Connect with us

Low cost spectrum analyzer <100MHz?

Discussion in 'Electronic Design' started by Joerg, Mar 3, 2007.

Scroll to continue with content
  1. Joerg

    Joerg Guest

    Question: Is there a low cost spectrum analyzer that allows low level
    noise measurements? Roughly like this:

    1-100MHz or more.
    Narrower filter than the usual 9kHz (but not steep).
    USB, can be "faceless" and laptop driven.
    RS232 ok but not so cool.
    Must be able to zoom in well.
    Reasonable dynamic range, 80dB+.
    Does not have to be "calibrate-able".

    Or is there a good deal on used ones?

    Reason I ask is that the only analyzers I am used to and know are the
    "big irons" for hard EMI cases, upwards of $20k. That would clearly be
    overkill for this client. They never have to measure any GHz stuff.
    Renting gets expensive over time, they will need it longer.
     
  2. John E.

    John E. Guest

    Joerg sez:
    I know very little about such test equipment, but I was at one time
    subscribed to a Google group focusing on such. They were constantly finding
    used Tektronix scope plug-ins on e-Bay for cheap. There is a constant --
    though thin -- supply of several different models (both portable and non-) of
    spectrum analyzers needing either little or no repair. The group collectively
    is quite a good resource of parts and repair knowledge when it comes to this.



    or go to Yahoo groups and join the regular way...

    Good luck,
     
  3. Some time ago (say, 10 years or so) there was a series written on a
    "homebrew" spectrum analyzer that used a TV tuner as an LO. I know the
    series of articles was googleable, at least a couple of years ago. You
    could have pretty much any bandwidth you wanted as a function of what
    ceramic or crystal filter you chose.

    Jim
     

  4. It's this one, isn't it?
    http://www.nitehawk.com/rasmit/sa50.html
     
  5. Joerg

    Joerg Guest

    Agree. TV tuners are among the worst performers in terms of dynamic
    range. Lately it has gotten worse. Sometimes I have the feeling the
    young lads at those design houses don't really know what a dynamic range
    is. After re-roofing our house I had to move the antenna onto a side
    mast. I figured, well, I might give it the perfect spit shine, a nice
    amp, distribution and so on. The old set we used to have in the garage
    really liked that. Tons of new stations. Not so the newer set in the
    living room. It completely fell off the rocker even after putting a
    notch on the two strongest channels. I had to attenuate almost 10dB to
    make that set happy and lost one station in the wake. Pathetic.

    Anyhow, back to the analyzer issue: It's for a client and they don't
    really want to make a science project out of it. What's a nice low cost
    "staple" among low end analyzers? I only know the top notch equipment
    but, for example, we do not need any precise frequency indication. It
    could be one of the old non-PLL analyzers. No tracking gen, just a good
    dynamic range. I am just not familiar which ones represent a good deal.
    Any hints?
     
  6. doug

    doug Guest

    Having built this circuit many years ago, I can tell you to avoid
    it completely. The dynamic range is very low (maybe 40db) and the
    results are unpredictable with frequency. You are much better off
    getting a used hp on ebay.
     
  7. vasile

    vasile Guest

    Depends. A lazy one, with a fake noise level measuremet (only for some
    frequency ranges and modulations) is here: http://www.aaronia.de


    Roughly like this:
    Breaking the 1Ghz range in multiple 100MHz ranges, with an overlap
    between adjacent ranges (see how works Tektronix analyzers)
    Designing a narrower filter is difficult, and could be useless.
    USB does not mean it will be for sure in "real time"... think at ISM
    band (2.5Ghz) and how fast are
    WIFI signals
    This will require a stable reference, preferable digitally
    synthesable.
    The maximum available dynamic range is finally a matter of bits
    numbers of the A2D converter:
    SNR = 6.02 x n bit + 1.76dB

    It means you need a fast ADC with more than 12 bits.

    Of course the analyzer arhitecture it may be zero IF or with one (or
    more) IF stage...
    Without a complete calibration (amplitude and frequency) a spectrum
    analyser is useless.
    If the frequency calibration could be easy done based on the odd/even
    harmonics of a stable 10Mhz crystal
    (you'll see up to 100MHz all harmonics), the amplitude calibration is
    not easy. An 1Ghz Tektronix scope, DPO4000, has about 2% amplitude
    nonlinearity, and it's factory calibrated.
     
  8. Joerg

    Joerg Guest

    I wouldn't shy away from whipping up a filter myself. Done it many
    times, including really steep crystal filters. One issue with the
    simpler analyzer models is that they don't have a nice low noise PLL but
    a YIG oscillator. I found those to be a bit noisy. But OTOH it'll be
    enough for us.
    If it needs a few seconds to transmit it's data that'll be fine. I grew
    up with Polaroid cameras hooked to side hinges on the screen. Worked.
    Except when you got the goo from the back foil onto your clothing.
    I've done it with really old ones. The trick is to turn it on, make
    coffee, do some other design work for an hour or so and then it's warmed
    up. You just have to make sure that nobody touches the thermostat on the
    wall or opens a window. Would be no problem here. Believe it or not but
    this office is usually heated by a wood stove. You can keep the room
    temp at 69.0F and it stays there. To the point where I sometimes thought
    the digital thermometer had frozen up on me.
    No ADCs in the old ones :)

    The impedance analyzer here in the lab uses a slope converter that (in
    theory) has almost infinite resolution. I get north of 90dB out of that.

    True, but my experience with I/Q demod scheme is, well, mixed.
    This is how I've done it: Take (or borrow) a generator with known output
    amplitude over the whole range. Or make one from a few PECL gates. Then
    plot the amplitudes on the analyzer. Now turn on your 10MHz harmonics
    thingie and compare. From then on you know the amplitude of the
    harmonics. In the 80's I did that stuff on a notepad sheet because all
    they had was a daisy-wheel printer and the accounting folks had that
    occupied all the time.

    On really old analyzers I sometimes let the harmonics generator run into
    the input line via a 20dB splitter, to see where I am on the frequency
    scale. I mean, in the olden days they navigated big airliners in a
    similar fashion by following headings to regular AM band stations. It
    always got them home.

    Anyhow, looks like we might just buy a real oldie like the HP141. I've
    used that a lot when I was young and it has never failed me. It is just
    very big and heavy. And no automatic stuff whatsoever. That was why the
    others back then hated this thing so I could have it as long as I wanted
    to ;-)
     
  9. john jardine

    john jardine Guest

    If you were in the UK I'd strongly suggest a Marconi TF2370.
    20Hz-120MHz, 100dB on screen, 50kHz to 5Hz filters. +30 to -150dBm range.
    Cost maybe $200.(weight 40kg!)
    john
     
  10. Joerg

    Joerg Guest


    I've only seen one of those for sale. Unfortunately it is in Australia
    (I'm in the US) and it is in pretty deplorable shape. H-deflection looks
    dead, digits missing etc. We'll probably go for the HP141T mainframe
    with the respective modules in there. That'll feel like stepping into my
    old Citroen again.

    While at it I was looking for Tek 7L13 modules for the 7704 mainframe
    here in the lab. Surprisingly those are still rather expensive, almost
    $1000. Same for the older 7L12. Something like that would be just fine
    for the task at hand. But you can get a complete HP141T set-up for less
    money and it comes with the display mainframe.
     
  11. maxfoo

    maxfoo Guest

    With those specs (80dB DR) you won't find nothing under $3000.00 even on ebay
    That said, this looks nice 50K-3.3GHz for abound 3 grand brand spanking new...
    http://www.bkprecision.com/www/np_pdf.asp?m=2650
     
  12. Joerg

    Joerg Guest


    It's been too long ago but I remember guys in the 80's saying that I'd
    be lucky to squeeze more than 60dB or so out of the old HP141. Then I
    did. They still didn't want to use it because everything had to be
    hand-dialed and hand-tuned and it was drifting a bit. AFAIR you just had
    to be quite careful not to fry its input.


    That is indeed nice. Thanks, might be an option for us.
     
  13. The 141T system still has much to offer.

    TMT
     
  14. Robert

    Robert Guest

     
  15. vasile

    vasile Guest

    Joerg, I have bought an old YiG Tektronix T494. Is useless for most of
    the digital signals
    requesting to be measured nowadays. Forgot about that and focus on:
    http://www.tek.com/site/ps/0,,37-19513-INTRO_EN,00.html

    This sort of old spectrum analyzer, like T141 is good only to see the
    TV stations and repetitive signals.
    You can't analyse anything is digitally transmitted and is switching
    fast (like WIFI, WIMAX, WCDMA, GSM, etc)
    I have no problems with DDS. One of my beginning designs with DDS is
    here:
    http://www.surducan.netfirms.com/top-ds1.jpg
    Now I'm designing 2.5Ghz and 5GHz OFDM sytems (WIMAX)

    And infinite laziness as well. I think you are familiar with all
    single-dual-multiple slope converters, right ?
    Then you know how much it takes for one measurement and why.

    Ha-Ha !
    We've just returned to Agilent a new bought $350.000 network analyser
    just because
    it comes from factory with a problem above the 22Ghz range...
    As long you can't prove the generator is working well, you have no
    reason to
    believe is working well... a calibrated certificate means NOTHING.

    Or make one from a few PECL gates. Then
     
  16. Anything of interest here?

    http://test1.contenttest.net/Spektrumanalysator_en.shtml


    martin
     
  17. Joerg

    Joerg Guest

    It sure does. However, most of the units available today are equipped
    with the 8554B plug-in that goes up to 1.3GHz. I am used to the 8553B
    (to 110MHz max). Do you remember how well the 8554B does when most of
    the stuff you have to measure is between 10MHz and 100MHz? I know it
    covers that range but it'll be at the bottom 10% of the dial range.
     
  18. Joerg

    Joerg Guest

    We are only after low level noise measurements. It's all going to be
    plain vanilla AM and phase noise but for the latter we already have
    external discriminators. Can't tell you details here but they operate in
    the Tera-Hertz region.

    Thing is, my client needs to make measurements for maybe a couple hours
    per week, then optimize a production process which requires days of
    re-tooling, measure again, re-tool again and so on. I doesn't make sense
    to spend $30k for an analyzer or shell out $500/week in rent if the job
    can be done with a vintage analyzer. In the same way that I wouldn't use
    a Ferrari to pick up lumber at the hardware store.
    I've built a few, too, but not at frequencies that high ;-)
    Time is usually not such a big factor. My impedance analyzer can take 15
    minutes on some scans. But it's programmable and does it while I can
    continue on a schematic, or take the dogs for a walk (they love that).
    $350k? Wow. Well, this ain't a big or gvt funded company but a start-up.
    They don't have that kind of budget :-(

    And I am absolutely certain we can do it with a 141T.
     
  19. Joerg

    Joerg Guest

    Maybe but I don't know how well they'd do noise floor measurements. I
    know the old machines from HP and some from Tek would. The only one from
    Aaronia that would be able to measure the required range is the big one
    (1MHz-7GHz). And they still don't seem to have a dealer in the US as far
    as I could see. On the online form it says overseas order can take 6-8
    weeks. Too long, and if it doesn't do noise floors for us the return
    could be a hassle.
     
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day

-