The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> RF Design >> How much Q we have on CMOS for a 10nH spiral @ 450MHz?
https://designers-guide.org/forum/YaBB.pl?num=1291185173

Message started by neoflash on Nov 30^th, 2010, 10:32pm

Title: How much Q we have on CMOS for a 10nH spiral @ 450MHz?
Post by neoflash on Nov 30^th, 2010, 10:32pm

Can people suggest a paper or TSMC RF library that roughly how much Q we can have for this 10nH spiral @ 450MHz?

Thanks,
Neo

Title: Re: How much Q we have on CMOS for a 10nH spiral @ 450MHz?
Post by rfcooltools.com on Dec 1^st, 2010, 9:54am

neoflash,

I do not have a paper to reference but from experience the Q will be well below 10. Many factors play into this like metal thickness of metal , type of metal and physical size of the inductor. Simply the Q is set by the (imaginary series impedance)/(real series impedance) so for 450MHz the imaginary part of the impedance will be around 30 Ohms.

The following is a very coarse approximation of resistance:
assume for example metal resistance of the conductor is around .05 Ohms per square. a straight line of on chip metal has around 1nH per mm. so around 10mm for 10nH if the conductor is straight, but if it is coiled then the overall length will reduce by say a factor of 4 (to ~2500uM) due to the magnetic flux concentration. So calculate the number of squares for example if the conductor width is 10uM 0.05OhmsPerSq*2500uM/10uM equates to 12.5Ohms at DC. now include skin effect and this will increase the effective reisitance. But even at 12.5Ohms the Q is only a little greater than 2.

http://rfcooltools.com

Title: Re: How much Q we have on CMOS for a 10nH spiral @ 450MHz?
Post by vp1953 on Dec 1^st, 2010, 11:25am

Hi Neoflash,

I dont recall the exact values for TSMC but just based on experience with other fabs, for a 4 layer process with thick metal, it could be less than 4-5 (close to Rfcooltool's rough but intricate calculation). You could get Q's of 9-10 at higher frequencies (>1GHz)

Title: Re: How much Q we have on CMOS for a 10nH spiral @ 450MHz?
Post by rfcooltools.com on Dec 2^nd, 2010, 10:01am

vp1953,

I agree a Q of 2 does seem a little low.

Neoflash,

why are you wondering about the Q?

http://rfcooltools.com

Title: Re: How much Q we have on CMOS for a 10nH spiral @ 450MHz?
Post by jayce3390 on Dec 2^nd, 2010, 12:23pm

Q = 0.2 usually.

Title: Re: How much Q we have on CMOS for a 10nH spiral @ 450MHz?
Post by ACWWong on Dec 2^nd, 2010, 2:08pm

jayce3390 wrote on Dec 2^nd, 2010, 12:23pm:

Q = 0.2 usually.

I can't see how possibly how this could be true.... unless "usually" the inductor design is very poor...

Anyway the Q value at 450MHz for a 10nH will tend to be low (maybe upto 5 in most technologies/designs), but its heavily dependant on your design. The only real way to find out do is to do some analysis. Either hand calculations like in Tom Lee's RFIC design book or via a quick simulation. I would recommend using ASITIC (free from the web if you google it) for a quick simulation. After you enter the technology information, then draw your spirals and simulate the L & Q. So if you use a large area/fewer turns/more hollow inductor for your 10nH design you will tend to get a better Q, buts a trade off of area for modest increases in Q may not be worth it.

Title: Re: How much Q we have on CMOS for a 10nH spiral @ 450MHz?
Post by loose-electron on Jan 27^th, 2011, 3:12pm

You should be able to derive this knowing the RLC characteristics of the metal layerss - In my experience, you are doing verrrry well if you get as high as 6-8, and that is copper layers, thick oxides underneath and thick metals.