Having spent a lot of time working on my alphas over the past couple of months I have come up with quite a few irregularities and want some advice from the gurus about validity and possible cause.

Here is an alpha which Realspeed shows in doppler without a spike but a spike in trackpoint. GT31
First the trackpoint graph, then doppler, then both.




You can clearly see the spike in the trackpoint data both in the graph and table. If you delete invalids the doppler value is even higher which in my mind confirms a problem with the data.

I have mainly been looking at only the doppler graph with trackpoint not displayed to check alphas but I guess this means I should be looking for a match between trackpoint and doppler to ensure validity?

In this case looking only at the doppler graph it looks fine.

A second case and the value doesnt change after deleting invalids but the trackpoint shows a w shape that I seem to remember Daffy saying indicated a problem (searched but couldn't find the post). I see this in quite a few of my alphas and wonder if it is because of where I wear my gps? (on the shoulder strap of my weight jacket).





So in conclusion, for alphas.

Should there be a very close match between trackpoint and doppler to pass alphas? No W shape in either or is a no W in doppler ok?
Should data be inspected before and after deleting invalids to find errors(even in doppler)?

Thanks for any help with this one.

Wasnt this the whole reason Doppler was supposed to be used for Alpha's instead of trackpoints, and the reason Doppler was recommended to be the standard for all Alpha's?

Unless you are using an old Garmin, Track points are a WAFTAM

Use the same analysis method you normally use and that way you're not fudging numbers

Mark,
Run them through KA72 & get another version

Sorry, I have given up tryng to analyse too much as there always seems to be some discrepancies somewhere, the deeper you go the worse it gets

It will do your head in

Hardie, I am using Doppler but in this case I think the result may be suspect.

Elmo, I am using the same method (doppler, realspeed) but I am keen not to give myself an inflated result. No fudging I have posted next lowest.

Paul, ran through ka72 and it gives the same as realspeed
http://www.ka72.com/TrackRecalc.aspx?fid=11298
The track looks very angular through GE but does it use trackpoints?


Anybody got the latest version of other software and prepared to run the alpha through it? Tried a trial version of GPSResults but it has expired and will not give alpha result. It does not show the spiky data point though.





Paul, you are right about making your head hurt. I have never been certain about alphas but if the position of gps is a problem then I will need to change.

Thanks for the comments, guys. Keep em coming.

I think the trick with fast alphas is being able to gybe.

Mark, I had a GT11 behind my shoulder on the vest strap, and a GT31 on my arm, no spikes but a strange result.
Used GPSarPro
GT31 gives 24.82 doppler-trapeze, 24.76 doppler and 25.05 trackpoint.
GT11 gives 25.01 doppler-trapeze, 24.95 doppler and 24.90 trackpoint.

I use doppler-trapeze for everything as it gives the same result as realspeed.
The strange thing with this file, is the GT11 behind my shoulder gives higher doppler than trackpoints. Not sure if that's a position or 11/31 thing.

Mark, I wouldn't be too worried about over analysing your alphas as they are very inaccurate and should only be treated as a fun subject.
IMHO, unless all the alphas are from the same type of meter, mounted in the same type of position on the body and run through the same analysis programme then they are not to be taken seriously, particularly while we have only 1Hz sampling.
At the moment I'm looking at a file in which the sailor did no alphas but which KA72 gave him a 23.5kt alpha for. GPSAR got 9.5kt for the Alpha.
Makes me suspicious of any alpha data coming from that particular source.
When we get 5Hz sampling there will be many sailors having lots of trouble matching their old 1 Hz alphas.

whats with the red thumbs guys? Mike has a good point. If you look at alphas from 2 different GPS units worn at the same time they are very often different by more than a knot regardless of the program used. Part of the problem is that the current crop of GPSs are not very good at resolving speed (including direction) through a turn.

There is no standard method between software versions to evaluate the alphas either, hence the different results with different software.

Alphas arent a very accurate measurement using GT31/GT11s. It is what it is. Unfortunately the software has to figure out when the GPS is sending junk or not, and there isnt enough information stored in the GPS file to accurately do so.

I'm sure most of the time its a reasonable estimate, but if you see differences of more than a knot between software versions or GPS units, then how do you know which one is correct? Is it always the highest figure

Can't find the earliest thread on this topic as it is more than 12 months ago.

When I started writing the code for ka72.com, I was kind of dreading the Alpha part of the process. There are lots of calculations involved. You need to know the distance between two points, which isn't a simple calculation (and there are multiple ways of working it out too) and you need to be able to detect a turn too. Both are not trivial.

Then there is the issue of what actually constitutes an "Alpha".

Many people mistakenly assume that an Alpha 500 must necessarily be 500m long. In fact, my first version of ka72.com applied that rule. Straight away I realised something was wrong, as my Alphas were a lot slower than those calculated by other programs.

In actual fact, there is no stated minimum distance for an Alpha, but there has to be one, otherwise any track of less than 50m length that includes a turn constitutes an Alpha. (That was my next problem.)

The main reason for the large discrepancy mentioned above was that ka72 and GPSARPro use different minimum distances for Alphas.

Consider the Alpha a "fun" division. They are very hard to calculate, and there are so many marginal cases that filtering becomes a game where knocking out one bad gybe also means knocking out one good gybe.

All us GPS developers do our best.

In fact there is no minimum distance for an Alpha. Mal Wright invented the category and set the parameters. There used to be a set of rules/guidelines on his website. Not sure if it is still there. The main controlling factor as far as I can remember is that the run must start and finish inside the 'proximity' circle and that the turn must be more than a certain amount that is roughly 180 degrees (but I don't know what exactly). I will ask Mal next time I see him which could be tomorrow.
Obviously, the way it is implemented in RealSpeed is the 'standard'. I am pretty sure Yann used exactly the same parameters in GPSAR-Pro but I know Manfred tried using Doppler for the 'proximity' and it led to problems so he changed back to trackpoints to determine the proximity location.

Chris is quite correct, out GPS's are currently not ideal for working out the Alphas with a high level of accuracy. It is nowhere near as reliable or accurate for Alphas as it is for normal speed runs. This on the main reason the WGPSSRC has not implemented Records in this category yet.

It was noted when the GT-31 was released that there were inconsistencies between the GT-11 and the GT-31 (there were always inconsistencies between trackpoint and Doppler in the GT- 11 and these continue in the GT-31. The first generations of firmware for the GT-31 were unworkable for Alphas but Tom Chalko worked out some changes with Locosys that got the Doppler Alphas reading really well, as they do now most of the time. I continue to see a lot of spike in the trackpoints in the GT-31 in some circumstances, especially from GPS worn on the arm (as distinct from in the helmet), probably partly due to the sudden movement of the arm during the Gybe.

Bottom lines:
-Alpha data from the GT-11 and GT-31 are pretty close but not always directly comparable when worn side by side.
-Trackpoint data for Alphas from the GT-31 is pretty useless and should be pretty much ignored. It is not usually an indicator that there is any errors in the Doppler data for the same run
-Just use a GT-31 and post the Doppler data.
-Sometimes strange crap happens!

Thanks, Daffy
KISS post doppler alpha data and don't stress!

Scusi for late response...been out in the sticks, incommunicado.... sailing.

Daffy said......."In fact there is no minimum distance for an Alpha. "

On the GPSTC the category is called 'Alpha racing 500m'

I guess that means there is a minimum distance and it is 500metres.

Hopefully, we are all entering alpha data covering 500 metres.

i think alpha 500 means a maximum of 500 meters
usually the longer the distance (calculated) the better chance of a faster result.
i.e i have regulary recorded alpha 1000's above 28kn but have never got close to this
result in the alpha 500 category.


some info here from the creator

http://www.intellimass.com/RealSpeed/Alpha%20Racing.htm

Select to expand quote

Te Hau said...

>>>>

On the GPSTC the category is called 'Alpha racing 500m'

I guess that means there is a minimum distance and it is 500metres.

Hopefully, we are all entering alpha data covering 500 metres.

Yep, snides is correct, it's a max length of 500, and 99% of the time it will be less by several meters.
As at a 1hz sample rate, the distance between samples can be around 15 meters

Select to expand quote

decrepit said...

Te Hau said...

>>>>

On the GPSTC the category is called 'Alpha racing 500m'

I guess that means there is a minimum distance and it is 500metres.

Hopefully, we are all entering alpha data covering 500 metres.

Yep, snides is correct, it's a max length of 500, and 99% of the time it will be less by several meters.
As at a 1hz sample rate, the distance between samples can be around 15 meters

Sometimes my best alpha 500 for a session is calculated well under the total distance i.e. my 50metre proximity only allowed for 300metres of distance covered but still gives the best average speed over that (300m) distance. Of course it is preferable to cover the 500metres within the proximity to give you the best chance of a fast alpha.

Snides, thanks for that link to Mal's original ideas.....all is clear