One detail about the EMI of a PWM oscillator. The base frequency and harmonics are only fully predictable at a duty cycle of 50%. Any other duty cycle will contain two (variable) base frequencies, and resulting harmonics. So the noise spectrum of PWM is not as easily deterministic as indicated.
Hello. This video is great. Could you do more videos on how to design such SMPS controllers? I am interested in designing a COT/PFM controller for Boost converters but any lessons are welcome. I am interested in choosing the on time and minimum off time and how the monostables that produce them are connected together.
It would be great to have a video or series about EMI/EMC and how to simulate in LT SPICE for the different standards (MIL-STD-461, DEF STAN 59/411, MIL-STD-704, MIL-STD-1275, MIL-STD-1399, FCC Part 15 class A/B, EN 61000-6-1/2/3/4)... Especially input/output filters!
Hi! First of all: Awesome channel! Thank you for all your educational work! Could you please make a video about popular communication buses like USB, HDMI, Display port, etc.? Esp. the DP is very poorly documented on the internet. Again, thanks for all your excellent work!
Interesting. For ham radio applications, I've used switching regulators where you can supply an external clock! They way, I know accurately where my harmonics will be and can push them out of the receiver bands of interest. 😜
Clock inputs are also used to drive more than one supply in phase opposition - in effect making it work like a multiphase supply. This works really well when the supplies are physically close. By applying synchronization you can reduce the noise, or be able to use smaller capacitors.
8:22 WHOA... "PFM is more efficient [in switching-losses]." (Sorry, maybe I'm nitpicking, as the context is explained better, later... Otherwise, highly-informative vid, thank you!) That statement is very much reliant on the duration of the PFM pulses... which depends very much on the load, as well as the design. E.G. In comparing a PFM-Only supply with a PWM-Only supply, if 50% duty-cycle is necessary for the output, and, say, the PFM-update-clock is eight-times the other supply's PWM-frequency, then the PFM-Only supply would have four times as many transitions. If the duty-cycle was 25%, it'd have twice as many transitions. I see at 9:30 your explanation, here at 8:30, is based on a specific chip that chooses PFM for low duty-cycles, and PWM for high. That makes sense, based on your observations, as well as mine. This chip optimises each usage. But, in general, without careful consideration of update-frequency, typical loading, etc. PFM is probably not as efficient in terms of switching-losses. (Interesting unrelated-ish hypothesis about these devices that choose PFM or PWM... A PWM-Only controller might inherently *appear* to be PFM at low duty-cycles if, e.g. the PWM duty-cycle has a specific step-size. e.g. if the PWM duty-cycle can only be controlled in 1% steps, then if 0.5% duty-cycle is needed, it'd just output 0% for one cycle, and 1% for the next. Wherein, it's quite plausible the same PWM control circuitry just *inherently* appears to be PFM when run at very low duty-cycles... No special PFM circuitry needed. But, that's just a theory.)
You are right that if the frequency for the PFM is higher then of course its less efficient; the point that I was trying to make was that if the power stage is the same, especially at light load, PFM can be made to be more efficient, but this is usecase dependent to some extent; Its easier to make a PFM converter to be efficient over a wide load range than it is to make a PWM converter; at relatively high loads, the switching losses are no longer the main contributor so the exact modulation scheme is not as important as with light loads. The other point that you made about the PWM controller, sometimes is called Pulse-skipping mode; where the controller skips a beat, but all the transitions are still synchronized to the base oscillator.
Excellent video as always. 🙂 As an amateur radio operator, i would LOVE to learn general techniques to to minimize switching noise in power conversion circuits. I live in a remote area with a -120db noise floor at HF frequencies, and every time i get a new device, I have to dispose of the crappy, unshielded, usually hideously noisy "wall wart". Manufacturers for consumer goods just don' seem to care about EMI. ☹
This was highly educational. A related cross-over is Pulse Density Modulation where the switching and thus the power density of the output is completely determined by the input signal. PDM is often used by MEMS microphones and feels like a cross between PWM and PFM with the switching circuitry being controlled by the input signal rather than a feedback signal. This got me to wondering what it would take to make a video of this sort could be made for a PDM source.
@@ericwazhung I suppose. The PDM circuit descriptions I've looked at have an internal feedback process that seems to be half PWM and half PFM. A bit weird but they work.
Please include all about spread spectrum pwm ,current controlled frequency foldback , & H.E.pwm in some other presentation/s and if possible something on two/three level pwm techniques , centre aligned dual edge etc .
not sure why you say that PFM can be used only for bucks?? I'm using PFM all the time, including for boost/flyback and even with valley switching or peak current limiting, smoothly transitioning between CCM and DCM. PFM makes all of this super simple, as you only ever have a single summing node on the hysteretic switch, into which you can integrate all kinds of functions one could want. But maybe i am misunderstanding something.
I was trying to say that the implementation where you have a single trigger-Schmidt comparator (and nothing else) will only work like a buck; once you add a delay timer and other digital logic, you can build any topology - including boost/flyback as you mentioned.
@@FesZElectronics I don't see why the flyback/boost would be any different from the buck. You take feedback from the output and sum it back into the comparator. No other parts needed.
With the buck, to get a higher output voltage, you need to keep the switch on for loner periods; once the target is reached, the switch is turned off; with a boost, to get an increase in voltage, you need to turn the switch off - keeping it on will keep the output at 0; without a mechanism to force the switch to reset from time to time, the boost will not work - this forced reset is not mandatory in the buck; there the reset occurs naturally.
@@FesZElectronics Ahh I see. Yeah but this can be resolved super easily by taking feedback from the FET gate, as well, to form a relaxation oacillator that naturally keeps going and where you only tune the duty cycle using the other feedbacks.
One detail about the EMI of a PWM oscillator. The base frequency and harmonics are only fully predictable at a duty cycle of 50%. Any other duty cycle will contain two (variable) base frequencies, and resulting harmonics. So the noise spectrum of PWM is not as easily deterministic as indicated.
Very good!
Great ...more videos on power supply
I will try to sprinkle some in from time to time
Hello. This video is great. Could you do more videos on how to design such SMPS controllers? I am interested in designing a COT/PFM controller for Boost converters but any lessons are welcome. I am interested in choosing the on time and minimum off time and how the monostables that produce them are connected together.
It would be great to have a video or series about EMI/EMC and how to simulate in LT SPICE for the different standards (MIL-STD-461, DEF STAN 59/411, MIL-STD-704, MIL-STD-1275, MIL-STD-1399, FCC Part 15 class A/B, EN 61000-6-1/2/3/4)... Especially input/output filters!
This is an excellent topics
Hi! First of all: Awesome channel! Thank you for all your educational work!
Could you please make a video about popular communication buses like USB, HDMI, Display port, etc.? Esp. the DP is very poorly documented on the internet.
Again, thanks for all your excellent work!
In Google Or On TH-cam, Do A Search With "DISPLAY PORT THEORY", You'll Be Amazed!
Interesting. For ham radio applications, I've used switching regulators where you can supply an external clock! They way, I know accurately where my harmonics will be and can push them out of the receiver bands of interest. 😜
Clock inputs are also used to drive more than one supply in phase opposition - in effect making it work like a multiphase supply. This works really well when the supplies are physically close. By applying synchronization you can reduce the noise, or be able to use smaller capacitors.
8:22 WHOA... "PFM is more efficient [in switching-losses]."
(Sorry, maybe I'm nitpicking, as the context is explained better, later... Otherwise, highly-informative vid, thank you!)
That statement is very much reliant on the duration of the PFM pulses... which depends very much on the load, as well as the design.
E.G. In comparing a PFM-Only supply with a PWM-Only supply, if 50% duty-cycle is necessary for the output, and, say, the PFM-update-clock is eight-times the other supply's PWM-frequency, then the PFM-Only supply would have four times as many transitions. If the duty-cycle was 25%, it'd have twice as many transitions.
I see at 9:30 your explanation, here at 8:30, is based on a specific chip that chooses PFM for low duty-cycles, and PWM for high. That makes sense, based on your observations, as well as mine. This chip optimises each usage. But, in general, without careful consideration of update-frequency, typical loading, etc. PFM is probably not as efficient in terms of switching-losses.
(Interesting unrelated-ish hypothesis about these devices that choose PFM or PWM... A PWM-Only controller might inherently *appear* to be PFM at low duty-cycles if, e.g. the PWM duty-cycle has a specific step-size. e.g. if the PWM duty-cycle can only be controlled in 1% steps, then if 0.5% duty-cycle is needed, it'd just output 0% for one cycle, and 1% for the next. Wherein, it's quite plausible the same PWM control circuitry just *inherently* appears to be PFM when run at very low duty-cycles... No special PFM circuitry needed. But, that's just a theory.)
You are right that if the frequency for the PFM is higher then of course its less efficient; the point that I was trying to make was that if the power stage is the same, especially at light load, PFM can be made to be more efficient, but this is usecase dependent to some extent; Its easier to make a PFM converter to be efficient over a wide load range than it is to make a PWM converter; at relatively high loads, the switching losses are no longer the main contributor so the exact modulation scheme is not as important as with light loads.
The other point that you made about the PWM controller, sometimes is called Pulse-skipping mode; where the controller skips a beat, but all the transitions are still synchronized to the base oscillator.
Excellent video as always. 🙂
As an amateur radio operator, i would LOVE to learn general techniques to to minimize switching noise in power conversion circuits.
I live in a remote area with a -120db noise floor at HF frequencies, and every time i get a new device, I have to dispose of the crappy, unshielded, usually hideously noisy "wall wart". Manufacturers for consumer goods just don' seem to care about EMI. ☹
This was highly educational. A related cross-over is Pulse Density Modulation where the switching and thus the power density of the output is completely determined by the input signal. PDM is often used by MEMS microphones and feels like a cross between PWM and PFM with the switching circuitry being controlled by the input signal rather than a feedback signal. This got me to wondering what it would take to make a video of this sort could be made for a PDM source.
As I understand, PFM is just a special case of PDM, where no pulses can be longer than one cycle (thus no duty-cycle greater than 50%).(?)
@@ericwazhung I suppose. The PDM circuit descriptions I've looked at have an internal feedback process that seems to be half PWM and half PFM. A bit weird but they work.
Thanks, FesZ 👍
Thanks, I always wondered what spread spectrum feature means:)
Please include all about spread spectrum pwm ,current controlled frequency foldback , & H.E.pwm in some other presentation/s and if possible something on two/three level pwm techniques , centre aligned dual edge etc .
not sure why you say that PFM can be used only for bucks?? I'm using PFM all the time, including for boost/flyback and even with valley switching or peak current limiting, smoothly transitioning between CCM and DCM. PFM makes all of this super simple, as you only ever have a single summing node on the hysteretic switch, into which you can integrate all kinds of functions one could want. But maybe i am misunderstanding something.
I was trying to say that the implementation where you have a single trigger-Schmidt comparator (and nothing else) will only work like a buck; once you add a delay timer and other digital logic, you can build any topology - including boost/flyback as you mentioned.
@@FesZElectronics I don't see why the flyback/boost would be any different from the buck. You take feedback from the output and sum it back into the comparator. No other parts needed.
With the buck, to get a higher output voltage, you need to keep the switch on for loner periods; once the target is reached, the switch is turned off; with a boost, to get an increase in voltage, you need to turn the switch off - keeping it on will keep the output at 0; without a mechanism to force the switch to reset from time to time, the boost will not work - this forced reset is not mandatory in the buck; there the reset occurs naturally.
@@FesZElectronics Ahh I see. Yeah but this can be resolved super easily by taking feedback from the FET gate, as well, to form a relaxation oacillator that naturally keeps going and where you only tune the duty cycle using the other feedbacks.
Hello, have you considered to use Qspice?
I did a review of it a while back - th-cam.com/video/snXUpFJpXGA/w-d-xo.html ; its not a bad simulator, but for the time I will stick to LTspice