How t choose BJT EmitterSize?

[anhnha]

Hi,
BJT transistors pnp in tsmc18rf has three options for EmitterSize: 10x10, 5x5 and 2x2.
Is 2x2 the best because it is the smallest one?

 

[johnjoe]

It depends on your application and condition of operation. Give us some more details about what you want to do.

 

[anhnha]

Hi, thanks for the reply!

I am designing a opamp based bandgap using bjt. At present, I am using bjts with EmitterSize 10x10 but because the area problem, I am thinking about choosing 2x2 instead.
Is that OK?

 

[erikl]

As mentioned above, this depends on the BJT current: each BJT size has a bit different B vs. Ic current characteristic. For collector currents in the µA range or below, the 2x2 sized should do.

For good matching, larger-sized BJTs would be better. But as you probably use a 1:n array, the small one will be fine.

 

[dick_freebird]

If you are using these BJTs as reference diodes then you
would want to look at how much Rc contributes to Vf at
your chosen set-current. Probably the area of the bandgap
will be resistor dominated in a low power design and maybe
Rc does not matter at super low currents. But you want
both diodes, big and little, to be on dead-log-linear regions
of the log(If) vs Vf curve to get classical bandgap operation,
ant very low and very high current densities your tempco
will be harder to get repeatable. But this is seldom, if ever,
expressed by a model. Has more to do with the real silicon
and where it does, and doesn't, agree.

In forward-active operation, emitter geometry can also be
driving collector contact geometry and Rc; emitter current
density also greatly affects the hFE and you would like to
be near peak, especially if this device is "peaky" to begin with.
And again the operation as a diode (transdiode) can be
affected by how much the collector-emitter conduction
accounts for If, transdiodes "idealize" the diode curve if
beta is high.

 

[anhnha]

Thank you.
I use 1:8 and the current is 2uA. So, from what you said, it is fine to use 2x2 instead of 10x10?

 

[dick_freebird]

You want to look at the device curves to make that
decision.

 

[erikl]

... it is fine to use 2x2 instead of 10x10?

I don't know your process. Here's the Beta vs. Ie curve of a 2x2 vertical PNP transistor of a 0.18µm process.



The small Beta is rather constant over many decades of current, and still fine for your 2µA.

 

[anhnha]

Thank you.
I am using tsmc18rf. I will try to plot the curve and see the result!