TL431 supply current/gain Q

That is to say one can build a multiple-feedback active filter out of one in a pinch, but I have noticed it is kind of a suck one. As you might expect...

I would ask you to run the math, but you can't cope with loop and nodal analysis >:-}

I have used TL431's as "one-legged" OpAmps for years, like since the late '70's... very nice! ...Jim Thompson

TLV431s and LM4041-ADJs have much lower anode current when not in regulation than the vanilla TL431. The 4041 is especially useful because it works the other way up, i.e. it holds Vref between K and ADJ. That makes it just the thing for negative regulators--it saves you a transistor. Only goes up to 10V, unfortunately.

The adjustable LM385 is like the 4041 but with higher anode current and only goes up to 6V, weirdly.

Cheers

Phil Hobbs

Reconsider that statement. ...Jim Thompson

-- | James E.Thompson | mens | | Analog Innovations | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | STV, Queen Creek, AZ 85142 Skype: skypeanalog | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at

I'm looking for work... see my website.

Thinking outside the box... producing elegant solutions.

Yep. I had some "circuit-dyslexia." The diode-connected transistor is on the left, not on the right!

Right transistor is the gain stage with the right PNP acting as active load, and the 20p wrapped around it from the output for compensation.

The diode connected transistor to the "REF" pin pulls current out of the current source when the "REF" input falls below the threshold to keep the comparator snappy (or as snappy as a diode-connected transistor can be, I guess)

Pretty sure it's older than that--the non-adjustable 1.2V version is in the 1980 National Linear book. Maybe they weren't initially intending to make other voltage options, and the collector doping is way too high or something like that.

Cheers

Phil Hobbs

It doesn't look too bad now that I consider the circuit more carefully.

There is a DC feedback loop around the band-gap that keeps the current through the pair constant and the defines the reference voltage at the emitter of the transistor connected to the "REF" pin. If everything is balanced, ideally no DC current flows through the 4k resistor (which would be bad as its emitter is grounded.) The collector of the DC feedback transistor connected to the 4k resistor is at virtual ground for signal.

If the voltage on the "REF" pin increases the current mirror enters conduction and sources current into the transistor connected to 1k. But the base of that transistor is connected to a stable reference and can only sink so much, so whatever excess flows into the output Darlington to pull the load down. Vice versa if the reference pin voltage drops below the virtual ground, the gain transistor will sink current out of the base of the Darlington and shut them down.

I will attempt to present an anal ysis later...;-)

My guess, before an actual analysis, is that the 1K is just there to prevent a current-hogging problem during start-up. ...Jim Thompson

The time constant in the equivalent RC circuit of the DC feedback loop and capacitor might be enough that in some situations (like PWMing a light load) the band-gap is staying continually active, but the current source is starting up and shutting down over and over again, putting the full unregulated current through the gain transistor. Could get toasty, maybe...

"Toasty" would be determined only by the load in the "cathode" path. ...Jim Thompson

Comparator is the wrong word. ...Jim Thompson

Investigate your theories...

includes subcircuit library.

So just draw up what you want to test in LTspice and let 'er rip.

Nodes are labeled allowing you to probe the innards. ...Jim Thompson

Cool. I think I have a TL431 Spice model in my files, but IIRC it is behavioral and "Not-ah so good..."

What's the ratio of the emitter areas of Q4 to Q5 in the chip?

Measure it yourself :-)

Lots of great die shots on there -- excellent site!

Tim

It depends. It's quite possible to write a behavioral model that duplicates the actual device-level model, but converges easily and simulates faster... I do that all the time ;-)

Learn to read netlists... all NPN's are unity except Q10, which is 5X. ...Jim Thompson

I just save the plots of the chips I design, easier to see what's going on...

Lousy photo, plus the plastic overlay has yellowed, need to reframe a whole bunch of these as I redo my office and library decor. ...Jim Thompson

"Jim Thompson" wrote in message news: snipped-for-privacy@4ax.com...

Gee, sure must be nice. . .

But what about the rest of us? And for that matter, how much of that stuff is out of NDA, or copyright? -- Is it ever public? Are you violating anything by posting that..? (Not that there's necessarily anyone left to do so, if it's obsolete.)

Tim