[SOLVED] Need Advice on Measuring ENR

[cmontoya]

Hello Folks,
I have a Noise source I would like to determine it's ENR. I have spent some time look at this issue and here is what I have come up with ENR=(Thot-Tcold)/290k. I can measure Thot by looking at Output power when the source is on.

How do I measure Tcold when it's less than the noise floor of my spectrum analyzer?

What if any other problems do you see in Measuring ENR using a Spectrum analyzer ?

What options exist for determining the ENR of a noise source ?

 

[RCinFLA]

T_cold is easy. It is the physical temperature of the termination impedance.

Measuring T_hot is more difficult. You need to know the NF and gain of the amplifier you use to takeover the rather poor NF of the spectrum analyzer.

Spectrum analyzer input noise figure is in the range of 15 db to 25 db without any input attenuation engaged. Some analyzers have a low noise input amp that can be engaged when needed, at expense of lower dynamic range on analyzer.

For accuracy, you should have greater then 10 db noise rise to analyzer noise floor with external amp gain and its input terminated with proper termination to establish T_cold floor. Noise rise when noise source is turned on should should have greater then 10 db rise over Tcold noise floor. For good Y factor measurement this means the noise source should have an ENR greater then 10 db more then the NF of device being measured.

 

[cmontoya]

Thank you for your reply. I have a non standard setup with out much flexibility I am unable to measure the temperature of the cold source. And I have a horrific amount of gain Built into my "noise source" as well.

I have noticed that the that Th-Tc/290k ≈ Th-290k/290k, Due to how small Tc will be.

I am wondering if I can use this aprox. value for ENR to preform Y factor measurements, with any sort of accuracy.

 

[jiripolivka]

To determine an unknown ENR with a not well known receiver/power meter/spectrum analyzer is difficult.
The best way is to use another noise source with a known ENR, and compare the indicated power.
In microwave radiometers, a calibration noise source is used which inject a known noise temperature to the instrument. It is typically compared with the ambient noise temperature, ~290K. The best way is to add another noise temperature; liquid nitrogen cooled lad is best but there are other options.
Noise gas-discharge tubes are usually generating ENR ~15 dB but over frequency, variations due to tube waveguiding structure, can be > +/- 1 dB. Avalanche noise diodes can generate ENR ~30 dB but they are combined with attenuators to get ENR ~15 dB as a comparable source for noise-figure meters.

 

[RCinFLA]

I have 15.4 db ENR noise source and a high power amplified noise source.

The amplified noise source is rated for total dbm power output over 1 to 1500 MHz which is relatively useless spec. Rough dividing output dbm spec by bandwidth of amp yields about 85 db ENR. I have a precision step attenuator that I place on output of the high power amplified noise source. I calibrate it against the 15.4 db ENR precision noise source by feeding it into a 4.5 db NF, 30 db gain broadband amp feeding a spectrum analyzer. Just adjust attenuator until it matches the known precision noise source. From that point, the more atten you take out, the more the ENR the output, db for db, from reference calibration point.

You need to be sure you have at least a minimum of about 20 db of attenuation after an amplified noise source to ensure the output impedance stays good when amp/noise diode is switched off.

A high power amplified noise source is a very useful device. You can eliminate the need for a tracking generator for doing passband frequency response measurements on filters.

 

[cmontoya]

Thank you for your help, it is much appreciated.

I have another related question. I know if I change the Resolution Bandwidth of my spec an, I am simply collecting more noise per bin.

What if any other effects do I need to be aware of when I use different RBW's to measure the noise floor ?

What considerations should I be making when picking the RBW to preform this measurement ?

 

[jiripolivka]

In noise radiometry, noise bandwidth is NOT taken as -3 dB bandwidth. It is taken as "linear" bandwidth to < -1 dB (1 dB means 26% down from 100%).

So the RBW change is not directly proportional to noise power change.
This is why I suggested (and one of the followers above confirmed) to use at least two well known noise sources, and handling the noise power in ratio terms (dB).
Also, the notice of an attenuator between the noise source and load is important, to conserve output match between noise on/noise off states.

 

[RCinFLA]

Modern noise sources are just a noise diode followed by an attenuator, typically 20db or 30 db pad.

Amplified is same thing with a broadband amp following noise diode. The diode and amp is switched on and off. There is not usually an attenuator after an amplified noise source so it is very important to have a minimum attenuation pad to establish a good termination impedance to device under test.

Amplified noise sources have more gain variation over the useful frequency range so calibration needs to be done across the frequency range. You also have to be careful about turn on response with amplifed noise source if using with a noise figure meter which typically chops the noise source on and off at 1 kHz rate.

---------- Post added at 22:21 ---------- Previous post was at 22:09 ----------

Thank you for your help, it is much appreciated.

I have another related question. I know if I change the Resolution Bandwidth of my spec an, I am simply collecting more noise per bin.

What if any other effects do I need to be aware of when I use different RBW's to measure the noise floor ?

What considerations should I be making when picking the RBW to preform this measurement ?

The best setup for calibration is a relatively wide RBW to minimize jumping around of spectrum analyzer display, 100 kHz should be fine. Put the analyser in manual sweep mode with knob control of center frequency with a narrow video filter to get a stable 'dot' on the screen. The more jumping the dot has the more error in making the measurement.

 

[cmontoya]

Hello Folks,
Your help has been fantastic I had to step away from this project due to other obligations but I am back now. I have successfully measured the ENR using the method suggested by jiripolivka, that is comparing the ENR of a known source with my own; here is a write up on it I found here Untitled. I then preformed a Y-factor measurement as specified in the agilent app note 57-2, with my Noise source and it correlated very well with the results from the Reference noise source.

cmontoya