To size mosfet (22nm technology), I need to know the on-resistance (Rdson) of the mosfet. Now How can I find the one resistance (RDSon) in cadence virtuoso for the MOSFET? DC operating point is not giving me any value for it. Is there any way to simulate? For example, for nmos, if I use voltage source at drain as supply and also use voltage source at gate voltage Vgs as input, then draw Vds/Ids across Vgs should give me Rdson (on resistance) - Is that the way to find out the on resistance for specific gate voltage or Vgs? I am working in Cadence Virtuoso.

Thank you for your teaching. I recently started trying to use SpectreRF simulation.

I may not have clearly expressed the points of confusion I had. The issue I encountered was that when I used PAC in Specialized Analysis - Sampled to simulate.

When I plot the result, I try to use two different settings that I mentioned. 1. Sweep - spectrum - select eventtime. 2. Sweep - sideband - choose specific sideband

I found that the magintude value difference between these two settings is about 97 times on the same frequency. I don't know why. Later on, I used Sweep - spectrum - time averaged, the magnitude values are same with 2.

The beat frequency I set in PSS is 100MHz. I don't know what operation caused this approximately 97x difference in the "time average"

First of all, this is not a design question, and so should not be posted in the design section of this forum.  I will move it to the RF simulation section, where it belongs.

Second, ADE tends to needlessly use a lot of confusing terminology with RF simulations.  For example, on the PSS analysis form it requests the Beat Frequency, which naturally assumes the presence of multiple fundamental frequencies.  But PSS analysis only works if you have a single fundamental frequency.  When it requests the beat frequency, it really is requesting the fundamental frequency.  Rarely, people apply PSS analysis to circuits that contain drivers at more than one frequency.  In this case the drive frequencies must be co-periodic (they must be integer multiples of a single fundamental frequency).  When they do, the fundamental frequency also happens to be the beat frequency.  But what makes this needlessly confusing is that while the fundamental frequency is defined and well understood in both situations (with a single drive frequency and with multiple drive frequencies), the concept of a beat frequency makes no sense in the more common case of a circuit that has only a single drive frequency.

Now, concerning your question. The term spectrum in the direct plot form means "all sidebands", where as sideband means "an individual sideband".  What makes this needlessly confusing is that the collection of all sidebands does not make up a spectrum.  Something else that contributes to the confusion is that spectrum is the default choice, but is almost never desired.  When used it tends to put up a whole slew of transfer functions that most people do not really want or understand.

Imagine that you want to measure the frequency response of a high-side down-conversion fundamental mixer.  That implies that the -1 sideband is of interest.  So you would select sideband and then specify -1.  If you instead selected spectrum and specified maxsidebands=7 you would get the transfer functions from the input to the output at each of the first 7 positive and negative harmonics.  Nobody wants that because all sidebands except the -1 sideband are all removed by the filter that always follows the mixer.

Having said that you are given the same choice on the PXF direct plot form, and there it is useful.  For example, imagine analyzing a clocked circuit such as a sample & hold or a switched capacitor filter.  Here you sweep the PXF analysis over the normal output range of the circuit, from DC to the Nyquist frequency.  Now observe the transfer function from Vdd to the output and select spectrum. In this case each sideband displayed is useful.  For example, the -3 sideband gives the transfer function for signals on Vdd near the third harmonic of the clock to the output at baseband.  Very useful.  On PXF analysis it makes sense that spectrum is the default, where it doesn't for PAC analysis.

Bottom line:
- when you see beat frequency on the PSS form, thing 'fundamental frequency',
- when you see spectrum on the PAC or PXF direct plot forms, think 'all sidebands'
- when you see sideband on the PAC or PXF direct plot forms, think 'individual sideband'
- when you use PAC, you should be careful to pick sideband rather than spectrum, then specify the sideband that makes sense in your situation.

This Topic has been moved to RF Simulators by Forum Administrator.

what is the difference between "spectrum" sweep and "sideband" sweep in the PAC Direct Plot Form?
I thought that the only difference is that "spectrum" plots mag. in whole spectrum. "sideband" plots each frequency band.
but it makes me confused that the absolute values in "spectrum" result and "sideband" result are not same.

for completeness: I've updated my calculations with the sinc behaviour due to sampling. Now, the phase rolls off much closer to fs frequency as per attached pic.

As for question 1, the origin of simple accumulating jitter is a white noise source.  A white noise source is uncorrelated.  Thus the value at any instant is uncorrelated with the value at any later instant.  The length of a cycle is related to the noise from the white noise source.  Thus, the length of any particular cycle is independent of (uncorrelated to) the length of any previous cycle.  The accumulating nature of simple accumulating jitter comes from the fact that the cycles follow one after another, thus the start of one cycle is delayed or advanced by the variation in the preceding cycle.  This is referred to as a random walk.

The situation is analogous to a person that, starting from the origin, repeatedly flips two coins to determine whether to take one step either north, south, east or west during each interval.  The motion on each interval is independent (uncorrelated to) the motion on the prior interval, yet the man slowly drifts from the origin because each of the steps accumulate.

As for questions 2, J from (35) is the variation in the length of one cycle.  Since the variation of each cycle is independent, the total variation in k adjacent cycles is J√k, which is (34).

Thank you for the feedback, all is clear now.

Dear all,

I am reading a guide document called "Modeling Jitter in PLL-based Frequency Synthesizers" in the Analysis menu
And I have two questions as below.

Question 1.
There is the following sentence on page 12.
"For systems that exhibit simple accumulating jitter, each transition is relative to the previous transition, and the variation in the length of each
period is independent, so the variance in the time of each transition accumulates"

In this part, It is said each transition is related to the others,
On the contrary, I do not understand how the variation in the length of each period becomes independent, can it be explained??

Question 2.
On page 12,
I don't understand how Equations (34) and (35) came out, can anyone explain how Jk is induced ?
I thought it was like the following, is the following equations wrong?

Jk =√(var(ti+k -ti))
  = √(var[jacc(ti+k)-jacc(ti)+k∙T])
  = √(var[jacc(ti+k)-jacc(ti)])

J = √(var[jacc(ti+1)-jacc(ti)])

And, I don't understand why T goes into jacc in J = √(var[jacc(ti+T)-jacc(ti)])---(35).



Thanks in advance
Jin

Presumably you performed a PAC analysis to see the sin(x)/x behavior.  So your input is not DC.  It is a sinusoid swept over a wide range of frequencies.  The sampiing comes from the clock on q1.  The voltage on the capacitor will be a sampled-and-held version of the input.  It won't be a simple sin(x)/x because of the influence of inductor and imperfections in the switch.