+24 to +-15

I did something like that to make a couple of isolated 5V supplies--the 'flybuck' topology. Problem was that when the non-isolated positive supply was loaded less than the isolated ones, every sync buck chip I tried went into burp mode and the isolated supplies collapsed. That included the MCP16312, whose datasheet claims that it keeps PWMing down to zero load. The datasheet lies.

Cheers

Phil Hobbs

I don't think the little TPS chip burps, but I guess I should try it. I do have the eval board.

fredag den 11. maj 2018 kl. 17.43.23 UTC+2 skrev John Larkin:

use two ?

I did that with two of the CUI three terminal switching regulators.

I did have issues with the negative starting up and had to add a fet to turn on the +15V after -15V. (I also powered the -15 from the +15... 'cause reading the spec sheet the voltage drop was a little out of spec... though it did work fine from +24.) I blew at least one up... but I'd have to check my notebook to find out at what voltage.

George H.

George H.

We've done that, or used a Cuk to make the negative rail. I really like that little TI spread-spectrum switcher, but it's only good to 32 volts, so it can't do 24 to -15 in that mode.

A variation would be to use two of the TPS chips, one as a buck to +15 and another one to make -15, but some topology that doesn't put 39 volts across the chip and doesn't have huge peak currents.

I could maybe buck to +15, and use another TI chip to flip +15 to -15 as in your circuit. I could tolerate the efficiency hit.

Yeah, circuits like this tend to have all sorts of nasty hazards.

Unfortunately it does burp.

Quote page 9 "The TPS54302 is designed to operate in pulse skipping mode at light load currents to boost light load efficiency. When the peak inductor current is lower than 500 mA typically, the device enters pulse skipping mode".

Or use a ball and chain circuit in your original idea that starts to load down the main output when the -15V begins to sag.

I have +24 and need +-15 at a couple hundred mA. My regulation requirements aren't extreme (I'll be powering opamps in my capacitance/fuel level simulator) and the load will be pretty well balanced.

This might work:

I haven't seen this particular topology before, but it's unlikely to be original.

For better efficiency and cross-regulation, which I don't need now, D1 could be a p-fet synchronous rectifier.

The cute little TPS54302 soft-starts and current-limits, so the initial surge into C1 should be OK.

This might do goofy things at low loads... gotta Spice it. I don't have an LT Spice model for TPS54302, but something else will do for the sim. Maybe just a voltage source, or some similar LTC part.

I didn't see it burp. Maybe my peak inductor current never got that low; inductor current can be high in a synchronous switcher even at low load.

John, John, You put 2 devices in series just to reestablish a new dc level. Remove T1 and replace its primary with a filtering inductor. Tie C1 to A. Send me a check and modeling is for girls.

Cheers, Harry

I don't think that would make -15. It need inversion (the transformer) to flip the waveform and preserve the duty cycle.

Not as fun as DIY but you could always just plunk down something like this and move on with your design:

or

Yikes, 6 watts in a SIP for $10. That would work.

I was going to recomend the astrodyne modules. But those would work, as well as CUI MEGA etc. They are not 1% but if 5-10% regulation is all you need then whynot.

Cheers

John, John, You put 2 devices in series just to reestablish a new dc level. Remove T1 and replace its primary with a filtering inductor. Tie C1 to A. Send me a check and modeling is for girls.

Cheers, Harry

I don't think that would make -15. It need inversion (the transformer) to flip the waveform and preserve the duty cycle.

John Larkin Highland Technology, Inc picosecond timing precision measurement

jlarkin att highlandtechnology dott com

Picky, picky, ok..I thought you wanted - 9 V... So flip the tranny so its output is back in phase, toss the cap, and use the clamp diode as a negative peak detector into a small PI output filter. And I love your sketches, even without values.

Cheers, Harry

Sorry to be so late to this thread. Does your 24V input return have to be the same as your system ground? Else you could just boost 24V to 30V and then rail split a virtual ground midway (you did say the both rails had similar loads)?

piglet

Yes, the supply has to be grounded. And 24 volts is pretty standard... customers often have it, or we can supply a wart.

Since my load is going to be well balanced, a rail splitter would be OK.