www.ka72.com/

The site now returns all results used for the GPS Team Challenge (doppler only, and still no support for anything other than SBN files.)

I'm still refining some of the results. I'm very happy with them generally, but still getting the odd unusual figure for Alpha, and the total distance is a little bloated.

Please give it a go and get back to me if you spot significant oddities.

Since the first beta went live last weekend, there have been more than 40 SBN files (plus a smattering of SBP and GPX files) submitted.

Every one of the SBN files that was uploaded was successfully read, and results returned. I haven't finished cross-checking the results for all of them, but on the whole I'm very happy with them so far.

Dylan.

Here are my results for the 2nd November.
The numbers I posted on the GTC site are in bold after the KA72 analysis:

KA72 reader:
2 Second Peak (kts): 28.48 (28.482)
5x10 Average (kts): 27.31 (26.942)
1 Hr (kts): 15.78 (15.772)
Alpha 500: 17.34 (17.656)
Nautical Mile (kts): 21.39 (21.386)
Total Dist (dop): 38.01 (38.002)

Looks very promising!

The difference in Alpha 500 is due to your using trackpoints for the Alpha. My site insists on using doppler. That's probably going to be a little bit controversial for existing GT-11 users, but since there probably won't be any more of them, I decided to be a bit more conservative and go with the measure that is better supported in the GT-31.

When I open your file in RealSpeed, I get 27.31 knots for the 5x10 too. So the site result is consistant with Realspeed, even though what you entered was a bit lower.

I'd call that one a complete success.

I should also mention that I tightened up the 1hr algorithm last night too. One thing that Realspeed does is to ignore speed for trackpoints that are over 200 seconds in duration. This means that time spent sitting around on a sandbank or swapping out gear won't be recorded as "700 seconds of travelling at 1.38 knots" and instead will be recorded as "700 seconds of standing still."

That's a good tip for you guys, if you are going to hang about and have a chat, keep it under 3 minutes and you will improve your 1hr speed!

I also tightened up the time based calculations (1hour, 2sec, 10sec) to use the creation date of the first and last records as the endpoints, whereas previously I was calculating cumulative time in seconds for each result.

So whereas previously my test data was within 0.3kts of the Realspeed results, now it is within 0.1kts.

That 200 seconds thing is the "time threshold"... It shouldn't come into play for the 1hr.

Select to expand quote

mathew said...

That 200 seconds thing is the "time threshold"... It shouldn't come into play for the 1hr.

But it does. I'm not passing judgement on whether or not it is valid, only trying to reproduce the calculation methods of Realspeed for these specific measures.

Here is how the calcs work in Speed Reader B4
General
No filters are applied to any of the measures apart from the 200 seconds alluded to above.
There is no concept of one-per-run filtering. If your two top 10sec speeds occur on the same run, they will be included in the 5x10 generated at the site.
However, having said that, overlapping sections are not allowed except where the first point in one speed is the last point in another.
All speeds are averaged so that n seconds is calculated over n+1 seconds. This is the default MO of Realspeed, and optional in other systems.
All speeds are based on doppler.
2 Sec
Taken over three points (usually) so there is a total elapsed time of at least two seconds.
5x10
All 10sec speeds (based on 11 points, usually) are put in an ordered list, then the list is deduped (any overlapping speed removed as per rule above.)
The top five remaining track segments are averaged.
No filtering for spikes or bad data.
1 hr
As per 2 sec above, except with 3600 seconds instead. Speed average removes any points over 200 seconds and assumes 0 speed for those periods.
nm
Works essentially the same as the time measures, except over a minimum distance of 1852 metres. The distance is calculated by multiplying the average speed for a point by the elapsed time of the point, and summing the results.
Alpha 500
Maximum segment length of 500m. Minimum segment length of 100m. Minimum course change of 90 degrees. Maximum distance between start and end point of 50m calculated using the Haveline formula. Doppler (like everything else.)
Total Distance
Calculated by sum of each point's (average speed x elapsed time)

Comes up good for yesterdays file. I updated GT31 to latest yesterday V1.2(B1218T) but still shows may not be optimum.
Surely this will make Nebbs life heaps easier.
Well done, Dylan.

Just an ethical question. I know this will make many peoples lives easier but how about the revenue stream for the likes of Mal to keep improving the software? Surely this will mean that newcomers will no longer contribute to what is already really a labour of love. I know Dylan is doing his own calcs but isnt the Alpha idea Mal's baby?

Not judging just putting up the point for discussion and concerned that those that have done a huge amount don't get cut out.

Select to expand quote

lao shi said...

Comes up good for yesterdays file. I updated GT31 to latest yesterday V1.2(B1218T) but still shows may not be optimum.

Well spotted. I've actually got an older firmware as my "optimum". I will update it tonight to recognise the newer firmware.

Select to expand quote

lao shi said...
Just an ethical question. I know this will make many peoples lives easier but how about the revenue stream for the likes of Mal to keep improving the software? Surely this will mean that newcomers will no longer contribute to what is already really a labour of love.

The site is designed to produce only a limited subset of the rich functionality that applications such as RealSpeed provides. I know that I would never be satisfied with an online calculator, and frankly a big part of my buying a GPS (two GPSs now!) for Kellie was so that I could tinker with the results.

On the other hand, I know there are a number of people out there that would join in the competitions if only it was a bit less of a learning curve to get started with the devices.

I hope that by providing basic info, it will then be possible to direct people to the other solutions for when they want to go further with their data.

I have a number of thoughts on how to achieve this, and the site will evolve a bit as I get to working these things out.

My next step will be to start storing the results and allowing others to verify and "correct" them. To do this, the site will allow others to download and process your .sbn file for a period of about two weeks after it is uploaded.

I'll also be providing a registered user login, which I hope will eventually allow me to automagically post your results to the GPS team challenge (including corrections.)

(I might add that all this uploading and downloading sbn files makes the site much more bandwith-intensive than the GPS Team Challenge site, so I will probably be adding a PayPal-type "donation" button to help cover the hosting costs once I start going over my limits. However I don't anticipate charging an outright fee for the service at the moment, and hope I never have to!)

Select to expand quote

Dylan72 said...

The difference in Alpha 500 is due to your using trackpoints for the Alpha. My site insists on using doppler. That's probably going to be a little bit controversial for existing GT-11 users, but since there probably won't be any more of them, I decided to be a bit more conservative and go with the measure that is better supported in the GT-31.

When I open your file in RealSpeed, I get 27.31 knots for the 5x10 too. So the site result is consistant with Realspeed, even though what you entered was a bit lower.

I'd call that one a complete success.

I think we still need to use one result per run for the 5x10 number. I can't see it being all that hard to implement.

Also, we definitely need to be able to use trackpoints -- most of my problems are to do with garmins, and they don't record a doppler figure at all!

Select to expand quote

nebbian said...
I think we still need to use one result per run for the 5x10 number. I can't see it being all that hard to implement.

I see what you are getting at. I wanted to do something as idiot-proof as possible, and that means having no options, if possible, on the calculations. So I would want to know that a one-per-run rule would have to apply to everyone or no-one, and which divisions to apply it to.

Also, it's probably not that simple to implement. It means creating some run detection logic that is capable of coping with wind shifts and gear changes.

I don't know how the guys do it in Realspeed and similar.

Select to expand quote

nebbian said...
Also, we definitely need to be able to use trackpoints -- most of my problems are to do with garmins, and they don't record a doppler figure at all!

No Garmin support at present, though the core engine just works with points, and doesn't care if they are trackpoints or doppler. I've tried hard to set it up so that it can cope with many different devices as required.

Select to expand quote

Dylan72 said...
I see what you are getting at. I wanted to do something as idiot-proof as possible, and that means having no options, if possible, on the calculations. So I would want to know that a one-per-run rule would have to apply to everyone or no-one, and which divisions to apply it to.

This will only apply to the divisions that use more than one run... which is only the 5x10 value

I agree with making it have no configuration options -- you upload your file, you get your numbers, end of story. No mucking around with various options to (intentionally or not) inflate your figures.

Select to expand quote

Also, it's probably not that simple to implement. It means creating some run detection logic that is capable of coping with wind shifts and gear changes.

/****** START PSEUDOCODE ****/
$highestHeading=0;
$lowestHeading=0;
$currentRun = 0;

foreach($tenSecondValue as $currentValue)
{
if( ($currentValue["Heading"] > (lowestHeading + 90))
|| ($currentValue["Heading"] < (highestHeading - 90)) )
{
// We're on a new heading, so increment run counter and reset headings
$currentRun ++;
$highestHeading = $currentValue["Heading"];
$lowestHeading = $currentValue["Heading"];
}

if($currentValue["Heading"] > $highestHeading)
{
$highestHeading = $currentValue["Heading"];
}

if($currentValue["Heading"] < $lowestHeading)
{
$lowestHeading = $currentValue["Heading"];
}

saveCurrentTenSecondRun($currentValue, $currentRun);

}

// Now sort by currentValue, grouping by run number.

/****** END PSEUDOCODE ****/

This will probably take some tweaking, and you will have to take care of the case when the heading in degrees wraps around the 359 / 0 degree mark, but you get the idea.

Select to expand quote

No Garmin support at present, though the core engine just works with points, and doesn't care if they are trackpoints or doppler. I've tried hard to set it up so that it can cope with many different devices as required.

Sounds very promising.

I'm impressed with how accurate it appears so far, and I'm sure we can work out the kinks

Dylan, just uploaded my tracks, was close but some numbers not same.
You can view the difference by checking out my numbers on GPS team site data under Pinnas and what the program gave me.
The file was uploaded raw from card file, no invalids deleted, no slow points deleted. I delete all slow points under 3 knots.

Mike

Select to expand quote

mineral1 said...

Dylan, just uploaded my tracks, was close but some numbers not same.
You can view the difference by checking out my numbers on GPS team site data under Pinnas and what the program gave me.

Cool.

I had a look at your file, your submission, and what RealSpeed (raw, with default settings for the comp) told me about your file.
2 Second Peak
You said: 23.95
ka72 said: 23.27
Realspeed said: 23.27
Analysis: I don't know. But since I agree with RS on this one...
5x10
You said: 21.21
ka72 said: 21.53
Realspeed said: 21.21
Anaylsis: This is because ka72 doesn't have a one-per-run rule yet. Keep watching this space...
1hr
You said: 9.37
ka72 said: 9.26
Realspeed said: 9.43
Analysis: Looks like we both lose!
Alpha 500
You said: 12.94
ka72 said: 12.8
Realspeed said: 12.95 for GT-11 and 12.8 for GT-31
Analysis: I use doppler (GT-31) measure for all Alphas.
Nautical Mile
You said: 19.05
ka72 said: 19.06
Realspeed said: 19.06
Analysis: On the nose, if you ask me.
Distance Travelled
You said: 26.17
ka72 said: 25.87
Realspeed said: 23.25 or 25.87 depending where you look.
Analysis: I really don't know about this measure.

Thanks for the input, Mike. I've seen lots of your files arriving.

Dylan.

Hey Mineral

I've still got a worse alpha than you - I'm with Elmo - what is this obsession with corners & turning[}:)]

Select to expand quote

Wineman said...

Hey Mineral

I've still got a worse alpha than you - I'm with Elmo - what is this obsession with corners & turning[}:)]

Wish I knew Have been at it for a while now to learn how to gybe, I crash most of them and resort to @#$$@#$ water start. But its getting slowly ever soooo slowly better.