LDO problem

the

Yeah, anything that goes all the way to +12 seems to be a nuisance. For a feature that I'm not sure anybody wants.

ftp://jjlarkin.lmi.net/LDO_ef.jpg

No, each pullup will probably be a 1K resistor in series with a small schottky diode. I'll just fudge the pullup by about 0.3 volts to make up for the diode drop.

Around 350 k/w is a more usual number for SOT23s. What part is it?

John

That could zener the B-E junction.

Best regards, Spehro Pefhany

Add a diode in series with the emitter?

Yep. If he's truly not worried about someone else trying to pull it too high he could put a diode e-to-b. That saves a diode drop.

But, then a goon with an external widget could try to pull the pin to

-- Cheers, James Arthur

I meant it might be a problem if the user has a pullup, like if this were programmed as an open-drain output and he had a device to +12v. (You said 'i/o', which could mean bidirectional, i, or o.) That would back-drive your NPN. Same applies if it's an input.

It's a FET. The quoted 60K/W theta is to the solder point, but that's still pretty hard to believe.

I'm only dissipating about 200mW w/c, so no problemo for me.

James

Not in my application. Each output is an open drain to ground (TPIC6595) and a 1K pullup with a schottky in series. The user will be able to program the pullup rail from 0 to +10 (+10.3, actually, to fudge the schottky drop). So this "regulator" doesn't need current limiting or reverse protection.

I guess it might zener the b-e junction for a millisecond or so, from the output cap holdup, if the user changes the output setpoint from high to low. Interestingly, LT Spice shows no negative base current even when the output slams from +10 to zero; the b-e junction is indeed back-biased 10 volts for a while. I tried a 2N2219 and a 2N3904 too, but neither zeners at -10 Vbe. Strange.

The other strangeness is that this simulates in about a second, but if I add a base resistor it takes hundreds of times longer. LT Spice has a max time step parameter, but doesn't seem to have a min time step.

We could slew limit the dac value, or just use a big lowpass filter if we go PWM or delta-sigma, so we don't drop the input signal faster than the output can track.

John

I'll put a small schottky diode in series with each pullup resistor, so they can't back-drive my pullup bus.

They learned everything they know from IR.

Good. That spec is silly.

John

John Larkin a écrit :

The spice GP BJT models doesn't support breakdown. You'd have to build a subckt with additional junctions for that.

Ah, now I see. I hadn't visualized the full setup.

-- Cheers, James Arthur

You can't bury it inside the feedback loop then. ...well, unless you use eight op-amps and NPNs. ;-) Hmm, maybe eight op-amps (two quads) and no follower? There are some cheap op-amps out there that will take some current.

;-)

se

d no

urrent.

Why not just add the schottky to the emitter, like you said first, and feedback around it? That saves a lot of schottkies and protects the transistor.

-- Cheers, James Arthur

That's probably what I'd do, but that doesn't keep the loads from scorching each other. Where's the engineering spec for this project?

Then the pullups that share the excition supply could interact.

John

current.

I'm still playing with that. This is the approximate circuit for one pin of 64...

ftp://jjlarkin.lmi.net/V250_chan.gif

I figured it might be cool to allow the user to program the pullup voltage from 0 to +10 or +12, in banks of 8 or maybe 16 pins. So I was considering the various LDO things to power the pullups.

Since I will have the diode drop, and the VME supply is +12, I can't honestly claim to get to +12 on the pullup. If I claim +10, I have enough headroom to fudge up to 10.3 and correct for the diode drop and still be fairly honest.

What would be great is a power, 100 mA or so, c-load r-r opamp. Sort of an LM8261 big brother.

Or an LDO with a built-in DAC.

John

current.

What are you using for the receiver (the triangle thingy)?

And lose the opportunity to do some engineering?

current.

Probably an HC14. There's debate over whether to add another resistor across the cap, to form a threshold divider. If the low end of that new resistor went to a negative voltage, we could even have an adjustable threshold.

One never knows whether a feature is worth it.

Sigh. We keep moving up the abstraction stack. Any good circuit eventually gets integrated. I used to design my own flipflops and opamps.

John

current.

What sort of HC14 will take 12V?

Guilding the lily, perhaps. I think I'd use a comparator, at that point. The threshold of most gates is pretty loose.

Sure, that's why I like VHDL. ;-)

current.

One with a big resistor in its input!

The state of the art in quad comparators appears to be... the LM339!

John

et

all

ake

ould

u use

and no

e current.

How? If the output node of the "LDO" is stiff, that shouldn't happen.

(I rejected the idea at first for your same reason, but then realized the above.)

James