The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> RF Simulators >> Simulate Phase noise DLL
https://designers-guide.org/forum/YaBB.pl?num=1092399689

Message started by Nico67 on Aug 13^th, 2004, 5:21am

Title: Simulate Phase noise DLL
Post by Nico67 on Aug 13^th, 2004, 5:21am

Hello, I want to simulate the phase noise of my DLL.

In first, I use the simulation PSS and Pnoise with my VCDL. I obtain a result in dBc/Hz, and with some calculations I obtain the phase noise in ps.

But I don't sure that the result is just because the jitter is approximately 0.15ps. I think that I not use a good transformation for pass dbc/Hz to ps.

If you can explian the calculation for this transformation

Thank

Title: Re: Simulate Phase noise DLL
Post by Jitter Man on Aug 13^th, 2004, 7:39am

To characterize the jitter of a VCDL you would use the same procedure you would for a PFD, which is described in Modeling jitter in PLL-based frequency synthesizers at www.designers-guide.com/Analysis.

[glb]Jitter Man[/glb]

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Aug 13^th, 2004, 7:54am

Thanks, jitter man

but I have some questions:

I use the method describe in this guide, and I obtain a curve for the phase noise in dBc/Hz.

I have a clock at 62.5MHz and for the simulation, i want to simulate the phase noise of the buffers.

I obtain, for example -165dbC at 100KHz eg 0.001ps that I can't be exact.

1. How choose the value of delta f ?
2.Why don't integrated the curve of the phase noise?

Thanks

Title: Re: Simulate Phase noise DLL
Post by Jitter Man on Aug 13^th, 2004, 10:58am

Sorry, I misled you a bit. You should not follow the procedure for extracting jitter of the PFD, which has an analog output, you should follow the procedure for extracting jitter of a frequency divider, a purely digital circuit. This is section 4.1 in the given reference. Notice that it indicates you should integrate the noise density from -f₀/2 to f₀/2.

[glb]Jitter Man[/glb]

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Aug 16^th, 2004, 12:50am

Thanks

I simulate with Cadence RF Pnoise

In sweeptype I use relative and in Sweep range I use
min 0 max Fo/2 . Are you ok?

Input source none because I want simulate the contribution of noise of the buffers in my VCDL.

In positive output node i use a output of my VCDL.

To this output I see the contribution of all the buffers or just the buffer corresponding to the node?

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Aug 17^th, 2004, 6:48am

Please I need help to calculate the phase noise in ps.

When I have the curve in dBc/Hz I don't know how to transform this phase noise to ps. In the analisys , you use a c=L(f)*deltaf^2/fo^2 but how choose the value of the deltaf?

Title: Re: Simulate Phase noise DLL
Post by Frank Wiedmann on Aug 17^th, 2004, 8:35am

You should probably use the strobed noise analysis of SpectreRF. See http://www.designers-guide.com/Analysis/PLLnoise+jitter.pdf, in particular section 9.

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Aug 17^th, 2004, 10:28am

I have read all the analysis but I don't understand how choose the delta(f) in the formule.

I make the simulation Pnoise and I obtain the curve Phqse Noise dbc/Hz.

In the example of the analysis you explain that you obtain -95dBc at 100KHz. But I don't know that you choos e the 100KHz.

In other analysis they integre the curve from fc to 10MHz, and the result is not the same.

Thanks for your informations

Title: Re: Simulate Phase noise DLL
Post by Frank Wiedmann on Aug 17^th, 2004, 11:46pm

The rule how to choose delta(f) is given immediately after equation (77) in the paper I mentioned. However, this should not bother you at all because a DLL is not an autonomous system. There is no delta(f) in the equations in section 9.

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Aug 18^th, 2004, 2:38am

Thank Frank,

The phase noise of the buffers is in the Jitter type synchronous and I need use the period jitter corresponding for a PLL to the PFD.

It is just ?

Title: Re: Simulate Phase noise DLL
Post by Frank Wiedmann on Aug 18^th, 2004, 4:03am

Use the same procedure as for the divider (as already mentioned by Jitter Man), see section 9.1.1.

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Aug 18^th, 2004, 6:38am

Ok How determine dv/dt with the PSS simulation ?

Thanls

Title: Re: Simulate Phase noise DLL
Post by Frank Wiedmann on Aug 18^th, 2004, 6:51am

You can use the calculator functions deriv() to get the derivative and value() to pick the correct t_c.

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Aug 18^th, 2004, 7:32am

But with PSS simulation
I don't have a curve , I have curve with the harmonic range.

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Aug 18^th, 2004, 7:33am

Ok I have find in sweep I use time and not spectrum

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Aug 18^th, 2004, 7:44am

In Pnoise simulation where did I use the value of tc?

Are you sure that the phase noise of the VCDL is the same of a divider, because the noise is add in one clock period and when a new clock came then the noise is noun.

I have see in a these the expression J=(To/2pi)^2 * integ(Sn)

This is very important because I'm in training stage and you're the only persons to help me.

Thank

Title: Re: Simulate Phase noise DLL
Post by Frank Wiedmann on Aug 18^th, 2004, 8:03am

Set Noise Type to timedomain, enter a large number for Noise Skip Count (or 0 for Number of Points), select Add Specific Points and enter the value of t_c there.

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Aug 18^th, 2004, 8:11am

Thank frank,

tc is the value to the beginning of the curve obtained with PSS or the value absolute of the beginnig of the signal

Title: Re: Simulate Phase noise DLL
Post by Frank Wiedmann on Aug 18^th, 2004, 8:22am

t_c is the time where you want to measure the jitter, usually the time where the signal crosses the theshold level of the following stage. You have to find t_c by performing a PSS analysis. Then you activate the Pnoise analysis for the following simulation run, while leaving all parameters of the PSS analysis unchanged (so that t_c will not change).

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Aug 18^th, 2004, 11:47pm

Ok i have do the pnoise simulation

Now I make Results-Direct plot - PSs

In PSS results I choose tdnoise and what is the frequency to choose?
When I plot I have timeindex=0 and I think that is timeindex=tc?

Thank

Title: Re: Simulate Phase noise DLL
Post by Frank Wiedmann on Aug 19^th, 2004, 2:01am

Select tdnoise, Integ Output Noise, Total Noise and Magnitude. Set Start Frequency to 0 and Stop Frequency to f₀/2, where f₀ is the frequency of your PSS analysis. The result will be the square root of equation (55), which you only have to divide by dv(t_c)/dt according to equation (54) in order to obtain the jitter.

Your sweep range in the Pnoise analysis should also be from 0 (or a very low frequency) to f₀/2. If you have 1/f noise in your devices, you should probably use a logarithmic sweep where a lower limit of 0 is not possible.

Regarding the timeindex: If you entered a large value for Noise Skip Count, you should have for timeindex the values 0 (which you can ignore) and t_c (which you entered in Add Specific Points). In this case, you should of course use timeindex=t_c.

If you entered 0 for Number of Points, you will only have timeindex=0, which is really timeindex=t_c. Any additional points you might have entered will also have a timeindex value that is too low by t_c (if t_c is the lowest value you entered). This is a bug in SpectreRF that I reported to Cadence some time ago in SR 32827714, it is now handled as PCR 740140.

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Aug 19^th, 2004, 4:40am

Ok, I have just one point in the curve is is just ?

I have 5.23m this is possible?

For dv/dt I have 15G , thus J=2ps rms

I'm surprised by this result, this result it is the jitter in all my VCDL or for just one buffer, i have simulate with all my buffer

Title: Re: Simulate Phase noise DLL
Post by Frank Wiedmann on Aug 19^th, 2004, 5:39am

The result is the jitter of the output signal that you have specified in your Pnoise analysis. This must of course also be the signal that you use to determine t_c and dv(t_c)/dt.

This simulation gives you the jitter caused by the intrinsic noise of your devices. In practice, you will generally also have other effects that will cause jitter, like for example disturbances on the power supplies or intersymbol interference due to limited bandwidth for data signals.

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Aug 19^th, 2004, 5:55am

In some thesis about DLL, they calculate the jitter by using the (TO/2pi)^2*Sn . They consider the phase noise in the VCDL like a accumulatimg jitter.

I find like them in dBc/Hz but then I want the jitter in ps I don't obtain the same.

The jitter is accumulating in one period of clock, that why I think I need to use the rule of the analysis of VCO.

Title: Re: Simulate Phase noise DLL
Post by Frank Wiedmann on Aug 19^th, 2004, 6:23am

A DLL by itself is not an autonomous system, so I do not think that the analysis in section 10 of http://www.designers-guide.com/Analysis/PLLnoise+jitter.pdf applies here. However, the input signal to the DLL might come from a VCO that exhibits accumulating jitter.

Of course, the jitter will generally increase over the length of the delay chain, as described for the case of the ripple counter at the top of page 28. However, this "jitter accumulation" is fundamentally different from the accumulating jitter in an autonomous system like a VCO.

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Aug 19^th, 2004, 6:44am

The thesis are for the DLL use with a edge combiner and in the edge combiner the jitter accumulate in one clock.

I calculate the phase noise in the output of the VCDL maybe the jitter in the beginning or beetween is more important

Thank

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Sep 3^rd, 2004, 12:07am

Hello, I use the method with noise type time domain but when I see the results I have the noise in V^2/Hz but for jitter I need to have dBc/Hz.

Thanks for your explication

Title: Re: Simulate Phase noise DLL
Post by Frank Wiedmann on Sep 3^rd, 2004, 1:41am

Jitter is usually measured in seconds and not in dBc/Hz. I have already explained you how to calulate the jitter in reply #20. You seem to have applied this method correctly in reply #21 (at least, the values look reasonable to me).

As I mentioned in reply #22, there usually exist other effects that will cause additional jitter. In many cases, the jitter caused by device noise is not dominant.

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Sep 3^rd, 2004, 2:20am

I am ok but in the equation (54)

J=squart(var...)/dv/dt the unit are not good for me.

Because J is second

the unit of dv/dt is Volt/second and Squart(...) is V.s^(1/2)

thus J is in s^(3/2)
i thimk

Title: Re: Simulate Phase noise DLL
Post by Frank Wiedmann on Sep 3^rd, 2004, 4:37am

If you follow the steps given at the beginning of reply #20, you will get a result with a unit of volts. When you divide this by dv/dt, you will get seconds.

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Sep 3^rd, 2004, 6:21am

I simulate all my DLL

When I do the result PSS/ tdnoise /integ output noise/magnitude/Start 0 Stop f0/2 I obtain a point at 1.29m but not units.And I have dv/dt=10G that gives J=0.2ps and I think that J must be more important.

In the reply 20, you indicates to choose total noise maybe it is this becausee i don't have option total noise

or maybe when i indicates the specific point in the pnoise

Thanks

Title: Re: Simulate Phase noise DLL
Post by Frank Wiedmann on Sep 3^rd, 2004, 6:46am

The unit is not shown but it is volts. You might want to double-check if you have specified all parameters of your simulation correctly (e.g. the timepoints for which you perform the tdnoise analysis). The Total Noise option might not exist in older versions (I am using 5.0.33), you will probably get the total noise by default in your version.

The value you get is the standard deviation of the Gaussian distribution. The peak value will be higher, depending on the probability with which it occurs. Like I said, in many cases there are other effects that generate larger amounts of jitter than the intrinsic device noise.

Title: Re: Simulate Phase noise DLL
Post by Nico67 on Sep 3^rd, 2004, 6:58am

Ok , Frank if you have time can you see this ( it is in french)
'
http://pastel.paristech.org/archive/00000619/01/Manuscript.pdf

in page 89 , he have a spectral density Sn = -135 dBc si he have 6,4ps

Very thanks

Title: Re: Simulate Phase noise DLL
Post by Frank Wiedmann on Sep 3^rd, 2004, 8:22am

The formula given there applies to a sine wave with stationary noise but not for the general case of an arbitrary periodic signal with cyclostationary noise.