"As you probably know, a PFC voltage loop must provide reasonable rejection at all multiples of line frequency, and at the same time must pass components below line frequency with as little attenuation and phase shift as possible."
Practically, yes.
Theoretically, a PFC voltage loop should not even pass components below or next to line frequency, since every spectral component in the current waveform, that does not 100% match voltage waveform in amplitude and phase, will degrade the power factor. Do the calculation: PF = real power divided by (rms voltage times rms current).
Going up and down in current amplitude (e. g. two big half waves, two small, two big two small,.....) is creating sidebands in the current spectrum and therefore will degrade power factor. Theoretically, 100% power factor can only be reached if all current wave cycles are same in amplitude from the big bang to the time when sun eats earth or universe will stop expanding (not at all a practical implementation, but the more you can integrate / buffer the power demand, the better)
There is different approaches to this power averaging. - it is totally OK if Eva and choco take the "fast regulator approach!
But:
I cannot do math: wrong.
It is a legend / myth that capacitors can help ease the power demand: wrong
One needs to have 300000uF or 3F or 3 megafarads to get any effect: wrong
Practically, yes.
Theoretically, a PFC voltage loop should not even pass components below or next to line frequency, since every spectral component in the current waveform, that does not 100% match voltage waveform in amplitude and phase, will degrade the power factor. Do the calculation: PF = real power divided by (rms voltage times rms current).
Going up and down in current amplitude (e. g. two big half waves, two small, two big two small,.....) is creating sidebands in the current spectrum and therefore will degrade power factor. Theoretically, 100% power factor can only be reached if all current wave cycles are same in amplitude from the big bang to the time when sun eats earth or universe will stop expanding (not at all a practical implementation, but the more you can integrate / buffer the power demand, the better)
There is different approaches to this power averaging. - it is totally OK if Eva and choco take the "fast regulator approach!
But:
I cannot do math: wrong.
It is a legend / myth that capacitors can help ease the power demand: wrong
One needs to have 300000uF or 3F or 3 megafarads to get any effect: wrong
Reactance:
Looking for suitable chips to control AC<-->DC with integrated magnetics and inherent soft-switch (primary side), and the gadgets of a 4ch amplifier (secondary side), I had to discard both the whole PIC24 family and of course the dsPIC, apart from all stuff from TI, ST and others not meeting requirements. In the end I found a new 8 bit MCU having just the elements needed for AC<-->DC, and a new release of enhanced 32 bit MCU having the features needed for secondary side. A piece of cake is granted for the fastest ones to learn the lessons. Misery is granted for the slowest ones to learn the lessons.
ViennaTom:
Your attitude against product improvement through R&D guarantees a progressive loss of competitiveness until getting out of the business, if you ever were at it. Not speaking about people only, but about whole cultures disappearing and being absorbed by others due to a stubborn lack of R&D investment. R&D is the first thing to close when an activity is being shut down. Greece. Italy. Spain. Eaten by corruption. They all thought it could be fixed with more capacitors. Only for another 10ms. I=C*dV/dt. You are not considering the time needed for re-charging the capacitors, you are assuming the capacitors are always full at the start of the transient. The more capacitance the longer it takes to recharge, when transients have a repetition rate there is never complete recharge, and no difference adding more caps. Not qualified as an engineer.
Looking for suitable chips to control AC<-->DC with integrated magnetics and inherent soft-switch (primary side), and the gadgets of a 4ch amplifier (secondary side), I had to discard both the whole PIC24 family and of course the dsPIC, apart from all stuff from TI, ST and others not meeting requirements. In the end I found a new 8 bit MCU having just the elements needed for AC<-->DC, and a new release of enhanced 32 bit MCU having the features needed for secondary side. A piece of cake is granted for the fastest ones to learn the lessons. Misery is granted for the slowest ones to learn the lessons.
ViennaTom:
Your attitude against product improvement through R&D guarantees a progressive loss of competitiveness until getting out of the business, if you ever were at it. Not speaking about people only, but about whole cultures disappearing and being absorbed by others due to a stubborn lack of R&D investment. R&D is the first thing to close when an activity is being shut down. Greece. Italy. Spain. Eaten by corruption. They all thought it could be fixed with more capacitors. Only for another 10ms. I=C*dV/dt. You are not considering the time needed for re-charging the capacitors, you are assuming the capacitors are always full at the start of the transient. The more capacitance the longer it takes to recharge, when transients have a repetition rate there is never complete recharge, and no difference adding more caps. Not qualified as an engineer.
Last edited:
Eva, a small request please.
http://www.diyaudio.com/forums/powe...able-psu-attempt-average-current-control.html
Do you still have images for that project ? can we/I get those missing images.. By the looks of it, I don`t think you will ever share picture or content like that again given the circumstances.
What 8-PIC/ 32-Bit are you talking about specifically? are you using a ICD3? in your toolchain?
I am currently using a PIC24 PIC24FJ256GB210 its got lots of I/O for a DC load project im building (a diy thing), from all the PIC chips their 24 series are probably the most attractive, the PIC32 series I never touched yet.
"I=C*dV/dt. You are not considering the time needed for re-charging the capacitors, you are assuming the capacitors are always full at the start of the transient. The more capacitance the longer it takes to recharge, when transients have a repetition rate there is never complete recharge, and no difference adding more caps. Not qualified as an engineer."
I do not consider the recharging? BS. In case of the 150BPM bass beats I get 300ms for recharging. 1000W PSU means 1000J energy delivery capability per second, so it can pump an energy of up to 300J within 300ms into the capacitors. Only 175J have been used during the beat - even less than 175J have been sucked out of the caps because the PSU would have contributed also during the beat. To cover the energy budget emergency case that the drummer goes crazy (high repetition rate, low recharge time) I add the PSU overload capability. If all that does not work, scale down the whole signal (turn down the volume) by 2 db and power budget is roses and sunshine again.
I was typing fast yesterday and now I found some calculator "typo".
1400J / 8 = 175J is actually 10 pcs. of 2200uF going from 182 to 130 V, not 8 pcs. Hope that this time I got it right. So in this case you are right that I cannot do the math, but not in the case how you meant it...lol...
Greece , Italy, Spain - eaten by corruption: You forgot to add the US of A and many others... (I am not from southern europe just in case you want to say that those guys there are all corrupt / lazy and have no R&D, what is not 100% true - the people there have just been misguided into overconsumption / debt slavery by their corrupt politicians who are just puppets of the banksters behind. Not sure where you are from but it looks we are no more friends.)
I do not consider the recharging? BS. In case of the 150BPM bass beats I get 300ms for recharging. 1000W PSU means 1000J energy delivery capability per second, so it can pump an energy of up to 300J within 300ms into the capacitors. Only 175J have been used during the beat - even less than 175J have been sucked out of the caps because the PSU would have contributed also during the beat. To cover the energy budget emergency case that the drummer goes crazy (high repetition rate, low recharge time) I add the PSU overload capability. If all that does not work, scale down the whole signal (turn down the volume) by 2 db and power budget is roses and sunshine again.
I was typing fast yesterday and now I found some calculator "typo".
1400J / 8 = 175J is actually 10 pcs. of 2200uF going from 182 to 130 V, not 8 pcs. Hope that this time I got it right. So in this case you are right that I cannot do the math, but not in the case how you meant it...lol...
Greece , Italy, Spain - eaten by corruption: You forgot to add the US of A and many others... (I am not from southern europe just in case you want to say that those guys there are all corrupt / lazy and have no R&D, what is not 100% true - the people there have just been misguided into overconsumption / debt slavery by their corrupt politicians who are just puppets of the banksters behind. Not sure where you are from but it looks we are no more friends.)
Here's my take on the problem.
AC input -> PFC stage -> PFC output cap -> DC/DC isolation stage -> output rail caps -> amplifier stage.
The output rail voltage can't dive to the point of the amp clipping at rated output power. If you make those caps smaller, they'll dive faster.
You can get around this by putting more voltage headroom on the output rails, allowing them to dive further and thereby storing more energy in a larger delta-V. So you've saved power, but increased the quiescent current of the amplifier due to the larger rail voltage - and if you needed higher voltage rated rail caps to do this, you probably haven't saved a whole lot of cost/board space anyway.
Or you can make the DC/DC stage have a faster control loop so it more aggressively "recharges" the rail caps. However the energy to do that is going to come from the PFC cap, and you can't have that voltage dive below the rectified mains voltage or bad things are gonna happen. You can add more margin there by using a higher PFC output voltage for more energy storage. Again you'll increase the quiescent current, and if you need higher voltage caps, you're probably not saving a whole lot here either. Or you can make the PFC stage have a faster control loop, though you'll start running into crappier power factor/line harmonics/whatever.
There's probably techniques where you can sense the output rail voltage or current, and feed that all the way back to the PFC stage for feedforward compensation, letting it increase the power draw from the mains without waiting for the PFC cap voltage to drop out. Things like that might let you lower the rail capacitance without any bad consequences. But yeah, that's R&D work.
AC input -> PFC stage -> PFC output cap -> DC/DC isolation stage -> output rail caps -> amplifier stage.
The output rail voltage can't dive to the point of the amp clipping at rated output power. If you make those caps smaller, they'll dive faster.
You can get around this by putting more voltage headroom on the output rails, allowing them to dive further and thereby storing more energy in a larger delta-V. So you've saved power, but increased the quiescent current of the amplifier due to the larger rail voltage - and if you needed higher voltage rated rail caps to do this, you probably haven't saved a whole lot of cost/board space anyway.
Or you can make the DC/DC stage have a faster control loop so it more aggressively "recharges" the rail caps. However the energy to do that is going to come from the PFC cap, and you can't have that voltage dive below the rectified mains voltage or bad things are gonna happen. You can add more margin there by using a higher PFC output voltage for more energy storage. Again you'll increase the quiescent current, and if you need higher voltage caps, you're probably not saving a whole lot here either. Or you can make the PFC stage have a faster control loop, though you'll start running into crappier power factor/line harmonics/whatever.
There's probably techniques where you can sense the output rail voltage or current, and feed that all the way back to the PFC stage for feedforward compensation, letting it increase the power draw from the mains without waiting for the PFC cap voltage to drop out. Things like that might let you lower the rail capacitance without any bad consequences. But yeah, that's R&D work.
100% agreed.
My interpretation of the current situation is that Eva is working on something new (like single stage PFC PSU or the like) what is not sure to be cost-performance competitive overall. I somehow pierced her by having a critical opinion to the point of taking an opposite view on purpose.
Because she is not sure if her work will win economically over cheap designs, finally her "way of defense" was a weird rant about me and "capacitor corruption" in southern Europe etc. Funny psychological experiment.
Eva has been off from this forum for almost a year, she is definitely working has worked on something big or surfacing after spending time in the mcu/analog/dsp space, she hardly says anything lately... age I think? tired brain, rewind back to 2003~2005 things were HOT here, now its boring only a few engineers left (like Choco doing the DAMP Lite). lots of copy cats though. scanning.
ViennaTom Do you have any projects in the MCU/PSU space you would like to show us?
"ViennaTom Do you have any projects in the MCU/PSU space you would like to show us?"
Sorry, I am "not qualified as an engineer", so I cannot have any projects.
Sorry, I am "not qualified as an engineer", so I cannot have any projects.
Where engineers talk non-engineers listen or leave. This is a rule accurately followed in many successful cultures, but not followed enough in the ones which are currently sinking, where narcissism in action is not suffocated with enough contempt, because power structures are nepotistic and many public figures and leaders are narcissists too. Like what it happens in AC-DC converters when control loop tries to get from magnetics (or reactive branches) higher di/dt than what is physically possible at current voltage level, oscillation and loss of regulation. For stability, di/dt requirement of load must stay below di/dt capability of source.
A collection of images related to "0-15V 0-120A SMPS with average current control" project is attached. Please note this is 10+ years old.
A collection of images related to "0-15V 0-120A SMPS with average current control" project is attached. Please note this is 10+ years old.
Last edited:
Yeah, they pretty old, its like I said some tips and tricks are still useful not really interested in the whole system just the tips.
(I have no interest in your schematics we both know what your stance is on that, sadly now its even subjected up close footage)
zips 1 and 2 seems to be corrupt. even after doing a merge. might need to re-upload.
reactance:
It does not count what you say - there is just one queen here authorized to judge who is an engineer and who has to listen or leave.
It does not count what you say - there is just one queen here authorized to judge who is an engineer and who has to listen or leave.
Reactance:
The ".zip" after ".z01" and ".z02" was added to cope with forum restrictions on file extension. Please rename the files removing the ".zip" after ".z01" and ".z02". I tried and got errors too without renaming.
Additionally: The files are made with PKZIPC command line utility, with the "-span=976000" switch to deal automatically with attachment size limit. In case windows built-in extractor does not recognize them (mine doesn't either), get winzip trial. Or, alternatively in a more computer geek fashion, get winzip command line from same winzip site (contains PKZIPC.EXE, good place to put a copy is in windows directory so that it is found in PATH). Put files in a folder, without renaming, and in command line, within this folder, use "PKZIPC -ext 120V_150A_".
The ".zip" after ".z01" and ".z02" was added to cope with forum restrictions on file extension. Please rename the files removing the ".zip" after ".z01" and ".z02". I tried and got errors too without renaming.
Additionally: The files are made with PKZIPC command line utility, with the "-span=976000" switch to deal automatically with attachment size limit. In case windows built-in extractor does not recognize them (mine doesn't either), get winzip trial. Or, alternatively in a more computer geek fashion, get winzip command line from same winzip site (contains PKZIPC.EXE, good place to put a copy is in windows directory so that it is found in PATH). Put files in a folder, without renaming, and in command line, within this folder, use "PKZIPC -ext 120V_150A_".
This forces me to think -->> Do we really need a dsp to control the smps control loops when it has been greatly done by using analogue ASICs. For supervision purposes, I strongly feel that a uC comes very handy.
Last edited:
Another topic to ad to the works, on the subject of "auto correction" algorithms for microcontrollers..
I completed and still am, evolved with going studies and R&D in Computer Science (my primary work I do everyday) which is programming related. Now and then look for bridging topics with Electronics like this.
Here is a topic that touches on the subject of Auto Correction / Linear Regression. Of course its up to the reader to find a solution that fits or extract principles to fit their problem model or just learn its fundamentals.
http://www.ti.com.cn/cn/lit/an/slaa048/slaa048.pdf
Of course we/i don't expect Eva to comment on this in detail (to much IP red tape), However I am busy with a concept for DIY project that attempts to "Auto Correct" ADC sample measurements and auto calibrate it to some external reference using "Linear Regression" a self correcting algorithm on a microcontroller. (Some may even be used in MPPT solar chargers to find the Maximum Power Point instead of traditional methods like gradient slope scanning)
I completed and still am, evolved with going studies and R&D in Computer Science (my primary work I do everyday) which is programming related. Now and then look for bridging topics with Electronics like this.
Here is a topic that touches on the subject of Auto Correction / Linear Regression. Of course its up to the reader to find a solution that fits or extract principles to fit their problem model or just learn its fundamentals.
http://www.ti.com.cn/cn/lit/an/slaa048/slaa048.pdf
Of course we/i don't expect Eva to comment on this in detail (to much IP red tape), However I am busy with a concept for DIY project that attempts to "Auto Correct" ADC sample measurements and auto calibrate it to some external reference using "Linear Regression" a self correcting algorithm on a microcontroller. (Some may even be used in MPPT solar chargers to find the Maximum Power Point instead of traditional methods like gradient slope scanning)
Last edited:
- Status
- Not open for further replies.