MY QO100 SETUP

  • SV1BDS Promising video ! : -)

    It is now interesting to which your TC converges. I placed a 10 uF cap between the RESET-pin and GND to prevent rebooting the Nano when connecting a RS-232 terminal (e.g. Putty) (cap has to be removed when reprogramming the Nano)


    Here the TC 'wobbles' a little due to the statistics I apply on diff_ns as an attempt to 'restrain' the contraption, so that it can't drift away, read: prevent loosing lock

  • PA3FYM Yes I put also a capacitor to the reset pin with a jumper and prevents reset when usb is plugged. TNX. I have to work now on clearing the output signal from harmonics etc. The OCXO alone is clean but when others units powered is become dirty. Now it reaches TC about 250 but I am not sure that it can go more. Yes I noticed that you leave a lot of margin from the TC that can be reached compered to the operating one. What I have to see it loose lock if it is better to begin from 16 or slow down the TC as you do now.

  • I have to work now on clearing the output signal from harmonics etc. The OCXO alone is clean but when others units powered is become dirty.

    Yep, known issue ; -) That's why I always emphasize denoising power supplies when using this kind of applications (accuracy & stability), but apparently you have to experience it to place my comments into perspective ; -)


    In the end noise (reduction) influences the time constant (TC). I regularly see a TC > 500, so that must be possible for you too.


    Some other tips/hints I did:
    I placed a temp sensor (LM35) on (I think ADC2). When you don't use temp compensation, ground that ADC pin , otherwise it is left open and produces noise.

    It is not by coincidence that open ADC inputs are used to produce cryptographic keys ; -) Same accounts for the other T-sense ADC pin.


    I placed a 100 uF cap between Vref and GND on the Nano board (on the board is 100n, but I increased that ; -)

  • PA3FYM Using your software:

    time ns dac temp status diff_ns filtX10 tc filt timer1 temp1 dema


    31900 -11 33348 0.0 Locked -10 5067 821 410 25060 736 -6 96 5177 33349 0.0

    31901 -8 33348 0.0 Locked 3 5067 821 410 25060 744 -8 97 5245 33349 0.0

    31902 -4 33348 0.0 Locked 4 5066 821 410 25060 736 -8 98 5190 33349 0.0

    31903 -4 33348 0.0 Locked 0 5066 821 410 25060 748 -8 99 5113 33348 0.0

    31904 -4 33348 0.0 Locked 0 5066 821 410 25060 743 -8 100 5167 33348 0.0

    31905 -1 33348 0.0 Locked 3 5066 821 410 25060 752 -8 101 5117 33348 0.0

    31906 -4 33348 0.0 Locked -3 5065 821 410 25060 743 -6 102 5126 33348 0.0

    31907 -8 33348 0.0 Locked -4 5065 821 410 25060 736 -6 103 5143 33348 0.0

    31908 -1 33348 0.0 Locked 7 5065 821 410 25060 752 -6 104 5093 33348 0.0

    31909 -4 33348 0.0 Locked -3 5065 821 410 25060 737 -4 105 5073 33348 0.0

    When I open the aircondition it looses clock. I will send you the captured data to your email.

  • SV1BDS I saw the data, interesting ... (first read the mail, didn't understand, until you now tell me you switched on the air conditioning, haha : -)


    What happens is that the DEMA, the statistical (or some say economical parameter, especially when it's used in stock trading algorithms) parameter to control TC is floating around zero (0) when TC reaches a certain value, in your case 770 seconds.


    Because the DEMA didn't exceed max deviation (arbitrary value, chosen by me as intuitive/educated guess) -at the moment when the decision has to be made to increase the TC- the contraption gets rewarded with a longer TC, in your case:


    Code
     TC += TC>>4 = TC + (TC>>4) = 770 + 48 = 818

    (B.t.w., 770- 818 seconds is a good value, which means 'the flywheel' is running stable)


    'Normally', when DEMA drifts away (at least that is MY philosophy/addition/interpretation) in such a matter that it's risking 'flying out of the bend' this is a sign that the contraption cannot remain its lock with the current Time Constant. So, before it looses lock, the TC is lowered until it reaches a moment that the system remains in lock and that the DEMA is in control and you get 'through the bend' with your car ; -)


    However, when a 'severe' discontinuity during this process (i.e. while almost flying out of the bend with your car) is involved and you try to press the brakes, but somebody cut the brake hoses (in this case switched on the airco ; -), a crash (or loss of lock) will be inevitable. That is what happened here.


    Hmm.. let me think about it (was some time ago I designed this addition), nice mental puzzle ...


    Of course, the external (exterior of the OCXO etc) are of importance. When your contraption is not inside a case, the impact of the airconditioner will be larger than when it's inside a case.


    A quick and dirty remedy/experiment could be adding Temp-compensation (NTC or LM35 at (iirc) ADC2) and see how it behaves. Anyway, my recommendation is to box your contraption first.

    • Official Post

    Just a remark on the Tc: having the GPSDO optimizing its Tc is a great thing but the conclusion that higher Tc automatically means 'better OCXO' is only partly true.. it can also mean 'shitty GPS receiver' as the control loop needs to average longer to benefit from the GPS information.

    When developing such a system myself some time ago, I could easily go down 10x with Tc (while keeping performance) once I included the sawtooth correction from a proper timing receiver. See also www.gpstime.com from Rick Hambly W2GPS (former AMSAT-NA president).

  • DH2VA Agreed, but I am 'in the flow' or focusing on this parameter only for practical purposes. Fast lock, stability & accuracy (and cheap! : -)


    Concerning Saw Tooth Correction, had lots of experience with that while running Stratum 1 ntp-servers, but to go that far ... We only need 1E-10 precision/stability ; -)


    Variance / jitter on the PPS (which is an intrinsic issue with GPS ; -) is attempted to deal with by Lars, have to dig into that issue too. Now there is a low pass filter trying to smear out/average this jitter. The system is not punished hard when it misses a PPS.

  • Hi, the design phase of Lars GPSDO is completed. A *3 xtal filter is added to eliminate the harmonics. A 470 ohm resistor between the OCXO and the xtal filter increase the isolation of load, adapts the Z of the filter and give the proper level of drive to the MAV11, which gives full output.The first QSO with GPSDO is done. The result is increased SNR compared with initial xtal of LNB. No important jumps occurred when it loose lock for a while. A reset button it will be added to start from the beginning if it can not lock quickly. Also another HC390 was added and now can provide 1,10,100 KHz,1,5MHz at TTL levels for PLL reference. A 8 way splitter for 10 MHz is now in the design phase.

  • SV1BDS I saw the logs. Again interesting . . . The 'more frequent sampling' trick works to some extend but apparently your airconditioner does 'something' to which the system can't react fast enough. I also see from your time constant (TC) values that your system is (intrinsically) stable. I never had such large TC's with my setup. What is also interesting, is that your DAC value remains ultra stable, also when TC is lowered.

    Perhaps you can experiment with the temperature compensation factor (parameter 'c' in the terminal), it's zero (0) by default. Perhaps try to give it a value ? Point is, that with long TC times, it takes a long time before you notice something, so experimening with temp coeff values will be tedious . . .


    I am now in the process rewriting some of the c0de because the TIC software will be enriched with an NMEA parser to display locator and QO-100 AZ and EL angles on a LCD display. Just I leave it here for now and come back later on the 'airconditioner' issue ; -)

  • PA3FYM Yes I have the same opinion : the software reacts but not as fast enough as required. I have noticed the high TC that archived and it seems that it is very stable. Even if it is a breidboard construction it is stable. I have noticed the 1024 TC and I can not believe it ! will try to see if the c parameter have any result. A lot of time and a lot of aircondition on off HiHI. That what I will propose to you to put an LCD in the code but you write it allready ...TNX again !!!