how to increase the gain of a cascode common source

[Renjie]

Hi guys,

I'm having trouble with a very simple cascode common source amplifier. Before I cascode it, I can get about 30 db gain; however now, the gain is only 3 or 4db. I can see that two or more transistors are not in saturation; somehow I just can't push them all in sat region. And one more wired thing is, the Vout is so high that almost reach VDD. Hope somebody can come up with some ideas. It should be a very simple cascode common source....

statistics:

V(vncas): 0.848949 voltage
V(vpcas): 1.22708 voltage
V(vpmir): 1.48993 voltage
V(vout): 1.97749 voltage
V(vin): 0.58 voltage

Name: m4 m3 m2 m1
Model: nmos nmos pmos pmos
Id: 3.25e-05 3.25e-05 -3.25e-05 -3.25e-05
Vgs: 5.80e-01 6.06e-01 -7.56e-01 -5.10e-01
Vds: 2.43e-01 1.73e+00 -5.31e-03 -1.72e-02
Vbs: 0.00e+00 -2.43e-01 1.72e-02 0.00e+00
Vth: 5.42e-01 5.87e-01 -2.62e-01 -2.58e-01
Vdsat: 7.80e-02 7.19e-02 -3.47e-01 -1.38e-01
Gm: 6.12e-04 6.19e-04 7.00e-05 3.54e-04
Gds: 1.32e-05 6.26e-06 6.08e-03 1.76e-03
Gmb 1.22e-04 1.04e-04 1.54e-05 7.86e-05
Cbd: 1.16e-14 8.48e-15 4.94e-14 4.95e-14
Cbs: 1.26e-14 1.16e-14 4.95e-14 4.98e-14

 

[FvM]

And one more wired thing is, the Vout is so high that almost reach VDD.

We usually call it an incorrectly biased amplifier. :sad:

Besides possible optimization options for the internal bias voltages, we primarly would adjust the Vin bias to achieve mid supply output voltage, isn't it?

 

[RobG]

Sweep the DC value of VIN and look at the output. The slope of Vout vs Vin is the gain. Note the value of Vin that gives you the largest gain. Set the DC value at this point and do the AC analysis there.

A more "robust" way to do this is to AC couple the VIN to the gate with a 1F (yes one Farad) cap and set the DC bias of M4 with a low frequency feedback loop to force the output into the proper region. The proper circuit is a bit difficult to explain, but a simple 1e9 ohm resistor from the output to the gate of M4 will *probably* work but M3 might need more headroom.

 

[sutapanaki]

You will not be able to keep the transistors - either the pmos part or the nmos part - in saturation because there are two current sources working against each other at a very high impedance node. There was a recent discussion on this topic here:

https://www.edaboard.com/threads/229970/#post985309

 

[FvM]

I agree that DC feedback is the best way to achieve a reasonable bias point. But the output resistance of the cascode isn't that high, you should be able to find a better static bias point at least in a simulation. Apparently Renjie didn't yet try.

 

[dick_freebird]

Your gain is going to come from Rout more than gm (A=gm*Ro)
in this amplifier, and Rout crashes when either of the cascode
guards leaves saturation. Look for a common-mode feedback
scheme to peak the DC gain (and then any successor stages
need to like operating near the peak gain point as input).