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.

Register Log in

Why using transistor sizes larger than the minimum length?

Status
Not open for further replies.
V

vistapoint

Member level 5
Joined
Feb 20, 2005
Messages
91
Helped
1
Reputation
2
Reaction score
0
Trophy points
1,286
Activity points
2,125
transistor size

We usually calculate transistor sizes as W/L. But in real design I saw 100/2, 100/3 when the technology is 1um. Why using sizes larger than the minimum length? And how to decide that?

Larger transistors may can reduce noise. But when we start a design, how can we generate sizes like 100/2 instead of 50/1?
 

transistor size

Usually we set the length to 1.5~2 times of the minimum allowable length. The main reason is for fabrication.
 

Re: transistor size

1. Relative accuracy is better for nonminimal sizes.
2 Gds ( Rout) better for long channel
I agree with Nxing that 2 times blow up is a good compromise.
 

Re: transistor size

In analog design, the length of gate is chosen 2-5 times bigger than the minimum size is a common practice
 

Re: transistor size

Thanks. That makes sense to me.
 

transistor size

Choosing 100/2 instead of 50/1 is better for matching devices.
Mismatch depend on area of the transistor. 100x2=200, 50x1=50 so your transistor area is 4 times bigger, so your mismatch will be 4 times smaller.
 

Re: transistor size

hai dreamteam,
any document regarding mismatch. post it if u can.
 

Re: transistor size

I had a doubt thtt incresing of the gate and hence channel size wont have any effect on the spped of the transistor when using in RF fo 1 to 5 Ghz range in the Analog Design???
 

Re: transistor size

Maybe the large size transistor is indispensabe.

On base of some research , the minimum( the line wide) width transistor is used in the analog design for the advanced foundry that is a trend.

As the width reduced, the second effecs of transistor(MOS) is appear. which one of them is dominant is varying as the foundry changes and the line width changes.

So we can use the minimum width transistor for the analog design considering the speed and other second effects of tansistors.
 

Re: transistor size

you should simulate some bias points for different overdrive voltages (Vgs-Vth) and different L's. Make W e.g. 5 or 10 * L. Compute afterwards the gm and gout (=1/rds). You know that the max. amplification of a transistor is equall gm/gout=Amax. If you do so you will see that for modern processes the Amax will get better as you increase L. For min. L the A is about 50%, which is definitely unacceptable. Howerver, you are right --> it costs additional area.

Furhter you have to take care about noise performance, especially 1/f noise. The larger W*L the less 1/f noise.

Another pro for taking L=2*Lmin is mathing, especially for the input stage of a diff-amp.


I hope I could help you a little bit.
I suggest you to simulate some bias points for diff. overdirce voltages and different Ls. Also 1/f-noise in respect to W*L. This will simplify the first attempt in your design and you will profit a lot if you are "familiar" with your process.
 

Re: transistor size

In 1 micron technology you would not suffer any channel length modulation. If in lower techonology e.g. 0.13micron if i donot suffer from short channel effect I would prefer using small length devices.
Use good layout techniques
 

Re: transistor size

Unity current frequency is inversly proportional to the length L of the transistor. Hence with a smaller L, your circuit can go faster. That's also a main reason why we want to develop more advanced process with smaller alowable L.

In analog design, transistors with larger area, i.e., larger W*L match better and have lower flicker noise. Also with larger L the channel legth modulation effect will be weaker and hence the drain source resistor will be larger which normally brings you more voltage gain.
 

Re: transistor size

Are you saying that even W/L is the same, the longer L will make the output impedance higher because ro~1/lamda. Therefore the gain will be higher?

eda4you said:
you should simulate some bias points for different overdrive voltages (Vgs-Vth) and different L's. Make W e.g. 5 or 10 * L. Compute afterwards the gm and gout (=1/rds). You know that the max. amplification of a transistor is equall gm/gout=Amax. If you do so you will see that for modern processes the Amax will get better as you increase L. For min. L the A is about 50%, which is definitely unacceptable. Howerver, you are right --> it costs additional area.

Furhter you have to take care about noise performance, especially 1/f noise. The larger W*L the less 1/f noise.

Another pro for taking L=2*Lmin is mathing, especially for the input stage of a diff-amp.


I hope I could help you a little bit.
I suggest you to simulate some bias points for diff. overdirce voltages and different Ls. Also 1/f-noise in respect to W*L. This will simplify the first attempt in your design and you will profit a lot if you are "familiar" with your process.
Click to expand...

Added after 6 minutes:

Combining your explanation with eda4you's, I can almost visualize the effect of L on output impedance and gain. The mention of wT also reminds me of some concept I almost fogot.

Thanks and I really like discussing here.

Alles Gute said:
Unity current frequency is inversly proportional to the length L of the transistor. Hence with a smaller L, your circuit can go faster. That's also a main reason why we want to develop more advanced process with smaller alowable L.

In analog design, transistors with larger area, i.e., larger W*L match better and have lower flicker noise. Also with larger L the channel legth modulation effect will be weaker and hence the drain source resistor will be larger which normally brings you more voltage gain.
Click to expand...
 

Re: transistor size

I like here too, It's a good place for analog IC design novices like me.
Vistapoint:
It is true even W/L is the same, the longer L will make the output impedance higher
Sorry eda4you, I can not stop myself to answer this question for you :)
 

Re: transistor size

1. for better matching
2. for smaller lamda: lamda is proportional to 1/L.
3. for smaller Id current
 
transistor size

May be I can add a point for matching. Beside, put a larger W/L transistor. Layout draw also critical for the device matching.
 

Re: transistor size

you can find out the min W/L and max W/L for a particular process (tsmc,csm,umc) from its website. if you want to use W/L more than the specified ones, you can use multiple fingers.
 

transistor size

larger size may have a good match to the other transistors.
 

Re: transistor size

somebody pointed out about RF circuits and the small channel length used for such circuits. That's true. Using channel length 2 - 3 times the minimum should not be thrown around like that. For RF circuits, design parameters are different from conventional analog/mixed signal design reqs. So you have to be careful, when you say use channel length 2 - 3 times that of min length!
 

transistor size

hi
chinito can u please explain what are the considerations for min and max lengths in RF design.
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top