The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl Design >> RF Design >> Questions about injection-locked LC-VCO https://designers-guide.org/forum/YaBB.pl?num=1384141942 Message started by ccarrot on Nov 10th, 2013, 7:52pm |
Title: Questions about injection-locked LC-VCO Post by ccarrot on Nov 10th, 2013, 7:52pm Hi, I'm designing an injection-locked LC-VCO with f0=2GHz. The circuit is shown bellow, this is a PMOS-NMOS crossed coupled LC-VCO, the injected signal Vij is put through M1/M2 differential pair. The Vij comes form a LNA in the previous part and a capacitor is in between the LNA and the VCO. I have some questions: (1) The coupled capacitor between LNA and the VCO blocks the DC part, that means the Vij is a sine-wave with DC=0. Is that mean the M1/M2 need bias circuit? (2) Why M3/M4 is put there? (3) My design procedure is: first finish the LC-VCO, then design the injection differential pair. Am i right? (4) I know that increasing the injected current will increase the locking range wL (let's say the input amplitude is constant), but no matter how big the current source (for the injection pair) is, or how large the W/L of the injection pair is (big gm), the locking range is always below 10MHz and looks like constant --- the locking range didn't increase. I even changed to another low-Q inductor, but the range still didn't increase. I don't know why this happen? and How to design this injection part properly? (5) Is it true that if the VCO(freq is f0) is locked, when the Vij (freq is f1) comes, the output frequency is f1, but if Vij disappear, the output frequency will immediately come back to f0? Does anyone know this and can help me out? Thank you very much! |
Title: Re: Questions about injection-locked LC-VCO Post by aaron_do on Nov 10th, 2013, 9:14pm Hi, I have no exp with injection locked vcos, but I can answer some of your questions. 1) yes, you need a biasing circuit. 2) Maybe for reverse isolation. Also, maybe so that the output impedance of the injection portion doesn't affect the tank Q. 3) You can do it that way, but your injection portion will affect the VCO f0. You could also design your VCO with the injection circuitry connected but with no actual injection signal. 4) Can't help much here. I know there are papers which talk about the locking range and how it is calculated, but I can't really remember them off hand. I suggest sweeping the injection signal strength from very small to large, and see if there is any trend at all. 5) "immediately" is probably not right. I suspect there is some settling behavior. Anyway you can simulate this with a transient analysis. Aaron |
Title: Re: Questions about injection-locked LC-VCO Post by tm123 on Nov 11th, 2013, 9:51am My thoughts: 1) Yes you need to establish a bias on the VCO side of the AC coupling caps 2) Looks like a cascode buffer to provide isolation 3) You need to account for the capacitance added to the tank by the injection diff pair, I would not ignore it during the design of the VCO 4) You say you decreased the Q by reducing the Q of the inductor, but you do not say what the loaded Q of the tank actually is. I have never designed an injection locked oscillator but I have designed several VCO's using the cross coupled N/PMOS architecture you have here. The objective was always to maximize the loaded Q for best phase noise. Maybe you need to reduce the Q even further to increase the locking range, but at the expense of phase noise. 5) As Aaron said there will be some settling time but yes if the injection signal is removed the VCO should return to its free running frequency. |
Title: Re: Questions about injection-locked LC-VCO Post by ccarrot on Nov 11th, 2013, 10:14am aaron_do wrote on Nov 10th, 2013, 9:14pm:
Thank you Aaron for your advice! When I test the free running frequency, I have considered the injection part, setting the input amplitude Vij=0. Another question is: what is the current of one side VCO looks like (the ds current of M5 or M6)? Is it sine-wave current? I notice that when the current is small, the Ids of M5 is sine-wave, but if the current is big, it's not sine-wave, even though that the output voltage is always sine-wave. Is that right? Thank you! |
Title: Re: Questions about injection-locked LC-VCO Post by ccarrot on Nov 11th, 2013, 10:23am tm123 wrote on Nov 11th, 2013, 9:51am:
Thank you! What you said is right, I want to increase the locking range a little bit, I have tried to increase the injected amplitude but the range doesn't change. So, for confirmation purpose, I decrease the Q as low as possible to check whether the range will increase, but still it doesn't. Or maybe the bias condition is not correct? For the inductor's Q, I get it from simulation. I modeled the inductor simply as Z=Rs+jwLs, than do the AC simulation to find the Z at f=2GHz, the I got the Q=wLs/Rs. Initially, I chose the biggest Q=15, no matter how big the I_inj is, the locking range is around 10MHz. Then I decrease the Q to about 8, and the locking range didn't change a lot (12MHz). The locking range is: wL=(w0/2Q)*(I_inj/I_osc)/(1-(I_inj/I_osc)^2). where the I_inj and I_osc is the current amplitude of injected signal and the VCO, respectively. You mentioned that the total tank Q. I know that put tuning cap bank to the VCO will decrease the Q, but how can I test and know the total tank Q during the design? Any suggestions will help a lot! Thank you! |
Title: Re: Questions about injection-locked LC-VCO Post by aaron_do on Nov 12th, 2013, 12:42am Hi, Quote:
A periodic but non-sinusoidal waveform is simply a sinusoid + strong harmonics. Whether those harmonics show up in the output voltage or not depends on the load. Since your load is an LC tank, the harmonics will be attenuated, and therefore the output voltage appears sinusoidal. regards, Aaron |
Title: Re: Questions about injection-locked LC-VCO Post by tm123 on Nov 13th, 2013, 8:25am ccarrot, Loaded Q is basically the summation of the Q of the individual components. Just like you know the Q of the inductor, all the other components have their own Q. This includes the varactor, any fixed capacitance (maybe a MIM capacitor you have to help set the oscillation frequency), even the MOS devices that are part of the amplifier. One way to find the loaded Q is to find the Q of the individual components and add them together as Q_loaded=(1/Q1+1/Q2+...)^-1. In your equation for locking range wL I assume Q is the loaded Q not the Q of the inductor by itself. So if you say reducing the Q of the inductor didn't have much impact perhaps some other component is dominating the loaded Q. Hope this helps. Tim |
The Designer's Guide Community Forum » Powered by YaBB 2.2.2! YaBB © 2000-2008. All Rights Reserved. |