Favourite FFT analyzer?

A lens provides a two dimensional Fourier Transform for lighted areas. If your one dimension can be made into an image, a lens can show a transform. See "Fourier Optics".

I seem to recall hearing about that one time. ;)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

Am 11.11.2016 um 20:06 schrieb Phil Hobbs:

I have an Agilent 89441 vector signal analyzer. It is not the undisputed king of dynamic range, (12 bit ADC @ 24 MHz IIRC) but good enough for everything up to now.

It is undervalued since everybody thinks that is only useful for debugging smart phones. There are many 89441 on the market now because current generation cell phones need even more bandwidth and funny new modulation schemes.

It consists of 2 Boxes. The IF box goes from DC to 10 MHz vect. and the other, larger one is essentially a down converter for

10 MHz wide segments up to 2.7 GHz. The IF unit also works alone if 10 MHz is enough. One can transparently sweep over multiple segments.

Make sure you get both channels (for cross correlation etc) and also the signal source. You can then do Bode plots from DC to daylight. There is an optional up converter for the source that mine does not have. I don't know by heart if the network interface is an option, but it works nicely. You probably need a $20 converter box for BNC-Ethernet to twisted line. One can then simply open port 5000-sth. on

192.168.x.y and feed / read GPIB-like traffic.

That's how I did the 7 decade FFT plots I linked to last week. (sources avail.)

regards, Gerhard

(BTW my converter box needed both 50R-terminations even for the 5 cm coax LAN.)

Have you actually tried a good scope FFT with appropriate preamplifier? I hadn't until my R9211C boat anchor died. I found the scope quite usable, and works up to very high bandwidths of course.

--

John Devereux

Do you have a price range/ target. I've lusted after the cheap Rigol.. about $1.5k I think... which is too much for me.

On the very cheap end (and looking at noise) I use my digital 'scope, in the average mode, normal trigger, near the peak of the noise voltage. That works OK, but limited dynamic range. (And it cuts off the high frequency end of the noise faster than you'd guess.) (Or faster than I guessed anyway.)

George H.

Thanks, Gerhard, that's just the sort of thing I'm looking for. A DSA with 100 times the bandwidth is the ticket.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

Yes, that's the fallback position. It would be nice not to have to guess at the noise bandwidths of the various windows, though. Of course if it's really pure noise, then I can use a rectangle window, in which case the noise bandwidth is 1/(trace length).

My TDS784A does a reasonable job on noise, though it has sort of a weird user interface--every time you select the FFT trace, it re-starts the averaging, which is a bit irritating when it's 1k acquisitions. You have to save all the traces before you can flip back and forth without restarting.

Thanks

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

Pity because there might be some cost effective almost solutions there.

I favour the hiring a high end one as and when you need it for the extreme sensitivity requirements that only occur infrequently.

Or grab the data with a fast ADC and FFT offline if you can live without realtime results but need the accuracy/long term averaging.

--
Regards, 
Martin Brown

Do you use the average function? And trigger near the top of the noise? A more math agile colleague showed that at high frequency the spectrum drops off at twice the rate indicated on the display. (The 3dB corner of a low pass filter is 6dB down.) which is easily confirm-able with a known filter.

Here's a noise ACF ((Quasi) auto correlation function)

and it's FFT

(about a 1 MHz BW.)

George H.

Mine has a settable RBW. See

for example.

--

John Devereux

You have to average, because the FFT of noise is noise--when you do a single FFT power spectrum trace, the variance in each bin equals the mean.

Huh?

Well, it's a power spectrum, so it falls off at twice the rate of the amplitude, but that's all. The analyzer isn't lying to us.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

I have this one plus the matching tracking generator:

It's essentially a glorified SDR. The software is also available for Linux but I only use Windows. For me a key advantage is that it's small and leightweight so I can also use it at clients, without having to check it and waiting at the baggage claim with white knuckles. Despite them saying that it won't work with a 1st generation Atom processor (need that for long battery life) it did work with it.

[...]

--
Regards, Joerg 

http://www.analogconsultants.com/

I suspect Phil could write a book on that. Maybe he already has. :-)

I'm another user of the 89410A. I've had more than a few issues with mine over the years, and if you're thinking of purchasing one I'd advise you to cost in the price of a donor unit as well.

Although I don't have personal experience of them you might like to look at the 44xx VSA's as well, which generally sell for less than the

89410's. They are a much more recent design and I suspect would be more reliable long term. Units with baseband IQ inputs give you a 5MHz bandwidth.

Thanks, Jay. The E4406 looks like a win, though its measurement accuracy spec is a bit loosey-goosey (+-0.6 dB when it's in calibration). OTOH I can get one for $500, which is pretty amazing.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

Remember you need one with the B7C option, which is pretty rare.