Author Topic: High voltage buck converter  (Read 4943 times)

0 Members and 1 Guest are viewing this topic.

Offline PtzfTopic starter

  • Newbie
  • Posts: 6
  • Country: ee
High voltage buck converter
« on: July 28, 2018, 08:40:01 pm »
Hello!

Here is what I want to do: convert a steady 325VDC (rectified mains) to a variable DC voltage in the range of 250-300V, depending on voltage feedback. The current at the load will be around 1-1.5A. I would like to use a buck converter to do this, as it is simple and possibly cheap. Switching frequency would be say 300 kHz. The named voltages imply a duty cycle of roughly 70%-90%.

Are there any obvious flaws in this plan of using a simple buck converter? Should I use some other topology to achieve the design goal (simple, cheap and with voltage-mode feedback)? Why I'm asking is because search results on the web seem to give buck converter duty cycles at 50% max (perhaps because of PWM controller current mode operation difficulties) and max 100V input IC buck controllers. Will the high duty cycle be a nonlinear problem for the inductor-diode circuit? In short: I lack domain knowledge :)
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: High voltage buck converter
« Reply #1 on: July 28, 2018, 09:11:05 pm »
Current mode. Always current mode.

Other than that, yes, that's perfectly doable.  Consider an N-ch MOSFET, and a bootstrap gate driver.  Make sure the controller has some way to limit maximum duty cycle, so it doesn't stay latched on (100% duty), which starves the bootstrap supply.

Current mode is easiest if you don't need common grounds, between input and output.  You can place the shunt resistor there, in series with the output ground (upstream of the filter caps, so it reads inductor current ripple).  If you do require common ground, you can still use a "flying" sense method, but it gets more difficult (either a current sense amp, with quite high voltage rating and CMRR, or a current transformer to read switch current).

You could also use another topology, like flyback or SEPIC, to get a wider range of input and output voltages, and consequently, different positions for current sense (which may turn out to be more convenient).  These take more reactive power (bigger inductor, dual windings) so they are less efficient than the buck converter would be.  Sometimes the advantage is worth the cost, depends.  Just putting that out there.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 
The following users thanked this post: ogden, Ptzf

Offline langwadt

  • Super Contributor
  • ***
  • Posts: 4857
  • Country: dk
Re: High voltage buck converter
« Reply #2 on: July 28, 2018, 10:44:07 pm »
Current mode. Always current mode.

Other than that, yes, that's perfectly doable.  Consider an N-ch MOSFET, and a bootstrap gate driver.  Make sure the controller has some way to limit maximum duty cycle, so it doesn't stay latched on (100% duty), which starves the bootstrap supply.

Current mode is easiest if you don't need common grounds, between input and output.  You can place the shunt resistor there, in series with the output ground (upstream of the filter caps, so it reads inductor current ripple).  If you do require common ground, you can still use a "flying" sense method, but it gets more difficult (either a current sense amp, with quite high voltage rating and CMRR, or a current transformer to read switch current).

You could also use another topology, like flyback or SEPIC, to get a wider range of input and output voltages, and consequently, different positions for current sense (which may turn out to be more convenient).  These take more reactive power (bigger inductor, dual windings) so they are less efficient than the buck converter would be.  Sometimes the advantage is worth the cost, depends.  Just putting that out there.

Tim


if common ground isn't need doesn't it make more sense to turn everything up-side-down like many of the off-line LED supplies?
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: High voltage buck converter
« Reply #3 on: July 28, 2018, 11:35:18 pm »
if common ground isn't need doesn't it make more sense to turn everything up-side-down like many of the off-line LED supplies?

I prefer average current mode to peak current mode (which is easier in that case), but that's definitely an option too.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: High voltage buck converter
« Reply #4 on: July 29, 2018, 06:44:03 pm »
I am inclined to think that there are better ways now but back in the 80s, Tektronix used a bootstrapped sort of SEPIC converter to convert a 340 volt line into to 42 volts DC.  This allows driving an n-channel buck transistor without a level shift and easy current mode control although Tektronix just used current limiting.
 

Offline PtzfTopic starter

  • Newbie
  • Posts: 6
  • Country: ee
Re: High voltage buck converter
« Reply #5 on: July 30, 2018, 08:37:49 am »
Thank you for the replies!

@T3sl4co1l - why always a current mode controller? In my application, output response to load variation can be as long as seconds. Are there more fundamental reasons to prefer current mode in this application? From my viewpoint, voltage mode seems easier to implement, as I do not have to dwell into slope compensation for above 50% duty cycle.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: High voltage buck converter
« Reply #6 on: July 30, 2018, 09:33:41 am »
Thank you for the replies!

@T3sl4co1l - why always a current mode controller? In my application, output response to load variation can be as long as seconds. Are there more fundamental reasons to prefer current mode in this application? From my viewpoint, voltage mode seems easier to implement, as I do not have to dwell into slope compensation for above 50% duty cycle.

Yes, precisely!

The inductor current is a free variable: the integral of PWM over time* (yadda yadda plus a constant, y'know?).  Current blows transistors, so we have a direct and absolute interest in fixing this variable!

*Roughly speaking.  In the real world, there is inevitable resistance that makes it an L/R time constant, but one much larger than Fsw, so that inductor current can build up over many cycles.  We assume zero resistance for simplicity's sake, and adjust as needed later.

Also consequently, the inductor and filter cap resonate together.  Voltage mode has an unavoidable double complex pole in the loop, so you cannot compensate it very well.  At best, you can put a zero in the loop (capacitor ESR).

Current mode wraps inductor current in one loop, voltage in the other.  The poles are kept separate, and the compensated voltage loop is typically close to 2nd order: very easy to compensate, reasonably fast, and doesn't require hacks like cap ESR. :)

Current mode is hardly, if at all, more complicated, and better in all respects.  After adding hacks necessary to protect a voltage mode controller, you've done more work, and the response still sucks; why not save effort doing it right the first time? :D

You don't need slope compensation in average current mode, and I don't think it's needed in other timing methods, just specific to peak current mode.  Peak current mode is best for small converters where a high ripple fraction (and maybe low efficiency) is acceptable.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 
The following users thanked this post: Ptzf

Offline PtzfTopic starter

  • Newbie
  • Posts: 6
  • Country: ee
Re: High voltage buck converter
« Reply #7 on: August 23, 2018, 05:45:43 pm »
So, here is where I am with this buck converter. Problem description: I'm seeing a switch-off lag between GS pins of the highside mosfet. Or in other words: the HS mosfet is not properly switching off.

Details: I'm using a FAN73932MX half bridge driver. The PWM signal is coming from my Rigol DG1022U @350 kHz, at 50% duty cycle. The mosfets are SIHB12N60E-GE3. Input capacitance is 120 uF on the PCB (input voltage is for the images below 38V), the inductor is 1mH, output capacitance 82uF, after the inductor I have a 100 kOhm resistor from output to ground and a big load resistor of about 220 Ohms.

Here is the datasheet application of the driver (as said, for the load I have a inductor, parallel capacitor and a resistor).

On my PCB, gate resistors are 10 Ohm in series and 22kOhm between GS.

The lag I'm talking about:

The yellow trace measures low-side GS signal and teal measures high-side GS signal. Vertical cursors show a dead time of ~400ns, which is factory provided by the driver. What is interesting with the lag is that it remains even when I remove everything from the output, remove the load from the circuit altogether or when I replace it with a purely resistive load. A very low impedance load lessens the lag (with the cost of several watts of heat), but does not remove it. As you see, there is no switch-off lag at the low-side GS.


The PCB layout is really quite minimalistic, there should not be any serious leakage inductance. As you see, the mosfets have large heatsinks in place for further higher dissipation experiments.


Here is the same waveform and same drive settings, but all load removed from the output. Purple trace is from the junction between HS and LS mosfet (Vs pin of driver).


Here there is a 150 Ohm resistor at the output.

Question: what's up wit the lag and how do I get rid of it? I've run a similar circuit at similar frequencies with the same driver and have seen no problems, with mosfets that have several times more capacitance (Coss, Cgs, Cds etc).
« Last Edit: August 23, 2018, 05:47:50 pm by Ptzf »
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: High voltage buck converter
« Reply #8 on: August 23, 2018, 11:02:26 pm »
RTFDS? :)
https://www.mouser.com/ds/2/149/FAN73932-1006913.pdf

You've quite well measured the dead time at about 400ns.  This is not a driver you should've selected for 300kHz operation... :(

Gate rise/fall time ~30ns and output peak current ~2A should be enough to drive those transistors quite fast.  I would guess your stray inductance is 20-30nH, and with those transistors being that fast, you will get ringing.  Which is the observation.  I would not say inductance is "not serious", or, "not significant" at least.  We'll see if it gets worse under load.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline PtzfTopic starter

  • Newbie
  • Posts: 6
  • Country: ee
Re: High voltage buck converter
« Reply #9 on: August 24, 2018, 04:08:52 pm »
The mentioned FAN is a nice "tight" package, with a single input and internal logic to share it to HOUT and LOUT. 400ns dead time is a good point though. I'm going to try a overall faster driver with separate HIN/LIN channels. The mosfet cross-conduction loss in the previous case was a too big.

I'm really not sure on the stray inductance, but I'll measure it in the next experiment.
 

Offline ocset

  • Super Contributor
  • ***
  • !
  • Posts: 1516
  • Country: 00
Re: High voltage buck converter
« Reply #10 on: August 25, 2018, 06:24:41 pm »
your spec is ok for buck...you need slope compensdation if current mode and D>50%...or else you can just use constant off time and then you dont need slope compensation.
 

Offline PtzfTopic starter

  • Newbie
  • Posts: 6
  • Country: ee
Re: High voltage buck converter
« Reply #11 on: October 28, 2018, 11:03:44 am »
Right!

I now have this buck converter operating at 325 Vin, 500 kHz and 300 W, 80% and over duty cycle. I have a 600+ uH inductor at the output. I chose to go with a asynchronous design, as it dissolves any problems with shoot trough. They flyback diode is based on SiC technology and seems to be very low loss and bulletproof all in all (I've killed the HS mosfet and the driving circuitry several times now, but the diode does not want to die).

However I'm having trouble with blowing up my HS driver. Most likely due to a too large negative transient on Vs pin (Vb-Vs surpasses the allowed limit of 25V). The HS driver is FAN73711. Here is a microscope shot of the "uncapped" IC:

In the upper left quadrant of the uncapped chip is a deep hole that runs from the die to the pin - and it is on pin 6, which is the Vs pin (pin 5 on the very tippy-top in the left upper quadrant is NC).

The load of the converter is very inductive and a fast di/dt will definitely induce negative transients on Vs.

ON-Semiconductor AN-6076 describes methods for relieving negative voltages on Vs pin in bootstrapped drivers, most notably relocating the HS mosfet gate resistor to the Vs output-input pin. Doing this, however, starts to have pretty drastic effects on the switching waveform on the switch gate. Compensation drives up the component count and, well, development and testing time.

My question: are there high voltage HS drivers out there that mitigate Vs transients to begin with, non-bootstrap? I mean, some other technology, like isolated drives etc. I'm thinking that Infineon EiceDRIVER family (1EDCxxI12MH) might do the trick?
I need this converter to be very robust.

Thank you for the help!
 

Offline ocset

  • Super Contributor
  • ***
  • !
  • Posts: 1516
  • Country: 00
Re: High voltage buck converter
« Reply #12 on: October 28, 2018, 12:43:41 pm »
You can have a high side isolated supply and just  use say a digital isloator to drive a hi side gate driver.
But what is you input capacitor size?
I am sure you are aware that at 300W, you really need a boost pfc stage up front.
If its just for hacking about with, then yes, use a isolated module to give you a "ground" at the source of the hi side fet...then you can just drive it freely with an isolated drive signal.
Or you could use a gate drive transformer and pnp turn off circuit.

one of these is good to give you an isolated hi side supply....it doesnt have much pri-sec capacitance which is good.
https://power.murata.com/data/power/ncl/kdc_nxj1.pdf
double check the isolation voltage is ok for you.
« Last Edit: October 28, 2018, 12:45:23 pm by treez »
 

Offline Wolfram

  • Frequent Contributor
  • **
  • Posts: 414
  • Country: no
Re: High voltage buck converter
« Reply #13 on: October 28, 2018, 01:18:59 pm »

My question: are there high voltage HS drivers out there that mitigate Vs transients to begin with, non-bootstrap? I mean, some other technology, like isolated drives etc. I'm thinking that Infineon EiceDRIVER family (1EDCxxI12MH) might do the trick?
I need this converter to be very robust.

Thank you for the help!

Yes, fully isolated drivers will solve the problem of switching node undershoot killing the driver. There are plenty of devices available, including the Infineon drivers you've mentioned, the TI ISO5xxx series and UCC53* series, Silabs have a bunch of nice parts and Analog Devices also make some good devices. The only thing to watch out for is the common mode transient imunity of these drivers (this also applies for the non-isolated bootstrap drivers). With your MOSFETs and bus voltage, I would expect somewhere around 20 kV/µs, which should be well within the ratings of many (but not all!) of the available devices.

I've used the TI UCC53* series and the Silabs SI8271 in a 30 kW buck converter (750 V input, 400 V output), and the only problems I've experienced was caused by us exceeded the CMTI rating on an older generation of Silabs drivers.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: High voltage buck converter
« Reply #14 on: October 28, 2018, 05:51:05 pm »
If you somehow can't fix the inductance problem in layout, yes, you need a fully isolated system.

This should only be necessary in higher power industrial systems, where you won't mind the added cost anyway.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf