Thanks for covering this topic! It is indeed very important to understand the HF paths and source of oscillations in such a system as this is related to EMC, where the HF content can potentially radiate if the system is prone to it. Also, the lower f osc can introduce uncertainty in current measurements sampling. Two points from my side for better matching of real behavior: 1. series stray inductance in the HF loop consisting of C2 and C3, thus putting some Lx in between the two (similar effect to L3 is expected) 2. If the switches (low side diode) in the HB are Si, I would consider diode reverse recovery effect and the phase node voltage oscillations as its consequence. The resonant f would depend on the loop L-C of the HB and the diode’s snappiness. Thanks again
Great explanation of the differences between the waveforms that are presented in textbooks and those that we actually see in the lab! LT Spice is a great tool for evaluating changes in component values too!
You just made my week a lot easier. The explanation is clear and you used LTSpice so I can duplicate the experiment and try my own values. The discussion about the time constant was especially useful. Thank you again and I look forward to more videos
Very interesting and useful. I've seen this before in a full-bridge WPT converter due to stray capacitance of the coupling coils. Thanks and Shana Tova!
Excellent video, Professor Ben-Yaakov my beloved. It may be worth noting that in practice, although output capacitor parasitics decrease the "noise" on the shunt voltage, they increase the common mode component which may increase the "noise" of the system as op-amps have poor high frequency CMRR. However, examining the true voltage on the shunt as you did is illuminating on its own, and ignoring the measurement considerations better highlights the effects of parasitics in the power loop. Thanks.
Wow, this is a videolesson that comes on the perfect moment for me. I have to slow down a feedbackloop because of the unsteady "jumpy" waveform of the measured signal going into the opamp, but when it is finally steady enough the whole loop is so slow that it becomes an oscillator when loaded with the load for wich i am building this. Of course with a much steadier testload it works always fine enough! (Btw, i needed to start up my old phone to leave this comment because both my new ones give an error when trying to type a comment in youtube!(thus, i still enjoy each&every new videolesson!) Oh btw, since i can shortly type a comment again, i can finally ask what i want to ask for since a long time: would you use a bigger mousepointer? (for people like me who learn from your videolessons from the phonescreen? Mostly it is clear enough to follow because the explainings mostly do not need the pointer, but these little moving "flashes", or so it seems on a little phonescreen, they often underline your words as èxtra clearmaking help, or so i think?) Also a big Thank-You, again, for giving me, us, such a great chance to learn from you with these up-level-but-perfect-to-follow videolessons!
Dear Rob. I missed you😊. A larger pointer is indeed a good suggestion. As it is, the pointer is large but a lager one will certainty help to follow it better. Or I will try a different pointer. I believe my recording program has a "laser" pointer. I will give it a try too.
Thank you sir. I encountered the same problem with a boost converter. Till now I was much confident that a capacitor of low esr ,something like a ceramic disc capacitor, will fix this problem. But from your explanation a capacitor across sense resistor will boost existing problem. In my system current sense circuit senses overcurrent at 25% load. Adding an RC filter to the voltage across sense resistor be a good idea? Thank you 🙏
Normally engineers remove transients by adding filters, this is the first time I see anyone adding components to induce transients and simulate real physical effects 😂. I would have easily brushed off the ringing as Gibbs phenomena due to harmonics of Fourier series coefficients or signal reflection since some resistors have wound coils. But the signal applied was low frequency, professor you have done it again. I have questioned myself and got lost as usual.
Hello, could you explain from where you took value of parasitic capacitance for inductance, or it was from experience? Also what type of capactior you would assume to achieve ESL equal 1nH, package, more caps in paraller, type of capacitor? thanks for answer.
@@sambenyaakov Not sure what that means. But if the PWM signal is generated by a microcontroller, then it could be coded to not read current around that time. I'm sure there is a way to do this with analog electronics also.
Ben, Put in the resistive element representing core loss in parallel with L2. That is most likely the cause of the step. You can manipulate a model to show a voltage-divider effect of two inductors, but that does not mean that this resulting measurement error is the main cause of the step.
Thank you for comments. To begin with, I am not sure that a parallel resistor that emulate core losses is relevant to a step change. Do you have any knowledge on that? Irrespective of this question, you surely must agree that the sum of a triangular waveform (the voltage on the pure resistance) and a square waveform (the voltage of the parasitic inductance) produce a waveform with two steps.
@@sambenyaakov At my first job out of college 40 years ago, I designed a BH loop tester after hearing contradicting and simply incorrect ideas about magnetic effects from other engineers at work, and I have subsequently built several more over the years at different jobs to model effects such as inductance, core loss, saturation, etc. A square wave of voltage on a core always results in an approximate square wave plus triangle wave. It is not measurement error. Core loss is real. The width of the BH loop correlates with the loss as predicted by the Steinmetz equations very well as measured by the height of the square wave portion of the current measurement. These current measurements on a ferrite track the (Frequency)^Kf pretty well. Kf is typically in the range of about 1.3 to 1.5 for the typical range of frequencies that we use in power supplies. Remember that, for a given flux, the frequency is proportional to the applied voltage and thus, the rate of change of flux. On the oscilloscope you can see the width of the BH loop get wider as this flux-rate-of-change get higher. Furthermore, for each unique voltage (and thus, rate of change) the resultant square wave correlates to a unique fixed resistor value to model core loss. (It is slightly non-linear but, for practical modeling purposes, very good). The resultant current*voltage is power loss as we all know. The equation: H=N*I/L is very useful to generate a set of curve data for each core.
@@darrellhambley7245 I am not clear about "A square wave of voltage on a core always results in an approximate square wave plus triangle wave" square wave of current? Do you have any reference to a publications on that. This notion seem to be inconsistent with the more advanced core loss descriptions such as GSE (see www.ee.bgu.ac.il/~pel/pdf-files/jour144.pdf ).
@@sambenyaakov Reference to a "publication"? No. I'm describing actual lab data and correlating it to the shape of a B-H loop. Simply look at the width of a B-H loop, the "H" term. Flip the equation "H=N*I/lc" to get I, the current, and there it is!
Thanks for covering this topic! It is indeed very important to understand the HF paths and source of oscillations in such a system as this is related to EMC, where the HF content can potentially radiate if the system is prone to it. Also, the lower f osc can introduce uncertainty in current measurements sampling.
Two points from my side for better matching of real behavior:
1. series stray inductance in the HF loop consisting of C2 and C3, thus putting some Lx in between the two (similar effect to L3 is expected)
2. If the switches (low side diode) in the HB are Si, I would consider diode reverse recovery effect and the phase node voltage oscillations as its consequence. The resonant f would depend on the loop L-C of the HB and the diode’s snappiness.
Thanks again
Thanks for comment.
Great explanation of the differences between the waveforms that are presented in textbooks and those that we actually see in the lab! LT Spice is a great tool for evaluating changes in component values too!
Thanks for comment
You just made my week a lot easier. The explanation is clear and you used LTSpice so I can duplicate the experiment and try my own values. The discussion about the time constant was especially useful. Thank you again and I look forward to more videos
Thanks for comment.
Very interesting and useful. I've seen this before in a full-bridge WPT converter due to stray capacitance of the coupling coils.
Thanks and Shana Tova!
Thanks for comment and sharing your experience. Shana Tova and good health.
Great lecture as always prof, happy belated Shana Tova. :)
Thanks. Never too late for greetings🙂wishing you and yours Shana Tova and good health.
Excellent video, Professor Ben-Yaakov my beloved. It may be worth noting that in practice, although output capacitor parasitics decrease the "noise" on the shunt voltage, they increase the common mode component which may increase the "noise" of the system as op-amps have poor high frequency CMRR. However, examining the true voltage on the shunt as you did is illuminating on its own, and ignoring the measurement considerations better highlights the effects of parasitics in the power loop. Thanks.
Thanks for comment
The Long Presentation! Haha good one beavis!
Great explanation, thanks very much
😊🙏
Wow, this is a videolesson that comes on the perfect moment for me. I have to slow down a feedbackloop because of the unsteady "jumpy" waveform of the measured signal going into the opamp, but when it is finally steady enough the whole loop is so slow that it becomes an oscillator when loaded with the load for wich i am building this. Of course with a much steadier testload it works always fine enough!
(Btw, i needed to start up my old phone to leave this comment because both my new ones give an error when trying to type a comment in youtube!(thus, i still enjoy each&every new videolesson!) Oh btw, since i can shortly type a comment again, i can finally ask what i want to ask for since a long time: would you use a bigger mousepointer? (for people like me who learn from your videolessons from the phonescreen? Mostly it is clear enough to follow because the explainings mostly do not need the pointer, but these little moving "flashes", or so it seems on a little phonescreen, they often underline your words as èxtra clearmaking help, or so i think?)
Also a big Thank-You, again, for giving me, us, such a great chance to learn from you with these up-level-but-perfect-to-follow videolessons!
Dear Rob. I missed you😊. A larger pointer is indeed a good suggestion. As it is, the pointer is large but a lager one will certainty help to follow it better. Or I will try a different pointer. I believe my recording program has a "laser" pointer. I will give it a try too.
Helpful video
Thanks
Thank you sir. I encountered the same problem with a boost converter. Till now I was much confident that a capacitor of low esr ,something like a ceramic disc capacitor, will fix this problem. But from your explanation a capacitor across sense resistor will boost existing problem. In my system current sense circuit senses overcurrent at 25% load. Adding an RC filter to the voltage across sense resistor be a good idea? Thank you 🙏
Thanks for sharing your experience.
Perfect ! 😁😁
Normally engineers remove transients by adding filters, this is the first time I see anyone adding components to induce transients and simulate real physical effects 😂. I would have easily brushed off the ringing as Gibbs phenomena due to harmonics of Fourier series coefficients or signal reflection since some resistors have wound coils. But the signal applied was low frequency, professor you have done it again. I have questioned myself and got lost as usual.
🙂 Thanks.
Hello, could you explain from where you took value of parasitic capacitance for inductance, or it was from experience? Also what type of capactior you would assume to achieve ESL equal 1nH, package, more caps in paraller, type of capacitor? thanks for answer.
Watch the video. The cap value from resonance
nice
Thanks
Thanks Sir. It is an wonderful video.i am facing the same issue. Pls let me know on how to remove this spike.
If you cannot replace the inductor with one of a lower capacitance, put a filter at the output, preferably by using a a ferrite bead and a capacitor.
SO IF I HAVE CIRCUIT THAT MEASURES THE OUTPUT CURRENT OF DC/DC IN THE SAME TOPOLOGY, WHAT DO I NEED TO DO TO PREVENT THESE SPIKES?
To add a filter
Would not reading the current sensor at the times when the PWM signal is switching on/off be another way of filtering out the spikes and noise ?
You meant at the peal value?
@@sambenyaakov Not sure what that means.
But if the PWM signal is generated by a microcontroller, then it could be coded to not read current around that time.
I'm sure there is a way to do this with analog electronics also.
@@station240 Which time ? You mean at the peak value? but the value does not reflect the correct current!
👍👍💖🙏💖👍👍
😊🙏
Ben, Put in the resistive element representing core loss in parallel with L2. That is most likely the cause of the step. You can manipulate a model to show a voltage-divider effect of two inductors, but that does not mean that this resulting measurement error is the main cause of the step.
Thank you for comments. To begin with, I am not sure that a parallel resistor that emulate core losses is relevant to a step change. Do you have any knowledge on that? Irrespective of this question, you surely must agree that the sum of a triangular waveform (the voltage on the pure resistance) and a square waveform (the voltage of the parasitic inductance) produce a waveform with two steps.
See also ua-cam.com/video/6NpobetG3t8/v-deo.html
@@sambenyaakov At my first job out of college 40 years ago, I designed a BH loop tester after hearing contradicting and simply incorrect ideas about magnetic effects from other engineers at work, and I have subsequently built several more over the years at different jobs to model effects such as inductance, core loss, saturation, etc. A square wave of voltage on a core always results in an approximate square wave plus triangle wave. It is not measurement error. Core loss is real. The width of the BH loop correlates with the loss as predicted by the Steinmetz equations very well as measured by the height of the square wave portion of the current measurement. These current measurements on a ferrite track the (Frequency)^Kf pretty well. Kf is typically in the range of about 1.3 to 1.5 for the typical range of frequencies that we use in power supplies. Remember that, for a given flux, the frequency is proportional to the applied voltage and thus, the rate of change of flux. On the oscilloscope you can see the width of the BH loop get wider as this flux-rate-of-change get higher. Furthermore, for each unique voltage (and thus, rate of change) the resultant square wave correlates to a unique fixed resistor value to model core loss. (It is slightly non-linear but, for practical modeling purposes, very good). The resultant current*voltage is power loss as we all know. The equation: H=N*I/L is very useful to generate a set of curve data for each core.
@@darrellhambley7245 I am not clear about "A square wave of voltage on a core always results in an approximate square wave plus triangle wave" square wave of current? Do you have any reference to a publications on that. This notion seem to be inconsistent with the more advanced core loss descriptions such as GSE (see www.ee.bgu.ac.il/~pel/pdf-files/jour144.pdf ).
@@sambenyaakov Reference to a "publication"? No. I'm describing actual lab data and correlating it to the shape of a B-H loop. Simply look at the width of a B-H loop, the "H" term. Flip the equation "H=N*I/lc" to get I, the current, and there it is!