accuracy and output resistance of a current mirror

[nassim_el85]

Hi

In the architectures like "simple current mirror" and "High Swing Cascode" Vds of the mirror transistors is not constant and so we say that we have an output resistance and so the mirror is not accurate. But for the architectures like "Cascode Current mirror" and "Improved High Swing cascode", we immobilize the Vds of mirror transistors and so the mirror is accurate.
The question is that, is the second sentence means that the output resistance of these architectures are infinite?

Regards

 

[timo67]

That means that the dynamic output resistance become reletively high
dont confuse the output resistance Vout/Iout
and the dynamic output resistance δ(Vout)/δ(Iout)
wher it is supposed you work with small signals around the bias points.

 

[rockycheng]

The output impedance is boosted by the cascode structure, but it is not infinite.

 

[qutang]

the current source always have large impedance.

 

[liuyonggen_1]

In my opinion, for cascode current mirror, the accuracy of output current is dependent on the matching of the under two transistors. if the under two transistors have the same Vds, the output current is the same with the input current, even though the transistor operates in linear region.
for cascode current mirror, the dynamic output resistance is very large and the output resistance of the under transistor is small compared with the whole dynamic output resistance of the current mirror. the result is that the change of the output voltage of the current mirror has little effect to the under transistor. this phenomenon is just like the voltage divider.

 

[nassim_el85]

That means that the dynamic output resistance become reletively high
dont confuse the output resistance Vout/Iout
and the dynamic output resistance δ(Vout)/δ(Iout)
wher it is supposed you work with small signals around the bias points.

Of course you right
But when you fixed the drain-source voltages in such structures, the small transitions of output voltage will change the cascode transistors drain-source voltage but not the mirror transistors. So the current of mirror transistors will remain constant.
Please note to the special architectures named, in witch the problem introduced.

 

[timo67]

You wrote:
for the architectures like "Cascode Current mirror" and "Improved High Swing cascode", we immobilize the Vds of mirror transistors and so the mirror is accurate.


In my openion the Vds is not immobilized in these architectures. In contrary, Vds can largely very without affecting output current
δ(Vds) may be very high
but δ(Iout) remains near zero
That is why the ration δ(Vds) / δ(Iout) becomes infinite

 

[nassim_el85]

Hi

Thanks a lot for your response
As my reference "Allen Book on Design of analog CMOC ....." says, Vds of mirror transistors are fixed at Vbc-Vgsm; in which Vbc is the bias voltage of cascode trs and Vgsm is the gate-source voltage of mirror trs.
My mistake was that I though that this voltage is independent of the output voltage and only depend to the bias voltage and current. But now I know that this voltage is correspond to Vout inversely and so the answer is that the output resistance is not infinite.

 

[hr_rezaee]

hi
I think that VGS of output transistor varies by change of output voltage.
because:
assume output voltage changes and current is constant and output transistor is in saturation so VDS of output transistor changes and current is constant so it forces variations in VGS of output transistor. so the VDS of mirror transistor can vary very little.
therefor output impedance is not infinite.
Regards

 

[liuyonggen_1]

in practice, the output resistance of the current mirror never be infinite due to the early effect or channel length modulation effect.
the accuracy demand of the current mirror is depentant on your application. it's not necessary to get a very high accuracy current mirror at the cost of area and power consumption.