Phase Noise

Hi.

The simulink block Phase Noise uses noise with a spectrum characterized by a 
1/f slope. I presume that means if the frequency offset is specified at F 
Hz, the noise spectrum is 1/f for frequency offset >= F. Then what about the 
noise spectrum for frequency offset < F? Is it also 1/f or simply flat? If 
it is 1/f, then it should not make any different to set the frequency offset 
at F1 Hz with a phase noise level -P1 dBc/Hz than to set the frequency 
offset at 10*F1 Hz with a phase noise level -(P1-20) dBc/Hz. However, I 
think the Phase Noise block does not give the same level of phase noise with 
these 2 settings. Hence, is it correct to say that the noise spectrum if 
flat for frequency offset < F?

Thanks.

Regards,
Kian Chong 

I would assume (I'm not sure) the phrase 1/F refers to the slope being
-6 dB per octave.  It does not define the actual value, only the slope.
 I am guesing the simulink block is letting you define the actual value
at one offset namely F.  The slope is 1/F both above F and below F and
the value at F is as you define it.

Mark

I agree that 1/F refers to the slope being -6dB per octave. However, if the 
slope is 1/F above and below F, then won't it be the same to specify, for 
example, -60dB at 100Hz or -80dB at 1000Hz?

"Mark" <makolber@yahoo.com> wrote in message 
news:1111627621.031681.108840@l41g2000cwc.googlegroups.com...
>I would assume (I'm not sure) the phrase 1/F refers to the slope being
> -6 dB per octave.  It does not define the actual value, only the slope.
> I am guesing the simulink block is letting you define the actual value
> at one offset namely F.  The slope is 1/F both above F and below F and
> the value at F is as you define it.
>
> Mark
> 

yes,

I see the point of your question now,

my bad

I don't know if this is an issue in simulation, but in real test
equipment it sometimes is....the measurement bandwidth may change over
different frequency ranges.   If the result is always normalized to dBc
/___ Hz___  then it should be the same.   Are your numbers -60 and -80
both dBc/Hz  or are they  dBc in the measurment bandwidth?

 
Mark

"Mark" <makolber@yahoo.com> wrote in message 
news:1111675196.432749.236190@l41g2000cwc.googlegroups.com...
> yes,
>
> I see the point of your question now,
>
> my bad
>
> I don't know if this is an issue in simulation, but in real test
> equipment it sometimes is....the measurement bandwidth may change over
> different frequency ranges.   If the result is always normalized to dBc
> /___ Hz___  then it should be the same.   Are your numbers -60 and -80
> both dBc/Hz  or are they  dBc in the measurment bandwidth?
>
>
> Mark
>
Hi ! I suspect that the phase noise density is intended to go up as 1/f^2 
for frequencies less than the breakpoint ( but I don't have simulink , don't 
use it , never have, so this is just a guess). Why not test it and see? It's 
often worth having a look so you can find out what the model element 
actually does rather than what you think it is supposed to do. You might 
also want to see what happens if you put in values like a -12 dB breakpoint 
at 20kHz offset just to be sure that the integrated phase noise out of the 
model element behaves the way that you think it should.

Best of Luck - Mike

Mark wrote:
> I would assume (I'm not sure) the phrase 1/F refers to the slope being
> -6 dB per octave.  It does not define the actual value, only the slope.
>  I am guesing the simulink block is letting you define the actual value
> at one offset namely F.  The slope is 1/F both above F and below F and
> the value at F is as you define it.
> 
> Mark
> 
You'd better test that assumption, or double-check your terminology. 
Noise density vs. frequency slopes are almost universally stated as 
noise _power_ per root hertz, so nearly all communications or electrical 
engineers will interpret "1/f" noise to mean a slope of -3dB/octave, 
_not_ -6dB/octave.

And yes, 1/f noise is an important contributor to noise in real systems. 
  It's known as "flicker" noise when applied to physical amplifying 
devices, it has had engineers tearing their hair out since the advent of 
semiconductors (as I understand it's not nearly as prevalent in vacuum 
tubes), it _still_ has engineers tearing their hair out, and unless I've 
missed something it's still not fully understood.

-- 

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Tim,

yep,  you're right...

I was thinking 1/F^2.

1/F is a slope of 3 dB per octave or 10 dB per decade.

So that may explain the OP's problem as well.

Also typical 1/F noise becomes significant BELOW a certain frequency
point  so perhaps the simulink tool is providing a break point BELOW
which the 1/F noise begins to rise at 3 dB per octave.

thanks
Mark

Tim is right. 1/F is a slope of 3dB per octave. I simulated with the
phase noise simulink block and indeed specifying -60dBc/Hz at 100Hz is
the same as -70dBc/Hz at 1000Hz and also the same as -100dBc/Hz at
1MHz.

Now I have another question. At 0Hz offset, that is, the desired tone,
the level is 0dBc/Hz, am I correct? If so, there should be a
breakpoint where the slope rises to this 0dBc/Hz point at 0Hz. If this
breakpoint occurs at the specified frequency offset in the simulink
block, then again -60dBc/Hz at 100Hz should not be the same as
-70dBc/Hz at 1000Hz.

"Mark" <makolber@yahoo.com> wrote in message news:<1111686191.836431.294410@g14g2000cwa.googlegroups.com>...
> Tim,
> 
> yep,  you're right...
> 
> I was thinking 1/F^2.
> 
> 1/F is a slope of 3 dB per octave or 10 dB per decade.
> 
> So that may explain the OP's problem as well.
> 
> Also typical 1/F noise becomes significant BELOW a certain frequency
> point  so perhaps the simulink tool is providing a break point BELOW
> which the 1/F noise begins to rise at 3 dB per octave.
> 
> thanks
> Mark

The units of phase noise, dBc/Hz, specify a noise DENSITY at a given
offset.  The desired carrier tone itself is not a density but is a
discrete carrier value (it has infinite density if you want to think of
it that way, all the power of the tone is concentrated at one
frequency, ideally)

The phase noise units dBc/Hz compare the density of the noise (the
power of the noise in a 1 Hz BW)  at a given offset,   to the power of
the carrier tone.

I think what you may be thinking about is the fact that the density of
the 1/F noise (not the tone) also (ideally) approaches infinity at 0 Hz
offset.  This is similar to an ideal integrator that has infinite gain
at DC.  (Black hole of noise???)   In practice, there is a break point
at a very low frequency below which the gain or noise density no longer
rise as the frequency is reduced.   I don't think that is what the
break point in the Simulink model is for. Again I think the Simulink
model is specifying  the 1/F noise slope as  3 dB per octave and the F
parameter allows you to "offset" this line up or down as needed. 

Mark

ckiancho wrote:

> Tim is right. 1/F is a slope of 3dB per octave. I simulated with the
> phase noise simulink block and indeed specifying -60dBc/Hz at 100Hz is
> the same as -70dBc/Hz at 1000Hz and also the same as -100dBc/Hz at
> 1MHz.
> 
> Now I have another question. At 0Hz offset, that is, the desired tone,
> the level is 0dBc/Hz, am I correct? If so, there should be a
> breakpoint where the slope rises to this 0dBc/Hz point at 0Hz. If this
> breakpoint occurs at the specified frequency offset in the simulink
> block, then again -60dBc/Hz at 100Hz should not be the same as
> -70dBc/Hz at 1000Hz.
> 
> "Mark" <makolber@yahoo.com> wrote in message news:<1111686191.836431.294410@g14g2000cwa.googlegroups.com>...
> 
>>Tim,
>>
>>yep,  you're right...
>>
>>I was thinking 1/F^2.
>>
>>1/F is a slope of 3 dB per octave or 10 dB per decade.
>>
>>So that may explain the OP's problem as well.
>>
>>Also typical 1/F noise becomes significant BELOW a certain frequency
>>point  so perhaps the simulink tool is providing a break point BELOW
>>which the 1/F noise begins to rise at 3 dB per octave.
>>
>>thanks
>>Mark

"Phase-locked loop circuit design" by Dan H Wolaver, Prentice-Hall, 
1991, ISBN 0-13-662743-9, chapter 6, "Noise".  There are other good 
books on PLL design which I won't disrecommend, but I took the class 
from Dr. Wolaver when the book was in proofs.

-- 

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com