Tutorial Video: "Modern Switch Mode Power Supply Design, Closing feedback loops"

[rbola35618]

"Modern Switch Mode Power Supply Design, Closing feedback loops"

This video is presented by a Wendell Boucher who is a level 5 Electrical Engineer. He uses a simulator called Simplis that can extract an AC model from the switching model. In Simplis, there is no need to derive an AC model. Simplis is a powerful simulator and is also very fast. Wendell goes thru a series of simulations and shows how to use the simulator to verify your designs.

Comments or feedback are welcomed.

[rbola35618]

I have gotten e-mails that the video link is not working properly.  They mentioned that the picture freezes at about the 12 minute mark while the sound continues. Can someone please let me know if they are getting the same problem?
Thanks,

Robert

[Circuitous]

when trying under youtube, it says the video is not available.

[rbola35618]

The author asked me to take if off if it was not working properly. So I made it private and only whose with the link such as link on this blog can see it. I am using my iPad to view it and it seems to working ok for me when I select the link up above. Does it work for you if you if you click on the link up above?

Robert

[alm]

It's private, not unlisted, so even with the link we're not allowed to watch it.

[rbola35618]

Thanks Alms,

I will switch it to unlisted.

RB

[rbola35618]

Just switched to unlisted. Does the link above work for you?

RB

[rbola35618]

Thank you AcHmed99!!!

RB

[xygor]

What are the prices of SIMPLIS?
Why do these companies keep their price list on a need to know?

[rbola35618]

XYOR

The price was in the 5K range last time I inquired about it. I agree most simulators are too expensive. That is one of the reasons that LTSPICE has been popular because it is free. However, Simplis is one of two simulator that can extract a Bode or AC model using a switching model. The other software that has this capability is PSIM. PSIM also cost about the same. But I have been told that the demo version of SIMPLIS is available and you can do quite a bit. You may be able to use the DEMO to do power supply simulations for free.

Robert

[xygor]

Thank you so much Robert!

When you say "extract a model" does that mean you actually get the transfer function?

I found nl5, which can produce AC response plots for a switching circuit that has reached steady state in the transient simulation.  I don't yet know how well this works.

[rbola35618]

Xygor,

You can build a PS using its pwm and you can plot the AC characteristics such as the open loop response, the input and output impedance, the frequency response of the power section which consist of the output filter. Usually, to performs these simulations, you have to use what is called an AC average model such as the pwm switch by Vatche Vorperians. it seems to be fast; faster than LTspice, PSIM, or any other simulators. Wendell, the author of the video presentation, is an experts on this simulator.  He also uses LTspice to simulate his power supplies. I'll see if I can get him  to response to your posting here and answer any questions you may have.

Robert

[free_electron]

These are by no means 'modern' switchers... this is the classic pwm system where they have a current sensor and a voltage sensor. been around for 40 years.

a modern switcher has sample and hold amplifiers on board and can actually measure the coil current with the need of a sense resistor. to do that an additional resistor from the switch point to the feedback point is inserted  ( typically a few kiloohm) is inserted . During switch commutation (synchronous recitification ) there is a deliberate switchover gap. At that point the voltage at feedback is sampled. since both switches are open the coil cureent runs in the resistor and half the feedback divider. this gives a negative spike on the feedack node. sampling this spike tells us the current in the coil... when either the top fet is engaged, or the bottom fet we sample the feedback point as well. some simple calculation gives the system the divider fraction. the only 'hard' value is the resistr across the inductor. so you can still program output voltage using a resistive divider.

this system requires three sample and hold amplifiers and interaction with the commutation controller. this is how all the modern switchers work. combine that with sensefets end you have a hypermodern switchers.

good luck trying to simulate that one ( it can and is done with a mixed mode simulator , not with an abstract model )

[wendell]

Hello all,
There were comments by 'free_electron' regarding the use of the word 'modern'. Please understand that I used that word to describe the use of a very unique and modern software tool to simulate the power supply behavior. It was not meant to imply that the power supply models shown are 'modern' approaches in them selves.  However, I have been designing switch mode power supplies since 1980 and the designs shown are still live and well. Keep in mind that the free downloadable demo version of SIMPLIS was used and this means that it is necessary to use the most simple means to derive a circuit that captures the essence of it behavior. Thank you for taking the time to respond.    Best regards  Wendell Boucher

[xygor]

Thanks for checking in Wendell!
Thanks for posting this as well as your videos rbola.  I really find them helpful.

Question:  I notice that Venable and Ridley produce instruments that measure response in a physical circuit by injecting a frequency stepped series of sine wave disturbances and measuring the response.  Instead, what would be the problem with injecting a pulse disturbance (repeatedly if needed to improve SNR) and doing an FFT on the data collected at the response point.  The result would need to be convolved with inverse of the pulse to correct for the fact that it is not a dirac pulse.  This works in spice transient simulations, but I wonder if it would work in the physical world.

[rbola35618]

Hi Xygor

Yes, Wendell contacted me and sent me his video that he made a while ago. I was blown away with the simulations the he was able to perform. I asked asked him that other people might find his video usefull. He was kind enough in sharing  his simulation video.  Thank you Wendell.

RB

[sudheer]

Thanks for sharing this video. Introducing Simplis and its capabilities to switching power supply design by Wendell was great.

Can anyone please explain how the feedback compensation component values are calculated by setting pole and Zero frequencies in F11 window? This was brought out in this video with example to "Buck Average Current mode, Unitrode Sem 900 Pg C13".  The .VAR was used to set the pole and Zero Frequencies, but does not point out from where the calculations for compensation component values(C1A, C2A2, R2A) come from or is there a script that need to be pre-processed which is not shown.

Appreciate if any one can help.

Thanks

[Richard Head]

Remember what Bob Pease said: My favorite programming language is SOLDER.
Stop screwing around with simulations and build the damn thing.

Dick

[David Hess]

Thanks for checking in Wendell!
Thanks for posting this as well as your videos rbola.  I really find them helpful.

Question:  I notice that Venable and Ridley produce instruments that measure response in a physical circuit by injecting a frequency stepped series of sine wave disturbances and measuring the response.  Instead, what would be the problem with injecting a pulse disturbance (repeatedly if needed to improve SNR) and doing an FFT on the data collected at the response point.  The result would need to be convolved with inverse of the pulse to correct for the fact that it is not a dirac pulse.  This works in spice transient simulations, but I wonder if it would work in the physical world.

Yes, it works great in the real world and is the reason I lament that modern DSOs throw away the FFT phase data while older ones displayed it.  Having DSOs with built in signal generators only makes it worse.

Differentiating the output before the FFT allows using a step instead of an impulse.  Tektronix, Agilent, and LeCroy have application notes discussing this implementation of vector network analysis.

Low frequency swept vector network analysis is available through inexpensive USB based instruments:

http://www.syscompdesign.com/CircuitGear-Mini_p_29.html
http://www.syscompdesign.com/CircuitGear_p_8.html

Or expensive ones:

http://www.omicron-lab.com/bode-100/product-description.html

[rbola35618]

Here is a less expensive alternative. I have this unit and will be doing a review and showing some tips in how to use it. Bart the designer has put a lot of time in getting the software more user friendly and I am happy with the results.



Robert

[mark03]

This video is presented by a Wendell Boucher who is a level 5 Electrical Engineer.

Huh.  Nobody asked the obvious question yet.  WT_ is a "level 5" EE?  I don't know my level.  Does that mean I'm still on zero?   

[rbola35618]

At where I work, there are 5 levels of engineering title. Engineer, research engineer, senior research engineer, principle engineer, and staff engineer would be the top post. Wendell is probably at the top of his skill level since he has over 60 years of experience working as an engineer.

[tiziao]

One basic version (no limitation) of Simplis free:
http://www.monolithicpower.com/Design-Support/MPSmart