An interesting situation that sounds slightly familiar. 
I think I would trust the survey grade GPS on the boat. 
)
"VMAX says 39.46, 10s says 30.89, I'll just split the difference and call it a day. I was really feeling fast !" -- heard every time on every spot before consumer receivers included dynamic models.
Select to expand quoteCarindale said..
The real question is: were they using an approved device?
Yes..... and no. .
$50,000 Survey grade GPS V's $300 Velocitek 
It was just a spike!
There was also the good old "display showed 34.56 but GPSAR says 32.99, I just feel like 34.56 is more accurate". Good times. Someone just sent me a log from an old device and it's plain insane what we put up with. You can see the receiver is completely lost, under water, and yet it mentions 8 satellites and an excellent HDOP as if things were fine.
Select to expand quoteJulienLe said..
There was also the good old "display showed 34.56 but GPSAR says 32.99, I just feel like 34.56 is more accurate". Good times. Someone just sent me a log from an old device and it's plain insane what we put up with. You can see the receiver is completely lost, under water, and yet it mentions 8 satellites and an excellent HDOP as if things were fine.
I'd be careful about bashing the older hardware - the u-blox is better, but by no means perfect. With Locosys units, the SDoP values typically go way up for "swimming spikes". In contrast, the sAcc on the u-blox units seems to be filtered, and goes up much slower. I have seen Motion data where the unit keeps recording a constant speed with a reasonable sAcc (below 2) despite having 0 satellites. That's quite nonsensical.
That's probably due to "Sailquik's filter" which throws away satellites under a certain quality.
When things are all-well, the receiver would include many satellites in the solution matrix. This would drain the receiver resources. So I remove the lowest quality signal satellites. Valid and okayish : 28, valid and of great quality : 20, used : 20.
And due to this, when all signals drop below this certain quality point, solution deliberately goes invalid. Valid and okayish : 8, valid and of great quality : 0, used : 0.
This is neat, it makes issues, falls, marginals clear as day.
--
I'm not bashing old receivers but comparisons do raise some interesting stuff.
Full race replays now on Australian Facebook:
www.facebook.com/SailGP/videos/2351431431616105/
www.facebook.com/SailGP/videos/656525104797792/
