i want to do that kind of simulation with cadence, but i don't know how to do it ?

Dear Hagar Hendy,

Hagar Hendy said:

in the below screenshot, there are two signals are competing each other and i want to check the metastbility, so one of them are swept and the other is fixed 

that was done in hspice , however i want to do that in cadence, but i don't know how to do that  with mastero ? 

I am not sure I fully understand the metric you want to use to measure "metastability". However, if you are attempting to examine how closely in time the two transitions of input signals vxex1_1 and vxex2_1 may be in order that the output  of some gate (or combinational logic) has a deterministic output logic state, you might consider the methodology described in the following Forum post:

community.cadence.com/.../1384108

In lieu of setting your two inputs to two frequencies f1 and f2 that are offset by a small amount to allow their edges to cross every (f1-f2) Hz. the two inputs are set to the same frequency but posses different delays. Create a variable for one of the delays and vary its value in very small time increments over a range where the two input signals are nearly coincident in Assembler or Explorer. You can measure the propagation delay as a function of the delay value which will indicate the region where your desired output is no longer a deterministic value dependent on your two inputs. There are a couple of references and figures in the Forum post to better illustrate this concept if my words are not clear to you.

Shawn

thanks a lot, What I did is that I have an arbiter and I want to check the aperture time of that arbiter, so, I checked how closely in time the two transitions of input signals are to the inputs of the arbiter, I chose transient noise with fmax:1MHZ, but I didn't see any effect on the noise on the signal, The setup is below. as well as how to use parametric analysis in mastero to sweep one of the input and the other will be constant. 

Dear Hagar Hendy,

Hagar Hendy said:

What I did is that I have an arbiter and I want to check the aperture time of that arbiter,

The aperture time will depend on the nature of the two input signals. In other words, the aperture time is a function of the magnitude and transition times of the two asynchronous signals applied to its two inputs.  I did not read any description of your input signals to the arbiter, nor how you are applying signals to it, so I can not provide any specific comments. However, I believe the methodology I referred you to in the prior post is relevant and will provide an estimate of the aperture time. Measure the propagation time of your arbiter as one input is delayed relative to the other input (and vice-versa) in very small delay increments. Plot the propagation delay as a function of the delay between the two input signal edges and determine where the propagation delay increases by about 1% from its value where the two signal edges are far from one another. Do this for both signals.

Hagar Hendy said:

 checked how closely in time the two transitions of input signals are to the inputs of the arbiter, I chose transient noise with fmax:1MHZ, but I didn't see any effect on the noise on the signal,

Once again, I need more details on your simulation. The transient noise will serve to phase modulate the threshold crossings of your two input signals and modulate the gain of the arbiter.

If the input to the arbiter is differential and the two driving sources are identical, then the presence of common mode transient noise may not make any difference in the arbiter response. Secondly, the threshold crossings of your two inputs will be phase modulated in relation to the slopes of their transition times at the threshold crossing point. For example, if your inputs have a slope of 1 V/ns at their transition times and the noise on your vdd line has a maximum amplitude of 5 mV, the time of the threshold crossing will move by about 5 mV/(1 V/ns) =  5 ps. When you bandlimited your noise to 1 MHz, the 5 mV amplitude changes are limited to under 1 MHz - which means unless you are simulating for over 1 us, the 5 mV change will never be observed in your simulation.

Hagar Hendy said:

as well as how to use parametric analysis in mastero to sweep one of the input and the other will be constant. 

I did not see anything relating to this in your attached figure. My apologies if I am overlooking something.

Shawn

thanks a lot for your clear explanation, appreciate it,  yes The slope of the input signals is 10GV/s which is 10V/ns, and the the time of the threshold crossing will move by about 0.5 ps, so it will not be observed by bandlimited noise to 1 MHz with simulating over 1us. 

 I want to know how to set the noise level as I see the noise factor, for example, I set that to 10,  but I don't know what that corresponds to.

The arbiter is shown below: so from my understanding, there is no common mode in that design, is that right

i will update you regarding the other question that you asked, and really appreciate your help and support

here is the setup that i mean for the parametric analysis, as you stated  above " variable for one of the delays and vary its value in very small time increments over a range where the two input signals are nearly coincident "  

Dear Hagar Hendy,

Hagar Hendy said:

 I want to know how to set the noise level as I see the noise factor, for example, I set that to 10,  but I don't know what that corresponds to.

Do you means the noisescale? From the spectre transient noise manual, the noisescale is a common factor applied to all your noise sources to amplify (or attenuate if less than 1) the device noise. If you want to see its effect, perhaps add a noisy 1K resistor connected tn parallel  with a 1 pf capacitor whose second terminals are at ground. At 125C, the rms noise of the common node between the noisy1K resistor and 1 pF capacitor should be about 74 uVrms. (see Figure 1). With a value of noisescale other than 1.0, the rms value should increase or decrease accordingly.

Hagar Hendy said:

The arbiter is shown below: so from my understanding, there is no common mode in that design, is that right

Yes, this appears to be a single-ended circuit.

Shawn

FIgure 1

Dear Hagar Hendy,

Hagar Hendy said:

here is the setup that i mean for the parametric analysis, as you stated  above " variable for one of the delays and vary its value in very small time increments over a range where the two input signals are nearly coincident "  

This looks good as long as the two signals have edges that will be coincident at some time between delay = 49.995 ns. and 50.000 ns, You might also want to reverse the direction of the crossing to validate that the responses are the same and consider not only a low to high set of transitions, but a high to low set of transitions.

Shawn

Dear Hagar Hendy,

Hagar Hendy said:

here is the setup that i mean for the parametric analysis, as you stated  above " variable for one of the delays and vary its value in very small time increments over a range where the two input signals are nearly coincident "  

This looks good as long as the two signals have edges that will be coincident at some time between delay = 49.995 ns. and 50.000 ns, You might also want to reverse the direction of the crossing to validate that the responses are the same and consider not only a low to high set of transitions, but a high to low set of transitions.

Shawn