Hi Professor. Great video as always. This is the model that my mentor Jack Alexander taught me many years ago. It is a very simple model that can also be used for DCM flybacks and is a first order model. Simple but still powerful enough to set the loop. Again, thank you for the video. Your videos are the best!
Hi Robert, I was not aware of the teaching of Jack Alexander or your videos and posts on the subject that I have just googled. Impressive. Judging from the outcome (you) Jack must have been (or still is) an exceptional mentor. I have been exploring the average modeling subject for many years now (almost too many to count). It has been my conviction that intuitive anderstanding, to which average models belong, are a prerequisite for good engineering.
As someone who spent a lot of time trying to learn how to SPICE model switch mode power supplies about 20 years ago and got very frustrated with it... I thank you sir.
Hello, thanks for great explanation. I really always enjoy how you explain how to study stability of systems using LTSpice! Great work, you are making the world better place :)
Very nice presentation, also seen from a practical side of view. Thank's Professor for your contribution. It may also be very helpful if you could present more examples, including all relevant converter specifications, those which are necessary for Loop designs. If possible, for nonisolated and isolated voltage and current mode converters. Best regards Udo
Hi sam , Nice explanation , I use the same way to set my loops. In case of PCM and D>0.5 I guess this model doesnt predict sub harmonic oscillations ! Thankyou .
Hi Sam , Nice explanation . I use the same way to set my loops but I think this model doesn't predict sub harmonic oscillations in Flyback CCM ! Thankyou .
something I realized while simulating this basic concept in Spice, is that at least with the built in SR latch in Spice, the SR latch is not edge controlled. Is this typical for the SR latch inside this converter? It necessitates the need for very narrow pulse train from the clock otherwise.
This method for loop gain stability checking seems applicable to many typologies so long as the loop gain falls to zero dB well below the Nyquist frequency. Thank you
Hi Chris/ Just seen you intersting write up, can the Siglent SDS1104X-E plot the loopgain of a PWM converter. That it, can it extract the injected signal from the ripple?
@@sambenyaakov I have used it to plot the loop gain of an op-amp circuit by using a test transformer to inject the voltage as you show in this presentation, but I have never used it for a switching converter. It might be possible of one uses the Eres (extended resolution) function on the Siglent. The Eres mode boxcar averaging acts as a low pass filter that depends on the sweep speed. By reducing the scope memory depth to it's minimum (7K points) the low pass filtering of the Eres mode will be maximized. This might be good enough. The low pass filtering effect on my Siglent SDS1104X-E (hacked to a SDS1204X-E) is not nearly as good as it is on my Tektronix TDS784D. If that effective low-pass filtering of the Siglent Eres mode is not adequate one would need to put an appropriate steep low pass filter on the input of the scope, characterized it, and then back out the phase error in post-processing of the data.
@@Chris_Grossman Filters will introduce phase shifts. For good result you do need to do an FFT transformation to extract the magnitude and phase of the injected signal as done in network analyzers.
@@sambenyaakov There is no need to do an FFT. The frequency and phase response of the filter can be characterized directly in the frequency domain with the Bode plot function of the Siglent scope and function generator. I would measure at the same set of frequency points used for the loop gain measurement. In thinking about it I would most likely use 3 to 5 pole Chebyshev filter with a cutoff of ~4/5 of the Nyquist frequency. Both the ripple and phase shift of the filter would be accounted for in the in the pass band by the calibration, and data would be unusable above the cutoff frequency. This would very similar to my compensation for the current probe gain and phase shift I did in my impedance measurement video. I should have gone in to a bit more depth with my first response.
Hi Professor. Thanks for the nice presentation. Can you make one video on how to model the average model of primary side regulated, multi output Flyback Converter operating in DCM in LTspice tool.
Thank you professor! This is quite intuitive, and explains the essence of PCM. For Ridley's model which covers current loop as a second order system, I still don't have intuition, like why it's a second order system, why it's accurate up to half Fsw. Just wonder if you can provide intuitions on that including sub-harmonic oscillation. Thank you again!
Happy new year Professor! I'm wondering if you move the 450uF to the IN node if the AC simulation will still give the same results or will break...? (of course the circuit is still the same, but I'm not sure the AC sim can handle that change)
Hi, for proper loopgain evaluation, the AC injection should be mostly forward and not backward . If you move the caps you will get an answer, but a wrong one.
@6.06 sir you mentioned average model is correct for half of the switching frequency...could you please elaborate..is it only for average model or for any model?
This has to do with Nyquist sampling criteria. In a PWM converter the signals are sampled and the converter is controlled at the switching frequency. This is a sampling system.
G1 is the relationship between the output of error amp and the measured current. That is, the current sensor constant. I think this is explained earlier in video.
Hi Professor. Great video as always. This is the model that my mentor Jack Alexander taught me many years ago. It is a very simple model that can also be used for DCM flybacks and is a first order model. Simple but still powerful enough to set the loop. Again, thank you for the video. Your videos are the best!
Hi Robert, I was not aware of the teaching of Jack Alexander or your videos and posts on the subject that I have just googled. Impressive. Judging from the outcome (you) Jack must have been (or still is) an exceptional mentor.
I have been exploring the average modeling subject for many years now (almost too many to count). It has been my conviction that intuitive anderstanding, to which average models belong, are a prerequisite for good engineering.
As someone who spent a lot of time trying to learn how to SPICE model switch mode power supplies about 20 years ago and got very frustrated with it... I thank you sir.
Thank for comment.
Hello, thanks for great explanation. I really always enjoy how you explain how to study stability of systems using LTSpice! Great work, you are making the world better place :)
😊🙏
I was just looking for information regarding this mode. Thanks Professor!!
😊Thanks
Special thanks
Mr professor 🙏
Thanks
Very nice presentation, also seen from a practical side of view. Thank's Professor for your contribution.
It may also be very helpful if you could present more examples, including all relevant converter specifications, those which are necessary for Loop designs.
If possible, for nonisolated and isolated voltage and current mode converters.
Best regards
Udo
Good suggestion. I will try.
Thanks a lot for sharing
Thanks
Hi sam , Nice explanation , I use the same way to set my loops. In case of PCM and D>0.5 I guess this model doesnt predict sub harmonic oscillations ! Thankyou .
Hi Sam , Nice explanation . I use the same way to set my loops but I think this model doesn't predict sub harmonic oscillations in Flyback CCM ! Thankyou .
Hi, Thanks for notes. Average models can not predict subharmonic oscillations
🙏👍🙏
😊🙏
something I realized while simulating this basic concept in Spice, is that at least with the built in SR latch in Spice, the SR latch is not edge controlled. Is this typical for the SR latch inside this converter? It necessitates the need for very narrow pulse train from the clock otherwise.
It does not need to be edge controlled just set rest. Obviosity when the S or R are removed it should keep the state.
This method for loop gain stability checking seems applicable to many typologies so long as the loop gain falls to zero dB well below the Nyquist frequency. Thank you
Indeed. Thanks for comment.
Hi Chris/ Just seen you intersting write up, can the Siglent
SDS1104X-E plot the loopgain of a PWM converter. That it, can it extract the injected signal from the ripple?
@@sambenyaakov I have used it to plot the loop gain of an op-amp circuit by using a test transformer to inject the voltage as you show in this presentation, but I have never used it for a switching converter. It might be possible of one uses the Eres (extended resolution) function on the Siglent. The Eres mode boxcar averaging acts as a low pass filter that depends on the sweep speed. By reducing the scope memory depth to it's minimum (7K points) the low pass filtering of the Eres mode will be maximized. This might be good enough. The low pass filtering effect on my Siglent SDS1104X-E (hacked to a SDS1204X-E) is not nearly as good as it is on my Tektronix TDS784D. If that effective low-pass filtering of the Siglent Eres mode is not adequate one would need to put an appropriate steep low pass filter on the input of the scope, characterized it, and then back out the phase error in post-processing of the data.
@@Chris_Grossman Filters will introduce phase shifts. For good result you do need to do an FFT transformation to extract the magnitude and phase of the injected signal as done in network analyzers.
@@sambenyaakov There is no need to do an FFT. The frequency and phase response of the filter can be characterized directly in the frequency domain with the Bode plot function of the Siglent scope and function generator. I would measure at the same set of frequency points used for the loop gain measurement. In thinking about it I would most likely use 3 to 5 pole Chebyshev filter with a cutoff of ~4/5 of the Nyquist frequency. Both the ripple and phase shift of the filter would be accounted for in the in the pass band by the calibration, and data would be unusable above the cutoff frequency. This would very similar to my compensation for the current probe gain and phase shift I did in my impedance measurement video. I should have gone in to a bit more depth with my first response.
Hi Professor. Thanks for the nice presentation. Can you make one video on how to model the average model of primary side regulated, multi output Flyback Converter operating in DCM in LTspice tool.
Great suggestion! Will try
Thank you professor! This is quite intuitive, and explains the essence of PCM. For Ridley's model which covers current loop as a second order system, I still don't have intuition, like why it's a second order system, why it's accurate up to half Fsw. Just wonder if you can provide intuitions on that including sub-harmonic oscillation. Thank you again!
Thanks for comment. Have you seen ua-cam.com/video/fF-jFFOWSY4/v-deo.html
@@sambenyaakov Yes, watched that before.
Happy new year Professor!
I'm wondering if you move the 450uF to the IN node if the AC simulation will still give the same results or will break...?
(of course the circuit is still the same, but I'm not sure the AC sim can handle that change)
Hi, for proper loopgain evaluation, the AC injection should be mostly forward and not backward . If you move the caps you will get an answer, but a wrong one.
@6.06 sir you mentioned average model is correct for half of the switching frequency...could you please elaborate..is it only for average model or for any model?
This has to do with Nyquist sampling criteria. In a PWM converter the signals are sampled and the converter is controlled at the switching frequency. This is a sampling system.
hello professor.. nice video... how to determine the gain of G1 ? what value we should target for design ?
G1? Which minute in the video are you referring to?
Hello sir .. minute 10.16 in the simulation model
In the model u had a gain G1 of 10. I was wondering how to determine that value. ?
G1 is the relationship between the output of error amp and the measured current. That is, the current sensor constant. I think this is explained earlier in video.
Ok sir thanks