Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

OTA sizing by gmoverid approach

Status
Not open for further replies.

EmbdASIC

Member level 5
Member level 5
Joined
Feb 18, 2006
Messages
85
Helped
2
Reputation
4
Reaction score
1
Trophy points
1,288
Location
Polar Region, Mars
Activity points
1,825
Hello all,

In order to size my OTA, Im obtaining the popular gm/id vs. Id/(W/L) curve as shown in fig.1 in the attached paper. Im saving my operating points in a separate .scs file, attached in model libraries containing:
save M0:eek:ppoint

Problem is I obtain a straight line horizontal graph, rather than the actual curve, (not one like the figure)
Could anyone suggest where is the mistake ?
 

I think 1st just try to plot the gm vs the Id sweep that you might be doing. Check if that is working. Once that works, then plot gm/Id vs Id. That should give the same shape as the one they have in the paper.
If that works then in Cadence if you plot both gm/Id and Id/(W/L) then on the waveform window you can select the x axis to be Id/(W/L) and you should see the graph.
If you follow these steps at which point do you start seeing incorrect results?
 

    EmbdASIC

    Points: 2
    Helpful Answer Positive Rating
Ok,
gm/id vs Id is working, but plz have a look at the output waveform !
Output plot is attached.
 

no, i mean gmoverid vs log(ids*L/W)
 

Are you sweeping Id to 100K? Is that amps? I am sure the models will not be accurate for a small transistor above maybe a hundred milliamps or even less. I guess that is why you see the nasty corner in the gm/Id vs Id plot. Can you sweep it to something less like maybe 1mA?
 

Ok,
here is the initial horizontal curve due to which i thought i should seek help.
This is a plot of gmoverid vs. IC
Where IC = Id /(W/L)

Im sweeping IC currently, and plotting it against gmoverid through result browser data !
(Ive reduced the swept range as suggested by one of you)
 

@oermens
since im defining my IC = Id/(W/L) as a design variable before sweeping, its strange that both curve with (e.g IC = log10(Id/(W/L)) and without log look the same. Perhaps its not accepting log function here (in design variable section)
 

Hi ASIC,
Can you first just attach a plot of gm vs Id for your simulation, that would help getting a little background on what is happening and understand the problem better.

Also how is your setup? Can you just connect a MOS diode and put a current into the drain and then sweep the current value and plot the MOS gm? How are you doing it currently?

Thanks.
 

My mistake. Plot gmoverid vs id/(W/L), then double clikc on the x axis and change the scale from linear to log. this is alot easier to do in matlab with the spectrerf toolbox.
 

@oermens
OK,
The plot in the waveform window when changed from Auto -> log remains the same.
However, when changed to 'linear' has the following curve: scaled differently on x-axis
 

I might as well ask:
Has anyone of you successfully obtained the curve as in Jesper's paper, with the technique im using ?
 

Hi ASIC,
I don't understand how you are using that setup to sweep the Id? How do you vary the gate voltage to account for the sweeping Id? If the Gate voltage is fixed then your transistor will go into triode and saturation and it won't give you a proper gm value for always in saturation. The gm/Id methodology depends on the transistor being in saturation otherwise the curves are not consistent. So I think you have to change your setup to diode connect the transistor and put just a current source on the drain and sweep that current source and plot the gm then.
This is how I had done it in the past. And yes I had got the curves as shown in the paper.
 

    EmbdASIC

    Points: 2
    Helpful Answer Positive Rating
but how in that case is your gate voltage being changed with the sweeping current ?
are you connecting a fixed current source or a diode connected load on the drain?
 

Sorry if I am confusing you. What I did was just connect the gate to the drain of the MOSFET, ground the source and connect a DC current source from ground to the drain/gate of the MOSFET. Now the transistor is always in saturation and you can sweep the DC current value from say 100n to 100u or a number that you know covers your range for the OTA circuits. The gate voltage changes automatically with the current since now the transistor is diode connected and the current sets the Vgs value. In this case the Vgs is same as Vds, that can be changed if you want but for the gm/Id vs Id/(W/L) plot it doesn't make much difference as long as your transistor is in saturation you get fairly consistent plot for different sizes and currents.
 

Ok thanks this reply as it much convincing.
About region: perhaps its saturation but they boast it to be region independent, so i never bothered!

Have a look at the plots i have now. gmoverid Vs IC, and gmoverid vs. Id. Also attached is your circuit schematic as i made it.
The curves are still not very nice !
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top