Suggestions for DC-DC converter topologies with variable output polarity.

[grg183]

I'm working on a DC-DC converter design with the following requirements:
- 48VDC input
- single constant-voltage output that is settable to a value between -12VDC to +12VDV
- about 12A max output current

The main challenge here is that I need to be able to set the output voltage (via some user/mcu input) to any value in the range -12V to +12V. During use, the output voltage setting will be varied (slowly or at intervals) anywhere in that range, so I cannot pre-configured it to a single polarity.

The simplest approach would be a buck converter followed by a full h-bridge circuit which is not a bad idea considering that the losses in the bridge would be minimal as the FETs are not switching and overall, it is a simple design. However, I wonder if there is any better topology, maybe using fewer FETs, more compact, cheaper, ...

Any suggestions?

[JohnG]

Are you driving a floating load? If not, better think hard about the H-bridge.It probably won't work if both input and outputs have the same ground.

If the load is not floating, you will likely need a transformer or perhaps a Watkins-Johnson converter. IIRC, there is a 4-switch version that uses only a single winding inductor.

John

[grg183]

Good point, yes the load is completely floating.

[JohnG]

Good point, yes the load is completely floating.

You could just do a H-bridge buck converter, then. Fewer switches.

Best to use an inductor in each leg, since you don't want the switching waveform going out to the load or you will be in EMI hell.

John

[Wolfram]

A synchronous buck running from a split rail is also an option, using only two switching devices. This one gives four quadrant operation, with the load referred to GND, with the caveat that you also need a negative rail to power it.

[David Hess]

A synchronous buck running from a split rail is also an option, using only two switching devices. This one gives four quadrant operation, with the load referred to GND, with the caveat that you also need a negative rail to power it.

That is what I was thinking.  Generate a negative supply from the 48 volt input and then use a synchronous buck converter to generate a bipolar adjustable output.  Some ATE pin drivers work this way and I think Linear Technology made a switching controller to do exactly this, but it relies on having a positive and negative supply.

Apparently they thought of this exact application:

https://www.analog.com/en/products/lt8714.html

[JohnG]

Interesting, that is a very neat chip and converter. Thanks!

John

[nctnico]

Indeed but be aware that the feeding PSU must be able to sink power for 4 quadrant operation. Otherwise the circuit is useable for 2 quadrant only (still not bad though).

[grg183]

You could just do a H-bridge buck converter, then. Fewer switches.

Best to use an inductor in each leg, since you don't want the switching waveform going out to the load or you will be in EMI hell.

I think this is a good idea, it removes 2 FETs and adds an inductor, however it also removes the need for high-side driving hardware for the H-bridge FETs as the switching ones can be powered through bootstrap.

[grg183]

A synchronous buck running from a split rail is also an option, using only two switching devices. This one gives four quadrant operation, with the load referred to GND, with the caveat that you also need a negative rail to power it.

That is what I was thinking.  Generate a negative supply from the 48 volt input and then use a synchronous buck converter to generate a bipolar adjustable output.  Some ATE pin drivers work this way and I think Linear Technology made a switching controller to do exactly this, but it relies on having a positive and negative supply.

Apparently they thought of this exact application:

https://www.analog.com/en/products/lt8714.html

That's a really interesting chip, if only it were in stock
Not sure I would want to add the complexity of having a negative rail for this application though.

[David Hess]

A synchronous buck running from a split rail is also an option, using only two switching devices. This one gives four quadrant operation, with the load referred to GND, with the caveat that you also need a negative rail to power it.

That is what I was thinking.  Generate a negative supply from the 48 volt input and then use a synchronous buck converter to generate a bipolar adjustable output.  Some ATE pin drivers work this way and I think Linear Technology made a switching controller to do exactly this, but it relies on having a positive and negative supply.

Apparently they thought of this exact application:

https://www.analog.com/en/products/lt8714.html

That's a really interesting chip, if only it were in stock
Not sure I would want to add the complexity of having a negative rail for this application though.

With that chip no negative rail is required, but unfortunately that is the only one.