0
\$\begingroup\$

I am running a few very rudimentary cmos circuits in LTSPICE. For some reason, when designing stacked mirrors, I seem to get a much lower saturation turn on thresholds would be expected. I've tried many different combinations of short/long channel lengths and aspect ratios. In this case, I would expect a min vout of 4*(VGS-Vt) or 4*.5 = 2V, but the current gets saturated at the output at only about < 1.5V. When using smaller lengths, it goes even lower, like 2(VGS-VT) when the current should saturate at 4(VGS-VT).

I'm using very simple level one models VTO = .8 and forced gamma to 0, to ignore the body effect. The spice operating points seem reasonable. The DC sweep saturation point, however, does not.

Several different stacked configurations are saturating way lower than expected.

Note: One device mirror only matches perfectly, it's when the transistors are stacked that the unexpected results occur. Also, the .op VDS are very slightly lower then VGS-VT, but I can tweak up and there's not much difference. I separated the Vdd sources, but results are the same with using only 1 Vdd source and sweeping output voltage. One other point, the number of matched diodes connected sources to the left (4), and the problem does not change at all. Current still saturates much lower than expected. It is understandable that models vary, etc... But the model deck has very simple models that mostly match expected 1st order level 1 calculations.

enter image description here

enter image description here

enter image description here

Here is a cascode stack of 6 devices, vdsat = vgs-vt = 0.5 I would expect 6*.5 = 3V min before devices saturate, but the dc sweep shows only < 1.5V.

enter image description here

\$\endgroup\$
  • \$\begingroup\$ use at least level 3 for this \$\endgroup\$ – analogsystemsrf Mar 2 at 3:09
0
\$\begingroup\$

It is indeed working as expected for level 1. The trick is that you will not see the obvious saturation drops when fully zoomed out. However, if you zoom way in before current mostly flattens out (or look at output resistance), you can see that there is indeed a transition that is an expected number of VGS-VT drops, before both transistors are fully saturated. I used two diode connected transistors here (vt + 2*vdsat min), but the concept does not change. It is a fundamental one related to compliance of devices.

enter image description here

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.