The problem about current mirror

[rockycheng]

Hi,

I found that the currents of both sides of a current mirror (no matter whether it's P or N MOS) is not the same, even if they have exactly the same transistor dimensions. And the current ratio is not equal to W/L ratio. Thus I have to reduce the value of the left current source to get the right current in another side. It seems very strange. Could anybody help me out?
Thanks![/b]

 

[alchen77]

do you check the VDS is the same also ??
the VDS is also a parameter in the current equation.

 

[no_mad]

Hi rocky,

Pls tell me, u used what types of current mirror?? Is it a simple current mirror?? I would suggest u to use a high-swing cascode current mirror. Coz it give u a good load (gmro2) for gain

 

[Vamsi Mocherla]

Like alchen said, it is important to maintain the VDS equal for small length transistors because of channel length modulation effects. So, make sure you repeat this with bigger lengths and aspect ratios. It should get better.

Finally, simple current mirrors are not useful. So, try cascode if you want to have similar performance on both sides

 

[jiangwp]

I think there must be a error for the mirrored current for the current mirror.

In the BIpolar or Mos process, the relative error of mirror is improved by increasing the ouput resistance .

 

[Colbhaidh]

Is this for a simulation or actual Silicon?
If actual Silicon the layout is probably responsible for the offset.

 

[MLR67]

Lenght modulation is the guilty , so you need to make a cascode to increase the rds

 

[kwkam]

Have you try add resistor between source and the supply. It should be help to the current matching

 

[konqueror]

the reason is most probably due to different Vds.use a configuration that has high rout may be cascode or folded cascode

 

[ambreesh]

Simulate with same Vds across the 2 transistors.
You would not observe any difference in the currents