bandgap voltages

The AVR internal bandgap is 1.1V. So it should be possible with a discrete.

Possibly the band gap of silicon (1.12eV)?

A Ge one should be possible at around 0.7V.

The reference in the LM10 is about 200mV but it requires a higher voltage to power it (about 1.1V).

A (standard) straight vbe + NkT/q stack always gives around 1.25V ±

However, if you have a low-VDD/low-current requirement (as I often do), you can make them anything you want by doing current mirror tricks.

But there's still the limitation that it takes around 1.4V VDD to run it.

...Jim Thompson

Is the reference at the bandgap because of some fundamental physical reason, though, or just by chance?

That's where the Vbe drift cancels out the PTAT drift. It varies a bit with the current density in the Vbe.

Cheers,

Phil Hobbs

Not by chance.. but it's not exact. There's a simple derivation in this presentation:

As i understand it, yes; it is the "definition" (actually an un-achieveable limit point) that "sets" that value. Mr Pease has an elegant explaination.

One of my guys just pointed this out to me:

I wonder how they do it.

John

Interesting. The usual parabolic tempco curve for a bandgap is inverted. Looks like it would overcorrect a bit if you put it in series with the output of a TL431 to get 1.5V.

Best regards, Spehro Pefhany

Maybe something like US6380723...

That appears to describe 3-terminal refs that need about 1.2 volts to power them. The ADI thing is a 2-terminal, 1-volt shunt reg.

CMOS? SiGe? Some new trick?

They may have a patent, too.

John

The noise number makes it look like a bandgap reference based design.

Perhaps, but I doubt that Analog Devices has that particular configuration.

Note the 100uA _minimum_ current... there's a lot of circuitry in there.

"Shunt" allows the lower voltage operation.

...Jim Thompson

I'm surprised no one's mentioned the LM10 -- 200mV:

One of the Bobs wrote it up somewhere, I think.

Cheers, James Arthur

James Arthur a écrit :

I think Spehro did.

Ahh, right you are. Missed that. Thanks for pointing it out.

Both Bobs wrote it up, actually. Widlar's account is here:

Cheers, James Arthur

Minimum supply is 1.2V

I can make a 100mV reference if you want (at the same minimum supply :-)

...Jim Thompson

Ask the Czar of Bandgaps:

There are ones that are lower, 0.8V seems popular with low output voltage regs, like the MAX1806 I'm using at the moment.

Dave.

But that needs a 2.25 volt minimum supply. It's easy to make a 1.25 volt bandgap reference and then divide it down, to microvolts if required. I'm interested in lower-voltage 2-terminal shunt references.

John