In an equal test of 5 Hertz Alpha data accuracy versus 1 Hertz Alpha data accuracy, as one might obtain for instance by wearing two GW-60 watches on the same arm and set to record at the different rates. When both sets of data are submitted to KA72.com does the programme rationalise similar curves and hence correct to the same Alpha speeds, or will a 5 Hertz Alpha invariably be the slower, but more accurate score?

I think the accuracy of gw60 alphas, has much more to do with the over/under hand scenario, and what happens with the wrist in the gybe.
This can cause not only loss of signal, but reflected signals as well.
5hz will obviously record the curve more accurately. But as speed in both cases is doppler, not distance over time, the actual recorded speed through the gybe shouldn't be a lot different.

Also I think I remember locosys saying that 1hz is only the record rate, the calculations are done with 5hz data. so the speed average through the gybe should be the same.

The main difference with 1hz and 5hz is the 2s, because the software can get closer to the peak speed with 5hz. 1hz can be offset. If it's a very peaky top speed, this can make a difference

I can't say how KA72 handles alphas, but I think most software just averages all the speed points recorded through the alpha.
The GW60 itself can give different results, because it doesn't do alphas shorter than 300m and used doppler to calculate the 50m separation, whereas other software uses positional data for the 50m separation. This can lead to different alphas being included or excluded between the two methods.

Thanks for your reply decrepit. ..Regarding your first paragraph. In your experience and where GW-60 Alpha data has been corrupted by wrist movements etcetera, does the situation simply resolve to a lower than actual speed for the Alpha, that is only a nuisance to oneself, or can the situation just as easily expand an Alpha to a fictitiously high score that is also a nuisance to other competitors?

Regarding measurement of speed using Doppler shift, I have a basic grasp of the concept in respect to navigation along a straight line, but I am at a loss to understand how Doppler shift can be reliably measured around a significant curve, and especially so in the case of a half circle that is only 50 metres in diameter? ..Are you certain that is the case for measurement of speed in the gybe? ..I think I had tended to assume that the KA72 programme somehow brought latitude and longitude data and the relevant time to the fore to resolve the speed in the gybe segment of Alpha's.

Select to expand quote

Phromsky said..
>>> I have a basic grasp of the concept in respect to navigation along a straight line, but I am at a loss to understand how Doppler shift can be reliably measured around a significant curve, and especially so in the case of a half circle that is only 50 metres in diameter? ..Are you certain that is the case for measurement of speed in the gybe? ..I think I had tended to assume that the KA72 programme somehow brought latitude and longitude data and the relevant time to the fore to resolve the speed in the gybe segment of Alpha's.

That's why you need multiple sats in a good spread around you. Yes in a straight line you only need one source in the same line as your direction of travel. But in a varying course you need at least 5 to get an accurate calculation. Unfortunately the GW60 only uses the GPS system, so in bad reception times, 5 may not be achievable. This aggravates the loss of accuracy due to wrist movement in the gybe.

I've done some tests, and with a good continuous sky view, the doppler calculated tracks are more accurate than positional data. But as soon as the sky view is degraded that advantage starts to disappear. Positional data is only accurate to a meter, whereas 5hz doppler can be much closer than that, even around corners.

Yes I'm sure the gybe speed is calculated using doppler

Standstill, straight lines, curves, hill, up, down are all our constructs. To GPS, we're floating on a spinning ball, always in motion.

The speeds calculated using positional data in a sharp turn out be the small straight segments between data points. Not the actual path followed. In theory this should result in a minutely slower speed if the points are 1 second apart V's 2 Hz or 5 Hz. But in reality, those positional points can vary from the true path somewhat also so the speed calculation may be also slightly slower, or faster.

From a practical point of view, it is such a small amount, that it is virtually insignificant most of the time, and all 1Hz users are in the same boat so no advantage or disadvantage, however small.

For Doppler calculation, each data point has a velocity component but also a heading component. That data is for each point so it you had a mapped track at 1Hz, it should look essentially the same as the Positional data, i.e.. A number of points linked between by straight lines.

But the data point speeds for Doppler are far more accurate than the speeds for positional for two main reasons:
1. The lines between positional points zig-zag with positional inaccuracy, effectively adding to the distance measured and increasing the speed. Sometimes that error 'spikes' quite high.
2. Most of the factors that cause inaccuracy in positional data (eg. ionosphere interference, positional and time errors in satellites and more) either do not effect Doppler measurement, or the effect is vastly less.

Obviously, if you have more points from either type of data, you should have what appears to be a smoother curve drawn on the map. That is, it will appear to be more like the reality of the actual course. 5hz Doppler is therefore much more accurate than 1Hz and 10Hz is more accurate then 5Hz.

The only positional data used in the Alpha category is to locate the proximity circle. Doppler heading was trialed and rejected as having more chance of large errors for the proximity than Positional data. There could be a long explanation of why we found this, but the simple answer is that a very small error in bearing data during the turn, which is more likely in the tuen due to sunned tarm/hand/body movement, remains in the data and is carried forward as a new course direction. Whne this is extended for some distance, the calculated location at that distance can be many meters out. at least with positional measurement, errors can accumulate (perhaps 'project' is a better term). In the short time and distance between the sart and end of an Alpha 500m, the positional error is very unlikely to be more than a meter or two, and is usually in error in the same direction, so the proximity circle distance error is likely well under the absolute positional error of one to two (or even possibly more) meters.

Why is it that Garmin and other mainstream GPS manufacturers are not providing Doppler derived speed in their devices?

well some of them say they do, but I guess most of their uses aren't as fussy about accuracy as we are

As silly as it sounds, accuracy isn't attractive. Not to users. So not to manufacturers either. The current marketing highlight is battery life. Check the main brands pages, it's everywhere. Battery life and a beautiful colour screen. And loads of apps. To solve for these, they reduce the GNSS receiver active time and fill the blanks with inertial measurements. It's the magical "high accuracy algorithm". It isn't there to improve anything, it can't. It's there to hide the GNSS trick. Because an active IMU will consume peanuts compared to an active GNSS receiver, it made battery life skyrocket. The GNSS chip in most recent watches is the same and comes with the "algorithm" embedded. There is no competition in the broad consumer watches : same format, same chip, same "algorithm", same price.