Designer’s Guide Community

mobil

Junior Member

Offline

Posts: 20
Australia

phase noise of 30G Hz VCO combined a divider
Dec 14^th, 2006, 4:45pm

Hello everyone. I'm involving a GHz project to design a PLL. Now I've got a 30G Hz VCO and a 30G/2 divider based on CML logic. When they're combined together to be simulated phase noise by SpectreRF, I'm confused these questions:

(1). To simulate divider's phase noise, in my understanding, the divider is a driven component same as mixer or LNA. Hence, the 'sweeptype' should be 'absolute' in pnoise setup and the 'oscillator' option in pss should be inactive. But it actually consists of a VCO in circuit, so I still active the 'oscillator' option in pss and selected the 'oscillator node' of VCO's output node. Is that all right under this configuration?

(2). The divider is self-oscillating around 15G Hz without any input signal. The phase noise of this self-oscillation is not good about -70 dBc/Hz @ 1M Hz. But when it's driven by VCO, the phase noise of divider is around -100 dBc/Hz. Which one should be considered?

(3). Actually, the pss is not easy to converge (I've set tstab=40n(the transient shows the circuit is stable in 5n ), gear2only etc. for pss converge), unless the beat frequency is set as 15G(I don't how to set beat frequency correctly). I found the reason is that the pss always gives the stable frequency in 15G. Hence, the final parameters of pnoise I used for divider are: 'sweeptype' = relative, refHarmonic =1 (but 2 when simulate VCO's pnoise), because I found in spectre.out file show that
the pnoise is calculate as `pnoise':freq = 15 GHz*refHarmonic + (1 kHz -> 10 MHz). Anyone has a ideal about this?

I'll be appreciate your response.

Back to top

IP Logged

ACWWong

Community Fellow

Offline

Posts: 539
Oxford, UK

Re: phase noise of 30G Hz VCO combined a divider
Reply #1 - Dec 15^th, 2006, 2:54am

heres my 2ps worth...
mobil wrote on Dec 14^th, 2006, 4:45pm:

(1). To simulate divider's phase noise, in my understanding, the divider is a driven component same as mixer or LNA. Hence, the 'sweeptype' should be 'absolute' in pnoise setup and the 'oscillator' option in pss should be inactive. But it actually consists of a VCO in circuit, so I still active the 'oscillator' option in pss and selected the 'oscillator node' of VCO's output node. Is that all right under this configuration?

The pss form should be set to oscillator. The guess at the beat frequency should be the lowest common divisor of the present frequencies. In your case you should set it to 15GHz (as you divide by 2). Also I would choose the divder output as the oscaillation node. As for the pnoise sweep you can chosse absolute or relative as you deem fit. If you wanted look at the phase noise at 1MHz offset from the 15Ghz divider output, you could set it to relative to harmonic 1, at 1MHz, OR absoulte at 15.001GHz.

mobil wrote on Dec 14^th, 2006, 4:45pm:

(2). The divider is self-oscillating around 15G Hz without any input signal. The phase noise of this self-oscillation is not good about -70 dBc/Hz @ 1M Hz. But when it's driven by VCO, the phase noise of divider is around -100 dBc/Hz. Which one should be considered?

-100dBc/Hz is what is quoted. This is the phasenoise of your vco cascaded with your divider. This will only be the divider noise if you know the VCO noise is lower than -94dBc/Hz (@1MHz offset). So if your VCO noise was say -104dBc/Hz at 1 MHZ offset from 30GHz, and you subsequently measured -100dBc/Hz at 1MHz at the divider output, then you could say the -100 is the divider noise floor.
The noise of a divider self oscillating is always poor.

point 3 is answered above.

cheers
aw

Back to top

IP Logged

mobil

Junior Member

Offline

Posts: 20
Australia

Re: phase noise of 30G Hz VCO combined a divider
Reply #2 - Dec 16^th, 2006, 1:13am

Many thanks aw! especially your answers confirm my design.

But in Ken's paper http://www.designers-guide.org/Analysis/PLLnoise+jitter.pdf, in page 12, the equation (35) and following explanation indicate that one should first perform a PSS analysis to determine the threshold crossing times of divider and slew rate (dv/dt) at these times. Then use SpectreRF's strobed pnoise analysis to compute Sn(f) in (35), then compute phase noise Sphi(f) by using (35) (seems different from VCO's phase noise).

So, the way I used to simulate phase noise of divider seems the way of computing VCO's phase noise, is this fitted the method from Ken's paper? Thanks.

-Mobil

Back to top

IP Logged

ACWWong

Community Fellow

Offline

Posts: 539
Oxford, UK

Re: phase noise of 30G Hz VCO combined a divider
Reply #3 - Dec 17^th, 2006, 4:43am

I think the method for simulating the VCO & Divider using pss and pnoise as i descibed it in my last post is ok. I am not familiar with strobed pnoise analysis, but will now download ken's paper and take a look.
Anyway what is you are wanting to quantify ? The phase noise of VCO or divider or both ? SpectreRF pnoise analyses allows you to seperate out phase noise from the total output noise so that you only get the noise you care about if you are interested in the divider noise.

Back to top

IP Logged

jerrymelb New Member Offline Posts: 3	Re: phase noise of 30G Hz VCO combined a divider Reply #4 - Dec 17^th, 2006, 2:47pm Hi, ACWWONG Interesting topic. If I wnat to know VCO's phase noise, should I set beat freq to be 15GHz, and relative harmonics to be 2? If I want the divider's, then I should set harmonics to be 1, is this correct? Thanks!
Back to top	IP Logged

ACWWong

Community Fellow

Offline

Posts: 539
Oxford, UK

Re: phase noise of 30G Hz VCO combined a divider
Reply #5 - Dec 19^th, 2006, 3:07am

Hi jerrymelb,
if i just wanted the VCO's phase noise, I wouldn't be including the divider in the testbench (although I would put representative VCO output loading such as divider input Z in parallel with other circuits input Z & parasitics etc.). Then it would be a 30GHz pssfund with noise relative to harmonic 1.
if I wanted the divider's noise then you can simulate it driven by the VCO in which case the output for the pnoise is the divider output, pssfund is 15GHz and relative harmonic is 1. The result would have to de-embed the noise of the VCO, or use a behavioural VCO model of known noise as your drive.
if your testbench includes your vco and divider and you can of course set pssfund to 15Ghz, and look at pnoise relative to harmonic 2 at the VCO output node, but this would then be vco noise with additioanl noise conrtibutors from the divider (ie divider components which add noise to the VCO output over the 2*15GHz+sweep freq range).
cheers
aw

Back to top

IP Logged

jerrymelb New Member Offline Posts: 3	Re: phase noise of 30G Hz VCO combined a divider Reply #6 - Dec 20^th, 2006, 6:35pm Thanks, a lot! That explained very well.
Back to top	IP Logged