OT: SOT-523 PNP Question

Yup -- still fine when a wimp is all you need. And in the grander scheme of things, 0.5mA is hardly useless.

Any clue how the Rth(j-j) is between them? Twin packs are notoriously weak, but I wonder if having common metal makes them any better, even if it's two bond wires.

Tim

OK, for that you need to charge a fee.

$0.xx for chalking a mark on a pipe. $XXX,XXX.xx for knowing where to make a chalk mark on a pipe...

I have some germanium transistors which act like really slow latches. Once turned on, they bias themselves at half supply voltage, go figure!

Tim

GE built some radios and phono amps with crap like that in the '60. No bias resistors, just the internal leakage. Probably fallout from the military & commercial product lines.

In the classic low-level audio amp stage bias scheme (divider to bias up the base, resistor in the source) the base divider often took current *out* of the base!

One occasionally saw an amp circuit with the base of a Ge transistor driven through a cap... and nothing else.

The Infineon ones seem to be a bit better--they claim somewhat higher maximum dissipation before instability.

Cheers

Phil Hobbs

The synchronized-runaway trick can be used to enable multiple transistor array chips to be used much like a single larger array. That's potentially quite useful in my business.

Cheers

Phil Hobbs

I believe that. CMOS tends to be crappy that way, but OTOH you can bandage it with a whole lot of auxiliary transistors.

Cheers

Phil Hobbs

Theta_JA isn't the pertinent number, it's Theta_JC (junction to case), or even more accurately junction-to-mounting-slug. What did you get you PhD in... exaggeration ?>:-}

More in a couple of days... off to Yuma to the inauguration of the son-in-law. ...Jim Thompson

Yep, that's the (golden) rule >:-} ...Jim Thompson

The best of all worlds is BiCMOS. But CMOS isn't "crappy"... if you know how to design with it. With my proprietary cascode I can easily get to 0.2V of rail before it sags. ...Jim Thompson

So you believe that heat disappears once it gets to the case? I'm talking about the real theta_JA in the real system, not the marketing number. You can't usefully write equations containing marketing numbers.

Cheers

Phil Hobbs

You're missing the point. I'll address the math on Sunday+ ...Jim Thompson

Maybe, but I don't think that I am. The two dice are not meaningfully coupled thermally--the lead frames are completely isolated, and the plastic isn't much better at heat conduction than air is.

Cheers

Phil Hobbs

Or two.

There's a speed vs. accuracy tradeoff, though. Slowish mirrors made with quick op amps are great, but trying to make a fast, accurate mirror that way is surprisingly difficult, mostly because the op amp needs a big input error to reach maximum slew.

Cheers

Phil Hobbs

One interesting possibility is to let the current mirror mostly do its own thing, but use an op amp to null out the average error in the emitter circuit via a voltage divider. Hmm, might be useful.

Cheers

Phil Hobbs

There are better ways, think charge pumps used in PFD's. ...Jim Thompson

In a built-up circuit? I can see it in an IC, provided the switching frequency is sufficiently far out of band. (There's a trick used in high power audio amps, to equalize the supplies: you ground one end of an inductor, and switch the other end from VCC to VEE. Any DC imbalance goes away magically, and the switcher doesn't have to supply the whole load.)

My interest is primarily in making accurate replicas of photocurrents for use in bandaging nonlinearities elsewhere in the system, e.g. Figure

I've been working for awhile, on and off, to make a noise canceller good to an honest 70 dB from DC to 10 MHz. The analogue accuracy and noise requirements are pretty stringent, and it has to work over about a

It simulates great, and I have some PDs that by my measurements should be adequate, but we'll see when the boards are done. My beautiful PCB hunchback and opera singer got chucked in the deep end on that one--over

Hopefully Firmware Hunchback and I can get the CPU board done soon, and it'll be a product this year. (I'm like butter spread over too much bread at the moment.)

Cheers

Phil Hobbs