[time-nuts] Spectrum Analyzer Specifications

Esa Heikkinen tn1ajb at nic.fi
Fri Mar 25 07:15:36 EDT 2016


timenut at metachaos.net kirjoitti:

> I have another question about test equipment. When using a spectrum analyzer
> to examine the output of a frequency standard, what are the minimum
> specification needed? Bandwidth, resolution, sensitivity, etc?

It depends what you want to measure from there. In general, analysis of 
this kind of strong signals are not very difficult but if you want to 
find weak spuriouses there then the dynamic range of the SA may become a 
problem.

> Looking at spectrum analyzers on eBay, I see quite a bit of difference between
> various models. Some have a resolution of 10Hz but others are 30Hz or even
> 100Hz.

This usually means resolution bandwith (RBW) and it has nothingh to do 
with actual frequency resolution. RBW means the bandwith of the 
adjustable IF filter in the SA signal path. With low bandwith you can 
see signals which are very close to each other but the sweeping will be 
slow. With high bandwith you can sweep large frequency areas in 
milliseconds but any sinals closer than RBW will be shown as one signal 
peak only. That's why it's very important to select correct RBW 
depending of what we are measuring. Usually the RBW selection is done 
auotmatically based on frequency span setting.

What comes to actual frequency resoltion you can usually set the center 
frequency with few decimals, it's not very accurate and it even doesn't 
have to be. When you sweep, you can analyze the peaks found in the trace 
with marker, but this is not accurate unless your spectrum analyzer is 
equipped with frequency counter function.

With counter function you can select the peak to be analyzed with marker 
and then use marker count function to see the actual frequency with that 
peak with the resolution of the counter (can be 1 Hz or 0.1 Hz for 
example -not as precise as with universal couner). But the difference 
with universal counter here is that you can count very faint signals 
also (for example less than -100 dBm) and you can select the signal to 
be counted even when there's multiple signals mixed. You may have some 
strong signal where your universal counter locks but then there's some 
faint signal mixed with it. If you want to count this faint signal you 
can do this with spectrum analyzer's counter easily.

With the counter function you will also need option to feed external 10 
MHz reference to the SA. Without external ref the counter is mostly 
useless because it may have serious frequency error. This error is also 
stongly dependent of the frequnecy what you count. For example if you 
want to count something around 10 GHz, only 1 Hz error in the 10 MHz 
refernce means 1000 Hz error in the counted value at 10 GHz.

> Some have a minimum frequency of 0.01Hz, 100Hz or even in the kHz
> range. Some are only sensitive to 60dBm, but others over 100dBm.

It's -60dBm or -100dBm. Both are quite poor values for spectrum analyzer 
but may be adequate if you only analyze strongs signals like oscillators.

> Are any of the cheaper USB spectrum analyzers worth getting?

If you want to analyze only "easy" signals like oscillators etc. then 
yes. For anything more precise or very faint signals, I'd say no.

> Most of these are appallingly expensive, so knowing what is needed can
> certainly help guide a purchase or to minimize cost. And if a "deal" is found,
> knowing that it is or is not adequate can help.

Yes, good specrum analyzers usually costs more than 2000 USD even as 
used. Also, a "high end" used 2000 USD specrtum analyzer can be much 
better than brand new "entry level" 10000 USD specrum analyer. And if 
you need any kind of vectored signal analysis, demodulation capability 
of modern communications etc. then the price will be very high.

Hope this helps...

-- 
73s!
Esa
OH4KJU


More information about the time-nuts mailing list