Practical relation between Phase Noise and ADEV
Both ADEV and Phase Noise can be measured around 1 to 10Hz so for my own
understanding I made an empirical graph that relates a certain phase
noise with its impact on the ADEV to allow checking for consistency in
measurements.
All data mentioned is only correct for a signal at 10MHz
The table was made by generating a 10MHz test signal with just the right
level of phase noise at the wanted frequency and measuring the ADEV of
the test signal. Of course it should be possible to calculate this
relation but these calculations are not my strongest point.
And I did not verify how the summing of noise impacts the total ADEV
Phase noise with frequency f has maximum contribution to ADEV at
frequency f*2.
Phase noise of -60dBc at 1Hz contributes maximum 1E-10 to ADEV at 0.5Hz
Every 20dB of decrease in noise level decreases ADEV contribution with a
factor 10
Every factor 10 increase in frequency increases the ADEV contribution
with a factor 10.
A simple graph showing the relation can be found here:
http://athome.kaashoek.com/time-nuts/PNvsADEV.PNG
Erik.
Hi
Since ADEV is a “broadband” measure, mapping it to phase noise is … errr ….
non-trivial. In fact HP ran into a brick wall trying to do this on one of their devices
back in the early 1970’s. This was for ADEV -> Phase Noise.
If you want to go from xDEV to phase noise, it turns out that the Hadamard variance
is a better starting point. There still is a bit of math involved. You can’t just look at
1 sec and get to 1 Hz.
Going the other way ( phase noise -> xDEV ) is a bit more practical. The issue there
is still the broadband stuff. In one case, close in dominates. In another case far removed
drives what you see. The “general” answer is to push a wide range of frequencies
into the measurement.
No this is not as crazy as a jitter measure, but it shares a lot of the same sorts of issues.
Bob
On Jul 22, 2022, at 7:50 AM, Erik Kaashoek via time-nuts time-nuts@lists.febo.com wrote:
Both ADEV and Phase Noise can be measured around 1 to 10Hz so for my own understanding I made an empirical graph that relates a certain phase noise with its impact on the ADEV to allow checking for consistency in measurements.
All data mentioned is only correct for a signal at 10MHz
The table was made by generating a 10MHz test signal with just the right level of phase noise at the wanted frequency and measuring the ADEV of the test signal. Of course it should be possible to calculate this relation but these calculations are not my strongest point.
And I did not verify how the summing of noise impacts the total ADEV
Phase noise with frequency f has maximum contribution to ADEV at frequency f*2.
Phase noise of -60dBc at 1Hz contributes maximum 1E-10 to ADEV at 0.5Hz
Every 20dB of decrease in noise level decreases ADEV contribution with a factor 10
Every factor 10 increase in frequency increases the ADEV contribution with a factor 10.
A simple graph showing the relation can be found here: http://athome.kaashoek.com/time-nuts/PNvsADEV.PNG
Erik.
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
Erik,
On 2022-07-22 17:50, Erik Kaashoek via time-nuts wrote:
Both ADEV and Phase Noise can be measured around 1 to 10Hz so for my
own understanding I made an empirical graph that relates a certain
phase noise with its impact on the ADEV to allow checking for
consistency in measurements.
All data mentioned is only correct for a signal at 10MHz
The table was made by generating a 10MHz test signal with just the
right level of phase noise at the wanted frequency and measuring the
ADEV of the test signal. Of course it should be possible to calculate
this relation but these calculations are not my strongest point.
And I did not verify how the summing of noise impacts the total ADEV
Phase noise with frequency f has maximum contribution to ADEV at
frequency f*2.
Phase noise of -60dBc at 1Hz contributes maximum 1E-10 to ADEV at 0.5Hz
Every 20dB of decrease in noise level decreases ADEV contribution with
a factor 10
Every factor 10 increase in frequency increases the ADEV contribution
with a factor 10.
A simple graph showing the relation can be found here:
http://athome.kaashoek.com/time-nuts/PNvsADEV.PNG
You should consult Enrico's chart:
https://zenodo.org/record/6476977#.YtsgN8FBw6E
This is also included into IEEE Std 1139-2022 in a similar form.
It is key to understand that ADEV does not determine phase-noise slopes.
Yes, there is a relationship, but the troublesome issue is that ADEV in
practice is unable to make white phase modulation and flicker phase
modulation phase noise slopes into different slopes in ADEV, they both
show up as 1/tau slopes despite having different phase-noise slopes.
This means that you with the ADEV slopes does not know if the 1/tau
slope is white or flicker phase modulation. This annoyed David Allan and
his colleagues and it took another 15 years to resolve it and that was
only achieved with the introduction of MDEV.
Another aspect is that any systematic signal like a sine or even a
narrow-band noise-spike will add additional wide-band energy ontop of
the ADEV obscuring the view of the random noise slope behaviours.
So, You can put things in relation sure, but be careful not to
overinterpret your plots and think they replace phase-noise. It's very
hard to achieve that, even if ADEV was developed to serve the purpose.
It's another tool and serve another purpose. I therefore want to see
both, as they provide different information to me.
Cheers,
Magnus - now in summerhouse
I would say that Magnus's explanation is the best one
I have ever read. And he did it without writing any
equations! Yet another great time-nuts posting.
Rick N6RK
On 7/22/2022 3:26 PM, Magnus Danielson via time-nuts wrote:
It is key to understand that ADEV does not determine phase-noise slopes.
Yes, there is a relationship, but the troublesome issue is that ADEV in
practice is unable to make white phase modulation and flicker phase
modulation phase noise slopes into different slopes in ADEV, they both
show up as 1/tau slopes despite having different phase-noise slopes.
This means that you with the ADEV slopes does not know if the 1/tau
slope is white or flicker phase modulation. This annoyed David Allan and
his colleagues and it took another 15 years to resolve it and that was
only achieved with the introduction of MDEV.
The companion paper of the Enrico's chart is here:
https://arxiv.org/pdf/2201.07109.pdf
On Sun, Jul 24, 2022 at 1:38 AM Richard (Rick) Karlquist via time-nuts
time-nuts@lists.febo.com wrote:
I would say that Magnus's explanation is the best one
I have ever read. And he did it without writing any
equations! Yet another great time-nuts posting.
Rick N6RK
On 7/22/2022 3:26 PM, Magnus Danielson via time-nuts wrote:
It is key to understand that ADEV does not determine phase-noise slopes.
Yes, there is a relationship, but the troublesome issue is that ADEV in
practice is unable to make white phase modulation and flicker phase
modulation phase noise slopes into different slopes in ADEV, they both
show up as 1/tau slopes despite having different phase-noise slopes.
This means that you with the ADEV slopes does not know if the 1/tau
slope is white or flicker phase modulation. This annoyed David Allan and
his colleagues and it took another 15 years to resolve it and that was
only achieved with the introduction of MDEV.
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com