[time-nuts] Question about KE5FX Tools -> PhaseNoise Tool

John Miles jmiles at pop.net
Sun Oct 16 23:31:15 UTC 2011


> As I tried a test using my Anritsu spec I got strange results.
> 
> The following happened:
> I didnt connect any signal source to get the noise of the instrument only
> and set the carrier to 425MHz. As signal strenght I put in  -140dBm
> and connected a 50Ohm dummy resistor to the instruments input.
> Offset-range was set to 30-1000Hz.
> 
> As expected I got a horizontal line at -127dBm from 10-100Hz.
> Unexpectedly level changed from 100-200Hz as a falling curve down to -
> 132dBm
> and went on horizontaly further at this level to 1kHz.
> 
> During the measurement I observed the instrument and saw that it switched
> from
> RBW 10Hz to 100Hz so that I guess thats the reason for the descending
level
> between 100 and 200Hz f-offset. So the observed effect seems to be a
> property
> of the Anritsu instrument.
> 
> Similar happened between 1-10kHz and so on.
> 
> My question: How can I adjust that offset to get a smooth measuring
result?

Correct, you'd expect to see a flat line when measuring the spectrum
analyzer's noise floor with no signal applied.  Because the noise is
displayed after being normalized to a 1-Hz bandwidth, the RBW shouldn't
matter.  Some analyzer models behave this way, but some don't, and frankly
I'm not sure why.  

It's probably related to the proportion of the 'grass' that arises from
noise in stages beyond the last RBW filter.  If most of the analyzer's
internal noise passes through the IF filter(s), then the 1-Hz normalization
should result in a reasonably flat trace.  To the extent the noise floor is
influenced by stages beyond the last IF filter, it will appear to drop off
in the higher decades.  

Does your measurement look like the attached example, where a Tektronix
494AP is compared to an HP 8566B?  For this screenshot, I held down SHIFT
while opening the files to keep the program from overlapping and blending
between the decades, and I used ctrl-0 to disable all trace smoothing.   I'm
pretty sure that differences in gain distribution are responsible for the
fact that the 8566B trace is relatively flat compared to the 494AP.  

One obvious thing to check -- does the trace remain visible within the
graticule area while the measurement runs?  If it is ever clipped near the
top or bottom of the display, the results won't be valid.

Also, the recommended way to use PN.EXE to measure the RF noise floor is
described in the third-to-last FAQ entry at
http://www.ke5fx.com/gpib/faq.htm if you haven't already seen it.  An
amplitude of -140 dBm might be clamped by the analyzer's firmware to a value
that it considers more reasonable, so I'd use something like -60 dBm
instead.  The exact value isn't important as long as the noise trace isn't
clipped.

-- john, KE5FX

-------------- next part --------------
A non-text attachment was scrubbed...
Name: RF_floor.gif
Type: image/gif
Size: 38719 bytes
Desc: not available
URL: <http://www.febo.com/pipermail/time-nuts/attachments/20111016/ab8a953f/attachment-0001.gif>


More information about the time-nuts mailing list