The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Measurements >> RF Measurements >> ZIF NF Measurement -3db ?
https://designers-guide.org/forum/YaBB.pl?num=1164118091

Message started by DBRF1 on Nov 21^st, 2006, 6:08am

Title: ZIF NF Measurement -3db ?
Post by DBRF1 on Nov 21^st, 2006, 6:08am

Hi
We are using a standard NF meter to measure the NF of our ZIF reciever IC { 5GHz to analog BB }in the Lab.
Thus we can measyre from LNA in to either I or Q out {but not both simultaneously }
So my question isif the measured NF reading is 5dB does this mean the overall "Receiver NF" is 2dB or
how do we account for the un measured quadrature channel ?

Thanks

DBRF1

Title: Re: ZIF NF Measurement -3db ?
Post by Jess Chen on Nov 21^st, 2006, 8:48am

It does not matter whether you measure noise figure with one or both I/Q outputs. If you use both, you double the signal but you also double the noise; the NF remains the same.

The more relevant question is whether you are measuring single side band or double side band noise figure. The difference is 3dB. The SSB noise figure will be higher because noise from either sideband contributes to the I (or Q) output but the SSB noise figure calculation only accounts for noise on one sideband. The DSB calculation accounts for noise in both input sidebands and is therefore 3dB lower than the SSB measurement. Since for ZIF, both sidebands contain signal, the DSB NF is the appropriate measurement.

Title: Re: ZIF NF Measurement -3db ?
Post by DBRF1 on Nov 21^st, 2006, 11:41am

Jess thanks for the reply
A few questions to double the signal you assume we shift the I {or Q by } by 90 degs then we can add 2 signals in phase hence get 6dB increase in voltage.
Does the same apply to the noise . Doesn't it depend if there is a noise contribution from the mixers thus the I noise will not be totally correlated with the
Q noise so it is not true to say that the noise on the I branch is in Quadrature with the noise on the I branch ??/

Also wrt the DSB/SSB NF comments. My understanding was that for a ZIF rx there is NO image band so in fact for this type of receiver the DSB and SSB NF are the same as ther is in fact only one Signal band . { Image = Wanted +/- 2*IF , IF = 0 thus image = wanted } ??

DBRF1

Title: Re: ZIF NF Measurement -3db ?
Post by Jess Chen on Nov 21^st, 2006, 1:33pm

We are interested in noise within a relatively narrow bandwidth. Narrow band noise can be expressed as
x(t)*cos(w*t)-y(t)*sin(wt) where "w" is the carrier frequency in radians per second. These are the I and Q components of the input RF noise of interest by definition. The question is whether x(t) and y(t) are correlated. If we assume the RF noise is wide sense stationary (i.e. that the statistics do not depend on time), one can show that x(t) and y(t) are indeed uncorrelated. (That does not necessarily mean that x(t) and y(t+tau) are uncorrelated.) The paper on RF modeling posted to this website touches on this derivation. If you read that paper, note that equation 53 has some typos.

Do you really need to combine the I and Q outputs to measure noise figure? How are you driving the RF input when you combine the signals?

Regarding the DSB/SSB issue, I suggest that you determine whether your measurement injects noise on both sides of the carrier. Ideally, you should inject noise over a band that corresponds to your signal band. If your signal occupies bandwidth on either side of the carrier, your noise source should too. If it does not, your noise figure measurement will be too high.

Title: Re: ZIF NF Measurement -3db ?
Post by ACWWong on Nov 22^nd, 2006, 12:14pm

In ZIF recievers I have worked on (for GSM/EDGE) we always measured noise figure on I or Q (the results being the same). This is what we quoted on datasheets/what the baseband DSP guys wanted.

DBRF1 wrote on Nov 21^st, 2006, 11:41am:

Also wrt the DSB/SSB NF comments. My understanding was that for a ZIF rx there is NO image band so in fact for this type of receiver the DSB and SSB NF are the same as ther is in fact only one Signal band . { Image = Wanted +/- 2*IF , IF = 0 thus image = wanted } ??

I agree with Jess's comments, i have measured the NF of a ZIF transciever in two ways:
1) C/N method: Using a CW carrier one side of the LO, so had to adjust the NF by -3dB to account for the fact the modulated input signal would occupy both sides (the noise was both side of the LO... its a bit diffcult for it not to be).
2) Measuemnt with noise diode (Y-factor method) gives the same result... without the need for the 3dB adjustment.

The treatment of this in Razavi's "RF Microelectronics" p183 is quite good, and aglient do a good app note on y-factor noise measurement.

DBRF1 wrote on Nov 21^st, 2006, 11:41am:

Jess thanks for the reply
A few questions to double the signal you assume we shift the I {or Q by } by 90 degs then we can add 2 signals in phase hence get 6dB increase in voltage.

I used to think of it as adding signal I and Q in vector form, so you add as squares (I^2+Q^2).... noise also adds as squares... not sure if my loose thinking is strictly correct but i managed to convince myself quite easily ;)
Best way to convince yourself is if you have a dual channel VSA (Vector signal analyser) like HP89441 or R&S FSIQ... then you can measure NF on I or Q using C/N method with -3dB correction factor, and then put the VSA in I+jQ demodulation mode to see it give you the same/similar NF or C/N.

cheers
aw

Title: Re: ZIF NF Measurement -3db ?
Post by aaron_do on Nov 29^th, 2006, 8:01pm

Hi all,

I managed to prove to myself that the zero-IF receiver has a minimum noise figure of 3 dB but since you managed to measure differently i guess i'm wrong...

Anyway quadrature mixing is required since the phase difference between the LO and RF causes signal loss due to folding. The noise does not experience the same loss however and so the noise figure takes a 3 db hit. Its easy to prove using a frequency modulated signal multiplied by a LO signal of equal frequency and unknown phase. The signal amplitude peaks at a certain phase and equals zero 90 degrees later. The noise however does not experience the same loss and so NFmin = 3 dB. I think all the equations you need to prove it are found in Digital and Analog Communications Systems by Couch.

Anybody know where i went wrong? I can send equations...

cheers,
Aaron

Title: Re: ZIF NF Measurement -3db ?
Post by cheelee on Nov 30^th, 2006, 5:37pm

Maybe you meant (I have not worked out the maths) ZIF demodulator/mixer, but put a LNA with NF<3dB and infinite gain in front of your ZIF mixer/demodulator, you get receiver NF<3dB.

I have used 'similar' approaches described by ACCWong, except that I have not measured the NF on I or Q channels only, always I+jQ.

Title: Re: ZIF NF Measurement -3db ?
Post by aaron_do on Nov 30^th, 2006, 8:34pm

Hi,

according to the math, it doesnt matter whether or not you have an ideal LNA in front. If you halve the SNR (by multiplication not by addition of noise/signal) your minimum NF is 3 dB regardless of the preceeding stages.

I am not familiar with noise measurement with a noise diode (Y-factor method) so i don't know about the 3 dB adjustment.

See if i can rework the maths and post it...maybe theres a problem there.

Aaron