Noise contribution of VCO

[dustbin]

Hi.
I simulated a 3.6Ghz cross-coupled complementary CMOS LC VCO in spectreRF. The noise summary shows that the MOS are contributing most of the noise while I expect that the low-Q LC tank or current source should be the dominant noise sources. This happens even though I tried different sizing of the MOS. Am I wrong about the assumption that LC tank gives the most noise or is it just that SpectreRF pnoise summary is not accurate? (I have chosen PMOS for varacters and spiral inductor...)
Thanks.

 

[radiohead]

Look if your voltage output swing is high enough, so that mos transistors switch on and off all the time. If this is not the case, noise is bigger.

 

[dustbin]

In this particular case, this may not be the problem since I have sweep the current source for different values till the swing is close to the power supply and they still dominate the noise. But you are probably right in saying that I should look more closely the impact of the swing on noise.

Actually one of the main concern here is that I am not sure how accurate is the nosie summary for a circuit like VCO and it is hard to find out since I am not very experienced on SpectreRF.

Thanks.

 

[ianomaley]

As far as I know with a careful design, phase noise in a VCO is dominated by current source. I have seen a couple of designs were very large transistors are used in the current source to avoid flicker noise influence in phase noise.

Try to check the noise models you are employing. Sometimes particularly flicker noise parameters are wrong.

Ian

 

[dustbin]

Thanks for your suggestion. The reason for asking this is that as I read different papers, they give drastic different figures for contributions of noise.
The more I read the IEEE papers, the more unsure I am about the answer. In papers like "Tail Current Noise Suppression in RF CMOS VCOs", JSSC Vol 37 nos 3, current source is identified as the main culprit for noise while in papers like the one in [Hajimiri JSSC 01 Vol36], tail current noise source is very small. Maybe it is just the difference in optimization approach that yield such a great difference?

 

[devrimaksin]

Dominant noise contributor is also a function
of offset voltage (close-in phase noise or
high-offset phase noise). When you
check the phase noise graph with respect to offset
frequency, you will see different region of phase noise (1/f^3, 1/f^2, 1/f and finally white), the contributor of each region is different.
To see that check the noise summary at differrent
offset frequency

 

[dustbin]

Thx for your quick responce.
Yup. Indeed for the 2 papers I have mentioned, 1 quote the phase noise at 3Mhz offset and the other at 600kHz offset. But what troubles me is that many literature claim that the noise from current source would dominate in both 1/f^2 and 1/f^3 region, which I think is not what the data in the Hajimiri paper suggests. That's why I wonder if it is the result of different approach in optimization or I just miss something there.....
Thx.