LM117-type bandgap: Better for matched currents or matched densities?

[Ted Pavlic]

I teach an electronics lab to undergraduates, but I'm a control
engineer and so sometimes I feel a little out of my element. In one of
our labs, we discuss several different voltage regulators, and I put
together a document discussing the bandgap mechanism in the classic
LM317:

http://www.tedpavlic.com/teaching/osu/ece327/lab3_vreg/lab3_vreg_lm317_example.pdf

In that document, I discuss the method National Semiconductor uses --
it uses a current mirror to push two identical currents through
transistors that have different emitter densities. I also mention that
National *could* have made its mirror with different emitter densities
to setup ratio-ed CURRENTS through transistors with the same emitter
densities.

It seems like every example I've found uses the choice that National
used for its LM117. Is there a reason for this choice? Is there a
reason why everyone seems to pass a matched current through ratioed
densities instead of ratioed currents through equal densities?

Thanks --
Ted

 

[[email protected]]

I teach an electronics lab to undergraduates, but I'm a control
engineer and so sometimes I feel a little out of my element. In one of
our labs, we discuss several different voltage regulators, and I put
together a document discussing the bandgap mechanism in the classic
LM317:

http://www.tedpavlic.com/teaching/osu/ece327/lab3_vreg/lab3_vreg_lm31...

In that document, I discuss the method National Semiconductor uses --
it uses a current mirror to push two identical currents through
transistors that have different emitter densities. I also mention that
National *could* have made its mirror with different emitter densities
to setup ratio-ed CURRENTS through transistors with the same emitter
densities.

It seems like every example I've found uses the choice that National
used for its LM117. Is there a reason for this choice? Is there a
reason why everyone seems to pass a matched current through ratioed
densities instead of ratioed currents through equal densities?

Thanks --
Ted

Not sure Ted, probably ease of fabrication and better thermal
tracking. Ask the chemical engineers.

 

[neon]

On Sep 17, 2:46*pm, Ted Pavlic <[email protected]> wrote:
> I teach an electronics lab to undergraduates, but I'm a control
> engineer and so sometimes I feel a little out of my element. In one of
> our labs, we discuss several different voltage regulators, and I put
> together a document discussing the bandgap mechanism in the classic
> LM317:
>
> http://www.tedpavlic.com/teaching/osu/ece327/lab3_vreg/lab3_vreg_lm31...
>
> In that document, I discuss the method National Semiconductor uses --
> it uses a current mirror to push two identical currents through
> transistors that have different emitter densities. I also mention that
> National *could* have made its mirror with different emitter densities
> to setup ratio-ed CURRENTS through transistors with the same emitter
> densities.
>
> It seems like every example I've found uses the choice that National
> used for its LM117. Is there a reason for this choice? Is there a
> reason why everyone seems to pass a matched current through ratioed
> densities instead of ratioed currents through equal densities?
>
> Thanks --
> Ted

Not sure Ted, probably ease of fabrication and better thermal
tracking. Ask the chemical engineers.

I am afraid a chemical engineer would not help you there. I am afraid that is physics field not chemestry.

 

[[email protected]]

I teach an electronics lab to undergraduates, but I'm a control
engineer and so sometimes I feel a little out of my element. In one of
our labs, we discuss several different voltage regulators, and I put
together a document discussing the bandgap mechanism in the classic
LM317:

http://www.tedpavlic.com/teaching/osu/ece327/lab3_vreg/lab3_vreg_lm31...

In that document, I discuss the method National Semiconductor uses --
it uses a current mirror to push two identical currents through
transistors that have different emitter densities. I also mention that
National *could* have made its mirror with different emitter densities
to setup ratio-ed CURRENTS through transistors with the same emitter
densities.

It seems like every example I've found uses the choice that National
used for its LM117. Is there a reason for this choice? Is there a
reason why everyone seems to pass a matched current through ratioed
densities instead of ratioed currents through equal densities?

Thanks --
Ted

This is really an interesting question. Maybe a Monte Carlo analysis
might yield a difference. Generally if I had to do a precision current
multiplier, I would use some emitter degeneration due to VBE mismatch.
Maybe the current mirror accuracy is more sensitive to component
matching than the N:1 in the NPNs.

 

[Ted Pavlic]

1:1 Matched currents are easy to obtain with accuracy and stability.

1:N is not.

Okay. But I'm guessing that there are similar 1:N-type problems with
the delta_(VBE) generator. Right?

Is it the case that the problems you'd have with generating the
matched currents are going to be more than the problems you'd have
with generating the ratioed current densities?

Why do you talk of Q17 and Q19, but call them PNP's?  They're NPN.

That was a typo. I wrote the second part before the first, and I
copied and pasted and evidently didn't catch everything. I've made the
fix and updated the live docs. Thanks for pointing that out.

Do you know Todd J.... ?  We are collaborators.

I don't think so. Is he a buckeye?

When I'm not teaching undergraduates about circuits or control, I'm
usually collaborating with behavioral ecologists on bio-mimicry
projects. Do any work with leaf cutter ants?

Thanks for your help --
Ted