Revisiting the MHS5200A

I’ve gotten a lot of questions on the blog about the new version of the MHS5200A function generators available on eBay. Viewer Tolga was kind enough to send one in to me to review and tear down. Although some improvements have been made over the older models, there are some concerning issues with these new models too!

6 thoughts on “Revisiting the MHS5200A

  1. I bought the new 12bit version of MHS5200A …
    Started to analyse and modify.

    One finding is related to precision/granularity of the generated frequency
    – simple setup: connect “Ext In” with “Ch 1” and start the counter with 10s gate time
    – set generator to e.g. 20,000 000 00 MHz
    – frequency counter shows 20,000 146 3 MHz (7,3 ppm off), there is only one clock
    reference!!! Nothing to do with crystel tolerance or temeperature!
    This is a deterministic SW issue! Assume the error is in the handling of the frequency word (e.g. word length issue between user interface and DDS core)
    btw: You can change DDS frequency by up to a few Hz without changing the generated signal! Try to generate precise 20,000 000 0 MHz by tweaking the DDS frequency input … you will fail ;-(

    Signal integrity
    Observed slow 1 kHz wide sweep with shortwave radio … CW-mode sounds terrible, sidebands with 50Hz grid! … needs further investigation!

    Next step is adding Bessel filter … will not impact above issues ;-(

    In general:
    – nice device for a very attractive price
    – compared it with my 40 years old Wavetek 144 function generator
    -> the new DDS is the clear winner! (even with all the known impairments)

  2. Hi Craig,
    actual status
    – signal integrity is no longer an issue, replaced “wireless cable” by real coax 😉
    – added 9th order Bessel filters (new 12bit-DAC version!) … was no big challenge
    – have pictures
    – comparing results in time and frequency domain
    – how to implement the filters

    how to upload the pictures? can send them via e-mail …

    • You can upload the pictures to imgur or a similar picture hosting site and post the links here. I have this unit as well (the 12-bit DAC version) and am wondering how you did the mods (the Bessel filters, the coax cable etc).

  3. preparing pictures will take a while …

    adding Bessel filter
    I designed the filter for n=9, 30MHz @ -3dB and 100Ohm
    normalised coefficients
    L 0.2313 0.5108 0.7407 1.0863
    C 0.078 0.377 0.6306 0.8639 2.2649
    resulting in following components
    L 100n+22n 100n+150n 330n+68n 470n+100n (in series)
    C 3p3 20p 33p 47p 100p+22p (last one in parallel)
    – the filters are added on the “solder side” of the PCB (clearance 5mm)
    – implemented on a 5mm wide stripe of PCB (a very delicate 3d implementation)
    – I cutted the signal flow between the output of R/2R-network and the input
    resistors of the following Opamps … impedance is 100Ohm!
    – you will find both signals on short connections on the solder side …
    simply cut the short connections and add the filter (picture will follow)
    – time domain signals are looking excellent (pictures will follow)
    noise reduction is limited (due to 30MHz bandwidth)

    Meantime I found other issues
    – set the generator to square wave (e.g. 1kHz)
    – trigger the scope on falling edge
    – zoom in (e.g. 1us/div)
    -> you will see an “extra step” … I have no clue !!!
    is this the forgotten rest of a desperate approach to reduce ringing of no longer
    existing elliptic filters …. ???

    Power supply
    – observed sporadic problems with displayed data on LCD
    – measured the received 5V and could observe up to 700mV voltage drop
    – maximum current consumption is up to 1.3A … above rating of poly-fuse (1.1A)!
    -> fixed problem by adding a local DC/DC step-down converter and replacing
    the external power supply by a 12V version

    Thermal issue
    – the two final Opamps are very hot!
    -> replaced the small heatsink by a much bigger one

    Signal integrity
    – added supply-filter Cs (1n, 10n, 100n, 22u) to the output amplifiers
    and the IO- and core-voltage of the logic device (solder side of PCB)
    – benefit? don’t know … at least a good feeling!

    Comment regarding “coax cable”
    – first measurement was based on unwanted radiation … 😉
    the magnetic loop of my SDR could pick up the signal (including the unwanted
    50Hz harmonics)
    – 50Hz harmonics disappeared after replacing antenna by 50Ohm coax cable
    -> strange ??? … not MHS5200A related!

    ideas related frequency counter
    – trim the crystal to 16.000 000 MHz (optional with an added “crystal heater”)
    the existing device is a 9 digit frequency counter with a 5-6 digit precision …
    – this would impact the precision of the generated frequency
    you can enter a 10 digit number … maximum 6 digits are relevant!
    -> it’s your decision … what do you want?
    – “fake” precision for generator and counter or
    – improved counter precision and degraded precision for DDS-generator

  4. the pictures are available!
    see under http://imgur.com/a/s9uRz

    figure 1
    – 3d implementation of Bessel filters (looks not so nice … but works)
    – additional decoupling for logic chip and driver
    – comment
    inductors are close together, risk for unwanted coupling
    assume this is no real issue … performance like expected!

    figure 3
    – shows required cuts for the filters
    upper one for ch1
    lower one for ch2
    warning: there are many differnt layout versions, no revision numbers …
    … at your own risk!

    figure 2
    – comparison with a 400MHz/14bit DDS from AD, LP-filter @160MHz
    both generating 4MHz signal, single shot acquisition, zoom around zero
    – yellow: MHS5200A with Bessel, blue: AD DDS
    -> much wider band, still much better signal!
    the MHS-generator is limited by glitches generated by the R/2R converter

    figure 4
    – FFT spectrum with Bessel
    figure 5
    – FFT spectrum without filter
    -> improvement between 50 and 100MHz (as expected)
    -> a LP-filter with 30MHz cut off can not clean up spectrum below cut off!
    btw: the implemented Bessel filter should have 40dB attenuation at 100MHz

    figure 6
    – 15kHz sine, 2.5Vpp, zoom around zero, single shot
    yellow: with Bessel, blue: without filter
    -> visible improvement

    figure 7
    – sine/triangle @ 5MHz, Bessel filter
    single shot acquisition

    figure 8
    – sine/triangle @ 5MHz, Bessel filter
    average 1024
    -> to see filter response without noise

    figure 9
    – bigger heatsink … strongly recommended!

    figure 10
    – redesign for external 12V power supply
    – local DC/DC step down with additional LC-Filter (10uH, 330uF)
    – poly fuse for 12V replaced by 700mA type
    additional benefit: you have 12V in the box!
    – for an optional “crystal heater” (search for “Kuhne crystal heater”)
    – for a better input amplifier for the counter
    warning: don’t try to copy my wiring … there are different layout versions!

    the best at the end!
    figure 11
    – the magic “bonus step”

Leave a Reply

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