BGR current density ratio

[ambreesh]

Hi,

how does one decide the current density ratio in BGR'.
I have seen design having a current density ratio of 8 to 40. But what are the parameter on which we decide the upper and lower limit for the same.

Kindly help

 

[newton_wang]

the up limit is by the power consumption and area (BJT is much bigger than MOS)

the low limit is by matching (8:1 is the smallest ratio for pure common-centric
in layout, say 3x3 array, the 1 is in the center)

 

[hung_wai_ming@hotmail.com]

1:8 to be honest, is a common centriod but not really matched ratio as the 4 of 8 BJTs at four corners are facing different disturbance of the other 4 from the central single BJT. However, it is already the best approach.

 

[gte582w]

Plus ln(8) is roughly 2, easy for hand calculation

Added after 31 seconds:

Plus ln ( 8 ) is roughly 2, easy for hand calculation

 

[ambreesh]

The BGR is one of the most important blocks in the analog system.
So I would not get bothered with area of it. 4X4 i.e with current density ratio of 15 or 5X5 with current density ratio of 24 are also easily matched by centroid.

I have been getting info. on some relation to less effect of opamp offset and something to do with noise, yeild etc.

As for the power consumption I can only pump current in diodes that keeps them in exponential region of their IV curve.

 

[scottieman]

Hi,
how does one decide the current density ratio in BGR'.
I have seen design having a current density ratio of 8 to 40. But what are the parameter on which we decide the upper and lower limit for the same.
Kindly help

To minimize impact of opamp offset on bandgap accuracy, large current density ratio should be adopted. Of course, chip area is increased. One way to reduce the area is to inject ratioed currents into that pair of diodes/BJTs. For example, if my intended ratio is 1::80. I can use 1:8 diodes/BJTs + 10:1 currents. This idea, if I remember it correctly, has been published in journal. Still, this approach increases power consumption. So, tradeoffs are always the problems of analog design

In addition as I known, 1:8, 1:24 or in general common centriod layout is not a must for commerical BGR. Let you to consider this point

Hope it helps
Scottie

 

[ambreesh]

Hi Scottieman,

Current density ratio and current would be 2 different things, correct me if I am wrong.
If the diode models I have show a good ΔVbe characteristic for low currents I can still get away without burnig too much of power. Say they show good responce from 0.5u to 1u. We could use both more current and more BJT (with a trade off) and get good current density ratio.
I do agree that we need not have number of BJT to meet centroid criterion.

How does more current density ratio get us to minimise offset of opamp. It would be great if you could pass some technical materia to read.