Hi
Back in August i wrote the attached rapport on small-signal modelling of self-oscillating amplifiers. By computing the equivalent gain of the hysteretic comparator or just the comparator (for an UCD solution), one can find the transfer function of the system or even more interesting the noise transfer function of the system (Shows how good the system is to suppress errors like distortion). The reason why self-oscillating amplifiers outperform clocked amplifiers is because their equivalent comparator gain is the largest. This results in more gain to suppress errors, or in other words a noise transfer function with a higher bandwidth.
There is also an example of how to make a UCD with a higher order
feedback loop that gives 20dB of extra gain to suppress errors.
http://www.student.dtu.dk/~s042302/diy/Small-signal_modelling_of_self-oscillating.pdf
I've also written a second report on self-oscillating amplifiers (it was the first of the two reports I wrote). I deals with the following topics:
Design of passive phase oscillating (UCD) and hysteretic oscillating amplifiers.
Design of additional active poles to increase the loop bandwidth and thereby suppress distortion.
How to get a constant oscillating frequency for both phase and hysteretic self-oscillating amplifiers.
Some layout considerations.
Schematic for a dicrete comparator + gate driver with under 100ns of propagation delay.
http://www.student.dtu.dk/~s042302/diy/low_dist_amp_master.pdf
Back in August i wrote the attached rapport on small-signal modelling of self-oscillating amplifiers. By computing the equivalent gain of the hysteretic comparator or just the comparator (for an UCD solution), one can find the transfer function of the system or even more interesting the noise transfer function of the system (Shows how good the system is to suppress errors like distortion). The reason why self-oscillating amplifiers outperform clocked amplifiers is because their equivalent comparator gain is the largest. This results in more gain to suppress errors, or in other words a noise transfer function with a higher bandwidth.
There is also an example of how to make a UCD with a higher order
feedback loop that gives 20dB of extra gain to suppress errors.
http://www.student.dtu.dk/~s042302/diy/Small-signal_modelling_of_self-oscillating.pdf
I've also written a second report on self-oscillating amplifiers (it was the first of the two reports I wrote). I deals with the following topics:
Design of passive phase oscillating (UCD) and hysteretic oscillating amplifiers.
Design of additional active poles to increase the loop bandwidth and thereby suppress distortion.
How to get a constant oscillating frequency for both phase and hysteretic self-oscillating amplifiers.
Some layout considerations.
Schematic for a dicrete comparator + gate driver with under 100ns of propagation delay.
http://www.student.dtu.dk/~s042302/diy/low_dist_amp_master.pdf
Wonderful, that by using one of such approaches, i've got indeed .00075% at 15W with THD/Freq flat (measured in 20-20khz BW, however amp does sound very close to the usual UcD400, so sound aren't automatically better if more zeros after point in the THD specs), or .0003% (maybe less even, sorry -no APsys2) if 2-3Khz THD rising, also same approach used in the 400VDC H-bridge for .0017% at 50-100Hz>600W@230VAC. 
Thanks TOINO, IVX and BWRX
In the "Small-signal modelling of self-oscillating
switch-mode amplifiers" report I cite a paper by Lars Risboe. I've attached it here if some of you are interested. He explains very well how to model the small-signal properties of a comparator.
The feedback approach or what are you reffering to?IVX said:
In the "Small-signal modelling of self-oscillating
switch-mode amplifiers" report I cite a paper by Lars Risboe. I've attached it here if some of you are interested. He explains very well how to model the small-signal properties of a comparator.
I appreciate your interest.
I've found out that self-oscillating converters (audio amplifiers or power supplies) also goes under the name sliding mode regulators. I've found several papers about sliding mode control, that among others describe self-oscillating converters with other power stages than a buck half- or full-bridge. This is very interesting since
it suddenly turns out that there's much more information out there, than I thought previously.
Searching google or even better IEEE.explorer with "Sliding mode control for switched mode power supplies" or something similar, really gives a lot of interesting hits. http://www.hait.ac.il/jse/B/vol0236B/jse15.pdf is one example.
Does any of you know something about sliding mode control?
I've found out that self-oscillating converters (audio amplifiers or power supplies) also goes under the name sliding mode regulators. I've found several papers about sliding mode control, that among others describe self-oscillating converters with other power stages than a buck half- or full-bridge. This is very interesting since
it suddenly turns out that there's much more information out there, than I thought previously.
Searching google or even better IEEE.explorer with "Sliding mode control for switched mode power supplies" or something similar, really gives a lot of interesting hits. http://www.hait.ac.il/jse/B/vol0236B/jse15.pdf is one example.
Does any of you know something about sliding mode control?
Hi sovadk
It is not the some.
You could have the full picture reading this:
page-45 of: http://edu.lut.fi/LutPub/web/isbn9517649800.pdf
http://www.pages.drexel.edu/~vn43/t...ntrol/sliding_theory/sliding_mode_theory.html
http://www.pages.drexel.edu/~vn43/tutorials/sliding_mode_control/smcsp/smcsp.html
http://etd.caltech.edu/etd/available/etd-09222006-170253/unrestricted/Venkataramanan_R_1986.pdf
http://www.dei.unipd.it/~pel/Articoli/1995/Jcsc/Jcsc.pdf
It is not the some.
You could have the full picture reading this:
page-45 of: http://edu.lut.fi/LutPub/web/isbn9517649800.pdf
http://www.pages.drexel.edu/~vn43/t...ntrol/sliding_theory/sliding_mode_theory.html
http://www.pages.drexel.edu/~vn43/tutorials/sliding_mode_control/smcsp/smcsp.html
http://etd.caltech.edu/etd/available/etd-09222006-170253/unrestricted/Venkataramanan_R_1986.pdf
http://www.dei.unipd.it/~pel/Articoli/1995/Jcsc/Jcsc.pdf
I presume that you mean same ant not someTOINO said:
but anyway I have to disagree. On page 45 in Mohammad Ahmed paper the type of control is the same as in a UCD audio amplifier. But instead of saying comparator they explain the modulation action by a sign(S) function, where S is the sliding surface.
The sliding surface is what we call the feedback signal in an UCD.
Sliding mode control is close connected to state space control and feedback of all states. In an UCD amplifier you approximate the inductor current (state 1) with a lead filter and add it the the output voltage (state 2). So it works the same way as sliding mode state space control.
Where we use transfer functions when we design an ucd, the sliding mode people uses the phase plane (or at least start their analysis with a phase plane analysisi. Most likley because they have an background in nonlinear control).
So the thing is that it's all the same but with different words.
It's some interesting literature you've found there. Have you worked with sliding mode in eg. power supplies?
Sorry… my practical experience is resumed to an only one “One Cycle Control” amplifier prototype.
Sliding Mode Control is some thing I collect information for future reading… as for many other control techniques.
I must confess that some of my best literature is from your University… like this one: http://www.diyaudio.com/forums/showthread.php?postid=942546#post942546
As far as I understand, in SMC the system trajectory moves and oscillates around the ideal sliding line in a defined boundary who is the Sliding Surface.
The system dynamics is effectively governed by the boundaries rather than the trajectory.
In this respect, many control systems have the attributes of Sliding Control. One example is the Hysteric control.
Are you suggesting that UCD is another example?
Remember that UCD is not hysteric and if you have only one control loop, from where do you sense the current?
Sliding Mode Control is some thing I collect information for future reading… as for many other control techniques.
I must confess that some of my best literature is from your University… like this one: http://www.diyaudio.com/forums/showthread.php?postid=942546#post942546
As far as I understand, in SMC the system trajectory moves and oscillates around the ideal sliding line in a defined boundary who is the Sliding Surface.
The system dynamics is effectively governed by the boundaries rather than the trajectory.
In this respect, many control systems have the attributes of Sliding Control. One example is the Hysteric control.
Are you suggesting that UCD is another example?
Remember that UCD is not hysteric and if you have only one control loop, from where do you sense the current?
I am suggesting that UCD i a SMC together with all the other self-oscillating amplifiers. When the SMC was first used for mechanic systems, they used comparators (sign controllers) rather than hysteretic comparators. When I went to Elbert Hendriks, my upcoming supervisor, last month and said that I wanted to write a project on SMC and started talking about hysteretic comparators, he thought I had got it all wrong.
Your right, the inductor current is not sensed in an UCD, but it is approximated with a lead filter though. The lead filter differentiates the voltage on the output capacitor. This gives you a voltage proportional to the current in the capacitor which is the same as the current in the inductor if there's no load. If there is a load the two currents are not quite the same, but it actually turns out that you get a lover output impedance by using the capacitor current rather than the inductor current. So this solution is better.
In the end, classic controllers with one feedback note, actually turns out to work the same way as state feedback, but with a much cheaper design.
By combining the notation of state control with the classic compensator filters like lead and lag eg. together with transfer function analysis, your most likley better off than if you just used state space methods.
But again, they use different words for two very related things and it will not be easy to bring people from those two poles closer together.
About the constant switching frequency amplifier by Søren Poulsen. Have you made it work and measured it's THD? Michael A. E. said that their was some problems with THD not reported in the paper. As most papers it's more a show off than a source of practical information.
I saw that someone on this forum has made a LTspice simulation of that approach. How did it perform?
IVX have you seen the single opamp second order integrator that IRF uses in their development board for iraudamp3?
http://www.diyaudio.com/forums/showthread.php?s=&threadid=90405
Your right, the inductor current is not sensed in an UCD, but it is approximated with a lead filter though. The lead filter differentiates the voltage on the output capacitor. This gives you a voltage proportional to the current in the capacitor which is the same as the current in the inductor if there's no load. If there is a load the two currents are not quite the same, but it actually turns out that you get a lover output impedance by using the capacitor current rather than the inductor current. So this solution is better.
In the end, classic controllers with one feedback note, actually turns out to work the same way as state feedback, but with a much cheaper design.
By combining the notation of state control with the classic compensator filters like lead and lag eg. together with transfer function analysis, your most likley better off than if you just used state space methods.
But again, they use different words for two very related things and it will not be easy to bring people from those two poles closer together.
About the constant switching frequency amplifier by Søren Poulsen. Have you made it work and measured it's THD? Michael A. E. said that their was some problems with THD not reported in the paper. As most papers it's more a show off than a source of practical information.
I saw that someone on this forum has made a LTspice simulation of that approach. How did it perform?
IVX have you seen the single opamp second order integrator that IRF uses in their development board for iraudamp3?
http://www.diyaudio.com/forums/showthread.php?s=&threadid=90405
Hi sovadk
You mean that the filtered carrier at output capacitor node is the Sliding Surface and the audio signal is the ideal Sliding Line?
In that sense, all the modulators with feedback are SMC…
That paper is not the only thing accessible on the site. I have found also (surrounded by other things
) the Simulation files :http://server.oersted.dtu.dk/au/peg...er/Act/files/CD_new/ORCAD project files/BPCC/
You mean that the filtered carrier at output capacitor node is the Sliding Surface and the audio signal is the ideal Sliding Line?
In that sense, all the modulators with feedback are SMC…
sovadk said:
That paper is not the only thing accessible on the site. I have found also (surrounded by other things
) the Simulation files :http://server.oersted.dtu.dk/au/peg...er/Act/files/CD_new/ORCAD project files/BPCC/
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.