Probable error in sgp4init function

[rob-groundspace]

We have been using this library even though it doesn't seem to be actively maintained. I decided to test the propagation against a set of TLEs and results obtained from running SGP4prop. I found that 99.8% of tests were passing and the rest failing (283 out of 128313).

After a few hours of comparing the code with other implementations I spotted the problem. In the official C code (https://github.com/nasa/ECI/blob/2ccd1ba6f3ec2168816098c4912f449af74a0fc1/examples/sgp4Prop/fsw/src/SGP4.c#L1309) we see the following lines

if (perige < 156.0)
            {
                sfour = perige - 78.0;
                if (perige < 98.0)
                    sfour = 20.0;
                // sgp4fix use multiply for speed instead of pow
                qzms24temp = (120.0 - sfour) / satrec->radiusearthkm;
                qzms24 = qzms24temp * qzms24temp * qzms24temp * qzms24temp;
                sfour = sfour / satrec->radiusearthkm + 1.0;
            }

while in this library we have

		if perige < 156.0 {
			sfour = perige - 78.0
			if perige < 98.0 {
				sfour = 20.0
			}

			qzms24temp := (128.0 - sfour) / radiusearthkm
			qzms24 = qzms24temp * qzms24temp * qzms24temp * qzms24temp
			sfour = sfour/radiusearthkm + 1.0
		}

I think 128.0 should be 120.0

It seems this library is no longer maintained. There are some forks which seem to be used so posting this might help people who see errors.

[Psykar]

I'd suggest making a pull request as well for this. I see that @joshuaferrara is active on github and has commits from this month so he may get to it :)
And if you have the code that did the tests you mention, having those as unit tests would also be great?

[rob-groundspace]

I would love to contribute, but not sure it's worth wasting time on it. Happy to share the bug to help others. There are a few forks of this library which look promising, but they all have the same error and they have issues disabled so cannot report directly.

Here is one test that would fail (satellite_test_suite.go)

			{
				line1: "1  7004U 73107B   21204.77194991  .11004814  20894-5  95609-4 0  9995",
				line2: "2  7004  73.8846 307.9676 0013503 255.1531 104.8238 16.51259389470580",
				grav:  "wgs72",
				testData: `0.00000000 4010.7618166170605 -5139.529125966895 -0.0002716261448097565 1.711037560218839 1.3332480035056196 7.513463567006438
360.00000000 3710.4572584958582 -2657.148731901094 4632.682588653529 -2.3632641481089007 5.499349058479544 5.039998172803235
720.00000000 701.1391025769603 2022.2058112236543 6121.854701271255 -4.9206723993317105 5.932095437578289 -1.3974559827223034
1080.00000000 -3009.3581236352206 5257.640150453373 2275.5728050804305 -3.329603584080322 1.1337063394028264 -7.018594629222161
1440.00000000 -3771.114779073523 3266.214794434546 -4098.283189528379 1.6716158085910164 -5.181033626695302 -5.664457186515631`,
			},

It is extracted from the test data here (large file) https://github.com/shashwatak/satellite-js/blob/1d52cbae62e20a6388e0ea3a01c5cc9590609050/test/propagation/__snapshots__/sgp4Catalog.test.js.snap

These tests were generated from sgp4prop, but the snapshot format is not ideal and it took some processing to convert it into a large json file which I have uploaded here (https://github.com/rob-groundspace/sgp4-test-data). Ideally the sgp4prop C program should be used to generate the tests.