Monday, November 10, 2008

VOACAP reliability, the end of myth ? part 3

fig1: This plot shows the signal received by VK7KRW in Hobart, Tasmania, on 40m long path. I have also plotted DG0OPK signal (using a G5RV at 5m). In fact this plot alerted me on a problem on my vertical, as I was expecting more difference (the simulation shows -7dBi for the G5RV at 10deg -assuming it is broadside to the signal-, and about 0dBi for my vertical), Checking my antenna I discovered that one set of radials was disconnected. For this test, I was also using an auto-tuner, which is quite lossy for HighZ/50ohms impedance transformation (clearly visible on the antenna bandwidth).
Fig 2: This plot shows the VOACAP "RPWRG" output for the path F6IRF>VK7KRW. As explained in the text, all the parameters have been set to the most optimistic values... remains 40dB offset between the real world and VOACAP !

Are the WSPR reported levels accurate? I checked the reported levels with a "path simulator" (AE4JY's PathSim, a great tool !), using the same soundcard that I used for the above ploted files . Taking into account that the path simulator uses 3kHz BW for the noise and WSPR 2.5 kHz thus 0.8dB for the BW relation, I found the WSPR reported levels to be within 1dB from the Path simulator displayed SNR's. I also used several "Path algorithms" (ie CCIR good, moderate and poor) with no impact on the WSPR reported levels; so the WSPR reported levels should be considered accurate (let say +/-2dB depending on receiver and soundcard used). By the way, the only simulator path algorithms for which WSPR does not provide any decoding are the "CCIR flutter" and "High latitude disturbed"- this explains why it is so difficult in Western EU, to get a report from US-west coast (as the path goes through the aurora oval). Under the same "perturbed" conditions, JT65A works perfectly.

Where the VOACAP error may come from ?

- I think that the first source of error, especialy on low bands, is the RX antenna gain. If you reverse the TX and RX antenna you will see that the calculation provides moreless identical results in terms of SNR (+/-3db due to the noise variation along the day at the receiver location). The hams familiar with low bands DX'ing know that a receiving antenna may have an absolute negative gain, but a better receiving performance. In other words the gain of a receiving antenna is generaly meaningless in terms of SNR.The typical examples are the beverage or the K9AY loop, which have negative gain (or very negative gain for the K9AY), but provide better receiving performances than an omni antenna. A suggestion might be to use W8JI's RDF factor, rather than the absolute gain for the RX antenna
. The problem is to know if the source of noise is in a different or in the same direction as the useful signal... not simple! In any case, based on W8JI table, and unless a long beverage is used, this error can't be more than 3 or 4dB compared to the isotropic antenna. Here none of the receiving stations were using a RX-antenna, therefore using the isotropic as RX-antenna can't be that wrong ! ."A contrario" using a TX-antenna on the RX side may produce exageratly optimistic results... as explained by W8JI, on HF bands, a higher antenna gain does not mean a better SNR, as the software assumes!
fig3: NEC2 simulation of VK7KRW antenna. Using an Isotropic antenna in the VOACAP model is also exageratly optimistic !

- Another potential source of error might be the 3Mhz noise level at the receiver location. Measured here on a base-loaded 18m monopole tuned at the frequency, the minimum noise level I can get (depending on the time, and the meteorogical conditions) is around -80dBm in 3000 Hz bandwidth (S7.5 according to my S-meter calibration table). This corresponds to -115dBm in 1 Hz bandwidth, thus -145dBW (the recommended average level, which is used for example in HAMCAP). I know by experience (having transported my transceiver in several other contest locations), that my location is noisy (I have 380kV power line passing some 200m from the antenna). It means that in most of the cases, for the radioamateur service, it might be more representative to use a lower noise level than the default one (-145dBW/Hz). The minimum that can be applied with some effect on the output is -160dBW (although the contextual help says: range 100-200), which represents a maximum of 15dB gained on the signal to noise ratio, not taking into account that there are also interferences (ie pactor BBS's, SSB, ...) on the bands that we are using, especialy on he 40m band... so the 15dB here should be considered as an absolute maximum on an ideal QRM-free band...Obviously not the case of the 40m band ! On the other hand, the bilateral data with ZL2TLD, shows that the reports offset, between his station and mine is just 4dB at long path and 10dB at short path (here, the band is far more noisy in the evening than it is in the morning), therefore applying -150dBW/Hz to 155dBW noise level at ZL2TLD and VK7KRW location looks appropriate as they both reported having a quiet site...
My location can't be considered as "radiolectricaly quiet" !

- VOACAP does not take into account the K or A index, but just the smoothed monthly SSN. Personnaly, I think that using the SSNe might be more representative for day to day conditions on high bands, as the SSNe which is obtained from the Ionosonders data, takes into account those variations of the K and A index. Anyway VOACAP does not accept negative SSN values and the designers insist that only the smoothed monthly SSN should be used. IMHO, everything depends on the considered path. For transequatorial pathes on 40m band such as the ones considered here, the SSN has almost no influence on the prediction (you can verify by yourself that, on this 40m path, any SSN value provides moreless the same output result). In fact, it seems that the SSN is mainly used to determine the MUF, which on low bands can't be considered as a predominant factor...
- Another critical parameter, is the "required reliability" percentage. As shown in the earlier published plots, using a too low percentage has for effect to predict earlier band opening and later band closure, than the observed one. Anything >50% seems to match quite accurately the path opening. As also shown in those plots, the "SNR" output provides absolutely identical results than the SNRxx when set to 50%. For the fig plot I used 10%... also exageratly optimistic, for a path that proved working every day over the month, with an excptionnal stability !
- The multipath tolerance, also has an impact. For conventionnal analog ham-modes, I think that this parameter should be ignored (setting multipath tolerance to 0 as for effect to disable the impact of this parameter). Of course for digital mode, it becomes more critical, but on the pathes considered here, the probabilility of multipath effect is very low (might be different on higher band, where both short and long path might be open simultaneously - ie EU to JA during EU mornings).

- The minimum angle might also be a critical factor. I have a quite bad take-off to the W-SW sector (long path) with some 8 degrees blocking due to a hill. However even 8degrees does not impact the prediction result, as the predicted radiation angle is some 10 to 12 degrees for the considered path. This is of course different on higher band, where the vertical radiation angle might be as low as 3 degrees, on a certain number of DX pathes.

- Remains the F-PROB parameters... must say that this is above my competency... I can only use the defaults. However those parameters (at least on 2 of them) seems to be linked to the E-layer effect (E and E's), which on this particular path at this time of the day, on this frequency, are unlikely to provide any improvement of the SNR, but more likely a degradation due to higher absorbtion.

The fig2 output was produced using exageratly optimistic parameters...still remains a large 40dB offset on the F>VK7 long path. How can it be explained ?

Fig4 the grey line position at the path peak. None of us are on the grey line...

- Everybody has heard about the grey-line improvement effect... If you look at the path at the observed signal peak, you'll see that none of us is on the grey line at the observed signal level peak...
- Another Hypothesis, is some "exotic" propagation mode that has not been taken into account in the model: for example and as suggested by several authors, some ducting within the ionosphere (something similar to the troposheric VHF/UHF modes, but in the F2 layer), or some other mode (ie interaction between E and F layer, or direct F2-F2 (or F2-F1 mode), avoiding the signal to be returned to earth in the middle of the path), thus reducing the path attenuation. The problem, is that "exotic" does not apply to a propagation mode which occurs every day with a remarquable stability !
- The last one is that VOACAP has never been calibrated for those "long DX" pathes, as the main objective was broadcasting at distances not exceeding 10,000 kms. As shown in my first study, I personnaly still have doubts about the proper calibration of VOACAP, even at shorter distances, but this may be the topic of another study...
Anyway, and as a temporary conclusion, I think that this humble contribution shows that VOACAP predicted levels cannot be considered reliable for long DX pathes. Remains that VOACAP predicts quite accurately the best time for a given path and most probably the proper required min radiation angle, those 2 parameters being in fact the most importants for a DX-man or a contester... Anybody who is familiar with the 40m band knows that it is quite easy to work VK/ZL from EU on 40m at long path in the EU mornings, even using Low Power and a modest antenna. I personnaly had many CW QSO's with VK/ZL just using a R7 vertical and 100W. VOACAP tells you that it is not possible, unless you use "broadcast type" antennas and power... Of course VOACAP is completly wrong on such pathes, but it is the vocation and the beauty of ham radio to investigate unexplored terrains ! Just remember that the radioamateurs were the first one to explore the HF bands. At that time, all specialists were considering those frequency ranges as completly useless, so even if VOACAP tells you that a path is not possible, just try it, VOACAP can be wrong ! Patrick Destrem - F6IRF October 2008
Fig 5 Using what look "reasonnable parameters" for the path and stations setup, the offset can be as large as 73dB...

Acknowledgements: I would like to thanks, all the radioamateurs who have contributed to this article, consciently or not; between others: AE4JY, DG0OPK, G4ILO, K1JT, VE3NEA, VK7KRW, W1BW, ZL2TLD, ...

No comments: