super LDOs

I did a Stratix-4 design with 19 including the VTT supplies. Sheesh!

When I was kid, 5V and sometimes +-12V was all we needed.

Bob

Or abject insanity.

Good Luck! Rich

I thought you were allergic to LDOs. Buck down from a higher voltage?

BTDT, no way around it sometimes. That many supplies (most switchers) give me sleepless nights just before compliance testing, too.

BTW, I'll be taking the board with the 13 ADI Iso-Power thingies up for compliance testing in the next couple of weeks. Rev-01 (pre-RevA, they had me do a prototype pass for possible EMI issues) worked out of the chute (13 of 14 LEDs and the JTAG connector got put in backwards, but...).

No, that's Joerg. I'm just very, very wary of them.

I already have a mess of LTM8023s in the power supply. One of them makes 1.5 volts, so I figured I could linear regulate down to 1.2 and

Yikes. That will sing like a flock of birds.

John

There are 13 of them because it's one isolated input channel and twelve isolated outputs. It's going to get big, and expensive, if I have to build an isolated supply for each one.

That's why management gave me a "free pass" with the first rev. ;-) I'm hoping I did a decent job of containing it, though. We're not planning on selling the widget for another six months, at least, so there is time to fix it.

I did a board with 13 ARM processors, 12 of them in isolated channels. I used little SIP dc/dc converters to power the channels.

The SIP supplies are the black things. About $4 each. They switch at about 60 KHz, so the EMI probably isn't too bad. I did use the ADI digital isolators to talk SPI to the isolated CPUs.

Since each channel has +30, +15, and +3.3 supplies, that's 36 rails. Add the ones on the main VME board, and I'm up to 42 or so.

John

me

I've seen DIP isolated supplies that'll work but space is a problem. Got a pointer to the parts?

The ADI ISO-Power modules are about $5 each but that includes both the isolated supply and the logic isolators (two on the outbound side and an RS422 transceiver for the receiver). The board sits vertically, hanging from the thirteen DE-9 connectors, on a 1U chassis, so the whole plan will have to change if I use separate isolators.

Yikes! The isolators are 3.3V and you boost from there? How much isolation are you talking about?

me

Digikey 102-1344-ND.

I did a bunch of measurements on these. They're really pretty good. Rated at 1 watt and perfectly happy at 2. The only problem is the lack of output short protection.

Oh, much worse than that.

ftp://jjlarkin.lmi.net/22S220A_20.pdf

John

I've been using the TPS74801 LDO in the QFP package for that sort of thing. A little over $1 in volume. The package is small, but the exposed ground pad allows for good thermal performance.

Dropout voltage at 1A will be about 100mV, but the line regulation isn't all that good until there's about 300mV across the n-channel pass fet. This should be quite ok for your 1.5 to 1.2 or 1.1V applications.

Regards, Allan

Correction: I meant the QFN package, not QFP.

Nice part. The nfet source follow topology is what makes it nice and stable. Thanks.

John

give me

an

Interesting. I found some others that are single supply. A bit more expensive, strangely enough. The efficiency of these things is pretty poor, though. Not quite as bad as the ISO-Power things, but efficiencies in the 70s are pretty bad, these days.

At .100" centers, I don't buy their "isolation" numbers, either.

I think I could live without that.

Not all that bad. Isolated 12V switching to +/-15 with a linear to 3.3V. I suppose your 3.3V needs are minimal.

a search on analogs site finds a few LDOs that could work, most (all?) with a seperate +5V bias supply

-Lasse

give me

an

I measured around 85% betwen 1 and 2 watts out.

The other sheet of each channel has an ARM processor and the ADI data isolator. We tweaked the ARM clock rate to keep the power down.

I like these SIP dc/dc things. I use them a lot.

John

give me

build an

fix

The specs were in the low 70s. If the ADI things bomb out in compliance maybe I'll go to something like that. As I said, size will be an issue, though. Or I'll have to rotate the board, but that'll cause other problems, I think.

RS422

I'd have to go with an SMT version and they look to be pretty big (.5"x.5", or so).

You might like to consider other (almost) pin-compatible parts in the same family that have a 3A capability (e.g. the '401), or a 1% tolerance, or a lower Vdo (the '301), or a tracking input (to replace the soft start feature). The '801 is the cheapest though.

Make sure you get genuine ones:

What really surprised me was the comment from a TI employee that they use EPROM inside these things.

Regards, Allan

If you can do without independent current limiting, just use an op amp and PNP bipolar transistor. Collector current 1.5A means base current of maybe 30 mA, an LM358 can handle that. The transistor only sees the 1.5V input, the op amp will need a higher bias supply, though. Compensation and bypassing are left as an exercise.

The MOSFET as pass element is... not a big improvement.