"Dumb" LLC converter with NO feedback is the best way to go in this application?

"Dumb" LLC converter with NO feedback is the best way to go in this application?
Posted by ocset on 20 Jun, 2020 23:49
Hi,

We need a “rough” isolated rail of 45-53V at 200w. Vin is PFC output. It Undulates from 370v to 430v at the usual 100hz.

We wish to simply use an half bridge LLC converter switching at f0 all the time. We will make the NS/NP equal to 0.5 *(vin/vout).
..No feedback…who needs feedback? In fact, feedback in an LLC converter makes the LLC more self-destructable…because during a sudden heavy load transient, the feedback error amplifier would command a lower operating frequency…..and its lower operating frequencies which bring about overcurrents in an LLC converter……so scrub the feedback…its not needed, would you agree?

……….Actually the FSFR2100XS LLC controller looks like its good for this. It can provide a constant switching frequency of 100khz (94khz to 106khz) if a 5.2k resistor is connected to its RT pin. (94khz to 106khz).
Another resistor R_START could be placed in parallel with it and switched in only at startup, so that the frequency was f0*1.5 for a soft start. During this soft start, it would be ensured that the secondary loading was off.
Once the Vout gets up to 48V, then the R_START resistor would be switched out…leaving the controller to gormlously switch away at 100khz.
I really cant see any point in having the TL431 based error amplifier….i mean, it wouldn’t be needed…so just leave it out…would you agree?
The only threat would come from output overload……but that would cause the primary side sense resistor to trip the controller out…so even that’s not a problem.
Why doesn’t the datasheet suggest doing it like this?

The LTspice sim (attached) of it shows that even when the secondary is suddenly no_loaded , there is no serious overvoltage at vout….it just harmlessly goes up a few volts.
(when the load suddenly goes from no_load to full_load, then , there is no serious undershoot in vout, and no overvoltage on the resonant capacitor which would be 1kv rated.
“Dumb oscillator” switching away and does the job great…where’s the gotcha?

FSFR2100XS datasheet
https://www.onsemi.com/pub/Collateral/FSFR2100XS-D.pdf

Half Bridge LLC converter_openloop.asc

#1 Reply
Posted by David Hess on 21 Jun, 2020 19:17
Wouldn't an inverter be easier?

#2 Reply
Posted by jbb on 23 Jun, 2020 00:56
This sort of thing can be done, but there are some issues: startup, output short, output open and tolerances.

For startup you probably want to start with a higher frequency so you don’t saturate your magnetic. No big deal.

You might want some current sensing for output short circuit protection. This can be done by tapping some voltage of the resonant capacitor.

With an open output (no load), the output voltage might creep up.

Finally, tolerances are a big deal. Your resonant L and C will have tolerances, and the capacitors will age. So you need to check if this could put you in a hard switching region (if so, increase Fsw) and / or change the converter gain.

#3 Reply
Posted by ocset on 23 Jun, 2020 12:54
Quote
You might want some current sensing for output short circuit protection.
Thanks yes, the FSFR2100 has primary side resistive current sense and temporarily shuts down whenever its triggered...then restsarts with "high frequency sweep down".

As you know, The great advantage of "controlled open loop" method is that an HB LLC, if constantly switched at f0, will chuck out an output voltage of ~ Vout = Vin/2 *Ns/NP.

Open load........the Vout wont increase beyond a few volts in this case as their simply isnt enough voltage referred to the secondary......but yes that few volts has to be handled.

Short circuit output will initially result in the feedback loop trying to reduce switching frequency, which of course, in that case, is deffo not wanted.....so just limiting the FSW to fo is good here, and the FSFR2100 has primary resistive current limiting anyway.
In fact, we will have another resistor hanging off the RT pin so that we can increase frequency to max frequency in cases of output overload or output short circuit........this will be triggered by monitoring vout with a comparator....and in any case, the primary current sense would catch it and trip out the controller......we just have to set the sense threshold high enough that it doesnt trip on a 0-100% load transient.

Basically, switching at f0 can manage full load, so we dont want to go any lower in frequency than that......lower fsw can mean too high volts on Cres if its too low Fsw.

Vin: Its a PFC output with variation from 370v to 430v (at 100Hz) due to minimal DC bus capacitance. We dont want the LLC's feedback loop to try and reject that, as an output of 52v to 44v is fine.

#4 Reply
Posted by T3sl4co1l on 23 Jun, 2020 13:05
If the source and load are guaranteed never to vary, sure, you can get away with a lot of things. A fixed frequency inverter and whatever kind of load network you want, will do.

Considering a circuit has to be constructed and powered up, it would seem there always must be at least one event where source or load varies, so this isn't all that useful.

Tim

#5 Reply
Posted by ocset on 23 Jun, 2020 15:31
Quote
Considering a circuit has to be constructed and powered up, it would seem there always must be at least one event where source or load varies, so this isn't all that useful.
Thanks, and of course, very true..however, in this case, the loads are leds...and we will always know when they are about to be turned on or off.

We will actually have the facility, at startup, to have all the led loads initially disabled , and make the LLC switch at 2xf0 for long enough for vout to get up into regulation...then we will make it switch at f0 continuously...and then turn the led loads on.