Buck switcher chips

Have you checked the ON Semi site? It could be that the LM2594 is being discontinued.

Use an LM339 and be safe ;-)

Reminds me of a design of mine that produced both +5V and -12V...

What are the relative loads on your +5V and -15V? ...Jim Thompson

I looked on the ONsemi site, and they're discontinuing the DIP package but not (so far) the SO8.

The +5 load is relatively light--it's running a 3.3V LDO for a small ARM with a bit of external flash and sram, some optos, a display, and an FDTI chip. So maybe 200 mA.

The -15 can be at least that, because there are beefy op amps driving series-terminated 50 ohm BNCs that somebody might decide to terminate in

Cheers

Phil Hobbs

My parasitic scheme only works with loads on the negative supply that are lower than that of the +5V. In my case I had ~5A on the +5V and ~100mA on the -12V. ...Jim Thompson

We use LM2574/SO14 and LM2576HV/TO263. Nice and slow.

I tend to go the other way! Higher magnetics frequencies are easier to filter and are attentuated more by groundplanes. The fall in PSRR of the opamps gets compensated by the improvement in local filtering.

Works best when signals are relatively slow I suppose, slower that the SMPS.

We've seen fast switchers, things with a couple ns edges, spray spikes all over a board, through paths that aren't visible on schematics. Synchronous switchers are especially evil.

I recently spun the layout on this

to reduce switcher noise, for a customer who is doing really sensitive RF stuff. The major bad guys were LTC3411s.

And one of my favorite waveforms, from an LM3102. The substrate diode step-recovery behavior is clear.

This freaked out opamps clear across the board, with no obvious connection to this switcher. We're probably talking something over 1e9 amps/second here.

There's always the old MC34063, almost dates back to Methusaleh, which can be had for slightly above $0.10 in qties:

The BD9876 is around $0.50, more efficient, but you'd have to see if ok for the inverter job:

The others I have encountered are mostly over a Dollar. If you can stomach an external FET there may be some further penny-pinching possible in case the MC34063 is not to your liking.

I see that there's a 150 kHz version of the 34063 as well, the NCP3063, which might be a win at 37 cents in hundreds. The switch burden voltage is quite a bit higher than the TI part, and it has an external current limit, which is obviously more accurate, but bumps the switch drop up to

(I'm quite fond of the LM2594--I still wish I knew why the On Semi one is half the price.)

Cheers

Phil Hobbs

The current sense only adds 200mV. The switch drop is a pain but you'll have that in almost almost all slow and old regulators.

One way to find out would be to ask a TI app engineer :-)

That is elegant to a cruel level.

?-))

Glad you liked it >:-}

That design (actually using LM339's :-) is in several of the OmniComp/GenRad portable testers. ...Jim Thompson