Discussion: OPAMP rail IN/OUT, 0.5uA/0.9V~5.5V supply mA out

[night cat]

Un-believable, but it is true after the measurement !

See attached, it is the rail to rail IN/OUT opamp from SEIKOII.
points are:
1. 0.9V~5.5V working voltage
2. 0.5uA supply current even on 1/2 Vsupply output (unity gain) and even for 5V supply
3. several mA output capability
4. Only one pole for bandwidth response

Any idea to make such ASIC ?

My points are:
1. What process could work as low as 0.9V (0.18u ?) and as high as 5.5V ?
2. How to get so small current source and still can work as low as 0.9V ? (<0.1uA I think for the initial bias current)
3. How does it still be 0.5uA even 5V supply ?

My first question is:
I think it is impossible to get so large resistor (>50M for 5V) by using poly layer in small dice area even the salicide removed.
I tried to use MOSFET to do the simulation by using TSMC 0.18u BSIM3 spice model, L could be as large as 2500u (if W=0.2u).
So that it is impossible to only use one PMOS to do as active resistor.
It is also problem for physical layout (right ?)
I could need to seperate it to many many PMOSs by tying all gates to GND and cascad all D-S. But the minimun working voltage will be increased by lots.

If you can make such OPAMP, then you are the expert !
Any idea ? Interesting !

Welcome to join the discussion !

 

[adanshen]

can you tell me where is the pole,
does the pole change with the external loading?

 

[dudu]

it is possible

the important thing is the unit gain band width, if it is low, you can make the power comsumption very low, of course there are more consideration like distortion requirment will affect the bias in the class AB output.
the R-R input it is little bit complex when the VDD just above 1*Vth+1*Vds, at lest there is method to solve it, like some extrme method using the parastic Bopolar as input or use the substrate as input.

 

[night cat]

Hi dudu,

Any more details regards "parastic Bopolar as input or use the substrate as input" ?

And, any comments for the very low (100nA) initial bias current ?

Below attache the data sheet of this OPAMP.

 

[adanshen]

0.1u initial bias can be easily generated
using several stages of M:1 current dividing

 

[night cat]

Thanks to compaq3100, the example is very useful.

To adanshen,

Sounds good idea, did you mean M:1 then serial w/ 1:M ?
looks like it will increase the operating voltage ?