Level shift for buck converter

[b_force]

Ok, I am working on a buck converter and I am 80% finished with the design.
There is only one thing I need to change, and that's the level shifters for the half bridge output stage.
Now I have an of the shelf IC, but it's not really needed and therefore I like to see if there are alternatives.
(cheaper as well)

Unfortunately, I just can't get a working circuit.
The output is floating/connected to a symmetrical power supply (so the negative is not connected to ground but -V)
That makes the level shifting so much more a pain in the ass.

The bootstrap circuit is doing fine (VCC, 12V above the negative reference) The input voltage is going to be 3.3V with GND as reference.
Output voltage is ±50V.

The lo-side is not that difficult, but I need some ideas for the hi-side levelshifting.
(with the floating reference between the two MOSFETs)

Anyone ideas? I searched an awful lot, but strangely enough I can't find much information or a proper starting point.
(most articles are about the fact that the output is connected to GND, which is easy)

[jbb]

Hi

You probably want an isolated gate driver.  You can buy complete gate drivers off the shelf, but they're super expensive ($200 and up!).  You can however get something like a Silabs SI8233BB-D-IS1 (https://www.silabs.com/Support%20Documents/TechnicalDocs/Si823x.pdf), which is a dual-channel isolated driver.  You'll need to supply 12 - 15V power to drive the upper & lower MOSFETs.

Power supplies can be approached 2 ways:
1) transformer isolated power, the gold standard.
2) bootstrap supplies.  Cheaper but less reliable.

If you're going bootstrap, you'll need a supply of say (-V + 12V) for the low side driver VDDB.  Then use a Schottky diode to charge up the supply capacitor for the upper MOSFET gate driver VDDA.  The downside is you can't use 100% positive duty cycle - the switching node has to go low periodically to charge up the supply capacitor. See Page 31 of the datasheet above.

Hope that helps.

[b_force]

Well, thanks, but not really what I am looking for, lol.

I am looking for just a discrete solution with npn/pnp transistors.
I don't need to isolate it all the way.

Like i said before. The bootstrap is all sorted out and working fine.
It's purely level shifting the 3.3V and GND to floating reference of the hi-side output.

[ajb]

The output is floating/connected to a symmetrical power supply (so the negative is not connected to ground but -V)
That makes the level shifting so much more a pain in the ass.

Output voltage is ±50V.

(most articles are about the fact that the output is connected to GND, which is easy)

I'm confused, what's your actual topology here?  Negative input->negative output?  Positive input->negative output?  Symmetrical input->two-quadrant output?

Unless you're working on something with unusual voltage/current requirements where you have no other choice, I have a hard time imagining an engineering case for NOT using an integrated solution at least for the switch driver.  Sure, you might save a bit of money, but at the expense of board size, much higher component count, and more difficult layout--especially when you consider the niceties that come free with an integrated bridge driver, like UVLO and cross-conduction protection.

That said, a basic level shifter for a high side switch might involve using a ground referenced current sink to turn on a PNP to pull the gate up to Vbootstrap, and another ground-referenced current sink driving a current mirror referenced to Vbootstrap that sources current into an NPN that pulls the gate down.  If the supply voltages are particularly high you may want to incorporate a darlington stage so that you can reduce the amount of current required in the level shifter stage.  The main trick (with any method) is to ensure that the slewing of the switching node doesn't interfere with the operation of the level shifter and potentially turn on the high side switch at the wrong time, leading to shoot-through and cooking the bridge.  Another option could be to capacitively couple a differential pulse to set/reset a latch referenced to the high side switch--using a differential signal will reject the slewing of the switch node, again to prevent misbehavior.  And finally, there's the transformer option, which would eliminate the need for a bootstrap voltage, but is probably not going to be any cheaper than using an IC. 

[jbb]

The SI8233 is available in SOIC16 for NZ$2.40 from Digikey.  It's small, cheap & well-behaved. Depending on what you order, it even has dead-time built in.  See attachment for how simple the circuitry is.

As ajb says, gate drive transformers can also be quite effective.  However, there are important considerations about duty cycle, so careful design is required.

[rob77]

and what's wrong with half-bridge driver ICs ? they usually go up to 600V diefference between low and high side no problem.. level shifting is handled in the driver IC and you are either provided with single input (high switches high side, low switches low side) or with 2 separate inputs for low and high side but the chip handles the minimal dead time internally.

[b_force]

$2.40 cheap, are you kidding me?

I am very familiar with gate drivers. So thank you for the offers, but I have already looked at every thinkable option.
There are cheaper (and better) ones on the market.
irs2011 for example.

But that's still to expensive for what I am looking for.
Could also be done with a bunch of transistors.

That was also my question, how to do it.

[rob77]

you can have half bridge driver chips under 1euro from IR (now infineon as well).
a discrete solution is NOT cheaper... the extra PCB footprint will negate the lower cost of parts. most probably it will be even more expensive at the end of the day.