effect of noise figure on phase noise

[WimRFP]

oscillator phase noise was -110 dBm/Hz at 10 dBm carrier power.

Consider we take the noise of the switch into account when it reaches 1/4 of oscillator phase noise - 6 dB below.
So 110-6-(-174) = 48 dB.
We dont care about switch noise figure until it is 48 dB. Or about 20000 times more than given 4 dB... That's just ridiculous
So why do we tend to minimize it if it is so negligible?
Read carefully about SNRin and SNRout: https://en.wikipedia.org/wiki/Noise_figure

What did I miss from your reference? Who is "we"?

Acceptable Noise Figure depends on your application. If you are working with low level signals it does matter generally. Example: Switch between UHF antenna system and low noise amplifier.

When dealing with signals in the dBm range F is not important in most cases. Example Switch between a PA and an antenna system, here other things like power handling and IMD are of importance.

 

[biff44]

If you look at the added phase noise of the amplifiers given in the link they are all way below your -120dBc/Hz so you would not see any degradation there. You dont give the operating frequency or offset I guess offset about 100Hz.
As for a switch I have never measured the added phase noise, I've always been more concerned about other affects that could degrade the signal integrity. I have generally used diode switches but have never seen any PN degradation using either FETs or diodes as switches in relatively low noise signal sources at high frequencies.
I think you will find there are more important things to worry about in designing a local oscillator than the added phase noise of any switches, isolation for one.

@Mityan
In this case it is negligible. The signal levels are so much higer than the noise added by the 4dB loss. If it were a low noise front end it would be a completely different story, there the input noise (sky noise) is very low and the switch loss and its added noise are significant and can not be ignored as the signal to noise ratio will be significantly impaired.

Peter

As switches operate fully saturated (either fully ON or fully OFF), they do not really impart much phase noise to a signal. There would be a little white noise due to the effective resistance, but unless you were overdriving the snot out of the switch with high power, I would not expect to be able to measure its addition to a synthesizer's phase noise power.

 

[Mityan]

The question was:
Imagine that there is a signal with phase noise for example -120dBc/HZ.The signal goes through some microwave part for example switch with noise figure=4 dB.then what is the phase noise of output signal???

I still think it will be -120+4= -116 dBc/Hz according to the NF formula, NOT 120 or 119.999 dBc/Hz. As I showed in my previous post, according to your logic the degradation of carrier-to-noise may be ignored (or cannot be measured) even with 10-20-30 dB of switch NF. Is it true?
Or is 4 dB a real value for switch NF if you tell that never observed a degradation?

 

[volker_muehlhaus]

I still think it will be -120+4= -116 dBc/Hz

If this would be true, any attenuator in the oscillator output would also degrade the phase noise by that amount (NF = attenuation).

 

[WimRFP]

@Mityan: The formula for noise figure (F) is based on the maximum noise production of a resistor, please read your own reference. "standard noise temperature" is of importance here.

F = 10 log (1+Te/To) = 10 log (1+Pnadded/Pnresistor) as noise power is proportional to noise temperature. Pnresistor = -174 dBm/Hz. When using these formulas (see also your reference), you can forget the definition based on SNR and that may solve your confusion.

The switch just gives some attenuation and adds (additive process) some noise. If you are familiar with phasor representation, you just get a small vector in series with your carrier vector. The length of that small noise vector changes (representing the envelope, rayleigh distributed) and the phase of that vector changes randomly over 360 degrees.

As this vector adds to the carrier vector, both amplitude and phase of the resultant vector changes somewhat, in the same way as noise adds to constellation points in for example QAM systems.

When you double the length of the carrier vector (that is 4 times the power), the small noise vector does not change power, so rms length does not change. Now the phase variation due to the noise vector halves (so phase noise reduces with 6 dB).

So increasing the input power to the switch, reduces the phase noise contribution because of the switch's noise production.

 

[Mityan]

If this would be true, any attenuator in the oscillator output would also degrade the phase noise by that amount (NF = attenuation).

No. The formula G*A + (G+NF)*N also suits the G<1 (or negative in dB, as it is in attenuator). You said it: (NF = attenuation).
So SNR will degrade whenever you amplify or attenuate the signal.

About what WimRFP said, I should think a little.

 

[G4BCH]

@ Mityan
Going back to your post at 0902
While noise figure is defined as
IP SNR / OP SNR
Looking simplisticly at the problem
The input to the device is
IP = S + Ni
and the output is
OP = G*S + G*Ni + G*(KT+NF)
Where G is the gain of the device
S is the input signal
Ni is the input noise
KT is the noise of a resistor at 290K (-174dBm/Hz)
NF is the noise figure
All the above in dB

The device has no way of separating the input noise from the input signal and processing them differently; the are both signal as far as it is concerned.
The degradation in signal to noise ratio comes from the amplified KT + N

SNR O/P G * S /(G * (Ni + KT + NF) )
SNR I/P S / Ni

The SNR is degraded by KT + NF

As WimRPF says working in temperature can be useful to show the degradation as in increase in noise temperature. This is especially useful with sub 1dB noise figures and low sky noise systems where the signal to noise ratio does not follow a dB for dB change in noise figure.

I see you posted while I was composing this. The above can also refer to you most recent post.
Peter

 

[biff44]

Mityan, your mistake is that you are trying to add NF to phase noise in dBc/Hz. They are not the same thing. NF refers to an actual power (i.e. a 4 dB NF is +4-174 dBm/Hz =-170 dBm/Hz). Note the units are not the same. "dBc" is a measure relative to the carrier power.

To compute the actual additive phase noise, you have to convert everything into powers and add them there.

Also, keep in mind that only half of the power generated by the NF of a device is Phase Noise, the other half is AM noise.

 

[BigBoss]

Interesting...
Someone says Noise Figure would be added to Phase Noise...
It's completely wrong by definition.In the definition of Noise Figure, there isn't any frequency related terms nevertheless it's been related directly to Phase Noise.Besides, Phase Noise should have a offset frequency, tell me where it is ??
OK, NF of active device which is used in VCO will degradade the Phase Noise but there isn't 1:1 direct relation such as PN=PN@Fref+NF...
For more detailed explanation please look at Leeson's approximation.

 

[tony_lth]

But the RF switch surely should have a noise floor, I remember it is may be at the level of -140dBc/Hz@10GHz, my memory seems from some vendor's datasheet. So you should tell me the target specifications of PN at all freq offset, such as @1KHz, @10KHz, @100KHz etc.

 

[Mityan]

Well, I was wrong. This may be misunderstanding or inattentive reading of some papers such as swra030 from TI etc.
Thanks for all

I used my formulas for estimating a receiver sensivity, but actually sensivity assumes that input noise is nothing else but thermal