How do I measure the NF of an LNA without a noise source?

[brmadhukar]

Hi,
How do I measure the NF of an LNA without a noise source? Can it be accurate. I have other LNAs with known GAIN and not NFs.
TIA
brmadhukar

 

[marcomdd]

nf = 10 log (si/ni/so/no)

I think you are not able to measure the NF of a DUT without a reference noise standard.

sorry

p.s. have you the facility of LN2 ?

 

[mmwave]

hewlett packard hp8971c

You can get an HP 8970B for under $5k these days. If that is too much, then try measuring change in amplified noise from a 50ohm load at room temp and when dipped in LN2. Then use the noise formulas to get NF and gain.

 

[marcomdd]

hp 8975a

I think the HP8970B is better than the new NFA (such as 8975) of Agilent.

mmwave, where did you read this notice ?

 

[brmadhukar]

nf measurement

Hi,

NF = SNRin/SNRout = (Si/Ni)/(So/No) = 1/Gain * (No/Ni)

Say I know the temperature.

I measure the gain at some small level say -120dBm.

I can take Ni = kT * (Bandwidth) -- This is one unknown I have
No can be measured on spectrum analyzer.

With this I should get NF. But seems like there is a catch on this I always get a vely low NF when I measuer this way. There should be some problem in explanation or interpretation itself. I know this measure Tone NF and to get the total NF I have to average this over bandwidth.

Madhukar

 

[flatulent]

hp8970b noise figure level meter

With the noise levels of modern amplifiers being so low these days it is difficult to measure it accurately without modern test equipment. One approximate way is to use a spectrum analyzer if the lna gain is high enough to make its noise output higher than the analyzer input noise. Use a very wide IF bandwidth to get a high noise level display with a 50 ohm termination on the amplifier input. Then drive the amplifier input with a signal generator. Raise the generator level until the analyzer shows a 3 dB increase of power. Your generator output power will be the same as the integrated noise power over the analyzer IF bandwidth. Make sure you use the correction factor for getting the noise bandwidth of the analyzer at the given 3dB bandwidth setting. If you do not have any clue to this, multiply the 3 dB bandwidth by 1.5 to get a good approximation to the noise bandwidth.

 

[AL0117]

ham radio 10 ghz 8970b noise figure meter

You will need a noise source. Check the articles in the Dubus archive:

**broken link removed**

9002-1 Novel approach to automatic noise figure measurement
9002-2 Construction of a precision noise figure measurement
9003-3 Construction of a precision noise figure measurement system
9101-4 A simple solution for a PANFI


Also you can do a search on RFGlobalnet for "noise figure"
**broken link removed**

There are a lot of related discussions

Cheers, Al

 

[mmwave]

dubus panfi

I think the HP8970B is better than the new NFA (such as 8975) of @gilent.

mmwave, where did you read this notice ?

Search for 8970B and you will find a lot of used ones for sale. I just bought one for $4600. including instrument cal but not a noise source. Noise source depends on your application of course, but are about $500 new.

 

[marcomdd]

standard measuring sound level nf

HP NFM 8970B (fmax=1.6GHz) :
Instrument Noise Figure Uncertainty : +- 0.1 dB
Instrument Gain Uncertainty : +- 0.15 dB
Instrument Match (VSWR): 1.7:1
Instrument NF: 7 dB

AGILENT 8975A (fmax=26.5GHz) :
Instrument Noise Figure Uncertainty : +- 0.15 dB
Instrument Gain Uncertainty : +- 0.17 dB
Instrument Match (VSWR): 2:1
Instrument NF: 10.5 dB

:!:

 

[goxy]

проблемы nfa 8975a

For noise referent sources you can use cosmics radio sources like galaxies, stars, kvazars with mesured noise flux.

Here you some papers about this: https://www.qsl.net/yu1aw/CosmicSrcs.zip

Other articles from this Serbian radio amateur are on:
https://www.qsl.net/yu1aw/published_articles.htm

On this theme try with Google

 

[mmwave]

panfi noise

HP NFM 8970B (fmax=1.6GHz) :
Instrument Noise Figure Uncertainty : +- 0.1 dB
Instrument Gain Uncertainty : +- 0.15 dB
Instrument Match (VSWR): 1.7:1
Instrument NF: 7 dB

@gilent 8975A (fmax=26.5GHz) :
Instrument Noise Figure Uncertainty : +- 0.15 dB
Instrument Gain Uncertainty : +- 0.17 dB
Instrument Match (VSWR): 2:1
Instrument NF: 10.5 dB

It doesn't make a lot of sense to compare an 8970B to an 8975A. A
better comparison might be an 8970B with an 8971C noise figure test set, then at least they both cover the same frequency band. Performance wise the 8975A is faster and also has variable IF bandwidths, rather than just a fixed 4MHz like the 8970B. The user interface is also superior, but the actual sensitivity is not much different. If you can afford it, get the 8975A.

 

[boy]

how to make good nf measurements?

For a broadband noise source, there ia a parameter that i dont know what it mean. the ENR (Excess Noise Ratio), there are 2 of them one with 15 dB ENR and another one aith 5 dB ENR. which one is good for which applications? thank you

 

[marcomdd]

安捷伦8975a

@ mmwave

You are right ,

HP NFM 8970B with HP8971C
Instrument Noise Figure Uncertainty : +- 0.2 dB (@18GHz)
+- 0.4 dB (@26GHz)
Instrument Gain Uncertainty : +- 0.28 dB
Instrument Match (VSWR): 2.5:1 max
Instrument NF: 14 dB max

Regarding 8975:
It also allows automatic downloading of the ENR table inside the analyzer and automatic meaurement & correction of the ambient (cold) temperature of the noise source!

@ boy

ENR = 10 log10( ( Th - T0)/T0 ) where T0 (constant) is equal to 290K and Th is the hot temperature of the noise source.

Basically the noise source with lower ENR (5dB) is obtained from a standard noise source (15 dB) putting in also an additional 10 dB attenuator. For this reason, the noise source with lower ENR has better match and lower difference from the impedances in the two states (on and off or hot and cold).
In addition, a lower power level inside the noise figure meter means a lower error due to the non linearity of the detector.
Also, a lower power level inside the noise figure meter means it will use a low value of its internal variable attenuator reducing so the noise figure of the instrumentation and finallly the uncertainty of the measurement.
If you have to measure DUT with low noise figure (less than 10 dB) it's better to use a 5dB ENR noise source; to measure noise figure up to 30 dB (!) use a 15 dB noise source.

Bye

p.s. download the Agilent application notes; you will solve your doubts.

 

[mrm]

8975a agilent

I have measured NF by building several samles of the LNA, cascading them, measuring the overall gain, measuring the overall output noise with an FSEA (R&S) and by finding the total NF summing -173.8, the Gain and the Output Noise...Inverting FRIIS's formulas I found NF for LNA.
I validated this method with industrial NF Meter and It Works!

 

[mkassem]

nf measure

i am using an hp 8562 spectrum analyser and a noise source to measure NF using the Y method
but i cant get the right gain and right NF figure
i also use a noise figure meter to verify the result
i use hp noise figure meter
i use special function 9.1 and 9.2 to get the power and i calculate the gain and NF figure meter and i get a value
when i use a spectrum analyser i use the noise marker but when i calculate the nf and gain i couldnt get the correct value as the nosie fgure meter

can anyone help

 

[MSH]

noise figure measurement spectrum analyzer

You can use this formula for your masurement

NF(dB)=No-Gd-10 logB + 174 + 1.7
No= Level of Noise power in the output of LNA
Gd= Gain of LNA
B= Frequency Bandwiht

 

[sergio mariotti]

noise meter panfi

to mkassem
Correct!
For my knowledge, SA is involved by (almost) 2 "non-precisions".

1) poor accuracy, typical +/-1.5 dB. There is nothing to do.
2) "Mathematical" error, pheraps may by a sistematic and corrected. It involve the fact that SA perform the mean (average) on dB rather than numbers. There is a very useful AN from Agilent but i don't remember the N°. Pheraps you may reduce this kind of error simply placing SA on [linear scale].

Overall, it is important remember that SA generally have NF in 30 - 40 dB range !

 

[RF_beginner]

noise figure measurement with hot cold resistors

"How do I measure the NF of an LNA without a noise source?"---a crazy ideal......

 

[dinone]

hp 8970b application note

When you measure the NF with this method (using NF definition) it is better to measure the power density (dBm/Hz). This beacuse the RW band in the spectrum is not ideal (rectangular) but real and the spectrum calculate the density with the real filter shape. So the error is lower then with the power. If you don't have the noise source can use a 50 Ohm resistor in the input (the power density noise is p=kT -> -174 dBm/Hz at room temperature). It's clear that if you have to measure a LNA with very low noise (1 dB for example) you can have a error bigger than value.

By