Most amplifiers are instable with capacitive loading, including the so called CFAs..
At some point the usual LR network keep these amps insulated from the real world, certainly a convenient mean to squeeze the ppms but IMO not a good design habit.
Slew rate and THD/IMD are not all, stability from 0 to say 1uF seems to me desirable for an amp that pretend to be of quality, moreover given that speakers impedances are highly variables from a model to another, let alone from different brands.
At some point the usual LR network keep these amps insulated from the real world, certainly a convenient mean to squeeze the ppms but IMO not a good design habit.
Slew rate and THD/IMD are not all, stability from 0 to say 1uF seems to me desirable for an amp that pretend to be of quality, moreover given that speakers impedances are highly variables from a model to another, let alone from different brands.
I do think the fashion for uF load driving plots is a bit unreasonable - pure uF load simply isn't encountered in loudspeakers
Electrostatic Loudspeakers are going to be driven through a step up transformer with series R and L which dominate the load Z well below most audio power amps loop gain intercept frequency
Electrostatic Loudspeakers are going to be driven through a step up transformer with series R and L which dominate the load Z well below most audio power amps loop gain intercept frequency
From a sonic standpoint, I'm very familiar with the tradeoffs 🙂. The tradeoffs are vastly different when the output stage is not in the global loop feedback, but however many may disagree.
Colin
That's what I said. Loop gain not constant but going down with frequency.
So I made a mistake with the word linear. OK so be it, I wanted to say it's not constant over the frequency range of interest.
And that was the discussion.
I still stand by this:
1. Open loop output stage, well designed, has enough low output Z (say 0.25 Ohms with BJT @100mA) by any practical / real world requirements.
Differences in FR between Amps are only relevant in ancheoic room. In normal listening environments, this is simply irrelevant and only academic. Far more other influences the FR in a normal room.
Designs to meet academic goals? Well then, yes....you need feedback.
Oh and there is not really need for a output coil.
2. If you need to significantly reduce crossover distortion with feedback, something is wrong from the start, i.e. before applying feedback.
To me, feedback being better than non feedback is purely academic.
And in exactly those academically controlled listening tests, you won't hear a difference between two well designed units (let's say an Ayre amp for the non NFB variety, and a Parasound for the other).
Way than you prefer NGNFB over GNFB amp if there is no sonic difference? A GNFB amp is easier to design to good sound standards and NGNFB will be in most cases A class.
I do hope we are going to see how it does work and what we need to take account of in our amplifiers.
Control theory is pretty well known stuff. I suspect that most of the engineers here (and more than a few of us non-engineers) are quite familiar with it.
That s not as unreasonable as single digit ppm THD at 20KHz, or slew rates that would be justified only with 10km/s tweeter domes speeds..
On the other hand loading tolerance is downplayed since it s hardly compatible with said much more marketable metrics, as if what is annoying to achieve has been conveniently swept under a rug.
It will be very nice of you if you show some of your designs, not just the words.
Oh this tone...I'm tempted to say, search yourself for open loop designs, there are plenty.
But since I'm a nice guy, some hints and even a simple schematic.
http://www.diyaudio.com/forums/atta...ill-little-amp-might-100w-amp-board-match.pdf
Only the right side, i.e. the power buffer, starting from Q7+10.
Courtesy Steve Dunlap.
Or the F4 from Nelson Pass.
As easy as it gets, but a bit more power "waste".
For power, drive it balanced.
Both do not need a biasing circuit (VBE multiplier), i.e. have a single point input
For the voltage amplifier, take a AD844 simplified circuit and maybe change the four input bias transistors and current sources for a complementary JFet pair. And obviously without output stage.
There are schematics around which mîght be similar to Blowtorch:
Uskok preamplifier
Make it balanced if you want more power. Gives also more freedom for the voltage regulation of input power supply, as current draw is rather constant.
Think current, not voltage btw.
Biasing and offset trimming needs some more thinking (and resistor netwoorks), I do not share everything.
This has to suffice.
For my audio listening, such designs suffice. I can say that I do not need feedback, and therefore to me it is not "a huge benefit for audio", just unnecessary.
Attachments
Last edited:
Hi Guys
Wahab wrote:
"On the other hand loading tolerance is downplayed since it s hardly compatible with said much more marketable metrics, as if what is annoying to achieve has been conveniently swept under a rug."
I see all kinds of circuits tested into 8R which is not a real world load as far as I have seen.
You can get great THD numbers etc into 8R, then you change to 4R and things go south. 2R is even worse. I found this to be particularly problematic with class-G and class-H amps.
Generally, one simply adds more output pairs and eventually another drive stage to accommodate heavier current loading. In conventional amps, there is great advantage even when the load is not changed where the added devices help to reduce THD particularly at low signal levels. Self called this Load Invariance.
The low THD20 goal is one that reliably indicates and assures freedom from slew induced distortions. That has been well known since the '80s or so. I believe most of the better designs published on this forum use this metric and the bigger amps I've designed have low ppm20 even at 2R to assure that the amps can be bridged.
Have fun
Wahab wrote:
"On the other hand loading tolerance is downplayed since it s hardly compatible with said much more marketable metrics, as if what is annoying to achieve has been conveniently swept under a rug."
I see all kinds of circuits tested into 8R which is not a real world load as far as I have seen.
You can get great THD numbers etc into 8R, then you change to 4R and things go south. 2R is even worse. I found this to be particularly problematic with class-G and class-H amps.
Generally, one simply adds more output pairs and eventually another drive stage to accommodate heavier current loading. In conventional amps, there is great advantage even when the load is not changed where the added devices help to reduce THD particularly at low signal levels. Self called this Load Invariance.
The low THD20 goal is one that reliably indicates and assures freedom from slew induced distortions. That has been well known since the '80s or so. I believe most of the better designs published on this forum use this metric and the bigger amps I've designed have low ppm20 even at 2R to assure that the amps can be bridged.
Have fun
I just have to say that something is wrong with "100w amp-board match" schematic, I think it can not work as expected.
Last edited:
There are a Krill thread where people was endlessly arguing about the design, that it cannot work.
And construction threads.
Well possible that the schematic i posted had a mistake.
But OT anyway.
All schematics I am referencing are for inspiration of open loop designs and to show that they are not difficult or complicated or class A, and do not have 10% of distortion.