thinking more on current mirrors

[zychang]

I know it sounds like basic stuff, but is only the W/L ratio that determines the current gain or the absolute value of the W and L also matters? Let's say in a current mirror you have W/L=3/0.5 in the reference branch and W/L=2*3/1 in the mirror branch. Is it true that Imir=Iref? I thought it would be true but my simulation didn't...

 

[hr_rezaee]

Hi
lambda is inversely proportional to L.
So it is one of the reasons of deference.
regards

 

[zychang]

I agree with you. But the thing is the difference I saw is over 25% (256uA in the reference branch and only 200uA in the mirror branch). I was kind of surprised that the channel length modulation can make such a big difference and wondering if there is something else.

 

[gpwu]

You need to take into account lateral diffusion length.
That is, the real L and drawing L is different.

 

[sanish]

It also depends if your simulation is prelayout or post layout . The post layout netlists also model lateral stress and scattering effects causing further mismatch.

 

[WindRay]

Please use a small iref current (to 2.56uA) and do the simulation again. Do you get the same amount of variation that is over 25%. Thanks.

 

[supreet_95]

Try to use larger values of L in current mirrors, use 3 times the minimum value. Also take care of LOD and try using spares.
Hope this helps.