High-Latitude Errors


[This document is part of the help files integrated into the ITS HFBC software package.]

By George Lane, Greg Hand and Lorraine DeBlasio

Computer prediction programs which use the "expected excess system loss" term may contain an error. The programs suspected of having this error are the original IONCAP, ICEPAC, VOACAP and other IONCAP derived programs. The error is contained in the files used by ITU-R PI. Recommendation 533 but at present that prediction program does not call the parameter which is in error. ITSA-1 may contain the error. It does not seem that ITS-78 or HFMUFES has the error since these programs have different values for excess system loss.

Excess System Loss, first introduced in 1966 by Lucas and Haydon [1], or the Distribution of Transmission Loss, as termed in IONCAP [2], is the additional loss which is added after all the other losses are accounted for. It was suspected that this additional loss arises from sporadic-E layer effects and additional losses in the auroral regions. Whatever the source, there appears to be at least 9 dB of additional losses encountered in ionospheric propagation at all latitudes than can be accounted for by conventional theory. Thus, excess system loss is the term that adjusts the theoretical prediction for the least error when compared to measurements.

Interestingly, the ITU-R WP 3L (formerly CCIR WP 6/1) removed 9.0 dB from the excess loss table and then compared the predictions from Rec. 533 to the measurement data in Database D-1. It was found that even with a totally different database than was used in the IONCAP evolution that 7.3 dB of additional loss had to be added to bring the predictions back in line with the measured values.

The binary file that contains the excess system loss tables (array SYS) is ordered: Dl (lower decile), Median, Du (upper decile). The tables in ITSA-1 report (appendix C) [1] and IONCAP theory manual (Table 7) [2] are ordered: Median, Sl (lower standard deviation), and Su (upper standard deviation). Note that S * 1.28 = D. The erroneous value in the binary file is for the equinox on paths greater than 2500 km at geomagnetic latitudes greater than 75 degrees for the local times of 0400 to 0700. The value in the file is 115.2 dB for the upper decile. The correct value should be 11.52 dB which converts to a standard deviation of 9.0 dB (11.52/1.28) as is shown in the tables for these conditions. The error translates to the southern hemisphere, as well. It will affect the months of March, April, May, September, October and November at 04, 05 and 06 hours (local time) at the path control points.

An error of over 100 dB seems rather serious and we wondered why no one had ever complained about it. After making the correction in the table, we reran the entire VOACAP benchmark containing over 100,000 frequency-hours on 27 paths ranging from 500 to 20,000 km. No differences of over 1.0 dB could be found. We then ran a set of paths on a 10 degree azimuth which would transit the very high latitudes. Again no differences were found. Finally on a zero degree azimuth going over the pole we found some differences, which occurred during the equinox months of April and October at 04, 05 and 06 LT hours. On path lengths of 5000, 7000, 8500, 10,000 and 15,000 km, only the 10,000 km path had differences greater than 1.0 dB. The largest difference was 18 dB at an unusable frequency far above the MUF. For frequencies above the FOT and slightly above the MUF, the differences were less than 1.0 dB. For frequencies below the FOT, the differences ranged from 7 to 15 dB. In all cases, the corrected table provided lower required power gain values (i.e. less loss) than did the original table with the error. Median values of signal power and signal-to-noise ratio were not affected. Only variables Sig LW, SNR LW, REL and RPWRG (required power gain) showed differences. All of these variables have a dependency on the upper decile of the excess system loss.

It seemed rather strange that such a big error did not result in numerous errors of greater magnitude. In IONCAP, the error in the table is restricted to a very small area of the ionosphere so the chances of having a control point in that area are small. Also the decimal error in the SYS file causes the lower decile of the signal power to be much too low. IONCAP tests for this decile value in subroutine REGMOD and if it is more than 25 dB below the median, it is truncated to 25 dB. The maximum error introduced by the table error is (25 - 11.52) or 13.48 dB. At frequencies above the FOT, the lower decile of the signal power was 25 dB using the corrected table. Since this is the same as the truncation value, the erroneous table entry made no dif-ference. Significant errors of 7 to 15 dB were found at frequencies below the FOT. Thus IONCAP may indicate overly-pessimistic values of circuit reliability during the dawn hours on polar paths of more than 2500 km. A symptom of the error is that SIG LW is 25 dB at all frequencies.

VOACAP users should download VOACAP 94.1114 or any subsequent version. Note that the version number is: YY.MMDD. If you use ICEPAC, contact:

Mr. Greg Hand
Institute for Telecommunication Sciences Mail Stop S1
325 Broadway
Boulder, CO 80303 or Email: {gh@its.bldrdoc.gov}.


[1]	[Lucas, D.L. and G. W. Haydon, 1966] ITSA-1
[2]	[Lloyd, J. L., et al., 1978] IONCAP Theory Manual [draft]

Acknowledgement: The authors wish to acknowledge Mr. David Sailors at NRAD (San Diego) who found the table error and brought it to our attention.