Originally Posted by
Hyperveloce
With a scaling factor of 0.25 (instead of 0.15-0.2), the 1st plot of the 2nd page of my doc would basically be translated higher (the 120 min peak would reach approx 290 Hz instead of 250 Hz): it would still be possible to reproduce the 120 min turn but no longer the initial conditions (already out of the enveloppe) and even around 180 min (the min enveloppe would be at 125 Hz slightly below the BFO).
Consider this.
From the radar, I have the location of the turn roughly at 18:30 at 6.7N 95.5E, which gives me initial satellite elevation 53.2°. Final satellite elevation is 40°. We can calculate line of sight distances for these angles. I get the difference of 900 km (you can double check). To travel 900 km in 5.67 hours, we need mean line of sight speed 159 km/h and mean Doppler shift 235 Hz @ 1.6 GHz (241 Hz @ 1.64 GHz).
What we can do now is take BFO data points for 19:45..00:11, extrapolate them back to 18:30, and see what mean observed BFO we get, and what scaling and bias get us there. I get BFO 160.5 Hz. If I assume that the frequency bias is given by the BFO at 16:30 (~87 Hz), mean residual Doppler is 73.5 and scaling is 73.5/235 ~ 0.315.
In your case, you seem to be getting scaling below 0.20 because you're trying to fit scaling and bias at the same time, and you can only get low scaling together with high bias. Which is why you're getting bad fits to the first 30 minutes.