The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl Design >> RF Design >> Is the value gm = 0.019 S too small for a rfnmos1v? https://designers-guide.org/forum/YaBB.pl?num=1390760575 Message started by baab on Jan 26th, 2014, 10:22am |
Title: Is the value gm = 0.019 S too small for a rfnmos1v? Post by baab on Jan 26th, 2014, 10:22am Hi, I am simulating gm of rfnmos1v from TSMC 0.13. My method is to plot id vs Vgs and then using calculator to plot gm = did/dVgs. Here are pictures about the simulation. Please help me with the question in the picture. Here is the schematic for simulation: |
Title: Re: Is the value gm = 0.019 S too small for a rfnmos1v? Post by baab on Jan 26th, 2014, 10:23am Here is the set up: |
Title: Re: Is the value gm = 0.019 S too small for a rfnmos1v? Post by baab on Jan 26th, 2014, 10:24am Finally, here is the gm plot along with my question: |
Title: Re: Is the value gm = 0.019 S too small for a rfnmos1v? Post by aaron_do on Jan 26th, 2014, 4:51pm Hi baab, 0.019 is not necessarily too small. You should look at gm/Id not just gm, as it will reveal more information. gm/Id = 2/(Vgs-Vt), so for Vgs - Vt = 0.7, expect very approximately 2/0.35 = 5.7. So at 4 mA, your gm would be 0.022 S. That's assuming Vt = 0.4 V. That equation doesn't take into account a lot of things, so I usually use it just as a ball park figure. So for very strong inversion, I usually estimate gm/Id = 5, and closer to weak inversion around 10-15. The maximum possible is around 25, but at that point the transistor is cutoff, and not very useful. Also think of how much load you need to drive. Assuming you need to drive about 200 ohm, 0.02*200 = 4 (12 dB). So you have a decent amount of gain. regards, Aaron |
Title: Re: Is the value gm = 0.019 S too small for a rfnmos1v? Post by baab on Jan 26th, 2014, 8:27pm Thanks a lot, Aaron. It is really helpful. As Vgs is fixed, then the only way I can do to increase gm is by increasing id. And with given technology, for example 130nm, then all I can do is to increase W (width). I mean that with those constrains, we can only increase gm by increasing W. Is that correct? And in my case, the load is a mixer, its input impedance is 50 ohms. Therefore, voltage gain will be gm*Rl = 0.02*50 = 1 (0dB). We have no gain at all. Do I need to create an impedance transformation to transform the load 50 ohms to a larger value, says 200 ohms? PS. This is not a related question but I am wondering if there is any way to label a current that is similar to the way we label a wire in Virtuoso. |
Title: Re: Is the value gm = 0.019 S too small for a rfnmos1v? Post by aaron_do on Jan 26th, 2014, 10:00pm Quote:
If you are just talking about the transistor, then yes.. Quote:
In the first place, is the mixer integrated on the same chip as the LNA? If so, then its unlikely that your mixer input impedance is 50-ohm. Adding a transformation network would be one option. You also have a transformation network at the input. Quote:
I'm not really sure what you mean. if you label two wires the same name, then they are considered to be connected, and current can flow between them... regards, Aaron |
Title: Re: Is the value gm = 0.019 S too small for a rfnmos1v? Post by baab on Jan 27th, 2014, 7:32am Hi, Aaron. Quote:
Yes, I was talking about transistor. In cascode LNA, I see that the overall Gm will be gm*Q with Q quality factor of input circuit. Is that another way that you imply? Quote:
Yes. Quote:
Well, may I know why? I thought that it is a standard that applies to both on and off chip mixers. Quote:
Are the other options? I am curious to know that. Quote:
I didn't express it well. To label a wire, we only need to press "L" and then place it on the wire that we want to label. I am wondering if there is a similar method/way to do that with current. If so, it is great. I want to rename current points by meaningful names. |
Title: Re: Is the value gm = 0.019 S too small for a rfnmos1v? Post by aaron_do on Jan 27th, 2014, 5:08pm Hi baab, Quote:
the result gm*Q is a combination of the transistor (gm) and the matching network (Q). It doesn't need to be cascode. Quote:
As far as I understand it, the purpose of having a 50-ohm standard impedance is mainly for component vendors to have a standard impedance to work with. So if you aren't using any off-chip components in between the LNA and the Mixer, then you can use whatever impedance you want. Also, the concept of wave reflection doesn't really even apply on-chip as the distances traveled are too small (unless you are talking on the order of 100 GHz). Quote:
I was just saying there's more than one way to get gain. The input matching network gives you a voltage gain of Q. You could also design your load for a high gain. regards, Aaron |
The Designer's Guide Community Forum » Powered by YaBB 2.2.2! YaBB © 2000-2008. All Rights Reserved. |