I have recollections now from my own EC work, many years ago. Can't find the MathCad docs right now, but remember some issues I found.
What I found of interest is that if you want to use EC to decrease the distortion by say 40dB, there are a few requirements for the EC circuitry itself. For instance, the EC works with a form of cancellation between two signals. And to effect the 40dB reduction (cancellation) obviously the EC circuit must be at least 1% accurately set as to its gain or whatever. This appears no issue with precision opamp circuits.
But the thing is that this 1% accuracy must be maintained to 100 times the closed loop roll-off frequency of the main amplifier. If not, the EC will start to fail towards the higher frequencies. Basically that means a 1% accuracy maintained to 100x the closed loop roll-off. That, my friend, is not trivial! Do the numbers ...
Jan
Hello Jan and ALL,
In another thread at another time you posted one of Jerald Graeme’s feedback articles. I purchased some operational-amplifier prototype PCB’s from Pete Millett. A few LM4562’s and variable pots later my AP2522 was showing what I thought was magical THD results with a 12B4A tube circuit. I had not seen or read Bob Cordell’s article in V0 of Linear Audio at the time.
Today I went to the shelf and pulled down Linear Audio V0 and looked at Bob Cordell’s article. Bob gives credit for the performance of his DM to the distortion performance of the LM4562’s that he used.
To me that put the optimum operational limits of the DM at the distortion floor of the operational-amplifiers used to build the DM.
Texas Instruments puts the distortion performance of the LM4562 at 0.000003% or ~ -150dB. That is pretty low. The DM is not going to do better than this.
Jan, you are right on with your Mathcad analysis.
Thanks DT
I think we've been trying to say that it doesn't work that way. This kind of superposition when dealing with a non-linear system inside a feedback loop does not apply. You need to actually measure the result. Auto-correlation would only remove the noise of the DM (and you would need two of them). I would recommend a lock-in amplifier tuned to each harmonic as possibly useful too.
Well my insistence on the fact that it does work that way, at least to the point where I've currently shown is based upon the fact that when I do my testing where I make the main amplifier distort highly and then I apply the EC, the amount of EC improvement remains static as I shift the distortion of the main amplifier downward. This also applied back when I had the DM working.
In other words I haven't gotten to a point where I've noticed a reduction in EC effectiveness. But of course I simply cannot see the distortion if the main amplifier is not at least somewhat distorting and I've never been able to get the distortion to go above the noise floor even with the DM and a X1 gain on the amplifier without doing something crazy.
My headphone amp is actually only a simple version of my EC circuit on a PCB because there was simply no need to make it sophisticated as the distortion never crossed the -150db noise floor that the DM was showing me until I melted it way beyond spec. I uses a class AB output stage so even though it's supposed to be a headphone amp I used it as a speaker amp and pumped like 10W into 8 ohms. The distortion only showed itself when the solder was melting. It's a really annoying situation.
It was my understanding that your point about the EC not affecting the amp after a certain point was only due to the noise floor issue. As far as your recommendation on lock-in amplifiers or whatever, I'm over my head. I wonder if there is somewhere I can go or someone I can see that can help here.
Last edited:
It was my understanding that your point about the EC not affecting the amp after a certain point was only due to the noise floor issue.
No, you just can't observe it so there is no basis in any claims to it working "forever".
The bi-polar output stage still working at the melting point of the solder (183C package temperature) is also possibly extraordinary.
I thought the EC maintaining performance as you reduce the main amp distortion was a false assertion/assumption not a demonstrated fact?hellokitty123 said:Well my insistence on the fact that it does work that way, at least to the point where I've currently shown is based upon the fact that when I do my testing where I make the main amplifier distort highly and then I apply the EC, the amount of EC improvement remains static as I shift the distortion of the main amplifier downward. This also applied back when I had the DM working.
I predict that this thread will end in tears. When someone pops up proclaiming wildly impossible results etc. and gets the feedback which their results deserve they usually either eventually go sulky and silent or throw their toys out of the pram and the thread gets locked by the Mods. Congratulations on being persistent; persistent in not telling us exactly what you are doing so we are unable to poinpoint the exact misunderstanding driving your experiments.
No, you just can't observe it so there is no basis in any claims to it working "forever".
The bi-polar output stage still working at the melting point of the solder (183C package temperature) is also possibly extraordinary.
It wasn't bipolar, I used integrated complementary SMD power mosfets. I forget which ones. They slid right off their pads after moving the amp during the test. Killed the entire amp 🙁
So....what exactly do I do from this point then? You're saying that I cannot prove with absolute certainty that the EC still functions after a certain point if the main amplifier itself is already of low distortion because it can't be measured, which is true. Do I just leave it at that? What would I need to do to have some manner of presentable data presuming I were to meet up with a potential investor?
I thought the EC maintaining performance as you reduce the main amp distortion was a false assertion/assumption not a demonstrated fact?
I predict that this thread will end in tears. When someone pops up proclaiming wildly impossible results etc. and gets the feedback which their results deserve they usually either eventually go sulky and silent or throw their toys out of the pram and the thread gets locked by the Mods. Congratulations on being persistent; persistent in not telling us exactly what you are doing so we are unable to poinpoint the exact misunderstanding driving your experiments.
It is demonstrable. All I do is make the main amp distortion badly and then apply a low level version of my EC and then slowly back off the distortion of the main amp. Then I keep track of the relative improvement of the EC as far as the analyzer will allow.
Also, no tears here. I'd be happy with whatever conclusion. Just stress off my back either way.
Last edited:
Someone in PM just gave me an idea. Amplifiers suck a lot worse @ 20khz. Maybe I measured 20khz performance at hundreds of watts I can get an easier measurement. I don't know why I didn't think of this before. Probably because I can't afford to build such a set-up, but it does seem like it would work. What do you think?
This is the bit which, if true, is truly awesomely clever. To do any error correction you need a detailed model of how the main amp distorts, which will depend critically on the signal level i.e. you need either a whole family of models and interpolate between them, or you need one model which is parameterised. To get a model which maintains correctness near and into clipping is amazing. So amazing that I am not sure I believe it.hellokitty123 said:All I do is make the main amp distortion badly and then apply a low level version of my EC and then slowly back off the distortion of the main amp.
However, if you know in detail how the main amp distorts why not design it out rather than compensate for it?
You already have our conclusion: your system is not doing what you think it is doing. What we still have to find is what it is doing, and why you think it is doing something much cleverer. I suspect we will fail in this quest. There is then a danger that you might misinterpret our failure to explain your error as somehow being an endorsement of your technology.I'd be happy with whatever conclusion.
Someone in PM just gave me an idea. Amplifiers suck a lot worse @ 20khz. Maybe I measured 20khz performance at hundreds of watts I can get an easier measurement. I don't know why I didn't think of this before. Probably because I can't afford to build such a set-up, but it does seem like it would work. What do you think?
Signal generation and measuring harmonics as you go up in frequency gets tougher, too, so it's not a free lunch. IMD methods might be more fruitful in this regard.
Again, don't think of it as error correction. It's not really a true EC in the typical sense. That's why I abandoned error correction even though it was my focus for a long time. It's good, but ultimately too limited.This is the bit which, if true, is truly awesomely clever. To do any error correction you need a detailed model of how the main amp distorts, which will depend critically on the signal level i.e. you need either a whole family of models and interpolate between them, or you need one model which is parameterised. To get a model which maintains correctness near and into clipping is amazing. So amazing that I am not sure I believe it.
However, if you know in detail how the main amp distorts why not design it out rather than compensate for it?
Now of course you could be right and maybe some how, some way I'm getting false results but I find that highly unlikely because I've done so many tests from so many angles to ensure that the results I was getting were correct and it is always the same.
I can do a side by side video recording of the scope and the analyzer as I back off the distortion of the main amplifier to show the consistency of the EC into clipping if you want. But that will have to wait because I just put away my lab and I'm not taking it back out again at the moment. Also I need to find some cameras and set it all up. I would need some sort of justification for putting in the effort. I've run out of time to goof off on amp stuff for the next few weeks at least anyway, however I'm open to ideas until then.
I'm have one of victor's oscillators if that helps. However, I don't know how well the QA401 handles 20khz measurements.Signal generation and measuring harmonics as you go up in frequency gets tougher, too, so it's not a free lunch. IMD methods might be more fruitful in this regard.
Last edited:
No matter how well an EC function performs its task in suppressing distortion, it will never be able to correct its own noise. This noise sets the limit to measurable distortion reduction.
Hans
Hans
If you really believe in this circuit you should file a patent for it maybe, so then you can feel secure in letting it be reviewed.
It is demonstrable. All I do is make the main amp distortion badly and then apply a low level version of my EC and then slowly back off the distortion of the main amp.
How do you back off the distortion of the main amp without lowering the output level? You already said that's how you make it distort in the first place. You would make a better experiment if you put a known distortion into the amplifier like a diode shaper and kept all levels the same.
If you mean tuning the EC to the 8 Ohm load we have already shown that that does not work with a speaker where there is any change in angle between the voltage and current, let alone the large variation in impedance. (the infamous Krill thread).
Last edited:
It is demonstrable. All I do is make the main amp distortion badly and then apply a low level version of my EC and then slowly back off the distortion of the main amp. Then I keep track of the relative improvement of the EC as far as the analyzer will allow.
So, this is how it works? You have a circuit that compares the output (with distortion) to the input (no distortion) and feeds the distorting component back in some opposite phase so that it is nulled?
Sounds like EC to me.
Jan
But isn't that irrelevant? Because let's say the main amplifer measures at X distortion under Y condition. If the EC provides Z bonus performance and maintains a static improvement throughout the range of distortion adjustment of the main amplifier then the result is still valid because it shows that the EC will maintain its own performance independent of the main amplifier no? At least as far as the analyzer will show anyway. As for the diode idea, could you elaborate on how that test would work?How do you back off the distortion of the main amp without lowering the output level? You already said that's how you make it distort in the first place.
I never said anything of the sort. That is not the principle of operation. I was talking about how I could show the consistency of the performance of the "EC" while shifting the distortion of the main amplifier downward in order to see if it changes at all once the amplifier gets to a certain level of low distortion.So, this is how it works? You have a circuit that compares the output (with distortion) to the input (no distortion) and feeds the distorting component back in some opposite phase so that it is nulled?
Sounds like EC to me.
Jan
I made a thread about this once and it concluded with the fact that NDAs and Patents mean nothing if you aren't rich enough to drop the money to defend them. Plus china ignores them anyway. Otherwise I'd be signing an NDA with a qualified engineer that can help me like white on rice.If you really believe in this circuit you should file a patent for it maybe, so then you can feel secure in letting it be reviewed.
Last edited:
It's going to get reverse engineered and stolen within 5 minutes if it works anyway, you know. I don't think it's easy to make a living through high-end audio and certainly not selling equipment on technical merits. It's a salesman's business.
Yeah but at least I can sell a bunch first based on hype. I don't know about the speaker market as much but the headfi market lives on hype and technical merit.
Of course it sounds better than anything I've heard anywhere else as well, otherwise I would have scrapped it immediately anyway.
But that mostly lies in my dac design because any high performance amp I've designed has not been the bottleneck in sound.
At the very least I'd like to recoup the cost of my efforts.
My plan was to make 1000 high powered units and sell them for like a grand each. But that cannot be done without external investment anyway.
I could also undercut the market and sell cheap but very potent systems but that only works well as a long term strategy.
Of course it sounds better than anything I've heard anywhere else as well, otherwise I would have scrapped it immediately anyway.
But that mostly lies in my dac design because any high performance amp I've designed has not been the bottleneck in sound.
At the very least I'd like to recoup the cost of my efforts.
My plan was to make 1000 high powered units and sell them for like a grand each. But that cannot be done without external investment anyway.
I could also undercut the market and sell cheap but very potent systems but that only works well as a long term strategy.
Last edited:
Sadly data (so far) and even Theory does NOT back your claims, so you better keep it far far away from any potential investors.What would I need to do to have some manner of presentable data presuming I were to meet up with a potential investor?
In fact, I think you already shot your own foot, big time, by asking here.
This is a widely known Forum, specially attractive to anybody interested in Hi Fi, Esoteric and Advanced Audio, etc. so any investor who might be even remotely related to Audio will find this verhy thread, or *somebody* will point to it, any google search will point here, the works.
Unless you had gotten an enthusiastical reception, this thread would be a mixed bag.
As is, it´s investment doom.
Sorry.
Pity you can´t erase it now and proceed as if it never existed.
- Status
- Not open for further replies.
- Home
- Source & Line
- Analog Line Level
- -290 dB Distortion?