Frequency Stability of the QO-100 NB Transponder

  • "But even a 1kEur GPSDO might not work" and "Path irregularities will be an issue but this can only be overcome with a communication mode which can handle it." - these are exactly the points I want to emphasize.


    TELE Greenland has some 8 GHz radios links at 10 Gb/s over 50-60 km but uses 8 x 56 MHz.


    The Si5351A accuracy and stability is as good as its clock which might be anything from a sloppy crystal to an optical clock.

    • Official Post

    The Si5351A accuracy and stability is as good as its clock which might be anything from a sloppy crystal to an optical clock.

    Accuracy and stability are two separate things and your statement very much depends on whether we talk about inside or outside loop bandwidth of the SI5351A PLL. Also the internal VCO certainly does have an influence.

    Same reason why there are so vastly different performing GPSDOs while they all use the same input.

  • Indeed accuracy and stability are two separate things. Here is a very good page that describes and shows the differences: https://www.febo.com/pages/stability/


    Yes, the loop bandwidth impacts the performance. What I wanted to highlight, with the remark, was that the role of the earth stations' LO stabilities may be greatly overvalued when not looking at the total system end-to-end performance. From where I sit the latter seems to be downplayed or overlooked. The concern, that sparked my original post "High speed data use", was an understanding that a stability of at least 1e-12 was required. Later I found out that this number was over 10 000 s =:~|


    Thus there are several things to (re-)highlight

    1. 1e-12 is very GOOD, higher is adequate
    2. looking over a period of 10 000 s is nonsense for high speed data communication in the light of both 1/Tsymbol and http://g4jnt.com/QO100_Stab.pdf The "system" and decoder should be good enough, already, to sustain such a time frame otherwise they should be scrapped
    3. short term stability is important with respect to FFT bin size, 1/Tsymbol, e.g. if Tsymbol is 10 ms -> 100 Hz am effective LO wobble of 5 Hz at 10 GHz should not be a problem for a robust modulation with more than 1/Tsymbol tone spacing and a properly designed decoder
    4. a suitable modulation has to be used for the purpose, see
      External Content www.youtube.com
      Content embedded from external sources will not be displayed without your consent.
      Through the activation of external content, you agree that personal data may be transferred to third party platforms. We have provided more information on this in our privacy policy.
      and the attached comparison of FT8 and FT4 that clearly shows that FT4 is more robust than FT8 which is relevant for QO-100 use (I am not saying that neither FT4, FT8 nor PI4 should be used for QO-100, far from actually). It is all about Link Probability
    5. don't expect to use a narrow band and shoulder-to-shoulder tone location at 10 GHz
    6. using HF designed modulation(s) in the 10 GHz range is FAR from optimal irrespective if this is PACTOR IV or WSPR. Decodes may happen but Link Probability will be low even if the signal is loud especially due to path irregularities
    7. a balanced approach will be much more cost effective that an over-engineered and over-spec'ed earth station LO
    8. it all boils down to the end-to-end performance

    I don't think moving my original message into an existing thread of LO stability discussions was a good idea. It downgrades the issue of "High speed data" use to be only a question of the earth stations' LO stabilities, but, what I want to highlight is quite opposite.

    • Official Post

    Diagram courtesy EA4GPZ


    It seems that, beside the daily sinusoidal Doppler curve, we are seeing some additional effects on the frequency drift first observed by Hans DL2MDQ and Dieter DL6AGC. Obviously also visible in an earlier post by DJ0MY, but got undiscovered so far..


    Indeed, we are seeing the BPSK beacon uplinked from Bochum. Although we do have a much better GPSDO there than what most users use, there could also be some kind of influence there. Anything regarding the transponder itself is NDA anyway, but the reference in the satellite is known to be very good too ;)


    Still interesting to see this scientific research by radio amateurs on the QO-100 satellite transponder and quite amazing that people can measure such 1E-9 effect with such high precision nowadays... I guess a side effect of QO-100 :)

  • Hi Peter,


    Thanks for sharing this. I'm currently measuring the BPSK beacon frequency 24/7 for the next few days and posting updates on my Twitter ( EA4GPZ). Unfortunately I lost last night's worth of measurements due to a network problem. I find this weird wiggles quite interesting, since I haven't seen them before, and certainly they weren't present the last time that I made precise measurements, back in October 2019.


    Mario DL5MLO tells me that the GPSDO at Bochum is not to be trusted 100%, since it is reporting some error (control voltage at limit) and people can't currently get in Bochum to check, due to the lockdown. I might also measure the engineering beacon at 10706 MHz to rule out anything having to do with the Bochum beacon. I think that beacon can be received in Madrid, since we're on the edge of the MENA beam, but it's been a while since the last time I tuned to it. Another idea is to put up a constant carrier myself through the NB transponder (with the appropriate CW id nearby, of course).


    Interestingly it seems that the wiggles start at eclipse and half an orbit later. This is eclipse season, so the sun angle with respect to the orbital plane is small (this is the reason why there is eclipse in the first place). Depending on how the satellite attitude is controlled to aim its solar panels to the sun, it might need to do large yaw rates at eclipse and half an orbit later. The wiggles may or may not be related to this. We'll see how this evolves as the sun angle continues increasing and eclipses disappear (last night it was the last eclipse of the season).

  • By complete coincidence, last Sunday I happened to do the exact same measurement. It came about because I set up my QO-100 station again and I decided it would be nice to do a 24 hour doppler measurement to check that my receive chain is properly frequency locked. For the test I was using a GPS locked Octagon OTLSO, GPS disciplined Ettus B210 and SDR console 3.0.21. The plot shows the beacon offset for 36 hours starting on 5 April at 1938 UTC. I also thought the wiggles were intriguing. Please keep us updated Daniel!


  • Hi,

    This is the plot of my doppler mesurment using this GRC. A Lars GPSDO is used to drive Pluto and a twin Octagon LNB. A Python script is used to plot the data. Time is UTC. These values :

    2020-05-01 11:41:46 87.9629

    2020-05-01 11:41:47 101.05432

    2020-05-01 11:41:48 108.66935

    2020-05-01 11:41:49 123.858665

    2020-05-01 11:41:50 132.41484

    2020-05-01 11:41:51 109.19765

    2020-05-01 11:41:52 46.226505

    are excluded from ploting. An other Python script calculates the expected doppler using current 2 line data which is 39 Hz for my location for 10 GHz and Bochum for 2.4 GHz. Excluding these values the max difference is 39 Hz. What is not verified is the absolute values of doppler that are calculated that are sumetric to 0 Hz comparing these that are not.

  • Hi George,

    Clould you please publish a table of the measured values? I am interested in the shortest term variations, as these could be seen as phase jumps to a QPSK modem.

    73's Gerhard

    How many values you need? I take values every 1 sec. Of course I can post all the values but in order to have more accurate measurements I can use Cesium for reference. I have jitter +-2 Hz with my GPSDO. Cesium has a clean signal. Have you seen the GRC? Do you have any proposals? Do you need values in 5K sample rate for a little time using Cesium? The current data are low passed. In my plans is to do transmission also and compare reception with 750. Can you read raw data or I have to translate them to Ascii? I am also interested for any tests in digital modes.

  • Hi George,

    Thanks for the plots. What I am really interested in is the possible frequency change because of doppler within one symbol at 50 Baud, that is 20 Milliseconds.

    Background: I am very active in digital voice for years on HF and now also on QO-100. If you check http://www.freedv.org you will find different modes for DV, all optimized for small bandwith and noisy HF channels. David Rowe has done an outstandig work for the amateur radio community during the last decade.
    Mode1600 is using 14 carriers DQPSK each at 50 Baud over a FDM modem. This mode is very robust over the satellite with perfect synchronization and low bit error rate. See: http://www.rowetel.com/?page_id=2458

    Mode2020, a new mode using neural network technology for enhancing audio quality is using QPSK and OFDM modem. First tests over the satellite showed, that this setup is not working and a new modem should be developed. As a quick workaround David changed QPSK to DQPSK and now we can use mode2020 over the satellite. See: http://www.rowetel.com/?p=6920

    But with problems: A highly stable equipment on TX and RX using GPSDO is necessary. And there are sometimes problems with first synchronization and loss of synchronization during a transmission.
    I have made some single tone recodings over the satellite to find the problem.

    David Rowe analyzed the files an came to similar results regarding frequency change as you. See: http://www.rowetel.com/?p=6996

    At that time we did think the problem comes from a frequency change of the transponder or the earth stations LOs.

    My question to you mainly shall show the influence of doppler to that problem.

    I hope this discussion will also help you and others to enlight the problems with digital transmissions over the satellite.

    I think we still have to learn a lot.

    73's Gerhard

  • Not intending to kick up a lot of dust, I am asking:

    In some local QSO's there has been, for a number of days, a discussion on QO100 frequency accuracy.


    On my own station, i find that I typically need to tune 70-100 Hz down to be on frequency. I blamed that on my own equipment (even if every single component is GPS-locked), though I wasn't sure where it would come from.

    All that, until I heard in the discussion that others are reporting same.


    People using the beacons for reference are OK and don't report anything.


    We know that the LO in QO100 is spot-on (see the NDA discussion elsewhere).


    Without wanting to blame others: can anyone confirm the correctness of the frequencies of the beacons?

    I know there have been issues in the past and realize that maintenance in Corona days is a challenge, but a brief "yup, we know" helps me to stop searching my own kit.


    There also have been discussions on a small offset in BG7TBL references, but that offset is constant and for this issue, the numbers don't add up, so I don't think it is that.


    Clue appreciated,


    Geert Jan

  • Without irony and with the utmost respect for everyone ...

    but you really don't sleep at night for a 70-100 Hz deviation in a 2.4 / 10 GHz amateur system?

    Do you think interplanetary probes are more accurate than your system? yet they work, they go to other planets and they don't even have GPS. How will they sleep at night?