I trust you are finding time to mentor your replacement(s) - or you'll end up rather like Steve Jobs
I'll dig deeper into the patent. Unfortunately, patent format/speak seems to wrap novelty and clarity in a cloak of impenetrablility.
..sounding to me like a real catastrophy...
Well, I am not a hero, but I volunteer to take at least the first part of that burden - just let me know.
Hi Bruno
First, congratulations for the inspiring design.
I just went through the article and the thread as well as the patent and I believe I got the trick.
Moreover I suspect both you and me ended up in the same trick for incresing the feedback beyond the UCD limits, although started with a completely different modulator.
A question therefore is have you used somthing like I've described here:
http://www.diyaudio.com/forums/class-d/169376-english-class-d-pro-audio-amp-5.html#post2247640
?
It is also my observation that all this poles and zeros theory works fine on the paper, but in real life you must also consider which of these transfer function properties stop working when overloaded.
All active zeros are usually no longer here even at slew overload and you end up with pure 1/2piLC oscillation, so the best solution here is to detect that situation and modify transfer function by reducing the number of active poles (I also move a passive zero down in frequency) using bilateral switches or similar.
Triming attack/release timing of these overload detectors is a lot of fun
Kind Regards
Adam
First, congratulations for the inspiring design.
I just went through the article and the thread as well as the patent and I believe I got the trick.
Moreover I suspect both you and me ended up in the same trick for incresing the feedback beyond the UCD limits, although started with a completely different modulator.
A question therefore is have you used somthing like I've described here:
http://www.diyaudio.com/forums/class-d/169376-english-class-d-pro-audio-amp-5.html#post2247640
?
It is also my observation that all this poles and zeros theory works fine on the paper, but in real life you must also consider which of these transfer function properties stop working when overloaded.
All active zeros are usually no longer here even at slew overload and you end up with pure 1/2piLC oscillation, so the best solution here is to detect that situation and modify transfer function by reducing the number of active poles (I also move a passive zero down in frequency) using bilateral switches or similar.
Triming attack/release timing of these overload detectors is a lot of fun
Kind Regards
Adam
Last edited:
The aim is the same (to knock out poles during clip) but the patented bit of Ncore is about how to do this only when necessary and only as long as necessary. You may also be interested in looking at the AES paper, because when you say that "poles and zeros theory works fine on the paper", you imply that practice departs from theory at some point. That can only mean one thing: that said theory is oversimplified. The AES paper shows how to get an exact handle on what the potential oscillation frequencies are under what conditions, including how it jumps from the wanted oscillation frequency to something close to the resonant frequency of the LC filter.
How to get the AES paper not being the AES member?
I just meant the state variables follow nonlinearities of transfer.
And they all saturate the harder (opamp) or softer (choke) way.
In my solution I completely eliminated the inductor pole from the small signal transfer, but not from the 'overloaded' transfer function, so I need two different error amp transfer functions switchable when nessesary.
The bad thing about using any sort of window detector and bilateral switch is that it reacts when the bad thing has already happened and therefore the action needs to be suspended for some time after trigger as well to let the thing calm down again.
I just meant the state variables follow nonlinearities of transfer.
And they all saturate the harder (opamp) or softer (choke) way.
In my solution I completely eliminated the inductor pole from the small signal transfer, but not from the 'overloaded' transfer function, so I need two different error amp transfer functions switchable when nessesary.
The bad thing about using any sort of window detector and bilateral switch is that it reacts when the bad thing has already happened and therefore the action needs to be suspended for some time after trigger as well to let the thing calm down again.
Last edited:
Pay more: AES E-Library Globally Modulated Self-Oscillating Amplifier with Improved Linearity
Exactly. I believe a fairly generic analysis was done by a chap called Tsypkin. I only heard of him afterwards, but all I can say is that there'd be an awful lot of class D gurus if his work were more widely known.
Sounds like a good reason to do it differently, then
Well it's not strictly slew overload that you get when the signal exceeds the power bandwidth. It's just plain clipping before the output filter. And indeed the loop catches that, as well as current limiting.
Once I get working on them again. There's also a 30W FFT people were asking about.
That's not real German in your signature, is it?
Last edited:
Hi,
it is also possible (if the modulator has a new scheme) using a linear equation and apply it as the "actuator" (composed of 3 op-amp.then has not been on-off). remains the "detector" but this is the cherry on the cake.
is much better if the amplifier inherently responds totally differently, instead of "fix or force"
Regards
Roberto
it is also possible (if the modulator has a new scheme) using a linear equation and apply it as the "actuator" (composed of 3 op-amp.then has not been on-off). remains the "detector" but this is the cherry on the cake.
is much better if the amplifier inherently responds totally differently, instead of "fix or force"
Regards
Roberto
Bruno
You may actually publish the paper on the Hypex website or elsewhere (or send it to someone) one year after the AES conference that it was written for. you might even publish it here (not that I would need it - I already have it of course ;-) ). The only exception is if the AES is interested in publishing it in the JAES. That is at least how I understood the copyright small print.
I had a coarse look at your patent dealing with the loop stabilisation. Do I get it correct that the clipping- and integrator reset- circuits themselves are not patented but the way you control them ?
I guess you aren't a fan of Monty Phyton's Flying Circus, are you ?
Regards
Charles
You may actually publish the paper on the Hypex website or elsewhere (or send it to someone) one year after the AES conference that it was written for. you might even publish it here (not that I would need it - I already have it of course ;-) ). The only exception is if the AES is interested in publishing it in the JAES. That is at least how I understood the copyright small print.
I had a coarse look at your patent dealing with the loop stabilisation. Do I get it correct that the clipping- and integrator reset- circuits themselves are not patented but the way you control them ?
I guess you aren't a fan of Monty Phyton's Flying Circus, are you ?
Regards
Charles
I feel guilty about having abandoned the European Convention of the AES, starving it of the money it usually gets for the combined Grimm Audio / Hypex booth. I think I should at least let it make some money for what it does best: bring audio engineers together for an exchange of ideas.
Your query about the patent is one of those "moo" questions. I think "nearly yes" is a good answer.
Monty Python was before my time, or at least the time when I would have appreciated it. I promised myself to get a box set some day. After I get a TV. It's almost certain that I'll love them, what with being a consummate Douglas Adams and Terry Pratchett fan.
Your query about the patent is one of those "moo" questions. I think "nearly yes" is a good answer.
Monty Python was before my time, or at least the time when I would have appreciated it. I promised myself to get a box set some day. After I get a TV. It's almost certain that I'll love them, what with being a consummate Douglas Adams and Terry Pratchett fan.
- Status
- Not open for further replies.