Power Regulation Strategy

[NiGHTS]

It depends on many factors. Although the frequency of the switching cycle may be only 100 kHz, there are some very rapid voltage and current changes which have large frequency components well into the tens or even hundreds of MHz. These are the one that cause the problems when inductance and capacitance are present, and when current loops are not properly controlled and kept as compact as possible. That's all I know about it.

I can't visualize where in my circuit there would exist frequencies that high running through the breadboard. Do you mean slight fluctuations in power line noise and such originating from inside the PWM chip?

Yes, as long as the diode is fast enough - something which is also true of Schottky diodes.

If you don't mind me asking, how does switching speed play a part in heat efficiency? Is there a calculation I can use incorporating switching speed to better understand how it translates to wasted heat energy? Is it anything like Complimentary MOSFETs where during the switching of one MOSFET the other will be in a shorted state for nanoseconds before it is turned on and thus generate heat?

"Voltage through" isn't meaningful. Current flows through; voltage appears across. Voltage across a diode doesn't cause it to dissipate heat, apart from rapid changes which can produce current flow through the diode's capacitance, but that's not much of a factor. Current through a diode, in the forward direction, multiplied by the diode's forward voltage, is what causes most of the power dissipation.

It's the forward voltage multiplied by the mean inductor current multiplied by the duty cycle of the diode, which is 100% minus the duty cycle of the MOSFET. You can estimate the MOSFET duty cycle under load as V_OUT / V_IN × 100%.

Ah, of course. I see. Thank you for that clarification.

You can't "configure" a MOSFET as a diode. If you're talking about the body diode that appears between the source and drain, this diode is just parasitic and doesn't have any impressive characteristics at all. When a MOSFET is used as a synchronous rectifier, its gate is driven by the controller IC and it is not used as a diode; it is just used to provide the path to 0V that the diode would otherwise provide. It's a lot more efficient than a diode because a MOSFET with gate bias has no equivalent of forward voltage in its drain-source path and its ON-resistance (R_DS(on)) can be very low.

Would this configuration work? Note the MOSFET in place of where the diode would normally be.

That's OK. "Input should be on the left" is the main criticism. Personally I like to show the high-current ground path explicitly in my schematics. Here's an example from a product I designed that shows what I mean. Ignore the fact that the power input is on the right! There were reasons for this; I'm not sure whether I still think they justify drawing it backwards or not!
View attachment 18922

I like this strategy a lot. I've been using a freehand paint program to draw the schematics presented in this thread based on an original screenshot of my Eagle CAD project. I'll check to see if Eagle has a feature to visually indicate high amp traces on the schematic sheets for my production version.

 

[KrisBlueNZ]

I can't visualize where in my circuit there would exist frequencies that high running through the breadboard. Do you mean slight fluctuations in power line noise and such originating from inside the PWM chip?

No. Any fast edge has higher-frequency components. The steeper the edge, the more high-frequency components it contains. Do you know about Fourier analysis? You know that a sinewave contains only the fundamental frequency, but a square wave contains the fundamental plus a series of harmonics extending upwards in frequency, getting progressively smaller, theoretically forever? That's because it has steep edges. That's the kind of signal you're dealing with in a switching regulator.

If you don't mind me asking, how does switching speed play a part in heat efficiency? Is there a calculation I can use incorporating switching speed to better understand how it translates to wasted heat energy? Is it anything like Complimentary MOSFETs where during the switching of one MOSFET the other will be in a shorted state for nanoseconds before it is turned on and thus generate heat?

Kind of, in that it's current that flows when it shouldn't. It's current that flows through the diode as the diode switches between conducting and non-conducting. I don't know any more than that. Someone like Steve or Harald or Hop could explain it. Do you want me to ask them?

Would this configuration work? Note the MOSFET in place of where the diode would normally be.

Not just like that, because a MOSFET starts to conduct when its gate-source voltage is around 3~4V typically. This means that as the controller chip's output slews from (V_IN - 4V) down to 4V, and back in the other direction, both MOSFETs will conduct at the same time. The synchronous rectifier MOSFET needs to be controlled independently, using a controller chip that's designed for that application. Please feel free to look for synchronous buck converter controller ICs that can operate at the voltages you want. There may well be something suitable. I always use Digi-Key's selection tables because they have a huge selection, including, mostly, components I never would have heard of otherwise, and their selection filters allow you to eliminate a good proportion of the unsuitable devices. Then, when you have to go one-by-one through the data sheets, they're all linked from the selection table. It makes things very easy.

I like this strategy a lot. I've been using a freehand paint program to draw the schematics presented in this thread based on an original screenshot of my Eagle CAD project. I'll check to see if Eagle has a feature to visually indicate high amp traces on the schematic sheets for my production version.

The main point isn't the thick lines for high-current traces, although I agree it's a good idea, but to show how the high-current fast switching path should be contained and connected to the 0V rail at only one point.

 

[NiGHTS]

No. Any fast edge has higher-frequency components. The steeper the edge, the more high-frequency components it contains. Do you know about Fourier analysis? You know that a sinewave contains only the fundamental frequency, but a square wave contains the fundamental plus a series of harmonics extending upwards in frequency, getting progressively smaller, theoretically forever? That's because it has steep edges. That's the kind of signal you're dealing with in a switching regulator.

Actually I am familiar with this concept, but I didn't think to associate that to parasitic problems with a circuit like this one assembled on a breadboard. Since you are taking about such an infinitesimally small scale of frequencies likely manifesting as a very weak influence operating on a project board with such a small inherent capacitance & inductance, I still don't see a clear demarcation point where I should be worried about this for this application. Don't get me wrong, I am completely in agreement with what you are saying, though I wonder how much of a practical threat this is to my experiments for this particular application.

Kind of, in that it's current that flows when it shouldn't. It's current that flows through the diode as the diode switches between conducting and non-conducting. I don't know any more than that. Someone like Steve or Harald or Hop could explain it. Do you want me to ask them?

Ah, this jogged my memory on the physics of transistors. I think I know how this works now. So switching speed in a diode is a measure of how long it hovers around the "gray zone" of between the definite on and off states, and I'd assume that gray zone operates much like a resistor. Either way the physics of the diode is not incredibly important for me to know right now. I was just curious.

Not just like that, because a MOSFET starts to conduct when its gate-source voltage is around 3~4V typically. This means that as the controller chip's output slews from (V_IN - 4V) down to 4V, and back in the other direction, both MOSFETs will conduct at the same time. The synchronous rectifier MOSFET needs to be controlled independently, using a controller chip that's designed for that application. Please feel free to look for synchronous buck converter controller ICs that can operate at the voltages you want. There may well be something suitable. I always use Digi-Key's selection tables because they have a huge selection, including, mostly, components I never would have heard of otherwise, and their selection filters allow you to eliminate a good proportion of the unsuitable devices. Then, when you have to go one-by-one through the data sheets, they're all linked from the selection table. It makes things very easy.

Aha, I see. Well that makes perfect sense to me. I normally use Mouser for my searches but I'll try Digikey as well. I will also ensure that I carefully select a part which I feel is a good match for my project. Obviously my main priority is heat so efficiency is critical, but operating 12V with 12V is just a perk, not a requirement. So I just asked these questions to get a feel for where this limitation was occurring.

The main point isn't the thick lines for high-current traces, although I agree it's a good idea, but to show how the high-current fast switching path should be contained and connected to the 0V rail at only one point.

Say the PCB has a solder pool connected to ground. Would it still be wise to route this noisy ground line all the way to near where the I/O ground pin is wired into the board or is not a major concern for a solder pool?

And here is a new question. When I chose the coil for my switching regulator I did so because a related datasheet described the need for a 40uH inductor but I only had 50uH available. How does higher impedance affect the switching IC, efficiency, and other important factors? What would choosing a lower impedance inductor do to this board?


Thank you again for all your prompt answers. My next 24/48 hours heavily depends on many aspects of this conversation so its amazing that I've had this opportunity to have this level of detailed dialogue with you.

 

[KrisBlueNZ]

I normally use Mouser for my searches but I'll try Digikey as well.

I find Digi-Key have a lot more information in their product selector - i.e. more columns, which makes it easier to filter out the unsuitable parts. They're not perfect, and sometimes I wish they had a column for an important characteristic because I could eliminate a lot of possibilities based on that characteristic, but I still find their selector a lot better than Mouser's.

Say the PCB has a solder pool connected to ground. Would it still be wise to route this noisy ground line all the way to near where the I/O ground pin is wired into the board or is not a major concern for a solder pool?

I'm no expert, but I would say that it's most important to keep the currents that circulate in the 0V rail that connects the input capacitor, the catch diode, and the output capacitor separate from the main 0V rail. Download some switching regulator and controller ICs from manufacturers like Linear Technology, Maxim, ON Semi and Texas Instruments. They have specific guidelines for layout (and magnetics, diodes, capacitors etc as well).

And here is a new question. When I chose the coil for my switching regulator I did so because a related datasheet described the need for a 40uH inductor but I only had 50uH available. How does higher impedance affect the switching IC, efficiency, and other important factors? What would choosing a lower impedance inductor do to this board?

It will depend on many factors; you would have to compare a lot more than just the inductance and the DC resistance to get a fair comparison between the 40 µH inductor they recommend and the 50 µH inductor you currently have. The major specification for inductors in switching power supplies, apart from inductance, is the saturation current.

DC resistance shouldn't be a big factor in efficiency. Inductance is obviously important but I've never analysed what would happen if the inductance is too high. I just work out the best inductor to use, using the calculations in the data sheet, and specify that! You should be able to reverse the calculations and figure out what problem will occur if the inductance is too high.

I think one of the Linear Technology regulators I recommended has an adjustable switching frequency, which you should be able to use to optimise operation for a particular inductance.

 

[NiGHTS]

I have noticed that Bridge rectifiers have a generally high voltage drop across their diodes (1V) compared to the Schottky diodes. Should I use four of those instead of a bridge rectifier in order to lower the heat demand on the AC inputs?

Also, do you have any advice on the capacitor selection for the AC-to-DC rectifier so as to reduce the heat the capacitor generates? Some of my early tests were done using a single 1000uF capacitor on a 24VAC input and it got as hot as 90C. I've seen improvements using two of them in parallel, so now I plan on using two 1500uF/50V in parallel as shown in the schematic to see how well that works. Haven't tested this yet.

Thank you again for all your help.

 

[KrisBlueNZ]

I have noticed that Bridge rectifiers have a generally high voltage drop across their diodes (1V) compared to the Schottky diodes. Should I use four of those instead of a bridge rectifier in order to lower the heat demand on the AC inputs?

No, I wouldn't recommend that. Schottky diodes are too easily damaged by very brief overvoltage. Unless your incoming mains supply has aggressive filtering and suppression, I think that would be risky.

Also, do you have any advice on the capacitor selection for the AC-to-DC rectifier so as to reduce the heat the capacitor generates? Some of my early tests were done using a single 1000uF capacitor on a 24VAC input and it got as hot as 90C. I've seen improvements using two of them in parallel, so now I plan on using two 1500uF/50V in parallel as shown in the schematic to see how well that works. Haven't tested this yet.

No, I haven't had problems with input smoothing capacitors overheating. I agree that two would be better than one. They're low ESR capacitors, right?

Also, use low-impedance capacitors in parallel with them, located between them and the input to the switching converter. I see you have a 0.1 µF capacitor near the regulator input. I would add a MLCC of at least 1 µF as well, preferably higher. For example http://www.digikey.com/product-detail/en/UMK316ABJ475KD-T/587-2988-1-ND/2714181. That will absorb more of the high-frequency current that may currently be being finding its way to the electrolytics. Also in my design in post #20 you'll see that I used an inductor at the input and at the output.

Thank you again for all your help.

No problem.

Edit: Oops that capacitor isn't X7R. Here are some better options in order of increasing price:
http://www.digikey.com/product-detail/en/UMK316AB7475KL-T/587-2994-1-ND/2714187 (4.7 µF, 50V, X7R, 1206)
http://www.digikey.com/product-detail/en/GRM32ER71J106MA12L/490-9971-1-ND/5026476 (10 µF, 63V, X7R, 1210)
http://www.digikey.com/product-detail/en/FK20X7S2A475K/445-8469-ND/2815399 (4.7 µF, 100V, X7S, through-hole)
http://www.digikey.com/product-detail/en/C5750X7S2A106K230KB/445-13408-1-ND/3955074 (10 µF, 100V, X7S, 2220)

 

[NiGHTS]

No, I wouldn't recommend that. Schottky diodes are too easily damaged by very brief overvoltage. Unless your incoming mains supply has aggressive filtering and suppression, I think that would be risky.

My input voltage would be at most 30VAC and would include a bidirectional TVS right at the input (not shown in the schematic). Would this configuration along with Schottky diodes still seem risky to you?

No, I haven't had problems with input smoothing capacitors overheating. I agree that two would be better than one. They're low ESR capacitors, right?

This is the one that was overheating: http://www.digikey.com/product-search/en?vendor=0&keywords=ECA-1VM102B

This is the one I recently ordered which I haven't tested yet at full AC load: http://www.digikey.com/product-detail/en/UHE1H152MHD/493-1626-ND/589367

I wish the capacitor wasn't so big, but I like the 10,000h life rating. This can be very useful in my application where uptime is very important.

Also, use low-impedance capacitors in parallel with them, located between them and the input to the switching converter. I see you have a 0.1 µF capacitor near the regulator input. I would add a MLCC of at least 1 µF as well, preferably higher. For example http://www.digikey.com/product-detail/en/UMK316ABJ475KD-T/587-2988-1-ND/2714181. That will absorb more of the high-frequency current that may currently be being finding its way to the electrolytics.

Was this capacitance value chosen at random, is it just conventional, or is there a calculation that led you to choose this value?

Also in my design in post #20 you'll see that I used an inductor at the input and at the output.

I can see this. I suppose the combination of this high frequency capacitor and input / output coil is acting as a kind of band-pass filter? As I asked about the capacitor, are the inductance values something you calculated or a "safe value"?

 

[KrisBlueNZ]

My input voltage would be at most 30VAC and would include a bidirectional TVS right at the input (not shown in the schematic). Would this configuration along with Schottky diodes still seem risky to you?

I guess not, as long as your TVS is rated higher than the expected peak voltage, and the diodes are rated comforably higher than the TVS's maximum clamping voltage. If that means you have to use Schottky diodes rated at 80V or 100V, the V_F improvement isn't so dramatic, although the cost increase will be!

This is the one that was overheating: http://www.digikey.com/product-search/en?vendor=0&keywords=ECA-1VM102B

That's not a low ESR electrolytic. I'm not surprised it was overheating with all the high-frequency AC current created by the converter! Also, that one is only 100 µF not 1000 µF.

This is the one I recently ordered which I haven't tested yet at full AC load: http://www.digikey.com/product-detail/en/UHE1H152MHD/493-1626-ND/589367

That's a lot better!

I wish the capacitor wasn't so big, but I like the 10,000h life rating. This can be very useful in my application where uptime is very important.

Will you be able to check that you're operating it within its specifications? Ripple current, specifically. Connect a low-value resistor (e.g. 0.1Ω non-inductive) between its negative terminal and 0V and put an oscilloscope across it, then calculate the ripple current from that. If you exceed that specification, the 10,000 hour spec doesn't apply!

Was this capacitance value chosen at random, is it just conventional, or is there a calculation that led you to choose this value?

Rule of thumb and based on suggestions I've seen in data sheets. Grab some Linear Tech data sheets, at least, and have a good read through their guidelines. Those guys really know what they're talking about! And any of their relevant application notes will be even more useful.

I can see this. I suppose the combination of this high frequency capacitor and input / output coil is acting as a kind of band-pass filter? As I asked about the capacitor, are the inductance values something you calculated or a "safe value"?

The inductors just keep the high-current high-frequency signals away from the rest of the circuitry. You can calculate the impedance of the inductors at the operating frequency to check that it's reasonably high - for example, 22 µH at 100 kHz is X_L = 2 pi f L = 14Ω which is about 100 times higher than the capacitors' ESR, which is great. But IIRC, I just used inductors with appropriate saturation current ratings that were somewhat smaller (physically) than the main energy storage inductor.

 

[NiGHTS]

I guess not, as long as your TVS is rated higher than the expected peak voltage, and the diodes are rated comforably higher than the TVS's maximum clamping voltage. If that means you have to use Schottky diodes rated at 80V or 100V, the V_F improvement isn't so dramatic, although the cost increase will be!

Well as long as this improves the heat output of the bridge rectifier circuit I'm all for it.

That's not a low ESR electrolytic. I'm not surprised it was overheating with all the high-frequency AC current created by the converter! Also, that one is only 100 µF not 1000 µF.

When I click that link I see 1000uF on the page. Either way I agree this was the wrong part for the job.

Will you be able to check that you're operating it within its specifications? Ripple current, specifically. Connect a low-value resistor (e.g. 0.1Ω non-inductive) between its negative terminal and 0V and put an oscilloscope across it, then calculate the ripple current from that. If you exceed that specification, the 10,000 hour spec doesn't apply!

What you explained sounds like (A) on the following diagram. I'm used to measuring current in the (B) way with a multimeter, so I am just confirming that you mean (A) and not (B)?



Also, how am I measuring the ripple? Is this going to be the difference between the highest point and lowest point of the current wave? Come to think of it I don't think my little oscilloscope has a function to read changes in current. I'll have to check.

Rule of thumb and based on suggestions I've seen in data sheets. Grab some Linear Tech data sheets, at least, and have a good read through their guidelines. Those guys really know what they're talking about! And any of their relevant application notes will be even more useful.

I don't often read datasheets as a form of casual literature so I wouldn't know where exactly to begin finding expert suggestions for capacitor filters, decouplers, layout design etc. I suppose you mean that I should try to stick with Linear Tech parts for their very detailed and informative datasheets. Or maybe over time I with more exposure to their brand I will just stumble across enough tips and tricks from their datasheets that I will come to possess this knowledge passively. I suppose one of these days when I have time I can spend some time browsing their hundreds of datasheets trying to find some juicy tips.

The inductors just keep the high-current high-frequency signals away from the rest of the circuitry. You can calculate the impedance of the inductors at the operating frequency to check that it's reasonably high - for example, 22 µH at 100 kHz is X_L = 2 pi f L = 14Ω which is about 100 times higher than the capacitors' ESR, which is great. But IIRC, I just used inductors with appropriate saturation current ratings that were somewhat smaller (physically) than the main energy storage inductor.

Interesting. I'll keep that in mind.