VARA: required frequency stability vs. ground LO stability (long post!)

  • I have heard several reports about OMs struggling to reach the advertised speeds for the higher VARA speed levels. As people were suggesting the stability of the transponder LO frequency could be a problem I decided to investigate the required stability parameters a bit more.

    According to the specification by the developer (Jose, EA5HVK) VARA is based on an OFDM with modulations between BPSK and up to 32QAM for the higher speed levels. See also his blog post for more details. He also states in his VARA specification document that short term frequency stability for both TX and RX should be <0.5 Hz/sec.

    Many OMs have their ground setup fully locked to a 10 MHz GPSDO. While this seems perfect at first sight, I would like to remind everybody about the difference between stability and accuracy. Ulrich Bangert (DF6JB, SK) has written an excellent article about it (german only). A very popular GPSDO used in the QO100 community is the Leo Bodnar GPSDO. Several measurements suggest, that the stability at 1 second intervals is of the order of 1E-10:…eobodnar%20GPSDO_rev2.pdf…Frequency%20Reference.pdf

    1E-10 (@ 1s) is quite a respectable number for such a low cost device, but is it enough for the VARA requirements on QO100?

    At 1E-10 any carrier generated at 2400 MHz will have an instability of 0.24 Hz in 1 sec. The downlink is on 10.5 GHz and assuming a perfectly stable transponder LO, this results in a downlink drift of 0.24 Hz here as well. However, the LO for the RX chain is also locked to the reference and should therefore drift by 1.05 Hz. The difference of the two (1.05-0.24=0.81 Hz) would be visible on the downlink as drift in a 1 sec time interval. As this is larger than the required 0.5 Hz a degradation in performance is to be expected.

    Disclaimer: I cannot and will not give any technical details of the QO100 transponder due to NDA. But I think it is obvious that any space-qualified transponder oscillator will have very good stability values so for the sake of simplicity we assume it is non-significant compared to the numbers above.

    The solution would be now to use a more stable (not necessarily accurate!) reference oscillator. James Miller G3RUH has an example with (now obsolete) hardware. In a few words good stability on short timescales requires (among other things) good thermal stability of the 10 MHz master oscillator. On short terms the GPS control servo is (should be) non-effective and a good free-running OCXO might be even better (again: stability vs. accuracy). This probably means something like a HP10811 (which many still regard as one of the best references available for hamradio) or even one of the double oven HP10811 versions used in the Z3801. They reach 1E-11 and below. But they also need to be kept running 24/7 as letting them reach thermal equilibrium may take a few hours..

    73s Achim, DH2VA

  • Thanks for the detailed explanation.

    You subtract the transmitter instability from the receiver instability. I agree that would be true if I were to listen to my own signals. But, if I were to listen to your signal, would the total drift not be the addition of the drift of your transmission equipment added to the drift of my receiver equipment? .

  • pe1hzg fully true. The example given above is a best-case scenario as TX and RX drift are closely correlated. If I listen to somebody else's signal the two drifts are uncorrelated and the worst case can happen (according to Murphy's law, it will..) with both drift having opposite signs and the result would be the sum of the two. So something of the order of 1.2 Hz/sec.

    Again, this is for a 1E-10 oscillator at constant ambient temperature. If you open the window in the shack to let in some fresh air, hell will break loose. I have about half a dozen GPSDOs of different grades sitting here in my lab and you can clearly notice it.. Ulrich Bangert has described in his article linked above that he has placed his OCXO in a aluminum vault with 20mm (!) wall thickness. This is quite some kilograms of mass and therefore acts as a thermal lowpass. Do not try to pack the OCXO in an insulating box (styrofoam) as the oven will need a certain heat dissipation to actually work. The aluminium vault is actually great as the thermal conductivity is low but the thermal mass is high.. my best double oven HP10811 is placed in such a vault and then on top equipped with its own UPS to keep it running even in case of a power failure. Uptime 8 years..

  • OK I have to ask a question here regarding OXCO's if you are more knowledgable, I use a 10Mhz Oscilloquartz type 8661 which is a double oven style unit, it seems to be reasonable for my up converter. The board it is mounted on would have been internally installed in test gear and I have wrapped it in 5 or 6mm styrofoam not a lot of foam and I can feel heat though the foam. It dropped the power requirement by 0.5 watt. So is this a bad idea?


  • Related.. I wonder why the transponders in these sats (not ours, the regular transponders) need to have such a high stability?

    After all, regular TV transponders are 8MHz "bent pipe" frequency translators.

    For analog CATV on modern, I know that all the carriers of all the analog TV signals were phase-locked, the reason being that the intermodulation product of the carriers of these AM signals would then be a DC signal, and hence not visible, so CATV operators were able to push more channels on the bandwidth.

    I'm not sure the same argument applies for sat-TV: on CATV, the signals all have the same strength and there is quite some amplification / processing in the network. On sat-TV, signals are spread by direction and polarization so I'd expect intermodulation to be less of an issue.

    Yet, sat operators use these ultra-stable frequency references. What am I missing?

  • G8UGD it's hard to give a clear answer but if you think about how the oven internally works, it might give a better understanding what can be done from outside to help the temperature controller.

    Theoretically to keep the crystal oscillator at any given temperature, the controller would heat if the target is too cold and cool if the target is too hot. Heating is easy but for cooling the control loop requires that the energy has to go somewhere (no active cooling) so we need some way out.

    The prime reason for anybody to isolate his OCXO is to separate the temperature variation from the outside from the device. That unfortunately means that heat cannot go out anymore as well thus changing the control loop time constants for both heating and cooling. It probably will still work (operating temp according to datasheet up to 70 degC) but the sweet spot is 25degC.

    The reason for DF6JB's proposal to use thick aluminum walls is to keep the thermal conductivity to the ambient (aluminium being a good heat conductor) but averaging out any short term effects like opening the window or even the door. The total energy change will be the same with or without thick walls but the slope will be softened by the thermal capacity of the aluminum. Ideally, the slope will be slow enough so the OCXO temperature controller can fully compensate for it without the user noticing. Without this thermal capacity the temp controller might be just too slow and not be able to catch up resulting in a larger delta-T at the crystal and hence a delta-F.

    Does this answer your question?

  • Hi, I use Lars GPSDO and I have similar problems with lose lock when aircondition was opened or closed as I have described in my topic. It was solved when the algorithm being more agresive and now does not loose lock. Leos GPSDO loose lock in similar conditions? When I compare signals in X Y scope it was not drift.

  • I’m working with one of these Trimble GPSOs in horizontal orientation. After the thesis of DH2VA I must do something for it. I don’t know if it is better to be used vertically with some legs or I must put heavy aluminum fins. Your opinion please.

  • DH2VA Yes I understand it needs to cool.

    I guess it's a compromise between being in a piece of test gear with other electronics around it and being open on the desk top, so I covered it.

    I do wonder how to check stability of any osc for both accuracy/drift over a period of time, is there an ultra-stable signal on HF that one can lock and compare to or any software designed for the measurement one can get hold of easily?

    It does for now but your post on stability got me thinking.

  • Hi

    I'm using signal from the CW beacon as a mean for testing short term stability of my reference frequencies. Looking at the 5 sec dashes on a waterfall, I see straight lines (vertical on my SDR setup). But when the frequency scale is expanded to something like 100 Hz per cm - or more - the line is not always straight and with sharp edges. Sometimes - depending on stability of the reference - I see either wobble or fuzzy edges. This is not a strict measurement of stability, but it gives an visual indication, especially when comparing various references. (note that setting of the FFT BW must not be too narrow, as this will integrate/mask short term variations)

    Ofcourse the CW beacon itself has to be stable and clean - and it is - otherwise you don't know what is measured. I don't know just how stable the beacon is, but when I compare to other carriers (test transmissions) on the satellite, I almost never see signals as stable.

    Using this method I have tested OCXO's from KVG, Telequartz, TDK but also cheap TCXO's from the internet and a Rubidium standard. Clearly the best was an old TDK OCXO, followed by the Rubidium with OCXO's from KVG and Telequartz marginally worse. The cheap TCXO was very bad, but even this gave perfectly sounding SSB reception. Stability is only critical for certain modes!

    Ufortunately I have no Leo Bodnar to compare with.

    73 Ole

  • I do wonder how to check stability of any osc for both accuracy/drift over a period of time, is there an ultra-stable signal on HF that one can lock and compare to or any software designed for the measurement one can get gold of easily?

    No. We receive here DCF77 from Germany and due to variations in the propagation, its stability is only of the order of 1E-9 on short timescales.