Posts by PA3FYM

    Yes, the cheaper the GPS module, the better it works ;; -)


    With the normal software, when you have a long TC initially , then it takes ages for a lock.


    What I do is the following:


    1 warm up time = 10s (i.e. short)

    2 initially TC = 16 and wait for lock (usually few minutes)

    3 if lock, check every TC if there is a lock

    4 if lock and DEMA of nanoseconds is below certain value, increase TC with TC>>4 and goto 3

    5 if DEMA is above certain value, decrease TC with TC>>4 and goto 3

    6 if system once locked but lost lock (for whatever reason) halve current TC and go to 3


    DEMA = Double Exponential Moving Average (of the nanoseconds variable, i.e. differnece of the system vs GPS PPS

    George, I filled in your contact form on your website, i.e. check your mail ; -)


    What gain factor did you determine with your OCXO?


    Edit: Made a screenshot of my contraption a while ago, see below. Note the large TC (941 s), an indicator that the stability is good, which is confirmed by a non varying DAC value (in my case the DAC apparently converged to 24029).


    Being active on QO100 since day one (Feb 12th) -with a deviant setup- I managed to work 70 DXCC and (almost) 290 grid squares I reckon I know 'what I'm talking about'. Listening to various QSO's certain patterns appear (to me).


    So it may be an idea to publish some 'QO-100-talk':

    1. "I have a SG-labs transverter with a modified (Octagon) LNB
    2. "I use a single dish with a dual band patch feed"
    3. "CQ QO100, CQ satellite, this is X0ABC calling QO100"

    4. . . . anyone else . . . ? (fill in below ; -)

    I remember almost being banned from Earth here using this approach.

    In fact the first NB hardware posting here was 'transceive' (or simplex), look here (and read the reactions (in another thread).


    (that posting was the incentive to add the rule "Full-Duplex operation is mandatory (you must be able to monitor your own downlink while transmitting!)"


    It's always the same: people who scream the hardest .. you don't hear them ...

    SV1BDS My implementation of the Lars GPSDO is fed with 220V.


    A small 10V transformer -rectified, blah blah- goes into a 7808 (decoupled etc).


    The Arduino is connected to that 8V point, as well as the Ublox Neo 7M fed from 8V with a capacitance multiplier (with resistor divider in the base). The OCXO has a separate capacitance multiplier (with resistor divider in the base and the cap is 1000 μF, so the effect of the cap will be 100000 - 200000 μF or so). With a capacitance multiplier the capacitance of the elco is multiplied by the Hfe (or β) of the transistor and therefore denoising the power supply, which improves stability and/or phase noise.

    That are a lot of screws! My experiment screws the amp above with (if that amp delivers really 30W) by +10 dB.


    You have 30W out with 29 screws, that is 1W/screw.

    I have 20W out with 2 screws, that is 10W/screw.


    So my 'screw factor' = 10log (10/1) = +10 dB and/or
    my screw factor = 10log(1/10) = -10 dB.


    Screw factor ≠ 'screw factor' (note the quotes)


    ;; -)

    SV1BDS Perfect , did you calibrate it with h1 = 0 and h1 = 65535 ? From that you can calculate the gain of the algorithm.


    Your x10 looks nominal , just in the middle of the ADC range ( 0 - 1023 / 2 = 512 * 10 = 5120).


    Admit, that 12 U$ OCXO isn't bad at all, isn't it?

    Welcome into the strange world of QSL, same as the world of snooker with 'abject' glasses ; -)

    Perhaps I will write it down, but it has to do with the ETSI standard concerning what we now call WiFi (then called 'radiolan') and how 'we' obtained +20 dBm output/EIRP (whatever) power (at 2.4 GHz) (draft recommendations were far less output)