Plotting Phase Noise for Injection Locked Ring Oscillator

OK, the error (and it's really hard to tell from a screenshot of a small part of the log file) appears that you have a transient convergence error (time step too small). Any chance you can post the input.scs (the input netlist for the simulator) so that I can see what you have on the sources, all the analysis settings, and ideally components too? Providing the whole spectre.out log file would be good too.

Andrew

Dear Ashish,

I just had a chance to review your post, and if you don't mind, have a couple of questions that may provide some insight into your lack of pss convergence in addition to Andrew's requests.

An injection oscillator can be a real challenge to simulate accurately for a number of reasons: the amount of injection, frequency of the injection relative to the lock range; and the timing of the injected signal are all important factors that will have a first order impact on the accuracy. Since I do not know your circuit parameters, but only that you suggest in the pss GUI that the fundamental VCO frequency is 80 MHz, this leads me to the following concerns with respect to your circuit.

1. Have you characterized the lock range of your ring VCO? This is typically done using a conventional transient analysis.

2. If so, is the 80 MHz frequency of the deterministic source V1 in your circuit within the lock range of the VCO? (Please also note that the range of the VCO will be impacted by the new impedance the injection source introduces into its feedback - hence characterizing its range should include some representation of the driving source output impedance.)

3. If the 80 MHz is NOT within the ring VCO lock range, the use of a pss analysis is not appropriate as the actual transient simulation will contain more than one frequency and a multi-tone analysis. This will clearly result in a non-convergent pss analysis as detailed in the Cadence On-line support article at URL:

support.cadence.com/.../cos

It might be best to study the transient solution prior to the start of the pss analysis in your simulation to verify it is converging to a single frequency and, indeed, is locked. Otherwise, I would recommend running a conventional transient analysis to study the performance.

For me anyway, answers to these questions will provide added insight into whether the non-convergence you are experiencing is a result of a circuit topology or a simulator related issue.

I hope this provides some insight for you too Ashish!

Shawn

Hi Shawn, can you help me explain how to find the lock-in range using Transient Analysis. Is it like this let's say my free Running frequency is 1GHz and I am seeing 10 cycles in 10ns and if Inject a periodic current source of 0.8 GHz then I should see 8 cycles in 10ns like here in the pic shown below so Can I say 200 Mhz is within my Lock in Range? So you mean to say sweep the injection frequency and see till what point Locking holds. 

Or is there any other way as well.

Thanks in Advance

Dear Ashish,

Ashish Papreja said:

can you help me explain how to find the lock-in range using Transient Analysis. Is it like this let's say my free Running frequency is 1GHz and I am seeing 10 cycles in 10ns and if Inject a periodic current source of 0.8 GHz then I should see 8 cycles in 10ns like here in the pic shown below so Can I say 200 Mhz is within my Lock in Range? So you mean to say sweep the injection frequency and see till what point Locking holds. 

The 200 MHz is at least within your lock range....let me explain a couple or items -or try tp anyway!

There are actually two parameters of interest - the lock range and pull-in range which are typically different. If you were to apply an injection signal of some amplitude and find the VCO output frequency corresponds to the same frequency as your injected frequency and then vary the applied frequency, the VCO frequency will track your input frequency between two frequencies, say f_max_lock and f_min_lock. These values will change as the amplitude of the injected signal changes, but represent the lock-in range for a given injected amplitude. However, if your simulation starts with no input injection and then you suddenly apply the same injected amplitude at a frequency of  f_max_lock or f_min_lock, the VCO may not change its free-running frequency to exactly f_max_lock or f_min_lock respectively. This is because the pull-in range is the range of input frequencies at a given injection amplitude which will cause the VCO to injection lock if the VCO is initially running at its free-running frequency. Both the pull-in range and lock-in range are usually important in an injection locked based VCO and must be understood.

Hence, there are really two sets of range simulations that are likely required. The resultant minimum and maximum frequencies for each (in absolute value) represent the lock-in and pull-in ranges.

Does this help to clarify the simulation(s) required or did I misunderstand your question Ashish?

Shawn