dimensioning a cascode

[little-nemo]

Hi,

I've always used for my cascode a short L (0.5 µm), then increasing the gm.
I've heard that some peolple are using a long L for the cascode. I don't understand why, and I don't understand how it should works.

Have you some explanations for that ?
Thanks in advance

 

[malizevzek]

you normally use cascode for increasing rout. One transistor "shields" the other one, so its drain voltage experiences less variations. In that way, the output resistance is "amplified" by approx. gm x rout.
As for the L, the longer the transistor, the greater the output resistance. Cascode generaly has large outupr resistance, but sometimes you need to get as much as you can by increasing the length of your transistors.

 

[little-nemo]

But if you increase the lenght of the cascode, you will then decrease the gm and decrease by the same way the output resistance.
Or you mean, keeping the same W/L ratio and increasing L ?

 

[malizevzek]

Of course, if you change the length and want to have the same bias current, you have to change the width as well.

 

[little-nemo]

OK, I will try that (but that will increase the area a lot).
Thanks for your help.

 

[leo_o2]

If the width is remained same, reducing L will lead to the increase of gm by 2^0.5, but ro will be reduced by a half. So gm.ro will be reduced.

 

[fehler]

Leo, also the excess gain added by cascode in this case is reduced by 2^0.5, you gain some headroom for the transistor near power rail, to prevent it from entering triode region, then its rds will be enhanced.

So, it is very difficult to judge (by simulation) whether shorter or longer L is better.

For me, I prefer to short one, still, I have reduced output parasitic cap!

 

[raduga_in]

The total o/p impedance of the cascode structure is

ro = gm2.ro2.ro1

Now gm2.ro2 = intrinsic gain of the cascode transistor which is roughly constant across the length variation of the devices ..

So, there is no sufficient reason to take a longer length
and decreasing the gm and increasing the ro as the product of both matters.

Also, shorter length device would reduce the cap at the high-impedance node
which is beneficial.

Raduga