TSP Short #17 – Teardown & Experiments with an Anritsu MS9710B 0.6-1.75um Optical Spectrum Analyzer

In this episode Shahriar repairs an Anritsu MS9710B Optical Spectrum Analyzer. The instrument intermittently does not boot and when it does it generates a Grating Mirror error. Some investigation reveals that the problems may simply be poor internal cabling connections which can be resolved by cleaning all connectors and re-building the unit. The LCD screen’s CCFL have also failed. An LED back-light replacement provides the screen with a vibrant and bright colors.

The unit is very compact (which is unusual for an Optical Spectrum Analyzer) and makes it ideal for bench-top working environments. The broad-spectrum built-in reference laser can be used to calibrate and align the instrument which proves to work perfectly. After calibration, a laser diode from an Agilent Lightwave Transmitter is measured. Interestingly it can be seen that the unit produces two tones at ~1nm apart which indicates the laser operates in two modes. A very interesting result given that the wavelength meter only reports one tone at the average wavelength of the two modes.

The Signal Path


Similar Posts


  1. Nice! I have the exact same OSA, if you want I can do the same test with the attenuator. I actually got a grating error once, freaked me out, I though my nice instrument was broken. Good to see it's probably just a cable issue.

    The laser you have can have 2 modes, actually laser diodes usually have more that one or skip between modes, having only 1 takes some effort. I also have an ANDO OSA which is a lot smaller but it has worse specifications and it's boken, it simply won't power up. So it can be made even smaller.

    I also have an HP wavelength meter, it occasionally gives a motor error. It has 1pm accuracy but not for peaks that close together probably, and the second peak is 12dB lower. I'd have to look up the specifications. you can actually acces the spectrum the wavelengthmeter measures, I though both graphically and via GPIB, then you can check if it cannot resolve it or that it's just the peak searching algorithm not separating the peaks.

  2. It is almost certainly an issue with the laser, and not an issue with the spectrum analyzer. It would be interesting to confirm by testing with a different laser (if you do not have another laser in the lab we could send you one). The second peak is not entirely uncommon for a DFB laser like that in your 83433A. Nearly all DFB lasers will have at least one side mode like that a few nm away from the peak, but typically the side mode suppression ratio (SMSR) would be >40dB for a high quality DFB laser, not the 12dB that you measured with your laser. Most likely they are simply different longitudinal fabrey-perot cavity modes which are competing for gain with the main peak.

    For an example of this behavior, you can reference the temperature turning curve of the LP1550-SAD2 (a typical 1550nm DFB laser sold by Thorlabs), as shown on page 2 of the datasheet https://www.thorlabs.com/drawings/7391120fee9f4d75-93F0E832-B94A-E8F0-BA14B4B4D54EAA58/LP1550-SAD2-SpecSheet.pdf

    As was mentioned elsewhere in the comments, you can probably reduce the side peak by tweaking the temperature of your laser. Just guessing, since the problem mode is at a longer wavelength then the peak ,if you increase the laser temperature (to shift the lasing peak to longer wavelengths) the SMSR will improve because your will be moving the main peak to better overlap with the gain of the semiconductor material.

    Furthermore – the 0.07nm bandwidth of your MS9710B corresponds to about 10GHz at 1550nm, so if you drive the modulator in your 83433A at its maximum 10.7GHz you should be able to just resolve the sidebands in the optical spectrum.

    Congrats on repairing your MS9710B, I wish mine were that easy to repair. It has an odd fault by which it will take a single trace but freezes while processing the data. I originally suspected a mechanical problem (sticky grating motor or the like), but the mechanical parts all seem to be working fine so I now suspect it is an electrical fault or corrupted firmware.

    I can at least confirm that the expected behavior of the attenuator is to greatly over-attenuate the visible region of the spectrum. This appears to be due to the attenuator being made of an absorptive glass with a non-flat absorption (likely something similar to Schott NG9 http://www.sydor.com/wp-content/uploads/SCHOTT-NG9-Neutral-Density-Filter.pdf). I am not sure why they chose a glass that has so much absorption in the 600-700nm region, but I suspect that it has to do with the non-flat sensitivity of the detector/grating and trying to find a balance between maintaining good performance at the 'important' wavelengths in the telecom band while not saturating at shorter wavelengths.

  3. Thanks for the great video’s. Gotta love them with great appreciation. I recall from spectroscopy classes at university that spectral lines such as laser light may be splitted by magnetism. I think it was called the zeeman effect. Hope this helps.

  4. I have a MS9710C that I repeared. The unit was pretty beat up (I bought this one from Ebay: http://www.benl.ebay.be/itm/ANRITSU-MS9710C-OPTICAL-SPECTRUM-ANALYZER-ANALYZER-OPTICAL-opt-05-G410/311929746096?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2057872.m2749.l2649) To answer your question, when I enable the optical attenuator, I see exactly the same result as you do (increased noise floor). So, or this is a common fault with these units, or it is normal.
    It seems you have the internal calibration source with two ports (SLD output and reference output, see https://www.upc.edu/sct/documents_equipament/d_125_id-630.pdf) , which one did you use for the calibration? I only have the reference output (1-port), but I was expecting to see the acetylene dips in the spectrum, but this is not the case (also not on yours, hence my question). On mine I can execute the wavelengt calibration with succes, but when I try autoalign I get a level error. I used to think the acetelyne cell had lost is content, but now when seeing yours, this is maybe not the case. When I see your video, your peak reference output is around -32dbm, where mine is -36.8dbm, so this is maybe the reason.Anyway, thanks again in making this interesting videa, it always make my day when a new one appears!

  5. I appreciate that you take out the boring bits but this time around you really lost me. Somehow it went from ripping out the HV backlight to booting with correct date and time.

  6. Most likely there is a DBR laser inside the Agilent lightwave transmitter. If it's the modulation it will cause symmetric sideband on both sides, however to create sideband this far away you have to modulate the laser at hundreds of GHz frequency. The most likely cause of the two peaks, besides a broken laser, is that the laser hasn't worm up properly and the temperature of the DBR is not yet stabilized which cause the diode to mode hop.

  7. It is perhaps a side mode from laser source, looks like there is also a small peak at the left within a same distance. There is a wavelength tuning nub at your source that changes the laser temperature, it is interesting to see if changing the temperature has any effect on these modes. BTW, you will have a ~110GHz oscillator in your lab if you send this laser to a high speed photodiode!

  8. I had no idea lasers could be polychromatic. Makes me wonder if frequency modulated LEDs or lasers exist too.
    I must consult the Google

    edit: Seems there is some fancy things and concepts, But nothing as a single device like an LED. Maybe one day.

  9. It may be a second order intermodulation distortion, but changing the fiber is a first step. If fiber and laser source is hunky dory, then may be the issue is on the analyzer, may be modulators.

Leave a Reply

Your email address will not be published. Required fields are marked *