2kW Step-down PWM converter for high voltage dc-BUS

[dado73]

Hello everybody.

I need a step-down dc/dc converter to be used as a front end for a 2kW HF-generator.

The input voltage is 320V.

The output voltage will be adjusted directly from the HF-generator CPU in the range from 100V to 300V: the goal is adjusting the HF-generator power just acting on its supply voltage.

I was thinking to design a simple non-synchronous buck converter with a 300uH inductor, that should be fine for about 50kHz switching frequency.

My concern is that currently I'm not sure that the microprocessor on CPU card will be fast enough to close the loop. And also I'm afraid that its PWM output will not have enough resolution at 50kHz to guarantee a decent regulation. Therefore I would use a simple analog controller ic to generate the PWM and close the voltage loop. Then the HF-generator CPU will simply give the set to such controller ic and close the power loop.

What do you think about this approach ?

Considering that any power change will be very slow I guess that a voltage mode operating controller will be ok: can you suggest an easy-to-use controller ic that will not require plenty of external components ?

Thank you

 

[Easy peasy]

remember, if the series pass buck element fails short - the load will see 320V ....

 

[dado73]

remember, if the series pass buck element fails short - the load will see 320V ....

yes I know, that’s not a big problem

 

[treez]

remember, if the series pass buck element fails short - the load will see 320V ....

This is correct....and you can put in an inline load fet switch...along with comparator, (and a TVS for good measure)...and switch out the 320V when the fet fails.........the load will not immediately see the 320v because the inductor is in the way for a bit....but obviously dont let it saturate.

Anyway, yes ...why not use a simple ucc28c43.....use a high side current sense transformer for current mode control ramp....and use a pulse transformer gate drive for the high side fet........put the current sense transformer upstream of the high side nfet...as its slightly less noisy than having it in the switching node.

If you want to spread the thermal load about a bit, then use a cascaded buck......both stages being operated from the single ucc28c43........being cascaded, you can make the upstream fet have low cds so it has low switching loss...and let the lower fet have low rdson...because it will only need a lower voltage rating.
ill get you a sim of it soon....ok now its attached..its in the free LTspice simulator which you can download

 

[dado73]

Thank you for your answer treez.

Please let me know: why should I choose a current mode solution and not a simpler voltage mode? The reference for the output voltage will come from the microprocessor, that can make it slowly ramp up to the desired value, based on power feedback. What should be the drawback?

About driving the fet I would choose an opto-driver providing an isolated supply: I'm more comfortable with this kind of approach that allows me go up to 100% duty cycle in case of need.

Also I would stay on a single fet solution, just because the converter works as a power regulator and at lower output voltage (larger input-to-output voltage drop) also the current will be low.

Rather, I am not that thrilled that the gnd of the IC is in common with the negative of the output voltage, with the risk of picking up some noise from commutation. What do you think about?

 

[treez]

SMPS Course _B.zip

drive.google.com

..GO TO THE "CONTROL MODE OF AN SMPS" folder of the above free smps course.
It will show you what is VM and CM in the ltspice sims.

But basically, even with VM, you need current sense in the fet source...for safety reasons.....so you may as well use that source sense for current mode control.

Read up on Dr Ray Ridley...he did his thesis on current mode control and can tell you why current mode is generally better than VM.

Eg with CCM buck, the dynamics are more complex in VM than CM.
In CM, you cancel out one of the poles of the double pole response of the buck inductor and cap...so its easier for you to get stability.

Read the pcb layout doc in the above course, to answer your concerns on that..i also attache here.

An opto for high side drive is not possible unless your f(sw) is ridiculously low. Optos are too slow.
Use either...
Pulse transformer.
High side supply & digital isolator into fet driver......or
Bootstrap high side drive.

 

[dado73]

But basically, even with VM, you need current sense in the fet source...for safety reasons.....so you may as well use that source sense for current mode control.

Thank you again, I'm a little worried because all this seems to me much more complicated than I thought, and I have just a few days available to design, protorype and test this converter....

Please do not forget that the goal of this converter is not to set the dc-link voltage at a specific value, but to regulate the power absorbed from the main based on the value set by the user, and this "power loop" can be very slow.

As already said at first I was thinking of using just the PWM generated from the microprocessor to directly drive the fet: in this case the only loop will be the power loop (input current and voltage readings).

But then I was a little concerned about the resolution of such PWM @ 50kHz, so I thought I'd foresee the possibility of driving the fet with a dedicated ic and use the microprocessor just to generate the set for such ic: that means a faster inner voltage loop managed by the ic and a slower outer power loop managed by the uC.

Do I really need to make a CM control with a further inner current loop ? Wasn't it much easier if I had enough resolution from the PWM of the uC and used it to directly drive the fet as in the first picture, saving me from using an external component ?

An opto for high side drive is not possible unless your f(sw) is ridiculously low. Optos are too slow.
Use either...
Pulse transformer.
High side supply & digital isolator into fet driver......or
Bootstrap high side drive.

Why you say that ? I normally use 10Mb optos like HCPL2611 to drive fets up to 500kHz (!) without any kind of problem, they have propagation delays in the order of a few tens of nanoseconds.