calculating MOSFET parameters K & Ve

Hi,

I'm having problems with measuring (through simulation) values for K
and Ve, in order that I can then use those values to design with in a
process which I have the design kit for. Something has gone wrong
though because when I try and use these values to design a
differential amplifier (following the methods in Sansens book "Analog
Design Essentials"), the gain and bandwidth I get are just rubbish -
nothing to do with the intended specifications I calculated device
sizes for. Can anyone tell me whether my procedure for calculating K
and Ve is sensible? It is as follows:

To calculate K, I just used a single nmos transistor (w=1u l=1u) with
a 50k resistor from vdd to drain. Then K'=Id/((Vgs-Vt)^2), and my
results for a sweep of vin from 0.75v -> 1.6V gave me a decaying
exponential with a knee at about vin=0.9v (ie. beyond (vgs-vt)~0.9V,
K' is flattish). So I have a range of K' values to use depending on
what the vgs-vt is.

To calculate Ve, I found an expression rds=VeL/(Ids), so I used the
same circuit as above but stepped Vin from 0.75V to 1.3V. My simulator
tells me the gds (=1/rds) at each Vin, so I just computed Ve=rds*Ids,
to get a result in Volts/micron. Plotting the results, I get a
parabola rising from Ve=30 at Vin=0.75, and peaking at Ve=52 at
Vin=1.2V (then decaying after Vin=1.2).

Part of my problem could be because my Vt value seems to change a
little for some reason, according to my simulators output. I've
connected the bulk to source so I don't know why this is. I'm not sure
if I'm doing something else fundamentally wrong though.

Thanks for any insight on this,

Oh, and thanks for the replys to my post the other day on what a Gm
stage is, that cleared some things up!

Ted

 

[Jim Thompson]

On Sun, 03 Jun 2007 11:29:38 -0700, tedthornton11@googlemail.com
wrote:

Hi,

I'm having problems with measuring (through simulation) values for K
and Ve, in order that I can then use those values to design with in a
process which I have the design kit for. Something has gone wrong
though because when I try and use these values to design a
differential amplifier (following the methods in Sansens book "Analog
Design Essentials"), the gain and bandwidth I get are just rubbish -
nothing to do with the intended specifications I calculated device
sizes for. Can anyone tell me whether my procedure for calculating K
and Ve is sensible? It is as follows:

To calculate K, I just used a single nmos transistor (w=1u l=1u) with
a 50k resistor from vdd to drain. Then K'=Id/((Vgs-Vt)^2), and my
results for a sweep of vin from 0.75v -> 1.6V gave me a decaying
exponential with a knee at about vin=0.9v (ie. beyond (vgs-vt)~0.9V,
K' is flattish). So I have a range of K' values to use depending on
what the vgs-vt is.

To calculate Ve, I found an expression rds=VeL/(Ids), so I used the
same circuit as above but stepped Vin from 0.75V to 1.3V. My simulator
tells me the gds (=1/rds) at each Vin, so I just computed Ve=rds*Ids,
to get a result in Volts/micron. Plotting the results, I get a
parabola rising from Ve=30 at Vin=0.75, and peaking at Ve=52 at
Vin=1.2V (then decaying after Vin=1.2).

Part of my problem could be because my Vt value seems to change a
little for some reason, according to my simulators output. I've
connected the bulk to source so I don't know why this is. I'm not sure
if I'm doing something else fundamentally wrong though.

Thanks for any insight on this,

Oh, and thanks for the replys to my post the other day on what a Gm
stage is, that cleared some things up!

Ted

Simple-minded text book equations are just ducky for getting a "feel"
for how things work, but rarely work very accurately in actual
practice.

Surely Willy (*) has some more advanced (and accurate) equations in
his book.

Willy Sansen and I shared the podium for a week in 1986 at the Royal
Melbourne Institute of Technology... Willy lecturing on CMOS analog
design, I on bipolar.

We both like Chinese food ;-)

...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona Voice480)460-2350 | |
| E-mail Address at Website Fax480)460-2142 | Brass Rat |
| http://www.analog-innovations.com | 1962 |

America: Land of the Free, Because of the Brave

 

On 3 Jun, 20:04, Jim Thompson <To-Email-Use-The-Envelope-I...@My-Web-
Site.com> wrote:

Simple-minded text book equations are just ducky for getting a "feel"
for how things work, but rarely work very accurately in actual
practice.

Surely Willy (*) has some more advanced (and accurate) equations in
his book.

He does make references to more exact equations but I don't think that
is the aim of the chapter in question; an introduction to the design
parameters for a mos transistor. I think he just wants to give the
reader some easy-to-use first-order expressions so that you can
quickly and easily see trade-offs in the system - ie. if rds=Ve.L/Id,
then for more gain(=gm.rds), increase the transistor length.
Additionally, the equations will give you first-order values from
which you can start your design. The exact equations, and refinement
of design parameters, are left to the simulator.

Willy Sansen and I shared the podium for a week in 1986 at the Royal
Melbourne Institute of Technology... Willy lecturing on CMOS analog
design, I on bipolar.

I'm on the other side of the fence - I was on one of his CMOS analog
design courses, hence the book I have

Ted

 

[Marra]

On 3 Jun, 20:20, tedthornto...@googlemail.com wrote:

On 3 Jun, 20:04, Jim Thompson <To-Email-Use-The-Envelope-I...@My-Web-

Site.com> wrote:
Simple-minded text book equations are just ducky for getting a "feel"
for how things work, but rarely work very accurately in actual
practice.

Surely Willy (*) has some more advanced (and accurate) equations in
his book.

He does make references to more exact equations but I don't think that
is the aim of the chapter in question; an introduction to the design
parameters for a mos transistor. I think he just wants to give the
reader some easy-to-use first-order expressions so that you can
quickly and easily see trade-offs in the system - ie. if rds=Ve.L/Id,
then for more gain(=gm.rds), increase the transistor length.
Additionally, the equations will give you first-order values from
which you can start your design. The exact equations, and refinement
of design parameters, are left to the simulator.

Willy Sansen and I shared the podium for a week in 1986 at the Royal
Melbourne Institute of Technology... Willy lecturing on CMOS analog
design, I on bipolar.

I'm on the other side of the fence - I was on one of his CMOS analog
design courses, hence the book I have

Ted

Why not build it and see how it goes.
Your gonna have to build one anyway?