The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Modeling >> Semiconductor Devices >> process variation of BJT Vbe
https://designers-guide.org/forum/YaBB.pl?num=1186734015

Message started by vivkr on Aug 10^th, 2007, 1:20am

Title: process variation of BJT Vbe
Post by vivkr on Aug 10^th, 2007, 1:20am

Hi,

I was wondering what process variation would affect Vbe most and how one could see this in simulations? I am assuming that the
BJT is diode connected and the temperature is constant.

Typical simulation models for BJTs group the process corners based on transistor beta into (low-beta, high-beta).
However, beta does not seem to have an obvious connection to Vbe.

Large Vbe => Small IS => Smaller area A, Larger base width Wb
Large beta => Smaller beta

There are other process control factors such as the doping and diffusion constants etc. which are also responsible in both cases.
However, area A has a very significant contribution in deciding Vbe, whereas it has no role to play in beta which is dependent
largely on Wb (recombination in the base).

So, how can I find out conditions in simulation which show me the largest variation of Vbe around its mean when I vary the process
parameters?

Thanks
Vivek

Title: Re: process variation of BJT Vbe
Post by ACWWong on Aug 10^th, 2007, 2:16am

This is achieved by variation of Is.
In BJT (Gummel-Poon, can't remember if vbic is the same) models I have used in the past, typically beta and Is are the process variables available in design kits to model process corners. Given fixed current, area and thermal voltage, vbe depends on Is (with a ln(I/Is) function i.e. Ebers-Moll equation).

Title: Re: process variation of BJT Vbe
Post by vivkr on Aug 10^th, 2007, 2:34am

Hi ACWWong,

It is clear that Vbe is a function of Is. My question is: given the process corner models which are typically betalo, betahi,
can the min./max. condition for Vbe also be simulated. Clearly, these corner models give min./max. beta, but not directly
min./max. Vbe.

So, if you were using these models to build a clamping circuit where the BJT-based diodes must clamp a cerain node to less
than a certain voltage at a given current, you would want to know exactly how much Vbe varies. How can one simulate this?

Thanks & Regards
Vivek

Title: Re: process variation of BJT Vbe
Post by ACWWong on Aug 10^th, 2007, 4:09am

I would say no, if only beta max/min is available, then vbe variation is not modelled...
If you have the process variation measured data for vbe for fixed current, temperature & emitter area, then you could modify Is or emitter area to match the data you have. Without the process data, you would be guessing... you'd need to seek information from your foundry...

Anyway, although nominal (at fixed temperature) Is variation is typically quite large, its impact in my experience is generally less than temperature over say a range of +/- 50K from 300K (vbe would be ~ +/-100mV)... but again it depends on the process.

Title: Re: process variation of BJT Vbe
Post by vivkr on Aug 10^th, 2007, 4:40am

Thanks for the feedback. I will check with the foundry. However, +/-100mV does seem a lot to me. I would have thought that Vbe for a BJT
(not a parasitic one, but a proper one) would vary much less than that, less than +/- 50mV even. But, let's see.

Regards
Vivek

ACWWong wrote on Aug 10^th, 2007, 4:09am:

Title: Re: process variation of BJT Vbe
Post by ACWWong on Aug 10^th, 2007, 4:50am

BJT vbe is often quoted in text books as -2mV/^oC. I have found in the techonologies I've used in the last decade its normally between that and -1.5mV/^oC

Title: Re: process variation of BJT Vbe
Post by vivkr on Sep 20^th, 2007, 1:55am

Hi,

I think the -2 mV/K is also a function of the exact BJT junction properties (junction area),
but more importantly of the current density. If the current is changed, then the Vbe drop with temperature
is different. Bob Pease wrote a very good article "What's this VBE stuff anyhow?". You can find it
on his website.

Regards
Vivek