Re: Pete Lefferts LED current source

JosephKK wrote:
On Tue, 28 Oct 2008 15:29:18 -0500, Phil Hobbs
<pcdhSpamMeSenseless@xxxxxxxxxxxxxxxxxx> wrote:


Laser noise cancellers can get up to 70 dB of cancellation (1 part in 3000 current error) with unmatched transistors running at very different current densities, which shows that BJTs follow the Ebers-Moll law very closely, until the extrinsic resistances become important. If your error is as much as 1%, you're not getting the performance available from your 2N3904s.

Ebers-Moll is not a law but merely model of how a transistor works.
Spice programs have moved to Gummel-Poon models for BJTs and have gone
through some 4 levels of MOS models.


The laws of thermodynamics are models as well. I'm really not intending to jail 2N3904s for logging across state lines. (Using 'law' to refer to a rule applying to any non-human thing is a metaphor in any case.)

Ebers-Moll is good to a couple of parts in 10**4 or better for small signal transistors if the CB bias is fixed, the transistor isn't saturating, and the collector current is low enough that the extrinsic resistances aren't important. (Even if they are, you can compensate for them accurately because they're real ohmic resistances.) This is the usual situation in logging applications, and is often true with diff pairs as well.

Phenomenological models with a bunch of adjustable parameters can fit just about anything you like, and for more general circuit simulations they are certainly superior. However, I rather doubt anyone has ever invented a circuit by using Gummel-Poon in his head.

One of the things I like best about BJTs is that they really really follow their simple models for noise and DC, so that I can toss them around however I like and be pretty confident how the circuit will work--generally to a fraction of a decibel. Accurate, analytically tractable models are key to all my best circuit hacks--I use about 10 times more algebra than I do SPICE. (I've used LTSPICE a couple of times, but my favourite SPICE is an ancient Windows 3.1 program, because I did a lot of work on its library back in the day and it has a nice schematic editor.)

If I were building SMPSes or discrete power amps I'd do a lot more modelling, of course. I care about the understanding and the results first of all--whether I use pliers or a Crescent wrench is a lot less important.

Cheers,

Phil Hobbs
.

Quantcast