Topology for 140W LED boost driver (two switch forward maybe)

[Zipdox]

So I previously posted on this forum asking for help designing a tapped boost converter driver for a BXRC-50G10K1-C-73-SE LED, and it did work, but it had some flaws.

Approximate ratings:
Input voltage: 12-24V
Output voltage: 70V
Output current: 2A

Obviously using an SMD MOSFET was a stupid idea cooling-wise, it should be replaced with a THT MOSFET that can be mounted to a heatsink.
Another issue I noticed is coil whine. Although the drive frequency is way above the range of human hearing, in the higher output power range the inductor seems to whine at an audible frequency, which I think is caused by the current feedback loop.
The output current ripple is also pretty bad, because it's driving an LED and as such a tiny amount of voltage ripple causes a lot of current ripple when the LED is just barely turning on.

I don't know if the efficiency of my design is optimal. I think that maybe creating a new design using a different topology will allow for a higher efficiency.

After some reading, namely

• https://www.digikey.nl/en/articles/dc-dc-converter-topologies-and-techniques-to-obtain-high-boost-ratios
• https://www.sunpower-uk.com/glossary/what-is-a-forward-converter/

I think that designing a two switch forward booster will mitigate a number of problems that exist in my current design.

First of all, using a two switch forward will eliminate the voltage overshoots and ringing associated with traditional boost converters. This is due to the clamping diodes, which also eliminate the need for a snubber network. This will allow for lower voltage ratings on the MOSFETs, which will allow for lower forward voltage MOSFETs and thus lower power loss.
I think the output inductor would also result in a reduction in ripple on the output, though that's just a guess, I don't know if it will make a difference.
I'm not sure about the power losses in the clamping diodes though, considering they will be conducting quite a bit of current due to the boost topology. Two switch forward converters are usually used in step-down mode so I couldn't find any documentation on such a scenario.
I also don't know if there will be any voltage overshoots on the secondary side, and what diodes should be picked for it.

Obviously a two switch forward is more complex than a tapped boost converter. I'd have to either use a different controller IC that is capable of high-side and low-side switching, or I'd have to add a separate gate driver.

I'm really an amateur when it comes to these kinds of things, so any insight would be appreciated.

[ogden]

Copy inspire from existing design. In your case I would try following or similar multiphase boost converter, converted to current mode regulation: AN-1820 LM5032 Interleaved Boost Converter (Rev. A). Diode rectification is good enough for 2A, no need for sync rectifier which will increase complexity and price. I do not see input fuse in your existing design - always have one.

[Zipdox]

Copy inspire from existing design. In your case I would try following or similar multiphase boost converter, converted to current mode regulation: AN-1820 LM5032 Interleaved Boost Converter (Rev. A). Diode rectification is good enough for 2A, no need for sync rectifier which will increase complexity and price. I do not see input fuse in your existing design - always have one.

What advantages would an interleaved boost converter give me besides the reduced ripple? It would increase cost and complexity, not to mention a significant chunk of board space.

[uer166]

Copy inspire from existing design. In your case I would try following or similar multiphase boost converter, converted to current mode regulation: AN-1820 LM5032 Interleaved Boost Converter (Rev. A). Diode rectification is good enough for 2A, no need for sync rectifier which will increase complexity and price. I do not see input fuse in your existing design - always have one.

What advantages would an interleaved boost converter give me besides the reduced ripple? It would increase cost and complexity, not to mention a significant chunk of board space.

I hate to say it, but Treez is right! In this case you have ~16A RMS input current, which translates to quite a high peak input current ( if you're doing a normal boost converter. Of course with a forward the peak current is not nearly as high, but it's still pretty substantial.

I would make a 2 or 4-phase interleaved boost (not tapped, just normal boost) for something like this. No way you could reasonably devise a single stage boost with this ratio and power levels. A multiphase solution will be much smaller and more efficient. I don't see why a forward wouldn't work here, but it's more work with the isolation and transformer design, boost is quite straightforward and here it's at the upper edge of reasonable step-up ratio.

[ogden]

What advantages would an interleaved boost converter give me besides the reduced ripple? It would increase cost and complexity, not to mention a significant chunk of board space.

Chip manufacturers do not agree to you. Main benefits I see in multiphase - smaller & lower cost inductances, option of higher switching frequencies, smaller switches, not to mention lower input/output capacitor requirements. All that may lead to actually lower volume of converter compared to single phase. If offered design is too complex for you - don't pick it obviously.

[Zipdox]

I hate to say it, but Treez is right! In this case you have ~16A RMS input current, which translates to quite a high peak input current ( if you're doing a normal boost converter. Of course with a forward the peak current is not nearly as high, but it's still pretty substantial.

I would make a 2 or 4-phase interleaved boost (not tapped, just normal boost) for something like this. No way you could reasonably devise a single stage boost with this ratio and power levels. A multiphase solution will be much smaller and more efficient. I don't see why a forward wouldn't work here, but it's more work with the isolation and transformer design, boost is quite straightforward and here it's at the upper edge of reasonable step-up ratio.

Using a non-tapped boost converter has the advantage of not having negative swinging voltage on the rectifier diode, but will it be able to achieve the desired output voltage?
On second thought, you are right about a forward converter being more complicated. I guess I'll make a multiphase booster converter.

One thing I have on mind me though, is what the best way to do current sensing. The LM5032 datasheet shows current transformers but I don't know if I wanna go that way. I'd rather just use a shunt and a current sense/instrumentation amplifier. The LM5032 isn't suitable by the way because of its minimum supply voltage, so I'll have to find something else.
But I need to ask, does the current sensing have a purpose besides making sure the MOSFETs don't exceed their rated current? Because the way that my LED drivers works is using the voltage feedback pin of the driver.

[ogden]

But I need to ask, does the current sensing have a purpose besides making sure the MOSFETs don't exceed their rated current?

Current sensing in current mode PWM converters is not about MOSFET but inductor current. Make sure to meet required loop bandwidth of current sense circuit, better use offered by manufacturer solution(s). Don't limit your options to LC5032. For example OnSemi NCV78702 could do the job. I would buy IC first, then design because you may get close to none availability for part of your choice.

[Zipdox]

Current sensing in current mode PWM converters is not about MOSFET but inductor current. Make sure to meet required loop bandwidth of current sense circuit, better use offered by manufacturer solution(s). Don't limit your options to LC5032. For example OnSemi NCV78702 could do the job. I would buy IC first, then design because you may get close to none availability for part of your choice.

I'm a little lost. I don't know what loop bandwidth means. The things I found aren't related to switching supplies.

The NCV78702 is controlled via SPI, which is outside of the scope of my project. I'd rather do it all analog. I'll look for alternatives.

[T3sl4co1l]

Right, loop bandwidth is not specific to SMPS.  It's any control system.  So, it has to do with the error amp between pins 1-2 of the UC3843, or, for the NCV78702, between VBOOSTDIV and COMP (note this is a transconductance type, so the feedback network is not required, just an RC network to GND; the effect is the same).  Or the LM5032 in terms of current loop maybe, but it looks like that chip actually requires an external error amp, so, a TL(V)431 say, with compensation in the usual way.

Tim

[ogden]

I don't know what loop bandwidth means. The things I found aren't related to switching supplies.

No worries. Then simply do not redesign/reinvent current feedback circuit. Leave as per reference design.

The NCV78702 is controlled via SPI, which is outside of the scope of my project.

Again - no worries. As long as you can get LM5032, you can ignore any other IC including NCV78702.

[Zipdox]

I did some searching on Analog's site and the LTC7840 seems a like a good contender. Thoughts?

[ogden]

I did some searching on Analog's site and the LTC7840 seems a like a good contender. Thoughts?

Yes indeed good contender, in 2-phase config.

[Zipdox]

I also found the LTC3862, which allows clock output and customizing the phase, so you can make 2-, 3-,
4-, 6- or 12-phase boosters. It's available on TME, so I might go with this one if there aren't any problems.

[langwadt]

Copy inspire from existing design. In your case I would try following or similar multiphase boost converter, converted to current mode regulation: AN-1820 LM5032 Interleaved Boost Converter (Rev. A). Diode rectification is good enough for 2A, no need for sync rectifier which will increase complexity and price. I do not see input fuse in your existing design - always have one.

What advantages would an interleaved boost converter give me besides the reduced ripple? It would increase cost and complexity, not to mention a significant chunk of board space.

I hate to say it, but Treez is right! In this case you have ~16A RMS input current, which translates to quite a high peak input current ( if you're doing a normal boost converter. Of course with a forward the peak current is not nearly as high, but it's still pretty substantial.

I would make a 2 or 4-phase interleaved boost (not tapped, just normal boost) for something like this. No way you could reasonably devise a single stage boost with this ratio and power levels. A multiphase solution will be much smaller and more efficient. I don't see why a forward wouldn't work here, but it's more work with the isolation and transformer design, boost is quite straightforward and here it's at the upper edge of reasonable step-up ratio.

if a suitable coupled inductor can be found or you have to wind it anyway, why not tapped? 

[uer166]

Copy inspire from existing design. In your case I would try following or similar multiphase boost converter, converted to current mode regulation: AN-1820 LM5032 Interleaved Boost Converter (Rev. A). Diode rectification is good enough for 2A, no need for sync rectifier which will increase complexity and price. I do not see input fuse in your existing design - always have one.

What advantages would an interleaved boost converter give me besides the reduced ripple? It would increase cost and complexity, not to mention a significant chunk of board space.

I hate to say it, but Treez is right! In this case you have ~16A RMS input current, which translates to quite a high peak input current ( if you're doing a normal boost converter. Of course with a forward the peak current is not nearly as high, but it's still pretty substantial.

I would make a 2 or 4-phase interleaved boost (not tapped, just normal boost) for something like this. No way you could reasonably devise a single stage boost with this ratio and power levels. A multiphase solution will be much smaller and more efficient. I don't see why a forward wouldn't work here, but it's more work with the isolation and transformer design, boost is quite straightforward and here it's at the upper edge of reasonable step-up ratio.

if a suitable coupled inductor can be found or you have to wind it anyway, why not tapped?

Because you run into leakage inductance, snubber design, and negative voltage swing of the output tap. It's a much harder thing to design than a boost, and only marginally simpler than a full-on flyback.

[Zipdox]

Actually the LTC3862 isn't suitable because it drives the MOSFETs at 5V, so I'd need a gate driver.

[langwadt]

Actually the LTC3862 isn't suitable because it drives the MOSFETs at 5V, so I'd need a gate driver.

LM5122 ?

https://www.ti.com/tool/PMP9595

[Zipdox]

LM5122 ?

https://www.ti.com/tool/PMP9595

That would require two chips. I'm already workings on a design with the LTC7840, which has lots of easily customizable parameters.

[Zipdox]

So I made a design.
It's supposed to run at 140kHz.

I probably forgot some things, but one thing I have no clue how to figure out is the I_TH pins. I know they're for a capacitor and a resistor for adjusting the loop response, but the datasheet doesn't really have a clear explanation of how to choose values.

[poorchava]

I would just go with a normal single phase boost. No need to overcoplicate it. I mean I have a 3.5kW boost pfc and it works no problem. 140W should be doable. Boost ratio of 5 doesn't call for a transformer, a normal boost will do.

At those voltages switching losses are usually not a problem either unless you use some really shit fet and diode.

[T3sl4co1l]

Two phase is nice because it lets you use much smaller capacitors.  For example I did a compact 1kW buck module a few years back, single phase would've needed massive capacitors like 2mF 160V at the input, literally taking up more space than the second phase.  Whereas a couple hundred uF is enough with 2ph.  Plus inductors of that rating aren't easy to get commercially, but halfsies, no problem.

Also very popular for PFC, even at fairly low levels (100W+).  It can run quite fast, saving even more component size; usually doing something like quasi-resonant boost to save on switching loss.  (QR requires BCM, so, 100% current ripple, so does incur a bit more cost in the inductors, probably being custom ferrite parts.)

For this, the tapped inductor is fine if you're doing a custom wind anyway, I mean, why not, right?  Makes the switching a bit easier.  Doing it without, just straight boost, also isn't a problem, as has been stated.  It's on the upper end of what's practical, so you will invest a bit more in component ratings to do it, but not by a ridiculous amount.  Good for builds using all off-the-shelf parts.

So I made a design.
It's supposed to run at 140kHz.

I probably forgot some things, but one thing I have no clue how to figure out is the I_TH pins. I know they're for a capacitor and a resistor for adjusting the loop response, but the datasheet doesn't really have a clear explanation of how to choose values.

Looks fine.  Ith pins are error amp outputs; they're also transconductance (gm) type, so an impedance to ground serves for compensation.  Also means they can be overdriven, say if you want to use an external error amp (like, connecting an opto here for an isolated converter), or wired in parallel (they won't fight each other, as would be the case for a voltage-mode output).  So, wire them together to get equal current setpoints for both channels, which is all you need to ensure current sharing between the two phases.  Fig.7 shows this.

Tim

[Zipdox]

Looks fine.  Ith pins are error amp outputs; they're also transconductance (gm) type, so an impedance to ground serves for compensation.  Also means they can be overdriven, say if you want to use an external error amp (like, connecting an opto here for an isolated converter), or wired in parallel (they won't fight each other, as would be the case for a voltage-mode output).  So, wire them together to get equal current setpoints for both channels, which is all you need to ensure current sharing between the two phases.  Fig.7 shows this.

Ok so I connect them together, but is nothing else required? Like I said I absolutely no idea what to do with them at this point.

[T3sl4co1l]

You need the RC compensation network...

Tim

[Zipdox]

I made a spice simulation. There seems to be a loop where the converter switches on and off at several kHz. Is this what the compensation network is for?

[T3sl4co1l]

You need to tie the Ith's and SS's together too.

Check waveforms and values on C9-C11, and value of R16 (assuming you retain those parts and remove C12, C13, R17).  Probably C9 has to be smaller and C11 much larger.

Don't forget to put the overvoltage clamp back in later!

FYI, units are incorrect (R doesn't exist, L is liter, C is coulomb), though I don't mind 'R' shorthand for Ω.  More importantly, SPICE ignores any suffixes that aren't unit multipliers (f (10^-15), p, n, u, m, k, meg, g, t (10^12)), so this complaint has no effect on the model; they can be left blank just as well for simulation purposes.

Tim