Efficient current limiting for a dc/dc SMPS?

[audiodane]

I'm designing a dc/dc SMPS (see other thread here), but wanted to ask a very specific question and thought a new thread would be appropriate..

In my dc/dc converter, minimum input voltage is around 11-12Vdc, and so I'll be drawing around 9A continuous worst case..

I would like a 11-12A current limit circuit, but I suspect a thermistor limiter is NOT the way to go because they have too high a resistance even when on. (even 0.5ohm at 12A is wasting 6W of power!)

I wondered if there were some other more efficient limiter circuits that anyone was aware of?

thanks,
..dane

 

[sawreyrw]

The LTC3780 has a built in current limiting function. Read the data sheet carefully to understand how it works and if it meets your requirement.

 

[audiodane]

Yup-- when configured as designed (buck/boost), yes, it limits the current nicely. However I have modified the circuit to essentially be boost-only, which cuts out the MOSFETs on the BG2 and TG2 gate drivers, and ties the inductor straight to VIN. The safety catch diode across the MOSFET connected to TG1 then effectively creates a short between VIN (through the switching inductor and the safety catch diode) to the Vout caps. In the case of initial connection to power with Vout caps fully discharged, this poses a problem...

Reason I went boost-only is that I don't really NEED boost-buck, but I DO need a synchronous drive system that can generate 30Vdc out without transformer isolation requirements. None of the LTC products that I saw in the boost-only category that supported my voltage in/out and current in/out requirements also supported synchronous drive mode for increased efficiency. This will be operating in an already HOT ambient environment so I want to add as little heat as possible..

I've found a Motorola appnote that talks about mosfet low-side inrush limiting.. I've made some adjustments to the circuit for a high-side inrush limiter... currently simulating it on the front-end of my system..

thanks!
..dane

 

[darkfenriz]

You can have a thermistor paralleled with P-channel mosfet, which you can open after inrush.

 

[audiodane]

pchan mosfets have higher RDSon resistance than nchan's, and given my high input currents (8.5Arms), I'd rather use an n-chan. Additionally I read that thermister based limiters are spec'd to operate at 25C and we're potentially operating in an 85C environment already, which could pose some problems for thermister based approaches.

I've got an interesting solution in place now, just doing some more testing on it ...

..dane

 

[audiodane]

Looks like the LTC3814-5 might be a good fit and is a boost-mode synchronous driver.. will have to investigate it also.

..dane

 

[sawreyrw]

dane,

I now understand that you are concerned about limiting the current during start up of the boost mode controller. A practical way to do this is to use a soft start circuit. This consists of a resistor in parallel with a low resistance MOSFET, both of which are in series with the input circuit. After the output charges to nearly the input voltage level, the MOSFET is turned on. You could use a time delay as well.

Rick

 

[audiodane]

dane,

I now understand that you are concerned about limiting the current during start up of the boost mode controller. A practical way to do this is to use a soft start circuit. This consists of a resistor in parallel with a low resistance MOSFET, both of which are in series with the input circuit. After the output charges to nearly the input voltage level, the MOSFET is turned on. You could use a time delay as well.

Rick

Bingo.. I've got a solution in place now fairly similar to your suggestion. I found an old motorola appnote on the technique and ran with it.. Using the PSMN2R0-30YL ... nice part, and fairly inexpensive, too!

thanks!
..dane

 

[Electrone]

Good that you are getting it worked out. I current regulate my projects with a very low value sense resistance like .005 ohms. It is just a wire, often one of the power supply wires with a sense return wire twisted around it. alternately, the sense wire can extend from the circuit and twist back on itself if that is more convenient.

The sense circuit is a transistor with the emitter connected to one end of the sense resistor and a 1N4148 diode connected to the other end. The other end of the diode goes to the base of the transistor. A resistor like 100k biases it about 60mV below the turn-on threshold.

When the wire drops voltage as it carries more current, the diode lets the voltage on the base reach the conduction point, turning on the transistor.

 

[star882]

You can connect a pair of sense lines across a length of wire and then connect it to a high gain instrumentation amplifier to use the wire as a current sense resistor. Or use a Hall effect current sensor.

 

[audiodane]

elecctrone and star882-- in both of those cases you still need some sort of 'disconnector' or 'slowdowner' device.. I'm using an nchan MOSFET since they have much lower RDSon resistance. pchan would have been easier (as would an NTC thermistor) to implement but the pchan's with the least RDSon (that I could find, at least-- from reputable MOSFET manufacturers) is still at about 9milliohm and the MOSFET costs around $3 in low quantities... (NTC thermistors are even worse from what I could find) The nchan I found (PSMN2R0-30YL) has an RDSon of around 2milliohm when fully on, and costs under $2 in low quantities. Price isn't a huge deal when you're only building a few, but the best pchan having over 4 times the RDSon resistance can be a deal killer.. I am targetting >90% efficiency at full load, so any little power loss (diode drops, transistor drops, etc) matters.

LTSpice simulations are currently reporting 93.1% overall efficiency at full load with the nchan limiter solution that I have in place..

cheers!
..dane

ps.- I'm pulling off a DC input, not AC mains.. that might be different from many of the folks' designs here..