Used Spectrum Analyzers for under $1k

[Rupunzell]

Another way of learning about this Super-het radio stuff is to get a copy of Ron Quan's book about radios. Read it and build a few of them to gain a basic understanding of how this Super-Het stuff and radios work. With this knowledge and understanding gained, how and what a swept SA does and works will begin to make a lot more sense. If radios are built, further understanding and knowledge can be gained by applying a SA to example radios. This entire journey can teach a lot about not only radio technology, but about instrumentation set up and what the resulting data could mean and how to interpreter it.

http://www.edn.com/electronics-blogs/anablog/4405264/Book-review--Build-your-own-transistor-radios


As for an FFT example, consider a sound card with FFT software. These are often good enough to provide a working example of FFT analyzers.

There were a good number of audio RTA (aka, bank-O-filters & display) produce and sold in mass for the consumer audio market. Get of of these if an example of an RTA analyzer is wanted.


Bernice

[Mr Simpleton]

What is the meaning of the letters (A, B, C, L, others?) in the HP 859X series?

What are the key tradeoffs between the 856X series and the 859X series?

The 859X do have 80 dB dynamic range on screen vs. 100 dB for the 856X... and the 856X have lower phase noise and a RBW down to 1 Hz vs 1kHz(?)...

[Wuerstchenhund]

How relevant are these distinctions and considerations to the selection of an "entry level / first-time / learning" SA?

It simply depends on what you want to do. Swept SAs were pretty much the oldest type of commercial SAs (FFT aka Vector SAs came much later due to the processing required), and as you stated correctly are generally the way to go if you need to cover large bandwidths (i.e. 2-25GHz).

If you don't need large bandwidths then a VSA becomes a viable option, and some of the advantages you listed (i.e. better for pulsed signals) can be pretty useful. And since VSAs are usually younger units they will statistically have more life left in them, and because they are fully digital they also support internal self-calibration and diagnostics which means they keep their specs for a very long time without requiring regular adjustments.

(On the one hand it's hard to imagine ever buying a second SA so it would be good to get a unit that will go the distance (5-10 years or longer?) but it's also hard to justify extra $ for something that is a nice-to have rather than a need-to have).

Honestly, how likely is it that any 30+ yr old SA you buy today, no matter how great, will still be in use by you 10 years down the line? Most of the stuff in those old boat anchors is already obsolete today, and most parts in it will also be pretty close to the end of it's useful life.

And frankly, with the relative low prices for some used kit it's in my opinion much more sensible to buy what you need now and, when you reach the device's limitations, to sell it on and buy a better device. This of course assumes that your first SA won't be a money pit which requires you to invest a multiple of the purchase price to keep it working.

Other considerations:

If you have a SA, a Tracking Generator would be pretty useful, yes?

If it's a swept SA, yes (VSAs don't have tracking generators because they're not swept). But again it depends on what you want to do. If you want to test cables, filters or transmission lines with a swept SA you'd want a tracking gen. Some VSAs like the E7495 which have an RF gen built in can do that as well.

I know, it would be good to start by pinning down some constant:  price, functionality, performance, ease of use, reliability, etc. 

A good start would be to think long and hard what you actually want to achieve, what bandwidth you need, and what parameters (i.e. specs, reliability, supportability) are most critical to you. And then you just filter the suggestions for adherence to your requirements.

As long as you haven't properly formulated your requirements I'd say follow joesmith's and Rupunzel's advice and try some sound card or TV tuner with SA software as a cheap learner. This way you're less likely to waste your money on something that's not right for you.

[Electro Fan]

Ok, still sorting out the big buckets of categories:

Large bandwidth (for example, 2-25GHz) might dictate a swept SA.
Less bandwidth (how much less or across what range?) might dictate or make practical a VSA?
Where/what is the distinction between VSA and real-time SA?  What bandwidths are typically provided by real-time SAs?

Is a Rigol DSA815TG a swept analyzer and not a VSA or a real-time analyzer since it uses a tracking generator?
In addition to the E7495 what are other less than $3k VSAs?
If there is a distinction between a VSA and a real-time analyzer, are there real-time SAs for less than $3k?

I saw the article below; it indicates that any SA that is a VSA is effectively a real-time analyzer; is that a fair statement?  Thanks

Realtime FFT
Most modern spectrum analyzers are now almost exclusively Hybrid Superheterodyne-FFT based giving a significant improvement in sweep time. However, even in such cases there is still processing time required to sample the spectrum and calculate the FFT. For this reason, both swept-tuned and FFT based analyser produce "blind time" meaning that while calculation of the spectrum is being performed, the instrument has gaps and misses information of the RF spectrum being applied to the RF front end.

A realtime spectrum analyser does not have any such blind time—up to some maximum span, often called the "realtime bandwidth". The analyser is able to sample the incoming RF spectrum in the time domain and convert the information to the frequency domain using the FFT process. FFT's are processed in parallel, gapless and overlapped so there are no gaps in the calculated RF spectrum and no information is missed.

Online realtime and offline realtime
In a sense, any spectrum analyzer that has vector signal analyzer capability is a realtime analyzer. It samples data fast enough to satisfy Nyquist Sampling theorem and stores the data in memory for later processing. This kind of analyser is only realtime for the amount of data / capture time it can store in memory and still produces gaps in the spectrum and results during processing time

[dom0]

Non-realtime vs realtime FFT analyzer is really a marketing distinction to highlight the higher processing power (and thus faster update rate) of the realtime ones. From a technical point of view it's essentially the same.

I dunno what you intend to do exactly and while realtime analyzers are really cool in general they're also really expensive (since they're newer)..

[Wuerstchenhund]

Is a Rigol DSA815TG a swept analyzer and not a VSA or a real-time analyzer since it uses a tracking generator?

It's a swept SA, like all the low end SAs. That's how they reach the low price. Swept SAs are generally simple to build and don't require powerful processing or much memory as a VSA does.

In addition to the E7495 what are other less than $3k VSAs?

I assume you're talking about 2nd hand prices (even the E7495B was north of $8k when new a few years ago)? There are several VSA only devices, like the Agilent E4406A. However, in my opinion some of the Wireless Communications Tster (which essentially are VSA + RF gen + other stuff in one unit) like the Agilent E7495A/B, the R&S CMU200 or some Anristu units are much more interesting as they can do more than just being an SA.

I saw the article below; it indicates that any SA that is a VSA is effectively a real-time analyzer; is that a fair

Yes. "Real-time" is more a marketing term for fast VSAs than a real distinction. It's generally used for high end VSAs.

[Electro Fan]

So, other than price are there any advantages to a swept SA vs a VSA?

And just to confirm, a 8560E/8590E are swept?

New question, any advantage to a 8560E with an external 85640A TG vs. an internal TG?

[Hugoneus]

The second my even cheaper and less useful SA that took longer to  build out of junk.   LEDs for the mixer diodes, a few old crystals make up the filter and a glass diode for the detector.    

Do you have any measured result for this you can share? What kind of conversion loss did you get from it?

[joeqsmith]

The second my even cheaper and less useful SA that took longer to  build out of junk.   LEDs for the mixer diodes, a few old crystals make up the filter and a glass diode for the detector.    

Do you have any measured result for this you can share? What kind of conversion loss did you get from it?

No.  This was only to show that you could build something very low cost to learn how an SA works. 

[avvidclif]

There was an HP 8560E in working condition sold on an auction site for $769.09 plus 16% buyers premium. Looked very nice and had the Cal seals intact.

Would I buy that over a Rigol? In a heartbeat any day. Besides the fact it goes to 2.9GHz.

[G0HZU]

There was an HP 8560E in working condition sold on an auction site for $769.09 plus 16% buyers premium. Looked very nice and had the Cal seals intact.

Would I buy that over a Rigol? In a heartbeat any day. Besides the fact it goes to 2.9GHz.

That's an amazing deal... We still use the 8560E in our design labs today. However, for that price I'd expect there to be a few niggles with it.

Deals like that don't happen very often but give it a few more years and they will become more commonplace.

I've not used the Rigol or the test sets like the R&S or the E7495A but experience has taught me to run a mile from cheapo test sets if you want decent RF performance on the bench. I'd expect the E7495A performance to be a bit woeful if subjected to a few basic analyser benchmark tests. Not something I would want unless it was for field/portable use where its performance limitations are less important.

One thing to add in favour of buying the big old HP8566 or HP8568 analysers is that spares are now plentiful on ebay and, contrary to what the 'hund claims, this type of analyser could therefore prove reasonably cheap to maintain as long as the user has reasonable diagnostic/repair skills.

I've got a HP8568 and a 8566 plus two TR4172 analysers and they all work and so far the repair/maintenace costs are less than the cost of a few drinks for all of them. I bought the 8568 for £125 many years ago from the company and I don't think it has ever failed since it was donated to the company on a contract over 25 years ago. It's a similar tale for the TR4172. It was donated to the company about 25 years ago and I bought it about 10yrs ago. It's repair costs have been close to zero in all this time. I bought a second TR4172 as a spares donor for it for £250 from someone in London and this cost me nothing to repair so it is a cheap spares/backup unit.

[alex.forencich]

So, other than price are there any advantages to a swept SA vs a VSA?

And just to confirm, a 8560E/8590E are swept?

New question, any advantage to a 8560E with an external 85640A TG vs. an internal TG?

Everything from that era is swept.  Well, actually, not quite.  The 8560E switches to FFT for resolution bandwidths below 300 Hz, IIRC, while the 8590E with the narrow bandwidth option uses analog filtering all the way down to 30 Hz.  This means that if you want to use a tracking generator with an RBW of less than 300 Hz, you cannot use an 8560E, only an 8590E.  Not sure if this is a concern or not.

As far as tracking generators are concerned, you can get external generators such as the 85644A (300k to 6.5G) and 85645A (300k to 26.5G) but these require a LO signal from the analyzer.  The 8560E series includes a LO out by default for use with external mixers.  However, the 8590E does not - you either need the LODA option 009 or the tracking gen option 010 to get a LO out SMA connector on the back panel.  However, the internal tracking gen only covers the low band up to 2.9 GHz.  You need an external tracking gen if you want to go above that. 

[Monittosan]

would an Anritsu MT8820A be suitable?  2.7Ghz bandwidth 25Mhz span and seemed to be quite fast and responsive when I last used it or are they more locked down to cdma/gsm type testing? I have one of these I am looking to offload. 

[Wuerstchenhund]

I've not used the Rigol or the test sets like the R&S or the E7495A but experience has taught me to run a mile from cheapo test sets if you want decent RF performance on the bench.

I'm not sure what you consider "cheapo" but the E7495 Series did cost somewhere excess of $18k which isn't exactly 'cheap' in my book.

The R&S CMU200 price varied a lot (there are tons of options for it) but some configurations went for in excess of $100k, The naked base was somewhere around $35k if I remember right.

As an example:
http://shop.rohde-schwarz.com/bargain/r-srcmu200-15.html
(look at the original price)

Not all Mobile Comms Tester are of the primitive kind of an Agilent E6392B or Racal 6103 which really are very simple devices.

Both the E7495 and the CMU200 were certainly not 'cheap' by any means, especially not for below 3GHz RF kit. And those prices aren't even ancient (the E7495B was EOL'd I think in 2012 and the CMU200 in 2010).

And that old SAs like the 8568 have more RF pipework in them is more due to the 30+yrs difference in technology than any performance differences.

I'd expect the E7495A performance to be a bit woeful if subjected to a few basic analyser benchmark tests.

Maybe, maybe not. Don't forget that this is a device that is designed for maintaining and repairing BTS systems, i.e. complex transmission systems which also are the cell provider's backbone. You can't get very far with just a scope and a detector diode on these systems, you need a proper SA and RF synthesizer for that.

If you have a certain test in mind (and if I can reproduce it in my limited home lab) then let me know and I'll try it on my E7495B.

[G0HZU]

By cheap I mean the RF components used in the E7495A spectrum analyser will be cheap. The RF synthesiser will be crude, noisy and cheap. The filters used in it will be cheap with less than spectacular stopband performance. The active devices will be low power and cheap resulting in reduced dynamic range.

I'd be interested to see how the E7495A performs in terms of signal handling and synthesiser noise. One basic test would be to put a small signal in at -120dBm and view it on a narrow span then and put in another much bigger signal at about 1MHz away but 120dB bigger at 0dBm and see if the analyser can still display the little signal without losing it in noise or blocking/overload. An old school lab analyser from 35 years ago can pass this test with some margin. Depending on how well the E7495A does then I can make the test tougher.

[Wuerstchenhund]

By cheap I mean the RF components used in the E7495A spectrum analyser will be cheap.

I wouldn't count on that.

I'd be interested to see how the E7495A performs in terms of signal handling and synthesiser noise. One basic test would be to put a small signal in at -120dBm and view it on a narrow span then and put in another much bigger signal at about 1MHz away but 120dB bigger at 0dBm and see if the analyser can still display the little signal without losing it in noise or blocking/overload. An old school lab analyser from 35 years ago can pass this test with some margin. Depending on how well the E7495A does then I can make the test tougher.

Sure, I can give it a try (when I found my box with RF couplers).

[G0HZU]

I wouldn't count on that.

Well, you have used the E7495B and I haven't.

But basic business sense tells me that the E7495B will have been designed by Agilent to be a cost effective solution that allows technicians to do fairly basic site surveys of base stations using a hand portable analyser.

The DSP back end will presumably have a decent capture BW allowing various kinds of signal analysis/verification. That's where the strengths of the instrument lie. But there really won't be much point making the RF downcoverter of this 'signal analyser' have the same RF performance as a top of the line lab grade spectrum analyser.

To do this would mean making it very big, expensive, heavy and power hungry and it would price itself out of the market.

I would expect it to have a fairly basic synthesiser and fairly modest IF filtering and a low cost construction to achieve the basic design goals.

Looking through its specs there is a spec for crossing spurs at -50dBc. Although this spec is a bit vague I suspect that this will be a clue as to the limitations of the IF filtering (SAW filter?) and some of these spurious terms will be alias terms. But I'm just guessing.

Here's a plot of my old HP8568B showing a deliberately added spurious term at about -90dBc on a 62MHz signal that is 400Hz from the carrier. The analyser noise is actually much lower than the noise on the plot as the 90dB log display is the weak link here. Normally I would alter the ref level to get the spur and the noise inside the log range. But the noise performance here is very good for such an old analyser.

Can you post up a plot showing a carrier on a 2kHz span on your E7495B? I'd expect it to be a lot noisier.

[G0HZU]

Here's an old plot of the input return loss of the 8566 RF port with the default 10dB attenuator selected.

The VSWR would therefore be about 1.05:1 or better across most of the range up to 2.5GHz where the input image filter begins to degrade the match. Having a low VSWR helps a lot with overall measurement uncertainty.

It's pretty impressive stuff from the big old dinosaur?

[Wuerstchenhund]

I wouldn't count on that.

Well, you have used the E7495B and I haven't.

So far I haven't used it much aside from a few basic measurements, though, mostly due to time constraints. I'm also not suggesting it performs as good as a HP 8568A/B, but I would be surprised if it's not noticeably better than the Rigol DSA830 or other SA's in that price range, so if these are considered good enough for hobbyist use then the E7495 and similar VSAs should be as well.

But basic business sense tells me that the E7495B will have been designed by Agilent to be a cost effective solution that allows technicians to do fairly basic site surveys of base stations using a hand portable analyser.

It's built for a bit more than just basic site surveys (there's another Agilent device for that), it's really designed for onsite fault finding and diagnostics. Also don't forget that with VSAs the RF part can be noticeably simpler than with an analog swept SA.

The DSP back end will presumably have a decent capture BW allowing various kinds of signal analysis/verification. That's where the strengths of the instrument lie. But there really won't be much point making the RF downcoverter of this 'signal analyser' have the same RF performance as a top of the line lab grade spectrum analyser.

Certainly not but I'd still expect it to show some good performance compared with a standalone unit in the same frequency band.

Here's a plot of my old HP8568B showing a deliberately added spurious term at about -90dBc on a 62MHz signal that is 400Hz from the carrier. The analyser noise is actually much lower than the noise on the plot as the 90dB log display is the weak link here. Normally I would alter the ref level to get the spur and the noise inside the log range. But the noise performance here is very good for such an old analyser.

It is, but also never said that RF performance was a problem with these old analyzers. It's age (and with it the high probability of failure), fragility, size, power consumption and the lack of basic secondary functionality (making screenshots for example ) that are the negatives in my opinion.

Can you post up a plot showing a carrier on a 2kHz span on your E7495B? I'd expect it to be a lot noisier.

Sure, I'll do some tests over the weekend. If anything it's a nice chance to compare this thing (and my CMU200) with a standalone swept analyzer.

[G0HZU]

It is, but also never said that RF performance was a problem with these old analyzers. It's age (and with it the high probability of failure), fragility, size, power consumption and the lack of basic secondary functionality (making screenshots for example ) that are the negatives in my opinion.

It is possible to take fairly decent screen plots/shots via the GPIB port with a USB to GPIB cable on all of my old analysers. My TR4172 is the analyser I use most and see below for a screenshot showing the 2nd harmonic distortion level of the analyser. I've used the freebie KE5FX plot utility to get this plot.

I've fed a very clean 28MHz signal in from a E4433B sig gen via a 30MHz LPF so the 2nd harmonic on the signal fed into the analyser is << -100dBc. So it's a very clean signal source.

The signal into the analyser is 0dBm and there is 10dB of internal attenuation so the analyser's mixer is seeing -10dBm and the displayed internal 2nd harmonic is -90dBc. I don't think there is any phase cancelling going on here to give a false result. Down at the VHF end of its range the analyser has very good signal handling properties. The local oscillator amplifier for the first mixer is a huge module that spits out about +22dBm to drive the high level 8 diode mixer.

Can you put a clean 28MHz signal into your E7495B analyser at 0dBm drive and 10dB attenuation and see how much distortion it shows as a comparison?

Note that the high noise level on the main black trace is caused by the relatively poor phase noise of the E4433B signal generator I used for this test.

[G0HZU]

You can also do higher resolution screenshots with these old analysers.

The display has 1000 x 1000 datapoints inside the grid area so each box in the grid has 100 x 100 data points in system memory.

The plot below is the highest resolution that the plotter program can go to but I think the analyser is better than this... The analyser can do 1000x1000 inside the grid area and the image below is only 1024x768 for the whole image so the plot still looks a bit grainy.

(making screenshots for example )

Can you show a screenshot from your modern E7495A analyser with finer screen resolution than this 35 year old technology?

I suppose I could try writing my own plotter capture program. The TR4172 allows full access to all system memory via GPIB so I should be able to grab the 1000x1000 grid data and produce a nicer image still...

[TSL]

I suppose I could try writing my own plotter capture program. The TR4172 allows full access to all system memory via GPIB so I should be able to grab the 1000x1000 grid data and produce a nicer image still...

You don't need to, its been done, you need the GPIB toolkit, go here....

http://www.ke5fx.com/gpib/readme.htm

Or alternatively if you want to part with a few $ and get serial coms too, try PrintCapture...
http://www.printcapture.com/

I use that to get screen captures from my HP 54720D scope and HP E6380 test set.

Here' some output I've recently captured using GPIB toolkit's plotter emulator from my HP8566B

cheers

Tim

[Wuerstchenhund]

It is possible to take fairly decent screen plots/shots via the GPIB port with a USB to GPIB cable on all of my old analysers. My TR4172 is the analyser I use most and see below for a screenshot showing the 2nd harmonic distortion level of the analyser. I've used the freebie KE5FX plot utility to get this plot.

Yes, I know, but that requires an external computer. It's much more convenient if you can just press a button and the SA saves a screenshot to the plugged in CF card.

Can you put a clean 28MHz signal into your E7495B analyser at 0dBm drive and 10dB attenuation and see how much distortion it shows as a comparison?

Sure. Unfortunately my only 28MHz source at the moment is a Rigol DG1062z, but I managed to do a really quick test today.

I fed the 0dBm (according to the Rigol's display, didn't check how accurate it is) via a Huber & Suhner BNC-to-N adapter connected to a Spectrum cable (2-18GHz) to a Mini-Circuits 10dB 6GHz attenuator sitting on the E7495B's input. The attenuator might be suspect, though, as I found it in a box and haven't had time to check it out.

It's not the best config but that was all that I could find in the short time available.

I'll see if I can do some more tests tomorrow, which could also include an R&S CMU200.

[G0HZU]

Here' some output I've recently captured using GPIB toolkit's plotter emulator from my HP8566B
cheers
Tim

That 8566 plot looks really good with very fine detail. I think I'm still using an old version of the KE5FX toolkit or maybe I'm not getting the best from it. I'll try downloading the latest version.

This is my HP3589A, shown with -100dB, spur 6Hz off the carrier. 

Wow, the 3589 looks good there!

The old 8568B can manage -112dBc/Hz phase noise just 300Hz from the carrier at the lower end of its tuning range but it can't go below 10Hz RBW...

[G0HZU]

Yes, I know, but that requires an external computer. It's much more convenient if you can just press a button and the SA saves a screenshot to the plugged in CF card.

USB/Flash is the modern solution but on my older gear I get pretty good results using a (highly portable)  mini EEPC netbook and a USB/GPIB cable as a wireless 'dongle' to my test gear that is mapped as a wireless network drive to my main PC. This way I can access the plots remotely as they appear instantly on my main PC. Usually I have the USB/GPIB cable connected to a common GPIB bus to all my test gear so I can get plots or control the test gear via the little EEPC and the wireless network link.

I use the same system when creating S2p models with my VNA as I can extract the s parameter data from the VNA using the EEPC and then it appears instantly on the main PC via wifi.

The EEPC holding the plots or S2p files pops up as a shared network folder on the main PC via wifi. The latest plots or files pop up first in the file list making it a reasonably powerful solution