Do it yourself Spectrum Analyser

[steve_w]

Has anyone seen this or had a go:

http://scottyspectrumanalyzer.com/

Looks impressive as a project but a big commitment in resources.  My worry is that if you built as described you could make a large junk pile with no useable product at the end.  Not because the project is neccessarily bad, but the skill and experience to debug and align to make a good product would not be for everyone. 

This is a complicated project with a wide range of skills required.

Does anyone have an opinion on viability?

[Wartex]

Has anyone seen this or had a go:

http://scottyspectrumanalyzer.com/

Looks impressive as a project but a big commitment in resources.  My worry is that if you built as described you could make a large junk pile with no useable product at the end.  Not because the project is neccessarily bad, but the skill and experience to debug and align to make a good product would not be for everyone. 

This is a complicated project with a wide range of skills required.

Does anyone have an opinion on viability?

I am a retired RF Design Engineer from Raytheon

Your skillset = |||
His skillset = |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||

I'm 99% positive it will be extremely out of tune and pretty much useless by the time you are done because mentally he is much, much better at seeing signal behavior in the circuit than you are. With high frequencies Z axis becomes very relevant and IMHO you will botch it unless you know what you are doing.

[steve_w]

I might be a retired RF design engineer

[sonicj]

Has anyone seen this or had a go:

i saw that a few months back... epic!

judging by the participation on yahoo groups, its not impossible.  boards are available through expresspcb & component kits through mini-circuits. the "MSA Summary" here is a good intro read.
-sj

[amspire]

I have looked into this project before, and I think it would be a really great project. I have been tempted.

But there is a lot of hardware, and unless you have some old RF gear to scrap for parts like connectors, plumbing, etc, it looks to me like an expensive project. Also not an easy project if you don't already have lots of RF test equipment.

Still, if you wanted to learn about working in RF, what a way to do it! Worth the effort for sure. Scotty Sprowls - you are a legend.

Richard.

[joelby]

A lot of advice and support is dispensed on the mailing list, so a long as you're reasonably competent I wouldn't be too concerned about ending up with something that doesn't work. The design is very modular so you can build and test each part separately. Perhaps the most complicated component is the cavity filter, which is built from metal pipes, but a recent innovation was a low cost SAW filter.

I bought a complete set of PCBs and aluminium boxes for about $130 from eBay (someone who is building their own manufactured extra boards) and you can get all of the Mini-Circuits parts at a significantly reduced price if you use the quote published on the site. Other miscellaneous bits like SMA bulkhead connectors and SMA cables are very cheap on eBay compared to the regular outlets.

The hardest part is finding the spare time to build it.. Especially when I already have two commercial spectrum analysers!

[amspire]

The design is very modular so you can build and test each part separately.

This is what I love about home made spectrum analyzers - they are basically built out of a collection of modules often with each module doing very simple jobs. So at any time, you can experiment improving one module without effecting the rest of the instrument much. You can try different oscillators, filters, mixers just by unplugging the old module and plugging in the new one to see what happens.

[steve_w]

When I read that web site I was convinced that you need an RF lab to construct the spec an.  I am seriously considering giving it a go, my only worry is that I don't have the support equipment to do the design justice.  Its a bit like the chicken and the egg. 

Once I get the Atmel out of my system and I finish off the DDS signal generator and the remote control on the pool lights etc etc I might get the motivation to kick off.

Also some of the new Chinese spec ans are so cheap.  what to do?   As one of my friends once said you can have a spectrum analyser or a wife but you can't have both...  :-)

[sonicj]

When I read that web site I was convinced that you need an RF lab to construct the spec an.  I am seriously considering giving it a go, my only worry is that I don't have the support equipment to do the design justice.  Its a bit like the chicken and the egg.

did you happen to read the "MSA Summary" i linked above? it sounds like you only need the basics (scope & dmm) plus a power meter and optionally a signal generator to complete the project. once a module is done, it can be used to test & tune the next module, etc.
-sj

[steve_w]

SonicJ,

I did read the summary, good pick up.  I' just so tempted to do this, I know it would be a lot of fun, but I know I don't have the time.

[saturation]

hi joelby,

Do you think once completed this SA will be on par with the lower end of those available commercially?

The website links a BOM of about $300-400.  Is this still ~ right? Availability and obsolescence not an issue?  Add PCB and housings  total parts cost ~ $540.

Looking at folks who've attempted it and the timeline, this one is better organized:
http://alternet.us.com/?p=1576

A lot of advice and support is dispensed on the mailing list, so a long as you're reasonably competent I wouldn't be too concerned about ending up with something that doesn't work. The design is very modular so you can build and test each part separately. Perhaps the most complicated component is the cavity filter, which is built from metal pipes, but a recent innovation was a low cost SAW filter.

I bought a complete set of PCBs and aluminium boxes for about $130 from eBay (someone who is building their own manufactured extra boards) and you can get all of the Mini-Circuits parts at a significantly reduced price if you use the quote published on the site. Other miscellaneous bits like SMA bulkhead connectors and SMA cables are very cheap on eBay compared to the regular outlets.

The hardest part is finding the spare time to build it.. Especially when I already have two commercial spectrum analysers!

 

[joelby]

hi joelby,

Do you think once completed this SA will be on par with the lower end of those available commercially?

I couldn't say for sure since I haven't compared the figures (maybe they're on the site?), but probably. In terms of features it wins, because some of the lower end commercial SAs lack tracking generators, and I don't think I've seen a cheap one with a network analyser mode.

The website links a BOM of about $300-400.  Is this still ~ right? Availability and obsolescence not an issue?  Add PCB and housings  total parts cost ~ $540.

I'd say that would be about right, and probably a bit less if you're already a well-stocked hobbyist. I presume that modules with obsolete parts are redesigned when necessary, or modifications are discussed on the forum.

Judging by the discussion group chatter, it seems like the project is nowhere near the stage where someone can buy all of the bits as a kit and put it together without any problems. Buy a second-hand commercial SA if you need one for your work. However this is great project if you're interested in RF design and are willing to tinker and improve it - due to the modular design, there is definitely a lot of scope for fiddling with this one if that's your thing.

Looking at folks who've attempted it and the timeline, this one is better organized:
http://alternet.us.com/?p=1576

Here's another good page: http://www.qsl.net/bg6khc/my_version2.htm . This is the guy I bought my PCBs and enclosures from, though he doesn't seem to have any listed on eBay at the moment. I think he redesigned the 'official' PCBs slightly to fit the dimensions of his enclosures.

[steve_w]

That takes a lot of effort and discipline to re-engineer I'm staggered at the commitment.

[saturation]

Thanks everyone, and particularly joelby.  That's a great link.  An irony is he has access to or owns commercial Chinese SA to confirm his build.  I guess the DMM or VOM Scotty suggests as a minimum leaves too much for the imagination, particularly when things go wrong. 

Even if a full kit of this SA were finally made so only assembly were required, testing and tuning [ if not calibration] will likely require an instrument of >= stature.   

Judging by the discussion group chatter, it seems like the project is nowhere near the stage where someone can buy all of the bits as a kit and put it together without any problems. Buy a second-hand commercial SA if you need one for your work. However this is great project if you're interested in RF design and are willing to tinker and improve it - due to the modular design, there is definitely a lot of scope for fiddling with this one if that's your thing.

Here's another good page: http://www.qsl.net/bg6khc/my_version2.htm . This is the guy I bought my PCBs and enclosures from, though he doesn't seem to have any listed on eBay at the moment. I think he redesigned the 'official' PCBs slightly to fit the dimensions of his enclosures.

[Conrad Hoffman]

Spectrum analyzers are hard to finish! I built The Poor Man's Spectrum Analyzer many years ago, tested the modules and made a nice case. It worked great on the bench. I've yet to finish the p2p wiring so I can use it. Still, I'd start with that design if you're not well versed in RF. It's a lot cheaper and easier to build.

[joelby]

An irony is he has access to or owns commercial Chinese SA to confirm his build.

I found this interesting - clones of expensive Agilent gear! I found it difficult to obtain single unit pricing, and when I did they weren't DS1052E-cheap.

[steve_w]

Thanks Conrad, where do you get the time to do stuff like that. 

That's pretty good design work and really neat idea to modularise but to verify each module you need more test equipment, like: Vector analyser, sweep generator, directional coupler, variable attenuators, loads, et al.

I did a micro strip design course at TAFE about 20 years ago where we did a lot of design and fabrication using PUFF and we made patch antennas, directional couplers, filters and amplifiers but that was very simple compared to what you did. Excellent work.

Pretty much everywhere I've worked for the last forty years has all this stuff, except where I work now. :-(  maybe I'll have to back to the old job.


regards

Steve W

[saturation]

Yes, its one of the most fascinating parts of your links.  I searched and found nearly identical looking devices with different brand names such as Modell.  But the price is puts it in league with the top end Rigol SA,  ~ $6000.

However, with the interest sparked by this thread you can get basic SA capacities for RF monitoring with the HP 8558b, typically $200-500 or less, for 0.1-1500 MHz.

An irony is he has access to or owns commercial Chinese SA to confirm his build.

I found this interesting - clones of expensive Agilent gear! I found it difficult to obtain single unit pricing, and when I did they weren't DS1052E-cheap.

[steve_w]

Hey Saturation where do I get a HP 8558b, for $200-500? sign me up

regards

SW

[saturation]

eBay ... the last sale I've been watching just ended:

http://www.ebay.com/itm/251000532621?ssPageName=STRK:MEWAX:IT&_trksid=p3984.m1423.l2649

Hey Saturation where do I get a HP 8558b, for $200-500? sign me up

regards

SW

[DaveW]

Hey Saturation where do I get a HP 8558b, for $200-500? sign me up

regards

SW

Keep watching eBay-I got one for £300, including shipping from Canada to the UK. Amazing performance for a low price

[saturation]

Yes, I wouldn't be surprised if the Atten 1GHz is a copy of this mostly analog SA.  It was made in fairly large numbers so its easy to find at reasonable cost.  The user/service manual is easy available, so know its capabilities and what it takes to keep the unit going.

It was modular, and CRT was sold separately, but not sure you can interface your own CRT to it as easily.    So the prices under $200 are typically just the SA module, and the $300+ is the full unit with varying types of CRT options. 

Good hunting and enjoy!

Hey Saturation where do I get a HP 8558b, for $200-500? sign me up

regards

SW

Keep watching eBay-I got one for £300, including shipping from Canada to the UK. Amazing performance for a low price

[tinhead]

An irony is he has access to or owns commercial Chinese SA to confirm his build.

I found this interesting - clones of expensive Agilent gear!

ehm, no idea what you guys talking about, Anritsu (and that's on the pictures) is not chinese manufacturer.
Beside this, Anritsu is producing since years much better SA than HP did or Agilent is doing.

Yes, I wouldn't be surprised if the Atten 1GHz is a copy of this mostly analog SA. 

ATTEN started to clone Hameg years ago, in the meantime they learned "much" about SAs. Their last
models are (analog part) still somehow very close to old analog Hameg SAs, but digital part is of coure new.
Good example is the "ghost SA - yeah, you can buy it, but manufacturer told me don't buy" - AT5010D

http://www.ebay.com/itm/110802550735

Sure, the expensive ATTEN models are not anymore based on these Hamegs, but probably
still electrical clones of existing gear (ATTEN is writing own firmwares, so not real clones).

Anyway, i would not recommend any ATTEN SAs - they trying to fix USB and some measrment issues on
these AT5010D since 6+ months, no go for me.

Now Rigol is having these DSA 815, they cheaper and light years better that this ATTEN crap,
even much metter as these HP 859x series (when 1.5GHz is enough for you).


For DIY SAs, the most critical path is (even if simple to build, but tricky to tune up without equipment)
the cavity filter. Therefore i used in my DIY SA out of the box GPS filters - 5pcs in series and
later more IF stages (1575.42 MHz, 868.3MHz, 10.7MHz) and each of them with 3-5 out of the box bandpass filters.
I even not tried to match them, simply used app notes.
The tuning of last filter stage (10.7MHz) can be done with scope or even multimeter (real RF or DIY RF meter),
all you need is to tune all bw filters to exact the same amplitude (10.7Mhz test signal) and of course max possible amplitude.
You don't need to know the amplitude (therefore DIY RF meter is god enough) later you can calibrate with known
signals in the software to get good linearity (which is anyway necessary because your mixers, filters and LOs are
not linear).

EDIT: or well, it make of course sense to have good signal generator to tune up DIY SA, signal detector is not a big
deal but ignal generator should be linear as much as possible.

[joelby]

ehm, no idea what you guys talking about, Anritsu (and that's on the pictures) is not chinese manufacturer.
Beside this, Anritsu is producing since years much better SA than HP did or Agilent is doing.

The page we're referring to is this one: http://www.qsl.net/bg6khc/test_on_filters.htm

[tinhead]

ehm, no idea what you guys talking about, Anritsu (and that's on the pictures) is not chinese manufacturer.
Beside this, Anritsu is producing since years much better SA than HP did or Agilent is doing.

The page we're referring to is this one: http://www.qsl.net/bg6khc/test_on_filters.htm

Yeah, this is for sure chinese SA (and here is bg6khc SA http://item.taobao.com/item.htm?id=13003940513 )

ok, i thought http://www.qsl.net/bg6khc/my_version2.htm

[saturation]

Thanks for the comments tinhead, your insights are always valuable.  What do you think of that Chinese SA?  Its about $2400 given conversion of yuan to dollars.

ehm, no idea what you guys talking about, Anritsu (and that's on the pictures) is not chinese manufacturer.
Beside this, Anritsu is producing since years much better SA than HP did or Agilent is doing.

The page we're referring to is this one: http://www.qsl.net/bg6khc/test_on_filters.htm

Yeah, this is for sure chinese SA (and here is bg6khc SA http://item.taobao.com/item.htm?id=13003940513 )

ok, i thought http://www.qsl.net/bg6khc/my_version2.htm

[tekfan]

This also might be an interesting spectrum analyzer project:

http://lea.hamradio.si/~s53mv/spectana/lcd.html

(navigate with the VCO, SA... buttons on top)

I know a lot of the local ham radio guys built his frequency counter and are very pleased.

[gregariz]

Years ago there were a number of kits doing the rounds in the magazines that used an old TV tuner as the Specan receiver section and a CRO for the readout. Just a quick seach shows this one - pretty cheap no cro but serial port instead.

http://www.mtmscientific.com/spectrum.html

If you want a good piece of gear - theres no shortage of 859x/856x HP/Agilent gear on the used market. I recently had to pick up a spare unit and picked up a 8594E for about 3.5k from one of the test rental places. That seemed like pretty good value to me for that instrument, warranted for a year, Cal'ed and NIST traceable.

[Neganur]

...
I bought a complete set of PCBs and aluminium boxes for about $130 from eBay
....

I did buy that PCB set as well and can not recommend it. Rather try to get a group buy going for those ExpressPCB panels or invest the time it takes to redraw the boards with software that exports gerber files (and send them to china). My reason is: even though the aluminium enclosures are excellent and allow for a nice modular design, the shielding is poor. The lack of fencing will punish you with a higher noise floor and a lot of spur.

Scotty & Sam spent a lot of time redesigning and verifying the boards.

[joelby]

Ah, good to know! Are you using his revised version, or an older set?

It's a shame that the 'official' site doesn't have a complete set of designs in gerber form that you can send off as a single panel or something (I find the site a bit baffling since details are scattered all over the place) - hence the attraction of buying a complete set.

I've just had another look at my boards and you're correct that the Chinese boards lack a mask-free perimeter fence area, or even clearance at the edges to remove the mask and add one in case you wanted to do that and still use the boxes. Maybe you could line the aluminium boxes with some sort of shielding material?

[Neganur]

The website links a BOM of about $300-400.  Is this still ~ right? Availability and obsolescence not an issue?  Add PCB and housings  total parts cost ~ $540.

It depends where you buy parts. I had the feeling that those in the US pay a lot less for parts, especially if they have access to surplus stores. (used coaxial SPDT switches etc., you need a few in case you want automatic band switching, or maybe a step attenuator)

The biggest part of my expenses were all the connectors, the Mini-Circuits VCOs and ICs like the 16-bit ADC and logarithmic amplifier. Also, I was really surprised how difficult it was to buy 1 inch (28 mm) copper tube in Helsinki. The local builder's store asked more than 50 EUR for a 3 m piece and I only needed 40 cm.

Next, if you don't have access to quality tools (drill press, fine saw, tube cutter) you may learn new swear words, at least I did On the positive side, the mechanical work is a lot of fun and you learn a lot.

I did not sum it up yet but I'm pretty sure I spent more than $400. You can save a lot by omitting connectors (or buy very cheap Chinese SMA connectors - I think they're OK for DC-3 GHz).

[Neganur]

Ah, good to know! Are you using his revised version, or an older set?

It's a shame that the 'official' site doesn't have a complete set of designs in gerber form that you can send off as a single panel or something (I find the site a bit baffling since details are scattered all over the place) - hence the attraction of buying a complete set.

I've just had another look at my boards and you're correct that the Chinese boards lack a mask-free perimeter fence area, or even clearance at the edges to remove the mask and add one in case you wanted to do that and still use the boxes. Maybe you could line the aluminium boxes with some sort of shielding material?

I was lucky to buy a spare panel from someone and it seemed to be the latest revision. If it wasn't for the high shipping cost (to Europe) I would simply order a panel from ExpressPCB. Minimum quantity is two panels though, a group buy of 10 panels would be ideal. Those are the basic boards for a complete vector network analyser, but there are additions like active reflection bridges, a DUT forward/reverse switch, special amplifiers and filters. Not sure why but those I asked (in Europe) would rather buy parts on their own than join a group buy.

The website holds incredibly much information and it takes several days of intensive reading to grasp what you need to look for. The great thing is, it doesn't take much to just get started. Even with a totally crappy IF-filter and a half-tuned cavity you have something that works well enough. You can tune filters with just the Control Board, the Log Detector, the ADC module, master oscillator and a DDS module and learn a lot about radio engineering in the process. It is a very laborious project. Keeping the website up to date must be a pain as well.

Regarding the aluminium boxes - I'm sure you can improve them to leak less but it's difficult with the way those lids are attached with screws. The problem inside them remains for a few boards that require fences like the DDS or the IF-filters (line of sight radiation). I decided it's not worth the risk compared to the amount of work invested and got myself a panel of those 'official' PCBs.