Understanding Voltage Reference

[mephi0544]

Can someone explain me how this circuit operates, for me to get the proper aspect ratio for each MOS devices. THANKS

 

[d123]

Hi,

I think if you search for bandgap reference material online, CMOS, there is a lot out there that should answer some of the aspect ratios.

 

[frankrose]

It is not a bandgap reference, there aren't diodes or BJTs. It is more like a constant-Gm reference, but I don't know. Anyway, I think the OPAmp's iputs are swapped.
At start-up the M1 pulls down the inverting input of the OPAmp. To switch on the top PMOS devices the OPAmp should pull down the gates, and for this the non-inverting input should be at a lower voltage than the drain voltage of M1, which is 0V from default. But the drain of M2 is a floating node, and can't be lower than 0V. If you swap the inputs probably it can work as ... something.

 

[dick_freebird]

I'm not sure I'd call it a reference. That would depend
on outcomes. To me, the MOSFET-as-resistor ruins any
chance of tight process or temperature tolerances.
Other than that it seems to be a pretty normal delta-VT
(quasi-bandgap, but with FETs) current "reference" core
with an attempt to obtain a voltage at the back.

I'd much prefer to use the substrate PNP in any application
where I had more than 2V of supply.

Not sure whether this schematic is (or is expected to be)
self-starting. This is often a problem when somebody hands
you a "simplified schematic" of a voltage reference - the
small matter of it not working reliably due to "special case"
bits being left off.

 

[BigBoss]

I think MR acts as a variable resistor to obtain PTAT current reference and this MOS is fed-backed from Vref.
M1-M6 should have the same size -except M3- due to same current.

 

[frankrose]

M2 have to be wider than M1, because of the degeneration, M6 has got an N multiplier by the figure, so it should be bigger too than M3-M4.
I found this thread, quite similar: https://www.designers-guide.org/Forum/YaBB.pl?num=1213674664
It says it is a constant-Gm, the OPAmp's pins are swapped there too, and the stability analysis gave back that the feedback is positive, so I think the pins are swapped.
There are much simplier and understandable reference circuits, I wouldn't recommend to design or build this.