Wanted: Spectrum Analyzer

[krenzo]

Hi, I'm in the market for a spectrum analyzer.  I've been developing an UWB transmitter, and I want to be able to measure it in the frequency domain accurately.  I've been using my sampling oscilloscope, but its FFT doesn't seem very accurate (amplitude changes based on the sampling frequency and memory depth).  I have also been messing with an RF switch to send the UWB signal through, but my readings do not match the datasheet (datasheet says 40dB of isolation when the switch is off, but I'm only measuring 10dB of isolation) and want to be able to verify this more accurately.

I need an analyzer that can do at least 6 GHz (signal is in the band of 3.1-5GHz) and don't want to spend over $2000.  From looking around, these appear to be my options:
an old HP-8566B off of eBay, Signal Hound USB-SA124A, and SPECTRAN HF-6060 V4 X.  I like going with a USB one because it's smaller, and the signal can be captured on a PC.  I don't like how expensive these are, but it doesn't look like I have a choice.  The Spectran appears to be the cheapest, but I can't find much information about it.

I'm looking for any advice on these or suggestions for others analyzers.

[amspire]

Hi, I'm in the market for a spectrum analyzer.  I've been developing an UWB transmitter, and I want to be able to measure it in the frequency domain accurately.  I've been using my sampling oscilloscope, but its FFT doesn't seem very accurate (amplitude changes based on the sampling frequency and memory depth).

You may find the low priced USB spectrum analyzers have the same problems as your sampling scope. Some, if not all of them take a wide bandwidth slice, and then digitize it, just like the scope did, to generate the FFT by calculation. It is cheap, but not nearly as good as a real spectrum analyzer that uses actual bandpass filters to discriminate between frequencies.

Richard.

[ejeffrey]

You may find the low priced USB spectrum analyzers have the same problems as your sampling scope. Some, if not all of them take a wide bandwidth slice, and then digitize it, just like the scope did, to generate the FFT by calculation. It is cheap, but not nearly as good as a real spectrum analyzer that uses actual bandpass filters to discriminate between frequencies.

All quality modern spectrum analyzers use digital IF filters as they are simply far better than any analog filter.  Analog filters are used for image frequency rejection and anti-aliasing, not defining the spectral resolution.  High end real-time spectrum analyzers digitize a wide band of frequencies and use FFT to generate the whole spectrum at once.  There is nothing intrinsically wrong with using an FFT for spectral analysis, and in fact it has many advantages over a traditional approaches both in terms of throughput and accuracy.

The problem with oscilloscope FFT features is primarily the poor software implementation -- mostly due to a lack of DSP resources and memory.  The analog front end and ADC is not optimized for a spectrum analyzer appliction, but it is good enough to give a much better result than what you actually see.

The problem with the cheap USB spectrum analyzers is mostly the lack of a high quality analog front end.  They both inject noise into your system and have a lot of spurs and aliases from the front end mixer that are not filtered out as well as in a "real" SA.

Anyway, I have the signal hound unit.  It performs OK for my application but there is substantial noise coupled from the USB signals onto the input port (most of the noise only appears during data transfer).  Likely it would be better off with a separate power supply rather than being bus powered.  It is definitely not a high end unit, but they seem to be fairly honest about the limitations.  The software can be a bit slow and unresponsive -- you can't cancel a sweep in the middle, and you can easily accidentally start a 30 second sweep.

The aaronia units look too good to be true on paper, so I assume they are.  Even so, I would have gotten one if I could see a review by someone who appeared to know jack all about RF.  Unfortunately, I could find no such review.  I would love to hear a review of either their high frequency or low frequency analyzers -- especially the low frequency units as there are very few comparable units on the market at any price.

[vk6zgo]

We had an IFR Spectrum Analyser at my last job.
It was a nice unit,but had a few bugs:

(1) It had an incomplete PC interface,in that you could control it from a PC,but you could not export live versions of the display at all,& the only way to do so with saved versions was to use a floppy disc----in a unit that was new in 2006!

(2)---And this is the big one!!

Due to a low sample rate,it could not display a carrier rapidly changing in frequency over a relatively large frequency range.

We had a PLL on 433MHz,which would suddenly lose lock,sending the VCO up to 470MHz
The IFR showed this as a series of individual carriers,depending on where the signal frequency was when it was sampled.

On an El Cheapo Chinese analog,the carrier could be seen to change frequency----quite a different fault to the parasitic oscillation we thought we had!

The "quality" Spectrum Analysers ejeffrey refers to would have addressed this problem,but it is annoying when the next level down exhibits this kind of behaviour,without any warning of the possibility in the operator's manual.

[nctnico]

Hi, I'm in the market for a spectrum analyzer.  I've been developing an UWB transmitter, and I want to be able to measure it in the frequency domain accurately.  I've been using my sampling oscilloscope, but its FFT doesn't seem very accurate (amplitude changes based on the sampling frequency and memory depth).  I have also been messing with an RF switch to send the UWB signal through, but my readings do

What kind of scope are you using? Given the way FFT works its logical that the graph changes based on sampling frequency and record length.

FFT works on a piece of signal. Because the signal is 'cut' it needs to be windowed so the beginning and end have less influence on the result. A longer record length gives more accuracy and less interference from the beginning and end of the piece of signal being analyzed. The sampling rate influences the highest frequency component. Every now and then I use the FFT function on my Tektronix scope. When I use long records lengths (say 15ksamples), use a proper sampling rate and use averaging for the FFT signal it works like a charm. The bottom line is: you need to know what you are doing. The same goes for using an advanced spectrum analyser. You need to know what you are doing otherwise you get rubbish on the screen.

[krenzo]

What kind of scope are you using?

Tek 11801B sampling scope with a 12 GHz module
My issue is that I'm working with an ultrawideband pulse with a width of 0.5-1ns.  When I increase my timebase (2ns/div), the FFT zooms in to where I want it to (1-8 GHz), but the dB drops because there's so much empty space surrounding my pulse (I think).  When I decrease my timebase (200ps/div), the FFT zooms out to 1-20GHz, and I can barely see the power of my signal while the displayed power also increases by something like 10 dB because there's nothing but the single pulse filling the entire screen in the time display.  When I don't change the timebase but increase the number of sample points, the FFT zooms out without changing the width of the signal in the time domain display.  Ultimately, I need to know the peak power of my signal because the FFC regulations state ultrawideband pulses cannot exceed 0 dBm, but with my power shifting with the change of my settings, I don't know how to find my peak power.  The FFC guidelines only talk about measuring it with a spectrum analyzer.

[nctnico]

My TDS744 allows to change the Hz/div setting for the FFT trace independent of the timebase. Anyway... If the FCC says it should be measured by a spectrum analyser you'll need a spectrum analyser which can be set to those specific parameters. The receive bandwidth, detection method (average, peak, quasi peak), sweep time, etc have a large effect on the trace you'll see on the display.

[steve_w]

You could build one, I am doing that now.  see this link http://scottyspectrumanalyzer.com/msaslim.html

I'll probably do a web site for my spec an if motivation takes me, and it works :-)

regards

Steve W

[vk6zgo]

I think you are stuck with using a real (commercially made) Spectrum Analyser,if the "Candy Company" says so!
The FCC may even specify the type!
They may be full of "suits" & lawyers now (like ACMA is),but they are probably still "hardline " enough with test equipment requirements.
(After all,it's not coming out of their budget!)

[krenzo]

I think you are stuck with using a real (commercially made) Spectrum Analyser,if the "Candy Company" says so!

The idea is that I would finalize my design to be within the limits verified by my own equipment, and when the time comes to submit it to the FCC, I would go through a local testing lab that does FCC testing for a fee.

[prestigetest]

I have an Agilent 8566B currently available. I can match the the price of the unit on Ebay which was $2000 but the unit would come as-is at that price with only a power cord and CD manual. This unit has not been fully tested because calibration is going to cost me $400. If you would like a 30 day warranty and a 5 day right of return, I can have it calibrated for an additional $400. If you or anyone else is interested please send me your zip code and I will get a shipping estimate for you. Have a great day everyone!

Best Regards,
Zach Young
Prestige Test Equipment, Inc.
www.prestigetest.com