Why are old spectrum analyzers so expensive?

[G0HZU]

If you want to play with decoding your Extech wireless meter, I don't know what's on the new or used market that a hobbyist could afford.

If you want to hack or spoof your neighbour's wifi or play with GSM phones then you could buy something from Ettus Research instead 



AFAIK their Rx+Tx SDRs cost less than the BB60C and you can do wideband Tx as well allowing all kinds of hacking and emulating.
The Ettus box isn't a proper spectrum analyser as it is a direct conversion device, but it is far more interesting than the 'listen only' BB60C and some of their devices cover a very wide real time BW. There are lots of people providing FPGA data for these to emulate various modern digital systems. I'd much rather have the Ettus box compared to a BB60C.

https://www.ettus.com/product/details/UN210-KIT

We have loads of these at work as versatile playthings but no BB60C boxes

[mmagin]

A good question these days is how can a manufacturer make a half decent Spectrum analyser so cheaply. They have historically been a niche product that cost significant sums of money to manufacture. Modern electronics miniaturisation has enabled hobbyists to have access to acceptable performance within a reasonable budget. Be happy with that rather than wasting time wondering why other people still buy old SA's and pay, what seems to some, high prices. They have their reasons. Be happy for them getting the object of their desires 😄

I think the 3GHz or 7GHz analyzers benefit from a lot of well-traversed engineering territory and the availability of cheap parts suited to automated assembly.  All of the mobile phone, wi-fi, etc products have driven the accessibility of products and techniques that put that stuff in the range of any manufacturer instead of a handful of industry-leading specialist manufacturers.  30 years ago that frequency range was pretty much only the domain of specialist designers of microwave equipment (telecom, military, space, test equipment).

[joeqsmith]

I don't own a cell phone.       

[Fraser]

@mmagin,

I agree. Direct to home satellite gave birth to the mass production of low noise 10 to 12GHz RF front end and down conversion technology at silly cheap prices. Having worked on SHF low noise block down converters over the years it has been interesting to see how the designs have developed and costs reduced, whilst still maintaining acceptable performance, and in the case of noise figures.... significant improvement. I am not suggesting that a typical SKY LNBD is lab grade though.... some are so crudely constructed that I am amazed that they actually work !

5GHz used to be considered advanced RF technology, along with 10GHz. Now such frequencies are easily manufactured in bulk for pennies. Pretty amazing advances in technology and production processes. I was working with 40GHz and 75GHz equipment last year.... that is still less common and is expensive for 'building block' prototype modules. Try finding a reasonably priced SA that can cope with a 75GHz Local Oscillator 😆

The way around having an inadequate bandwidth Spectrum analyzer, such as my Advantest 3GHz R3132 is not always to buy a very expensive higher bandwidth unit, but to use your RF Engineering knowledge. You buy a well specified down converter that contains an accurate PLL controlled Local Oscillator and comes with a calibration chart. Why buy a massively expensive Analyser when a block down converter for the frequency range of interest is cheaper ? I know it is not ideal but when working at 40GHz and above, you sometimes have to improvise in a carefully considered way to achieve the desired results. At least COTS >40GHz building blocks are now available at affordable prices.

Fraser

[G0HZU]

I don't own a cell phone.       

The point is that you can set it up to modulate or demodulate (emulate?) pretty much anything you like. Here's a video about setting up analogue FM demodulation. Maybe you don't listen to FM radio either and you just like decoding basic FSK from cheap chinese DMMs but this is a very versatile system. It supports GNU radio and there is a wealth of support out there.








In the next video above have a look at 4:26 onwards to watch the Ettus box sniff through the 900MHz ISM band. Presumably there will be GNU radio apps readily available to listen/emulate/interact with a lot of the ISM stuff or you can decode stuff elsewhere.

There's no way I'd spend $3k on a BB60C when I can get the Ettus box AND a decent old school HP spectrum analyser for the same money. At the minute I just use the HP8566B and the E4406A as the E4406A (with the Agilent 89600 SW) can do the FFT analysis and the data capture stuff I need to do but I keep meaning to buy my own Ettus box.

[mmagin]

@mmagin,

I agree. Direct to home satellite gave birth to the mass production of low noise 10 to 12GHz RF front end and down conversion technology at silly cheap prices. Having worked on SHF low noise block down converters over the years it has been interesting to see how the designs have developed and costs reduced, whilst still maintaining acceptable performance, and in the case of noise figures.... significant improvement. I am not suggesting that a typical SKY LNBD is lab grade though.... some are so crudely constructed that I am amazed that they actually work !

I forgot about the direct broadcast satellite stuff.
It's truly amazing how little there is in the modern LNBs.  I pulled one of the 'universal' ones apart last year and it was like a single crystal, a single LO synthesizer + downconverter IC, 3 FETs for the frontend, and passives.  The IC also managed the biasing of the FETs to select the horizontal or vertical antenna probe.  All on a (presumably low-cost) microwave substrate and in a cast zinc enclosure which also forms the feedhorn.
That application has the luxury of not having to be incredibly frequency stable, so I guess an ordinary crystal is good enough.

[joeqsmith]

I watched a few videos about it.  They were at NI Week 2016 and had a decent overview.     

Yea, it's pretty sad the best antique HPs can't decode some simple FSK from a cheap Chinese DMMs. I would think this would be a pretty basic thing to do.  I don't think it would take much to connect a PC to my old Tektronix and decode it. 

[wkb]

This has been an interesting thread.

You do not always need the best to get the job done. Way back when, people had to design and repair  without access to an SA, they were just too exotic and expensive. These people got the job done by being top of the line techs who could innovate and make do with what they had available.

Very true. That is the inspiration for the gliding aphorism in my .sig.

....

The analysers we really wanted to use simply didn't exist in those days. We explored the limits of the HP8568B and eventually went over to the TR4172 for some of the design work. I think this particular TR4172 cost more than £60k when it was new in the mid/late 1980s but there were rumours that it cost even more because of the strong Yen back then. Luckily it was paid for with public money by the customer and you could have bought a very nice 4 bed house in the south east of the UK for that money back in those days

Sounds like a customer who in days gone by built Colossus 

[G0HZU]

Yea, it's pretty sad the best antique HPs can't decode some simple FSK from a cheap Chinese DMMs. I would think this would be a pretty basic thing to do.  I don't think it would take much to connect a PC to my old Tektronix and decode it. 

It isn't really designed for stuff like that but you can set it into a 'static' zero span mode and select 'linear' mode and there is a 21.4MHz IF output at the back. This can be fed to a limiter/quad demod in a remote box if you want to decode FM. It's years since anyone tried stuff like that at work but somewhere I have a 455kHz SA604 based limiter/demod box with a mixer at the front of it. The idea was that it could be used with various IF outputs from various analysers. Today, you could stick a modern DSO on the 21.4MHz IF output and capture/demodulate stuff and send it to a PC.

Eg use the scope and the 89600 SW from Agilent to decode loads more stuff than FSK. Or maybe use a cheap SDR dongle on the IF output and run some freeware SDR SW on a PC with it. On the HP8566B this could presumably be done up to 22GHz.

[joeqsmith]

Seems like as long as you had the IF and a way to record it you could be set.   BB60C will stream the IQ directly and has a fair amount of support already built in for decoding.  But as we saw, it could not decode the FSK.  But at least no other hardware was needed to pull it off.

Attached is the old 496 directly looking at the digital data.  Zoom out I don't think you would have enough data to be able to decode the whole frame.  So it would require something external as well. 

It would be simple enough to put something together without the SA to do it.  The chipset they used is fairly common.  Lots of eval boards for it already.  For that matter, just use the radio that came with it.  Not very useful but inexpensive.   BB60C is a little more generic. 

[David Hess]

I think the 3GHz or 7GHz analyzers benefit from a lot of well-traversed engineering territory and the availability of cheap parts suited to automated assembly.  All of the mobile phone, wi-fi, etc products have driven the accessibility of products and techniques that put that stuff in the range of any manufacturer instead of a handful of industry-leading specialist manufacturers.  30 years ago that frequency range was pretty much only the domain of specialist designers of microwave equipment (telecom, military, space, test equipment).

I would have said Mini-Circuits (and maybe some others like Hittite?) made it possible.  When did they start making their standardized parts, especially MMICs, available to anybody who could pay in small quantities?

[G0HZU]

I think the 3GHz or 7GHz analyzers benefit from a lot of well-traversed engineering territory and the availability of cheap parts suited to automated assembly.  All of the mobile phone, wi-fi, etc products have driven the accessibility of products and techniques that put that stuff in the range of any manufacturer instead of a handful of industry-leading specialist manufacturers.  30 years ago that frequency range was pretty much only the domain of specialist designers of microwave equipment (telecom, military, space, test equipment).

I would have said Mini-Circuits (and maybe some others like Hittite?) made it possible.  When did they start making their standardized parts, especially MMICs, available to anybody who could pay in small quantities?

I can't comment on other peoples' experience but I regularly met the UK rep for Mini Circuits (JP N) in the early 1990s and there was no problem getting stuff in small qty. Their boast at the time was that you could have any component from their huge RF/IF Design catalogue inside a 3 week leadtime (USA to UK) even if it was not in stock anywhere. They would make it and deliver it inside 3wks. They also did custom versions of a lot of their parts even in those days. eg you could have a special (pregraded) mixer or amplifer to your own spec limits and it didn't cost much. They were one of the best suppliers around in this respect. Most stuff was available in a day or so if it was in the UK stock. They were great at providing samples and application info as well.

Other sources for decent RF parts in those days included WJ and Avantek for exotic MMIC amplifiers, also Stanford Microwave for amps etc and PLL chips were generally Plessey or Motorola and often this required a design using external dual modulus prescaling. RF design was much harder back then because they were very few chipsets available I also used to walk home 43 miles through the snow in my bare feet after each 72 hour shift.

[G0HZU]

Seems like as long as you had the IF and a way to record it you could be set.   BB60C will stream the IQ directly and has a fair amount of support already built in for decoding.  But as we saw, it could not decode the FSK.  But at least no other hardware was needed to pull it off.

There are hidden tricks the HP8566/8 can do as it does have some basic DSP capability built in for post processing captured trace data. But I've never done anything useful with this feature. But I have played with it a few times. You don't need any additional hardware as the processor inside the analyser can be set to do it all for you. Each trace can be treated as a 1K record length and the analyser can do some basic DSP on this. eg it can produce and display a frequency spectrum from the trace data and this can be useful in zero span mode. I wonder if anyone ever (properly) used this feature in this old analyser?

[R005T3r]

I hate to admit it, but at this point, why would you have to own a spectrum analyzer? I mean, you can use your oscilloscope as a "spectrum analyzer" for very basic stuff via the FFT command, to begin with, it will be slow, and you can't expect NASA resolution and whatever, but better than nothing.

I mean, For 2000/4000 Euros you can get other instruments as well: why not a milliOhmmeter, why not an electronic load, or even better: an impeadence analyzer, an SMU... As now, I don't need a spectrum analyzer, and I'm fine with it since I won't use it, and If I had to I'd consider Siglent or an USB one.

[joeqsmith]

You are right of course for those of us who play with electronics as a hobby.  I certainly did not need all the cheap handheld meters I bought.     

[G0HZU]

I hate to admit it, but at this point, why would you have to own a spectrum analyzer? I mean, you can use your oscilloscope as a "spectrum analyzer" for very basic stuff via the FFT command, to begin with, it will be slow, and you can't expect NASA resolution and whatever, but better than nothing.

I kind of agree and I'm surprised at just how many people on Eevblog seem to want a decent RF spectrum analyser. I need decent RF gear here to help me with research work and generally speaking, the better the analyser in terms of RF performance the easier it is to make reliable measurements with minimal uncertainty. Also, an 8 bit scope with FFT couldn't really do critical testing for linearity if you need to measure artefacts down to -70dBc or lower. There's other stuff the scope wont be able to do but I agree that most people probably don't need a decent spectrum analyser.
I'm guessing that some people use them for casual EMC checking and maybe for looking at RF comms waveforms and some users will be radio hams.

[SingedFingers]

I'm going the awkward sod route and spending an old HP / new Rigol's worth of cash on building an inferior fully analogue tracking SA from scratch with sweep range of 100KHz-500MHz. That's an illogical dedication to a lost cause.

I'm not sure if that's more or less healthy than spending a pile of cash on an old HP crate that is toast if anything breaks.

[tggzzz]

I'm going the awkward sod route and spending an old HP / new Rigol's worth of cash on building an inferior fully analogue tracking SA from scratch with sweep range of 100KHz-500MHz. That's an illogical dedication to a lost cause.

Excellent 

I'm not sure if that's more or less healthy than spending a pile of cash on an old HP crate that is toast if anything breaks.

A little bit of what you fancy does you good. Have fun; that's what life's about.

[SingedFingers]

Wise words actually and motivating. Thank you

[Fraser]

My experience of Advantest Spectrum Analysers has been very different. I had responsibility for over 50 Advantest SA's and the common faults were just age related after 20 years of service ! CRT would fail, PSU caps would fail and the odd 1st Mixer got fried. All, without exception, were repairable and I carried out some of the repairs myself. Unrepairable.... I think not, even for old models. If the BRL chaps thought that, they were not trying hard enough  As for failing calibration.... nope never had issues with that side of things provided the PSU was running OK. I can quite believe that some models are better in this respect than others though. The R3132 has a Local Oscillator that does drift off frequency due to age. That is not difficult to fix an is understandable due to their age. It is just an adjustment and not a component replacement.

Finally on the Advantest front. R&S became a partner with Advantest in Europe. I do not see R&S associating themselves with any company that does not produce high quality kit that they also stand by in case of failure.

I can see from the post that BRL have a love of  Agilent kit. I think many will agree that Agilent is the Gold Standard in test equipment. I do not disagree. Marconi in Chelmsford used HP test kit in the production of their test kit.....says it all really  Having used HP test kit, the only comments I would make are that it is often very large and heavy !...Well built ? absolutely, Custom parts used in the design ? Yes, Repairable ?..... only if the custom parts are still available. Great kit but you often need very deep pockets to purchase it new or used.

Fraser

[nctnico]

My experience of Advantest Spectrum Analysers has been very different. I had responsibility for over 50 Advantest SA's and the common faults were just age related after 20 years of service ! CRT would fail, PSU caps would fail and the odd 1st Mixer got fried. All, without exception, were repairable and I carried out some of the repairs myself. Unrepairable.... I think not, even for old models.  If the BRL chaps thought that, they were not trying hard enough

I agree. I have owned a few Advantest SAs myself and so far none of them have been unrepairable. Then again I have succesfully repaired lots of stuff which others had given up on.

[G0HZU]

Agilent is built to last and is repairable

My experience of HP test kit is that the stuff designed and built in the 1960s and most of the 1970s was built to last. Then they started using lots of plastic and also cheaper switches, knobs and cases. Where I work the test gear gets moved around a lot from engineer to engineer and sees a LOT of use and not all users are sympathetic towards fragile user interfaces.
Most of our HP kit from the 1980s just couldn't stand this intense use and most of the kit suffered from broken and missing buttons, knobs or worn out button membranes. Their scopes were the worst and also the portable HP8560 style analysers were very fragile. A lot of their older kit that used slider switches was notoriously unreliable. By contrast, we have quite a bit of Japanese test kit and this was designed to properly withstand heavy use and I don't think any of it ever suffered with broken buttons etc. HP/Agilent gear is very good technically but I don't think the build quality is that great any more.

Assuming the Rigol and Siglent analysers don't have some kind of weakness in the PCB manuf/reflow process (leading to poor joints etc) then I would expect that the reliability will be pretty good compared to an old school HP analyser. Even if my HP8566B or my HP8568B was (somehow) 15 years younger I would still expect the Rigol and the Siglent analyser to be more reliable.
I wouldn't swap my HP8566/8 for one of them but I would definitely allow the Siglent into my workroom and use it now and again. By contrast, there's no way I'd have/use an HP859x series analyser here or a Tek 496 or any of the test set analysers or the mid range HP/Anritsu/IFR/Advantest analysers from the 80s and 90s. Even if they were free. I'd rather have the empty space they aren't taking up in my house.

[G0HZU]

As for my big old Advantest TR4172, I've been using this analyser for over 25 years and it is still working OK. I've had to do a few repairs on it but none involved any new parts apart from the backup battery (or BATTLY as Advantest call it on the PCB silk screen) and I had to swap the main OCXO across from another 4172 when mine aged too much.

In terms of calibration we used to send it for a (£££) NAMAS calibration every year (same as UKAS calibration today) and it seemed to get through this OK.

As for my HP8568B, that has been reliable too but this was partly due to it being treated like royalty at work and only a few people were allowed near it for many years. I have it here at home and it is still working OK but I did have to sort out the Z drive board for the CRT when the display went a bit fuzzy and dim. Didn't cost anything to repair. It was always calibrated and serviced at HP at great expense when it was in its most productive years at work. The HP8566B has needed several repairs over the years mainly poor connections and dried out/leaky caps and the same issues with the Z board and dim/fuzzy display.

[G0HZU]

"Agilent is built to last and is repairable.  Rigol and Sigilent are use and through away.  We've never seen an Advantest that can make it through 2 cal cycles and it also cannot be repaired

I think that the build quality of my old Advantest is very rugged and impressive too.

I don't have many shots of the insides but here is one that shows the main OXCO. It's the big box in the centre of the image with PIN CONNECTION written on it. You can look around this OCXO and see how solid this analyser is built. Advantest had a different approach to HP in that the favoured method was to use lots of high quality metal enclosures for the RF and IF sections and proper semi rigid cable was used everywhere. It meant the analyser was a fair bit bigger and heavier than the 8568B. HP used a motherboard and a series of screened slots for a lot of the IF stuff and cheaper flexible cables.

See also the final amplifier module for the LO for the first mixer. This is arrowed in yellow below and is a huge high power amplifier. You can see how solid the build quality is for this amplifier. The first mixer in the 4172 runs with a +23dBm LO  and contains 8 diodes ( compared to just two in the HP8568B).
This was partly why we simply had to have the TR4172 back in those days. The mixer input IP3 and 2HI performance in the VHF region totally outclassed the HP8568B and we needed this performance for our design work. The build quality inside the TR4172 is very impressive although the construction methods are quite dated and inefficient as you can see by all the wasted space. I think the TR4172 was Takeda Riken's response to the fabulous HP8568 and I think they let their engineers off the leash to try and match the HP8568 in as many areas as possible. It managed it fairly well although I think most users will prefer the HP8568 as an all round package. The TR4172 has a few quirks and niggles and it doesn't feel as complete and polished as the HP8568. It is a big ugly thing as well although I do like using it

[cncjerry]

I have to jump in hear as G0, Nct and one or two others have the same gear.  I have an R3131A with TG that works perfectly, paid $750 for it two years ago.  9K to 3G or more.  Only hang up is the RBW doesn't go that low. 

I just had to have an 8568B.  Loved all the buttons, knobs and man, you can do some crazy things with it for surveillance using a little code.  Stacked traces, etc.  I then got a great deal on a matching 85685 preselector/preamp.  That combination is lots of fun.  I turns right on, I can get a measurement out in about 20 seconds, and I don't have to worry down to a .1DB.  Read all the Rigol and Siglent threads about glitches, etc.  That stuff would drive me crazy.  Yes, this stack weighs in at 150lbs at least.  I think the entire setup had a list price approaching 80k new.

This summer I found a sat engineer selling an 8566B.  I was the first of five that called him within a few hours of posting as he was asking $500.  I worried all the way over there, waited in his driveway for 30 minutes.  That one has near perfect cal constants but someone had been in it and didn't replace some of the metal covers.  One I machined, the other I got from ebay.  That analyzer is also near perfect.

But if siglent or rigol made one with a TG that went from 9k to 3G for $1K I would probably snap it up since I travel a lot and would like to be able to shove it in my carry-on.