Feedback Compensation in SMPS
- Thread starter Fethiyeli
- Start date
you look at basso book for control to output transfer functions of smps topologies.....then you plot it out in logs...then you get the error amplifier transfer function and plot that out over frequency...then add the two plots together cuz they are logs , and you have your closed loop feedback loop for your smps, and you can read off your gain and phase margin.
Seriously though, if you ever learn to derive small signal transfer functions for smps topologies (for the modulator and power stage), then tell your bank manager to expect a huge influx of dosh into your account....because few people in the world can do this.
wHy don't you just buy a AP300 analyser from ridley and hack the feedback loop, and then measure it......that's certainly the most robust way of doing it because you actually measure it.
I never forget this PhD guy coming up with bode plots for a hysteretic buck converter......I am sure that hysteretic bucks aren't suited to feedback compensation analysis.
Seriously though, if you ever learn to derive small signal transfer functions for smps topologies (for the modulator and power stage), then tell your bank manager to expect a huge influx of dosh into your account....because few people in the world can do this.
wHy don't you just buy a AP300 analyser from ridley and hack the feedback loop, and then measure it......that's certainly the most robust way of doing it because you actually measure it.
I never forget this PhD guy coming up with bode plots for a hysteretic buck converter......I am sure that hysteretic bucks aren't suited to feedback compensation analysis.
Ridley AP300 is like $15,000, that's probably a good reason why not to buy one 
You can rent one for only about $1000 a day though for final testing of a design. Good design in the first place saves a lot of money.
It's not that hard to derive transfer functions if you're good at maths, although it's not for the faint hearted. I've done it before with only an A-level in maths, no engineering degree and a potentially autistic obsession. You don't really need to derive transfer functions most of the time though as most designs are based on existing work with known transfer functions, you just have to know how any alterations you make will affect it, or when what should be stable on paper isn't stable in SPICE you might have to start from scratch to work out where things went wrong.
e.g. when I first did a feedforward voltage mode control, and I read in the textbook/app note that the effect it had on the transfer function was to divide by Vin, changing the existing Vin to Vin/Vin=1. I made a linear model, made a stable design, made a switching model and it was unstable. I didn't expect a textbook to be wrong, so after a week of head scratching I went and derived the transfer function and found that this part of it while mostly constant regardless of Vin was not =1. I put in the number it was equal to, adjusted my compensator and it worked.
I also derived a whole linear model for average current mode control (it was way overcomplicated until I changed from a Norton to Thevenin equivalent circuit). It was so satisfying to see where the inductor drops back into the loop at high frequency as it should without me having to do anything to the model for that behaviour as I took it as a strong suggestion that my model was correct.
This brings me to my point... I'd be grateful if you could tell me how to make a ton of money with this skill, because that's something I haven't figured out yet...
You can rent one for only about $1000 a day though for final testing of a design. Good design in the first place saves a lot of money.It's not that hard to derive transfer functions if you're good at maths, although it's not for the faint hearted. I've done it before with only an A-level in maths, no engineering degree and a potentially autistic obsession. You don't really need to derive transfer functions most of the time though as most designs are based on existing work with known transfer functions, you just have to know how any alterations you make will affect it, or when what should be stable on paper isn't stable in SPICE you might have to start from scratch to work out where things went wrong.
e.g. when I first did a feedforward voltage mode control, and I read in the textbook/app note that the effect it had on the transfer function was to divide by Vin, changing the existing Vin to Vin/Vin=1. I made a linear model, made a stable design, made a switching model and it was unstable. I didn't expect a textbook to be wrong, so after a week of head scratching I went and derived the transfer function and found that this part of it while mostly constant regardless of Vin was not =1. I put in the number it was equal to, adjusted my compensator and it worked.
I also derived a whole linear model for average current mode control (it was way overcomplicated until I changed from a Norton to Thevenin equivalent circuit). It was so satisfying to see where the inductor drops back into the loop at high frequency as it should without me having to do anything to the model for that behaviour as I took it as a strong suggestion that my model was correct.
This brings me to my point... I'd be grateful if you could tell me how to make a ton of money with this skill, because that's something I haven't figured out yet...
I have tried Pspice lots of time but it always gave me an error. Pspice is really annoying software. I connected all components but it always says that unmodelled connected pin etc, even unconnected optocoupler base pin and UC3825's Rt and Ct pins. Another one, psim is really nice but when i need to have 370VDC output, simulation gives me between 120VDC and 200VDC. I am an electrical engineer but i cannot design any f.king sh.t. I've been always reading textbooks but they gave me nothing i think. I dont know how to design feedback network with optocoupler and TL431 since i need isolation and i calculate output choke and capacitor but psim gives me really different waveforms than i expect. Sorry for bad words. I think i will able to understand feedback network and other s.its when i am 50 yo. Now i am 25 yo =)
I calculated the output inductor and capacitor according to pressman's book but when i put calculated inductor and capacitor, the graph of output is as like that:
View attachment 4486
View attachment 4487
I've tried many things but nothing changed. Output voltage is always so low. I would like to have 370VDC at output.
View attachment 4486
View attachment 4487
I've tried many things but nothing changed. Output voltage is always so low. I would like to have 370VDC at output.
Please have a look at this video, it will demonstrate everything required to design, simulate and prototype feedback compensation networks. Its part of a multi part video, this is video 8 that takes a individual through the design process of a buck converter.
http://www.youtube.com/watch?v=_QzzBDvHU5M
Here is another one.
http://www.youtube.com/watch?v=1KJjtK6jEkQ
Last edited: