Class D in detail.
Best wishes
David
Hi David,
I started in that direction some time ago, and hopefully will put it to a working amp. 200W Class A amp with high efficency
This is really the crux of it.
It's not just the small improvement in efficiency or headroom that interests me.
An EF can be considered as a CE with local 100% -ve feedback.
So the output impedance is lower than a CE.
But if we use a CE we can then move the extra gain into the global loop.
Where it not only reduces the output impedance back to the level of the EF version but also reduces the distortion of the now-included IPS and VAS.
At least that's the idea.
The distortion reduction may not be that substantial, the IPS is already low level and low distortion, but it looks a neat trick to improve efficiency and simultaneously improve the distortion.
And educational to see if it works.
Best wishes
David
Hi Damir
Thanks for the reminder, I see that I was post #4 in that thread! but it went quiet.
I think it's probably an excellent way to do it.
But I would need to study all the theory, see how Labgruppen do it, consider all the options, I can't just build an amp
Actually, I don't need to have any more amps.
I hope your project is successful, I just want to understand linear amps first.
Best wishes
David
Hi Damir
Thanks for the reminder, I see that I was post #4 in that thread! but it went quiet.
I think it's probably an excellent way to do it.
But I would need to study all the theory, see how Labgruppen do it, consider all the options, I can't just build an amp
Actually, I don't need to have any more amps.
I hope your project is successful, I just want to understand linear amps first.
Best wishes
David
Hi David,
The same for me, I've got to many amps, actually don't need more, but this is just a hobby and technical challenge so I have to make more.
Best regards
Damir
What I'm not sure about is whether that will work acceptably. The trimmer potmeters on the two VBE multipliers have to trim out two things: the mismatch on the P-side and the mismatch on the N-side. Can you combine that into one control?
By the way, my circuit circumvents trimmer potmeters by using well-matched transistor arrays for the parts that need to match.
This reminds me of an old paper from Cherry's research group advocating common-emitter outputs for that precise reason. I'll see if I can find it in the attic.
I just looked it up and remembered why I had mixed feelings about it. It's the article:
Edward M. Cherry and Gregory K. Cambrell, "Output resistance and intermodulation distortion of feedback amplifiers", Journal of the Audio Engineering Society, vol. 30, no. 4, April 1982, pages 178...191
They advocate common emitter outputs for the same reasons you mentioned, with lots of calculations to back it up. So far so good.
They calculate in an appendix that the inherent negative feedback of an emitter follower has an advantage for the overall loop stability that is the same as that of adding an extra zero in the loop transfer, but don't pay any attention to this result in the main text.
They don't say anything about how to bias common emitter outputs in class (A)B.
Edward M. Cherry and Gregory K. Cambrell, "Output resistance and intermodulation distortion of feedback amplifiers", Journal of the Audio Engineering Society, vol. 30, no. 4, April 1982, pages 178...191
They advocate common emitter outputs for the same reasons you mentioned, with lots of calculations to back it up. So far so good.
They calculate in an appendix that the inherent negative feedback of an emitter follower has an advantage for the overall loop stability that is the same as that of adding an extra zero in the loop transfer, but don't pay any attention to this result in the main text.
They don't say anything about how to bias common emitter outputs in class (A)B.
Your memory is excellent, I have only read Cherry's summary of the results of that paper.
I was probably biased not to check his claims because they supported my own conclusion, always a trap.
So I will read it carefully, thank you.
Yes, this may be a bit of an issue in a discrete implementation.
Even for an IC the manufacturers admit that rail-to-rail adds a bit of complexity to this area.
Best wishes
David
Last edited:
I am happy even if I work out a better circuit, I don't need to physically make it
I tried to send you an email but your mail box was full, maybe you have a new address since we last wrote.
My DIY mailbox is full, can you send an email to the address in my profile?
Best wishes
David
Rail-to-rail, wasted headroom, wasted power
These parameters are counter to each other. You can achieve close to zero wasted voltage by having a variable supply that tracks the signal. You can almost make this zero-power-wasting if the PSU is an SMPS, but to never waste power headroom is impossible unless you compress and gain-0ride the audio signal so its amplitude is constant. Somewhere in any of these approaches is wasted or unused capability for power output.
In a class-G or class-H amp with many supply levels, the PT still has to be sized for the maximum output of the amp. So, you are "carrying around" the capability of of say 50W when you only ever use 4W.
You can make the same argument for a switch-mode version of the above since the supply has to be designed for its maximum capability and that invokes a transformer rated for it albeit tiny compared to the linear supply.
Music is dynamic despite the massive compression used for CDs and DVDs. There is always headroom not being used and that is a good thing. As the third post stated, semiconductor capacitance goes crazy at low voltage drops across the device.
I have to assume that your query is purely academic? A practical amplifier must have headroom for voltage, current and power to be of any use let alone to be of decent performance. Considering how simple a circuit can be these days that has stellar performance, and that the cost of most components is actually quite low these days, it is of no practical value to take on unnecessary design challenges that will most likely impair performance and add cost.
These parameters are counter to each other. You can achieve close to zero wasted voltage by having a variable supply that tracks the signal. You can almost make this zero-power-wasting if the PSU is an SMPS, but to never waste power headroom is impossible unless you compress and gain-0ride the audio signal so its amplitude is constant. Somewhere in any of these approaches is wasted or unused capability for power output.
In a class-G or class-H amp with many supply levels, the PT still has to be sized for the maximum output of the amp. So, you are "carrying around" the capability of of say 50W when you only ever use 4W.
You can make the same argument for a switch-mode version of the above since the supply has to be designed for its maximum capability and that invokes a transformer rated for it albeit tiny compared to the linear supply.
Music is dynamic despite the massive compression used for CDs and DVDs. There is always headroom not being used and that is a good thing. As the third post stated, semiconductor capacitance goes crazy at low voltage drops across the device.
I have to assume that your query is purely academic? A practical amplifier must have headroom for voltage, current and power to be of any use let alone to be of decent performance. Considering how simple a circuit can be these days that has stellar performance, and that the cost of most components is actually quite low these days, it is of no practical value to take on unnecessary design challenges that will most likely impair performance and add cost.
I am happy even if I work out a better circuit, I don't need to physically make it
I have something similar; although I prefer to physically make them, I also like just working out circuits.
I am happy even if I work out a better circuit, I don't need to physically make it
I tried to send you an email but your mail box was full, maybe you have a new address since we last wrote.
My DIY mailbox is full, can you send an email to the address in my profile?
Best wishes
David
Hi David,
I answered your mail, but from my other email address (also in my profile).
I am out of my home until 27.
BR Damir
Two postscripts.
Someone wrote a reply to the Cherry and Cambrell paper.
AES E-Library >> -Feedback Amplifier Output Stages,- -Comments on Feedback Amplifier Output Stages,- and Author's Reply
Looks like they may have the same observation as you, I'll try to read it tomorrow, with the Cherry /Cambrell paper.
And I have a hazy idea that Cherry did make at least a brief mention of CE bias but I can't recall where, I'll have a search.
Best wishes
David
Thanks Damir, your reply did turn up a few hours after I received the "full mailbox" response.
Enjoy your trip.
Best wishes
David
No, I do intend to build amplifiers of this type if I can create a circuit that shows promise in simulation.
Most modern op-amps are of this type, presumably they find it does not impair performance or add cost
My objective is similar to any IC manufacturer, to improve performance and reduce cost.
The trade-offs for a discrete implementation are a little different so whether I can do it remains to be seen.
If I can't create a better circuit then I will build a conventional EF OPS more happily, not worried that perhaps I could have done better.
In either case I will have had fun to play with the circuit.
Best wishes
David
Not so simple.
Anyway you will need to have fast/wide “current” loop around OPS itself and comparatively narrow, say audio-band, global loop. Yes, you can pick additional 20-30 dB of gain, but they will be hugely load-current-dependent.
I’ve tryed very deep global loop, you will go mad around UGF at ones-tens of MHZ.
One case is load-dependance stability, second case is complicated output filter with audio-band bootstrap, third is poor low-impedance damping-factor.
It’s better to drop all your gain in local OPS negative feedback.
Dave, from my estimation, the drive to rail-to-rail output is the result of needing to squeeze rails down as far as possible. It's a compromise against other aims, let's be honest. Certainly the knowledge base and tools at the modern-day designer's fingertips allows him/her to push to spectacular performance levels within these constraints. In one sense, that's an admirable goal, on the other side, the technical tour-de-force isn't necessary in this application space, nor has any class-B OPS IC amplifier been designed rail-to-rail (that I know of, class D another story). This makes me wonder if it's a output stage limitation when you need that kind of power, even in the IC space, or if there just has not been the market demand for such a product (I suspect it's 25/75%, respectively).
This makes me wonder why boosted rails on the front-end is off the table? Winfield Hill recommended this here on DIYaudio in his high slew rate laboratory amplifier thread (which I might need to build for work, myself)
This makes me wonder why boosted rails on the front-end is off the table? Winfield Hill recommended this here on DIYaudio in his high slew rate laboratory amplifier thread (which I might need to build for work, myself)
I see a lot of comments about rail-to-rail outputs not being necessary, but then again, this whole forum isn't necessary: you can just as well go to a shop and buy audio equipment instead of building it yourself, can't you? I for one am curious what David will end up with, necessary or not.
Getting back on topic, chances are you will need some local feedback around the output stage to keep the overall loop stable when the output stage current varies a lot, but you will only need that at high frequencies (last few octaves before overall loop gain drops to unity).
Getting back on topic, chances are you will need some local feedback around the output stage to keep the overall loop stable when the output stage current varies a lot, but you will only need that at high frequencies (last few octaves before overall loop gain drops to unity).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.