Really interesting topic, hope you do more videos about it. Some practical measurements will be great. Thank you so much for your informative contents.
The fully enclosed choke, Fe powder, has high core losses due to the high current/flux ripple and the fringing of flux near the internal gap heating the wire. For the ferrite open bobbin construction, the wire is smaller which is better for Rac losses at this level of current ripple, while the ferrite core has significantly lower losses. As the freq goes up the ripple goes down, explaining the drop in losses with freq in the Fe-powder core ...
Don't be confused by the 14.0 Amp value that is shown, that particular value is directly from the datasheet and is the rated RMS current that will cause a 40°C temperature rise from a 25°C ambient. They only show it so you can compare between the various inductors they have recommended. It has nothing to do with the simulation per say. If you look at the datasheet for the DO part, you'll find the rated current is the exact 5.7A value shown from the "results" page as well. As for the drastic differences in core losses shown between the two inductors, we need to know more about how the inductors are constructed to make a determination. Just because both inductors have the same inductance value, doesn't mean the cores are experiencing the same conditions. For all we know, the distributed gap core of the XAL inductor may have 1/4th the number of turns as the discrete gapped ferrite to achieve the same inductance value. The lower number of turns leads to higher delta-B in the core for a given voltage and "on time" and thus significantly higher core losses. That's my guess as to the difference. One inductor has more turns than the other, which leads to one core experiencing a higher flux density change and more losses.
Thanks Shawn or sharing your thoughts. May I add: It is not just the number of turns that matters but nAe. Still, there is a big difference in core losses between ferrite and iron powder materials for same deltaB, so it is hard to tell what is the reason for the difference. But it is important to know that same inductance inductors might have a big difference in losses, something that does not show in the datasheet, and that is the point that I tried to make in the video.
Professor, 1) Could the Rs increase with frequency (slide 6) encapsulate the core losses besides the skin & proximity effects? I find it very compelling to say yes because you can see that Rs increases initially with f² and being a resistor of course the power losses in it will be proportionally to I² ∝ H² ∝ B². So the overall losses in Rs above ~30KHz are ∝ f² B² which agrees with the Steinmetz equation. 2) Also I'm sorry to bother you again, but on slide 24, though I totally agree with point 2, regarding point 1 I beg to differ. In my other comment (in the retracted video) I mentioned that in the range 100kHz to 1MHz the equivalent model of the inductor was an inductor L=27uH in series with a resistor with value R=(ωL)²/2500 (i.e. a resistor that scales up with f²). I calculated the above result by hand. But to better share the result with you I used Sapwin this time to analyze the inductor model in slide 16 (despite being a very simple circuit to analyze). After getting the equivalent complex impedance Z, I plotted in Matlab the Real{Z} and Imag{Z}/ω. The idea being to try to get a picture like the one in (slide 6) and understand to what RL series circuit (Rs + Ls) the inductor model corresponds to. The result is here: drive.google.com/file/d/1X5Ov4HFvcsw7J8pe1hFHsfg1x33DZQQO/view?usp=sharing Basically one can see that up to ~50KHz the inductor behaves has a Ls=27uH in series with Rs~45 mOhm (no suprises at low frequencies). From then on up to ~ 2 MHz the inductor behaves has a Ls=27uH in series with a resistor that scales up with f², a behavior exactly as shown in slide 6, reaching values of Rs~100 ohm @ 2 MHz. After 2 MHz the model does not captures well the Ls & Rs behavior shown in slide 6, so I would say the model is not valid anymore. But, given the typical range of frequencies used in PWM converters, in my opinion the very simple model of slide 16 does captures some frequency dependence of losses. Eventually what you're trying to say in point 1 of slide 24 is that the inductor model does not do any better job in modelling the frequency dependence of losses in a real inductor than the lumped circuit of slide 16. PS: I guess my point on the other video was to say that the inductor model was equivalent to the lumped circuit (which you have proved in this video) and that the increase in Rac or power loss with frequency seen there was not due to some hidden parameter or special thing in the model, but it was simply due to the 2500 ohm resistor and that it could be calculated .
Hi, concerning the inclusion of core losses (as I have shown in an early video) , the excitation during testing is extremely small so the core losses are negligible AS for the other points. I will have to defer as I can not spare the time now to plunge into it. Will try to do it at a later time. Sorry. Would you like to prepare a video to outlay your observations?
@@sambenyaakov Hi. I have prepared a video trying to explain my point of view. I don't plan to become a "youtuber", so the video is unlisted and I kindly ask you not to turn the link public. You can find it over here: th-cam.com/video/Gaew76gHkmY/w-d-xo.html Please let me know when you have seen it so that I can delete this comment to prevent others (although unlikely) from finding it. PS: this experience led me to appreciate much more the work you have with each of your videos. So, thank you.
So you really need to test the losses with physical inductors, and either a test circuit with variable PWM frequency, or the actual circuit you intend to use.
Interesting video although it is not surprising that all of the models are simplistic or highly questionable. At 3:05 the Bode plot shows the inductance and winding resistance peaking at the same frequency. Is there a reason for this, or is it a coincidence, or maybe they chose the winding so that the resistance went crazy at the same point as the inductance hit the curie point? Just seems strange that these two seemingly loosely related variables would peak at the same frequency...
Sir, in the derivation @7:08 voltage applied to inductor is taken as sinusoidal. But in the DC-DC converter under consideration, it's a square-wave. Can we still use Steinmetz equation to calculate core-loss in this case? Or does the f^a and B^b captures this non-sinusoidal input ?
Yes, this is a major issue. The information we have is on sinusoidal small signal. There are improved models such as RGSE. See: www.ee.bgu.ac.il/~pel/pdf-files/jour144.pdf
אהלן, רציתי לשאול האם יש לך ערוץ גם בשפה העברית ? יודע שזו רמת אקדמיה ולכן ההרצאה הינה באנגלית אך בכל זאת שואל מאחר ואני עוקב אחרי העלאות המדהימות שלך ולומד המון בעיקר בפענוח ההבנה בגלל קשיי השפה.. תודה ושנה טובה !
It is an honorable thing to admit personal shortcomings and redo the thing!
👍
Really interesting topic, hope you do more videos about it. Some practical measurements will be great.
Thank you so much for your informative contents.
Beyond my ability to spend the time. The conclusion is that one needs to test the inductor in circuit for temp rise.
The fully enclosed choke, Fe powder, has high core losses due to the high current/flux ripple and the fringing of flux near the internal gap heating the wire.
For the ferrite open bobbin construction, the wire is smaller which is better for Rac losses at this level of current ripple, while the ferrite core has significantly lower losses.
As the freq goes up the ripple goes down, explaining the drop in losses with freq in the Fe-powder core ...
Don't be confused by the 14.0 Amp value that is shown, that particular value is directly from the datasheet and is the rated RMS current that will cause a 40°C temperature rise from a 25°C ambient. They only show it so you can compare between the various inductors they have recommended. It has nothing to do with the simulation per say. If you look at the datasheet for the DO part, you'll find the rated current is the exact 5.7A value shown from the "results" page as well.
As for the drastic differences in core losses shown between the two inductors, we need to know more about how the inductors are constructed to make a determination. Just because both inductors have the same inductance value, doesn't mean the cores are experiencing the same conditions. For all we know, the distributed gap core of the XAL inductor may have 1/4th the number of turns as the discrete gapped ferrite to achieve the same inductance value. The lower number of turns leads to higher delta-B in the core for a given voltage and "on time" and thus significantly higher core losses. That's my guess as to the difference. One inductor has more turns than the other, which leads to one core experiencing a higher flux density change and more losses.
Thanks Shawn or sharing your thoughts. May I add: It is not just the number of turns that matters but nAe. Still, there is a big difference in core losses between ferrite and iron powder materials for same deltaB, so it is hard to tell what is the reason for the difference. But it is important to know that same inductance inductors might have a big difference in losses, something that does not show in the datasheet, and that is the point that I tried to make in the video.
Professor,
1) Could the Rs increase with frequency (slide 6) encapsulate the core losses besides the skin & proximity effects?
I find it very compelling to say yes because you can see that Rs increases initially with f² and being a resistor of course the power losses in it will be proportionally to I² ∝ H² ∝ B². So the overall losses in Rs above ~30KHz are ∝ f² B² which agrees with the Steinmetz equation.
2) Also I'm sorry to bother you again, but on slide 24, though I totally agree with point 2, regarding point 1 I beg to differ.
In my other comment (in the retracted video) I mentioned that in the range 100kHz to 1MHz the equivalent model of the inductor was an inductor L=27uH in series with a resistor with value R=(ωL)²/2500 (i.e. a resistor that scales up with f²).
I calculated the above result by hand. But to better share the result with you I used Sapwin this time to analyze the inductor model in slide 16 (despite being a very simple circuit to analyze).
After getting the equivalent complex impedance Z, I plotted in Matlab the Real{Z} and Imag{Z}/ω. The idea being to try to get a picture like the one in (slide 6) and understand to what RL series circuit (Rs + Ls) the inductor model corresponds to.
The result is here:
drive.google.com/file/d/1X5Ov4HFvcsw7J8pe1hFHsfg1x33DZQQO/view?usp=sharing
Basically one can see that up to ~50KHz the inductor behaves has a Ls=27uH in series with Rs~45 mOhm (no suprises at low frequencies). From then on up to ~ 2 MHz the inductor behaves has a Ls=27uH in series with a resistor that scales up with f², a behavior exactly as shown in slide 6, reaching values of Rs~100 ohm @ 2 MHz.
After 2 MHz the model does not captures well the Ls & Rs behavior shown in slide 6, so I would say the model is not valid anymore.
But, given the typical range of frequencies used in PWM converters, in my opinion the very simple model of slide 16 does captures some frequency dependence of losses.
Eventually what you're trying to say in point 1 of slide 24 is that the inductor model does not do any better job in modelling the frequency dependence of losses in a real inductor than the lumped circuit of slide 16.
PS: I guess my point on the other video was to say that the inductor model was equivalent to the lumped circuit (which you have proved in this video) and that the increase in Rac or power loss with frequency seen there was not due to some hidden parameter or special thing in the model, but it was simply due to the 2500 ohm resistor and that it could be calculated .
Hi, concerning the inclusion of core losses (as I have shown in an early video) , the excitation during testing is extremely small so the core losses are negligible AS for the other points. I will have to defer as I can not spare the time now to plunge into it. Will try to do it at a later time. Sorry. Would you like to prepare a video to outlay your observations?
@@sambenyaakov
Hi. I have prepared a video trying to explain my point of view.
I don't plan to become a "youtuber", so the video is unlisted and I kindly ask you not to turn the link public. You can find it over here:
th-cam.com/video/Gaew76gHkmY/w-d-xo.html
Please let me know when you have seen it so that I can delete this comment to prevent others (although unlikely) from finding it.
PS: this experience led me to appreciate much more the work you have with each of your videos. So, thank you.
@@justpaulo Please hold on a coupled of days or so until I see the video> Thanks for sharing.
So you really need to test the losses with physical inductors, and either a test circuit with variable PWM frequency, or the actual circuit you intend to use.
I guess at this stage there is no escape from testing the inductor yourself.
Interesting video although it is not surprising that all of the models are simplistic or highly questionable.
At 3:05 the Bode plot shows the inductance and winding resistance peaking at the same frequency. Is there a reason for this, or is it a coincidence, or maybe they chose the winding so that the resistance went crazy at the same point as the inductance hit the curie point? Just seems strange that these two seemingly loosely related variables would peak at the same frequency...
The plot assumes only L and R while when approaching the resonance the capacitance is significant, so the the L and R approximation is incorrect.
Sir, in the derivation @7:08 voltage applied to inductor is taken as sinusoidal. But in the DC-DC converter under consideration, it's a square-wave. Can we still use Steinmetz equation to calculate core-loss in this case? Or does the f^a and B^b captures this non-sinusoidal input ?
Yes, this is a major issue. The information we have is on sinusoidal small signal. There are improved models such as RGSE. See: www.ee.bgu.ac.il/~pel/pdf-files/jour144.pdf
Thank you Sir 🙏🏽 for sharing both video and the paper.. They are wonderful and need of the hour.
@@roja_p 👍
It seams that X inductor has more eddy current losses, and D is has less
Possibly, but what bothers me is that the 14A pops up and the simulated losses seem to be related to it.
אהלן, רציתי לשאול האם יש לך ערוץ גם בשפה העברית ? יודע שזו רמת אקדמיה ולכן ההרצאה הינה באנגלית אך בכל זאת שואל מאחר ואני עוקב אחרי העלאות המדהימות שלך ולומד המון בעיקר בפענוח ההבנה בגלל קשיי השפה..
תודה ושנה טובה !
הי , אין לי ערוץ בעברית אבל ישנן הקלטות נגישות של קורסים בעברית. ידוע לך?
@@sambenyaakov היי תודה על התייחסות, לצערי לא מכיר תוכן בעיברית נגיש
@@omridavidi8885 קישורים: www.ee.bgu.ac.il/~angcirc/
www.ee.bgu.ac.il/~dcdc/
כדי לראות הקלטות יש לפתוח באקספלורר לא כרום
@@sambenyaakov תודה רבה!
👍👍👍💖🙏💖👍👍👍
😊🙏👍
😀😉😐
🤔