The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl Design >> RF Design >> Injection lock VCO https://designers-guide.org/forum/YaBB.pl?num=1229048701 Message started by Mehran on Dec 11th, 2008, 6:25pm |
Title: Injection lock VCO Post by Mehran on Dec 11th, 2008, 6:25pm Hi all, im new here, and almost new in IC design. i am doing an Injection lock VCO. i am trying to do an experiment in cadence to find my lock range. basically i am using 2 ideal current source to inject a differential current in to a cross couple VCO. i am changing the frequency around the natural frequency of the tank and doing PSS to find the output spectrum. 1. i have seen the experiment results of prof Harjani in CICC, and here is the more complete one, http://www.cmoset.com/uploads/7a.3.pdf page 24 if you look at the results, there is a region which amplitude is almost constant, which we can find the lock range and the spur supression from the graph. but in my simulation, i am getting a curve like normal tank but with more supression than without cross couple mos, which i beleive its do to regenerative amplification. i dont see that flat region in my simulation. is there any special setup for simulating injection lock VCOs? 2. when i am running the transient simulation, for a free running VCO, if i look at the common source of cross couple pair, i can see the wave form with twice of frequency which is wellknown. but when i inject a signal, i see a wave form with 4 times of frequency. how can i make sure that my VCO is getting lock to the injected signal and how i can find the lock range in simulation? thanks in advance. Mehran |
Title: Re: Injection lock VCO Post by pancho_hideboo on Dec 19th, 2008, 5:39am Mehran wrote on Dec 11th, 2008, 6:25pm:
Mehran wrote on Dec 18th, 2008, 11:51pm:
If your ILO(Injection Lock Osc.) could be enough locked within your simulation accuracy setting, a driven PSS can be valid for ILO where driven signal is injection signal. If your ILO could not be locked, PSS can't converge. But you can't distinguish whether it is ILO unlock or PSS convergence fail due to PSS ill-robustness. Use driven envelope analysis and observe instantaneous frequency of Nth harmonics you desire. And using chirp FM signal or conventional FM signal as injection signal, you can evaluate lock range. Here frequency change speed must be enough slow compared to settling time of ILO. But I can never assure simulation performance of envelope analysis in Cadence Spectre. I use envelope analysis of Agilent GoldenGate simulator for this purpose. |
Title: Re: Injection lock VCO Post by Mehran on Dec 21st, 2008, 12:47am Hi, thanks for your reply. i dont have access to Agilent. but can you tell me more about what you are telling by FM signal? i do transient and then i am taking DFT? i can not get the correct spectrum for quasi lock. would it be possible due to my wrong setup parameter in DFT or tolerances in my simulation? reltol=1e-5 DFT window size = 1us number of points =2^13 Mehran |
Title: Re: Injection lock VCO Post by pancho_hideboo on Dec 21st, 2008, 1:09am Mehran wrote on Dec 21st, 2008, 12:47am:
What spectrum do you get ? I think you expect one remarkable peak spectrum at oscillation frequency but actual result is very wide spreaded where you can't find any remarkable peak in spectrum. Right ? Teach me followings. (1) Frequency of injection signal (2) Waveform shape of injection signal (3) Expected oscillation frequency Mehran wrote on Dec 21st, 2008, 12:47am:
Try a driven envelope analysis of Cadence Spectre. Then observe instantaneous frequency of Nth harmonics you desire. Mehran wrote on Dec 21st, 2008, 12:47am:
This is same as actual measurement. Do you have experience of actual measurement ? Not EDA Tool Play. |
Title: Re: Injection lock VCO Post by Mehran on Dec 21st, 2008, 1:28am Hi, thanks again. i am trying different frequency. tank resonance is 4434MHz. my injection frequency is 4444MHz when i expect it to be lock. i expect my lock range by (W0/2Q)*Iinj/Iosc my injected signal is sinusoid. what i expected is a remarkable power at injection frequancy when it gets lock and when it is not lock, i must see a bunch of tone as Razavi's paper experimental results. http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=1327738 but i cant get those bunch of tone. i expect the lock rage to be 70MHz , but even when i inject a signal 150MHz away of tank resonance, i still see remarkable power at injection frequency. i dont have experience on actual measurment. i actually eager to know how they get their graph on spectrum at this http://www.cmoset.com/uploads/7a.3.pdf page 24 thanks alot Mehran |
Title: Re: Injection lock VCO Post by pancho_hideboo on Dec 21st, 2008, 1:41am Your post is usage of simulator not Design issue. Please post to "The Designer's Guide Community Forum ≫ Simulators ≫ RF Simulators". Mehran wrote on Dec 21st, 2008, 1:28am:
Do you mean your FREQosc is not equal to N*FREQinj where N is integer number ? As far as I know, oscillation frequency of conventional ILO is N*FREQinj when it is locked. So I can use a driven one fundamental PSS for ILO if it can be locked. Have you invented a novel ILO whose oscillation frequency is not restricted to N*FREQinj ? If so, you can not use a driven one fundamental PSS and you have to use autonomous two tone QPSS although Cadence Spectre doesn't have autonomous QPSS yet. But if your ILO is conventional one where FREQosc is equal to N*FREQinj, autonomous QPSS is not completely appropriate. In unlock state, oscillation could be never stationary or could disappear in steady state. So there is no solution for QPSS. You have to use a driven envelope analysis. Mehran wrote on Dec 21st, 2008, 1:28am:
I can't access the above paper now but I think your result is natural since you do static DFT. Mehran wrote on Dec 21st, 2008, 1:28am:
Maybe this graph is a result of Max-Hold plot in Spectrum Analyzer. In simulation using EDA Tool, I can plot such graph by using simulation results such as envelope analysis with chirp FM signal. Or instead of chirp FM signal, I might use parametric sweep function for frequency of signal source in simulation. But it requires script to create such graph. Do you have a skill to write script handling simulation result ? Rather I'm using plot in page.16 for ILO in both simulation and actual measurement. It also requires scripts. But I use Agilent post processing environment, so I don't have to write script. |
Title: Re: Injection lock VCO Post by Mehran on Dec 22nd, 2008, 5:39pm Hi Pancho, many thanks for your help. I got your point about FM signal. now I can get the result like page 16 as you told me. My ILO is a conventional one. I expect output frequency to be 1*FREQinj. but I still can not get the result of page 24. I need that for my application. :( can you just give me a hint about writing a script? is there any special language in Spectre? Mehran PS: since this discussion might be useful for others, i will ask forum admin to move the whole topic to RF simulators. |
Title: Re: Injection lock VCO Post by pancho_hideboo on Dec 22nd, 2008, 8:55pm Mehran wrote on Dec 22nd, 2008, 5:39pm:
Did you understand an evaluation of an instantaneous frequency by using Envelope Analysis ? An instantaneous frequency can be calculated like following. In Agilent Tools. Inst_Freq=diff( unwrap(phaserad(out[N]), pi)/(2*pi) )+HB.freq[N] In Cadence Tool. Inst_Freq=-groupDelay(harmonic(v("/out", ?result "envlp_fd"), N))+N*fund_freq_of_Envelope_Analysis In your case, fund_freq_of_Envelope_Analysis=4444MHz and N=1. Mehran wrote on Dec 22nd, 2008, 5:39pm:
Do you understand "Max-Hold plot in Spectrum Analyzer", "parametric sweep function for frequency of signal source in simulation" in my answer ? Mehran wrote on Dec 22nd, 2008, 5:39pm:
Almost all EDA vendor's tool have some languages for database, simulation and framework, etc. You can use language called "skill(ocean)" in Cadence DFII environment. First you should try manual procedure without scripts. Here you have to know general procedure of actual measurement for VCO, PLL, etc. using actual instruments. Many measurements using actual instruments are intensively used in both papers you refered. Not EDA Tool Play. Mehran wrote on Dec 22nd, 2008, 5:39pm:
I don't think so. It is no more than simple and common usage of simulator. |
Title: Re: Injection lock VCO Post by Mehran on Dec 28th, 2008, 10:33pm Hi, Thanks for reply. I know the Max hold on the spectrum. If I plot with different frequency for injection above center frequency and plot the peak of the power at the spectrum, I don't see that flat region in my circuit.The power is decreasing fast, like a normal tank. Also, I don't understand why we should do this, because when it goes out of lock, in fast beat, the power at the frequency which is different from my injection frequecy is maximum which spectrum analyzer will show that at the max hold. |
Title: Re: Injection lock VCO Post by pancho_hideboo on Feb 9th, 2009, 5:12am Mehran wrote on Dec 28th, 2008, 10:33pm:
Mehran wrote on Dec 28th, 2008, 10:33pm:
See page-24 of http://www.cmoset.com/uploads/7a.3.pdf This Spectrum are measured by Agilent PSA or ESA Series Spectrum Analyzer of "positive peak" detector mode not "sample" detector mode. |
The Designer's Guide Community Forum » Powered by YaBB 2.2.2! YaBB © 2000-2008. All Rights Reserved. |