It is possible to configure a buck converter with a negative output?

[flote21]

Hello!

I am facing a power supply design where the requirements are:

Vin: 30V - 36V
Vout = -10V
Iout(max) = -5mA

For convinience I would like to use the followinfg buck converter: LT8607HMSE.

However the darsheet does say anything about negative output voltage.

I am wondering if exchangin the GND by the Output, I can get a negative voltage. See the following pic example for getting -5V0.

What do you think? Is it correct? Or should i have some issues?

Thanks in advance.

Greetings.

Enrique Perez

 

[KlausST]

Hi,

For sure you are free to connect your load reversed.

But usually input_GND is the same as output_GND, which isn't the case for your solution.
It depends on your requirements.
Neither we nor the datasheet can know.

Klaus

 

[flote21]

Hi!!!.

I have running the following connection and it seems it is working:

But as you said. I need to be careful with the GNDs....Any other issue on the previous simulation?

Thanks in advance.

 

[KlausST]

Hi,

I'm not sure about C2 and C6 referenced to _not_ the LT8607_GND.
I'd rather use LT8607_GND.

Klaus

 

[FvM]

Buck converter IC vendors published application notes about operation for negative output.
This points need to be considered at least:
- total voltage rating must be kept. Vin = 36, Vout = -10 e.g. exceeds recommended 42V Vin of LT8607
- stability might be affected. National/TI e.g. suggested an additional capacitor between Vin and Vout in an old application note.
- soft start parameters may need adjustment

 

[flote21]

Hi!!

Thanks again for your feedback. finally I have selected another buck converter with higher input voltage: LTC8618. Same working principle of the previous one but with higher input voltage. See below the simulations.

However I am wondering about how to do the routing of the GND. Because my design has only one GND plane and all the devices share the same GND. Could I have some problem on the GND plane when I am doing such of connection?

Thanks in advance.

Greetings

 

[dick_freebird]

For a true, negative buck you'd be working from a negative
supply.

I've done one chip design for a -48V from +5V source, a
negative boost architecture. It wasn't that bad, just some
odd compromises in order to use regular JI BiCMOS; the
feedback reference begame GND and the feedback return
was the 1.25V bandgap (buffered for mA instead of uA).
An error amp that doesn't freak out or go limp when
feedback input is below GND. Other than that, pretty
normal guts.

 

[Easy peasy]

The LM2575 data sheets show a similar way to get a neg o/p for a pos input - using essentially the ... BUCK-BOOST ( or flyback ) approach of store and release in the choke.

 

[FvM]

However I am wondering about how to do the routing of the GND. Because my design has only one GND plane and all the devices share the same GND.

The -5V node is not GND. It's a local common node of the converter, connected e.g. to LT8618 enhanced pad. It may be implemented as a power plane zone or a bottom side copper pour.

 

[flote21]

Hi thanks for the replies.

There is an evalkit of the LT8618 with the negative functionality (See attachement). On the schematics of the evalkit it is possible to appreciate the negative output:

I tried to run a simulation following the schematic, but at the beginning there is a positive peak of around 1.5V. Is it normal? If I want to make it zero, how can I do it? With a TVS diode?

Thanks in advance.

Greetings

 

[FvM]

Positive peak is expectable due to capacitive voltage divider. Clamping it with a schottky diode should be sufficient in most cases.

 

[Easy peasy]

expectable => expected.

 

[BradtheRad]

Vin: 30V - 36V
Vout = -10V
Iout(max) = -5mA

It hasn't been clear whether you need this negative supply referenced to the same common ground as your other devices? If so then you really need a converter that inverts polarity of the positive source.
Here's a popular method based on a charge pump with diode directions reversed. Low power.

The clock pulses can come from any oscillator which tolerates 36V. (A plain 555 does not.) A classic astable multivibrator is suitable using two transistors.

 

[flote21]

Hi!
Yes. The negative power supply must share the GND with the other positive power supplies. The is the reason because I am not sure if the approach with the buck converter could cause some issue in the GND plane....
To be honest your charge pump approach looks a smart way to get the -10v. But i am wondering if there is not any commercial solution based on that....
Greetings

 

[FvM]

I am not sure if the approach with the buck converter could cause some issue in the GND plane....

What do you mean?

 

[BradtheRad]

because I am not sure if the approach with the buck converter could cause some issue in the GND plane....

Buck-boost is the particular type which inverts polarity of the incoming supply. Which means you're not looking for a buck type, properly speaking.
Easy's post #8 refers to a data sheet which tells how a buck IC might be arranged to act as a buck-boost.

 

[flote21]

Hi again.

So it means that the simulations i run with ltspice on post #10 are wrong and on the real world, this buck converter will never output -10V?

Greetings

 

[FvM]

So it means that the simulations i run with ltspice on post #10 are wrong and on the real world, this buck converter will never output -10V?

No, you can trust the simulation in post #10. The converter is operating as inverting buck-boost here. Capacitor values of C1 and C7 as well as additional input filters are an object of detail optimization, but the circuit should work as shown. I've used different buck converter ICs in this configuration.