Haha! There is obviously not much interesting on the forums today.

I don't know how you can prove that ANY gps measures speed via doppler just by looking at the output files. For gpx as an example, all it says is SPEED and gives a number. How it got that number is an internal algorithm, hopefully by doppler. It is prettly hard to reverse engineer the firmware code to ascertain whether it is doppler or not.

The guy who told me that the original Garmin Legend was doppler was a retired Garmin software engineer.

In any case, we all benefit from the speed accuracy of doppler compared to that of ds/dt. Because modern electronics has evolved to the point of being able to resolve the speed of light to sub-centimeter levels, there is no reason not to use doppler. Can't prove it, but I think they all do.

I run my GW-60 at 5 Hz, so the the 2-second speed is a running average of the last 10 points, and the 10-second speed is a running average of the last 50 points. Pardon me, I'm still gonna call this "smoothing." This, and SPD GENIE, are, by far, the best features of the GW-60 over pretty much all the others.

Select to expand quote

segler said..

I run my GW-60 at 5 Hz, so the the 2-second speed is a running average of the last 10 points, and the 10-second speed is a running average of the last 50 points. Pardon me, I'm still gonna call this "smoothing." This, and SPD GENIE, are, by far, the best features of the GW-60 over pretty much all the others.

This is not smoothing of data by the gps. It is the averaging of the raw log data from the GPS by analysis software (KA72, GPSResults, Realspeed etc) to provide results for the six disciplines (2 sec, 5x10, hourh alpha 500, nautical mile, distance) for GPS Team Challenge.

Select to expand quote

segler said..
I don't know how you can prove that ANY gps measures speed via doppler just by looking at the output files. For gpx as an example, all it says is SPEED and gives a number. How it got that number is an internal algorithm, hopefully by doppler.

An indication is the difference between unit display and post calculation results.
If it's a doppler file it should give the same max speed as the display, if it's a trackpoints only file it will often give higher numbers from the file than the display showed.

Some gps units output both doppler and trackpoints, they make different speed graphs and different tracks, so it's obvious there's both. Other units only output one set of data, (trackpoints). The analysis software we use, has a check box, for "doppler" and "doppler if available". If this is set to doppler, any file that doesn't have a doppler output will give no results.
So we don't have to reverse engineer anything, The output itself, tells whether it's doppler or not.

Your Garmin engineer was probably just talking about the display.

Here's an example of what I'm talking about. The alpha starts at bottom right, the doppler and trackpoints are aligned, but during the gybe they start to diverge a little bit. So there are definitely two sets of data for the same alpha. On trackpoint only devices you won't see this.

Select to expand quote

segler said..
I don't know how you can prove that ANY gps measures speed via doppler just by looking at the output files. For gpx as an example, all it says is SPEED and gives a number. How it got that number is an internal algorithm, hopefully by doppler. It is prettly hard to reverse engineer the firmware code to ascertain whether it is doppler or not.

The guy who told me that the original Garmin Legend was doppler was a retired Garmin software engineer.

In any case, we all benefit from the speed accuracy of doppler compared to that of ds/dt. Because modern electronics has evolved to the point of being able to resolve the speed of light to sub-centimeter levels, there is no reason not to use doppler. Can't prove it, but I think they all do.

I run my GW-60 at 5 Hz, so the the 2-second speed is a running average of the last 10 points, and the 10-second speed is a running average of the last 50 points. Pardon me, I'm still gonna call this "smoothing." This, and SPD GENIE, are, by far, the best features of the GW-60 over pretty much all the others.

Again, no they don't. Not by a long shot. Almost all 'may' use the Doppler derived speed for the screen display, but it is very clear in the raw data if Doppler speed is recorded. There are specific fields in the data sructure that correspond to the various outputs. A gps that does not record Doopler speed, or Velocity calculations will only list positional information. (Lat. and Long. of each point.)

I think your confusion is caused by you mistaking what is output by analysis software, from what is actually input from the GPS. The difference is bleedingly obvious to all of us who have studied it.

Your Garmin engineer probably confused this too. He was probably correct that the screen display speed was Doopler derived data, we are pretty sure this was the case with the other Garmins we used back pre GT-11, but they very definitely did not record that Doppler data.

Some firmware in GPS's may internally convert the positional data into 'speed' when it writes to a proprietary file type. That by no means implies in any way that that data may be derived from Doppler calculations.

The average of a lot of points (2 second is actually more commonly derived from 11 points @5Hz, 10 sec from 51 points), is not what data analysts call 'smoothing'. As I said, GPS engineers refer to something like a low pass filter as 'smoothing'

Here is a graph of 5Hz Doppler speed from Locosys GW-52. It is not 'smooth', but shows minute variations in speed calculated that contains micro accelerations, many caused by vibrations/shaking probably. Note also that it is calculated over 11 points. Also note that the speed graph, and the point list, show both the positional speed and the Doppler speed.



Please send me your Garmin Legend data file and I will examine it to see if it includes Doppler speed data, By the way, what is the file output type from the Garmin Legend? Is it NMEA or GPX?

Steady on Daffy, Those of us who are not data analysts or GPS engineers would consider a moving average to be a type of smoothing.
As does Wikipedia en.wikipedia.org/wiki/Smoothing

True, I just sayin'. If you start your 10 second run at 35 knots, peak at 40 and finish at 35 again. The average may be 36,5, That suggests the run was smooth at 36.5knts. But that was not actually the case. Smoothing?

We all benefit from this discussion. Thanks.

All my handheld Garmin hiking units, including the aged Legend, output gpx files, with data taken at typically one point per second. Nice thing about gpx and nmea formats is that you can read them with a text editor.

If we had to depend on ds/dt for speeds, with s being ?10 meters most of the time, our speeds would be noisy all over the place. At our on-water windsurfing speeds, the standard deviations would be bigger than the moving means. Talk about noise. Since, however, our speeds are hugely less noisy than that, there must be a much better way to get speeds. Doppler is the only better one that I know of.

Since my old Legend, and later Legend HC, a couple different HCx units, the Oregons, the Montanas, and my Timex all report speeds that plot out with noise not too different from that of my GW-60, their speed points, even without moving averages, are definitely not ds/dt. Doppler has to be their method.

Again, just by looking at a gpx file, and looking at the number, I don't know how you can conclude whether it is doppler-derived or ds/dt-derived or somehow-otherwise-derived. You have to look at firmware code.

Have a look at the speed graph example I posted above. There are two lines. One is Doppler speed and one is speed derived from positional data. See that they are both about the same 'smoothness'. With 1Hz data, many GPS's already apply a 'smoothing' filter to to the positional derived speed, but even if they don't, there is often not a big difference in their appearance if there is excellent reception conditions and the atmospheric conditions are not changing rapidly. On the other hand, sometimes there are sudden changes.

yes you have to look closely at sailquick's graph to see there are 2 separate tracks overlaid, one doppler the other trackpoints.
It's a bit easier to see on my alpha graph above that there are 2 tracks. If there's doppler and trackpoints in a file it's very easy to tell. But I'm not sure about doing that with a text editor. May be later if I have time I'll try it with some ancient GPX files I know include doppler tracks.

Here's a Real Speed view of a GPX file from back in 2008, almost certainly from a GT11

This clearly shows the difference between doppler and trackpoints, (doppler green, trackpoints black). And how a doppler file is "smoother" than trackpoints, but doesn't have "smoothing".

Below is the same period in the GPX file, that this speed graph came from, via Real Speed.
It only has 1 speed reading with a max of 16.03m/s, (31.16kts)

Realspeed must be calculating the tackpoint speed from the positional data, as the below table doesn't show the huge spike above.
The speed in the below table must be doppler.
So the highest point in the file is 31.16kts, the highest doppler from the speed graph is the same.
Interestingly there is a gap of 5s here, I guess the old GT11s had problems!
The spike in the trackpoints is at 06:24:32, that's about 32kts, the GPX table's speed at that point is, 22.45. definitely doppler.


BUMMER, I can't post the raw GPX file, this smart page converts the HTML, so the first number appears to be "ele" no idea what that is.
Then there's date and time, speed in m/s and HDoP and lastly the number of sats.
-35 2008-01-26T06:23:56.815Z 13.05 1.4 6
-35 2008-01-26T06:23:57.815Z 12.53 1.4 6
-35 2008-01-26T06:23:58.815Z 12.12 1.4 6
-35 2008-01-26T06:23:59.815Z 12.37 1.4 6
-35 2008-01-26T06:24:00.815Z 12.93 1.4 6
-35 2008-01-26T06:24:01.815Z 13.02 1.4 6
-35 2008-01-26T06:24:02.815Z 13.57 1.4 6
15 2008-01-26T06:24:07.814Z 16.03 1.4 6
17 2008-01-26T06:24:08.814Z 15.58 1.4 6
20 2008-01-26T06:24:09.814Z 15.12 1.4 6
24 2008-01-26T06:24:10.814Z 15.05 1.4 6
27 2008-01-26T06:24:11.814Z 14.74 1.4 6
30 2008-01-26T06:24:12.814Z 14.61 1.4 6
34 2008-01-26T06:24:13.814Z 14.30 1.4 6
37 2008-01-26T06:24:14.814Z 14.17 1.4 6
41 2008-01-26T06:24:15.814Z 13.81 1.4 6
44 2008-01-26T06:24:16.814Z 13.37 1.4 6
48 2008-01-26T06:24:17.814Z 12.96 1.4 6
52 2008-01-26T06:24:18.814Z 12.41 1.4 6
55 2008-01-26T06:24:19.814Z 12.22 1.4 6
60 2008-01-26T06:24:20.814Z 11.67 1.4 6
64 2008-01-26T06:24:21.814Z 11.67 1.4 6
68 2008-01-26T06:24:22.814Z 11.53 1.4 6
71 2008-01-26T06:24:23.814Z 11.52 1.4 6
75 2008-01-26T06:24:24.814Z 11.46 1.4 6
79 2008-01-26T06:24:25.813Z 11.62 1.4 6
82 2008-01-26T06:24:26.813Z 11.46 1.4 6
86 2008-01-26T06:24:27.813Z 11.33 1.4 6
90 2008-01-26T06:24:28.813Z 11.26 1.4 6
93 2008-01-26T06:24:29.813Z 11.07 1.4 6
97 2008-01-26T06:24:30.813Z 11.06 1.4 6
101 2008-01-26T06:24:31.813Z 11.27 1.4 6
105 2008-01-26T06:24:32.813Z 11.55 1.4 6
-42 2008-01-26T06:24:46.350Z 12.05 2.0 5
-40 2008-01-26T06:24:47.349Z 11.94 1.4 6
-38 2008-01-26T06:24:48.349Z 11.89 1.4 6
-37 2008-01-26T06:24:49.349Z 11.72 1.4 6
-37 2008-01-26T06:24:50.349Z 11.26 1.4 6
-36 2008-01-26T06:24:51.349Z 10.99 1.4 6
-36 2008-01-26T06:24:52.349Z 11.17 1.4 6
-36 2008-01-26T06:24:53.349Z 11.15 1.4 6
-36 2008-01-26T06:24:54.349Z 11.02 1.4 6
-36 2008-01-26T06:24:55.349Z 11.38 1.4 6
-36 2008-01-26T06:24:56.349Z 11.53 1.4 6
-36 2008-01-26T06:24:57.349Z 11.70 1.4 6
-36 2008-01-26T06:24:58.349Z 11.62 1.4 6
-36 2008-01-26T06:24:59.349Z 11.99 1.4 6
-36 2008-01-26T06:25:00.349Z 12.16 1.4 6
-36 2008-01-26T06:25:01.349Z 12.23 1.4 6
-36 2008-01-26T06:25:02.349Z 12.08 1.4 6
-36 2008-01-26T06:25:03.349Z 11.96 1.4 6
-36 2008-01-26T06:25:04.349Z 11.68 1.4 6
-36 2008-01-26T06:25:05.348Z 11.43 1.4 6
-36 2008-01-26T06:25:06.348Z 11.27 1.4 6
-36 2008-01-26T06:25:07.348Z 11.27 1.4 6
-36 2008-01-26T06:25:08.348Z 11.01 1.4 6
-36 2008-01-26T06:25:09.348Z 10.72 1.4 6
-36 2008-01-26T06:25:10.348Z 10.58 1.4 6
-36 2008-01-26T06:25:11.348Z 10.81 1.4 6
-36 2008-01-26T06:25:12.348Z 10.95 1.4 6
-36 2008-01-26T06:25:13.348Z 10.84 1.4 6
-36 2008-01-26T06:25:14.348Z 10.42 1.4 6
-36 2008-01-26T06:25:15.348Z 10.45 2.4 5
-36 2008-01-26T06:25:16.348Z 10.57 1.4 6
-36 2008-01-26T06:25:17.348Z 10.50 1.4 6
-36 2008-01-26T06:25:18.348Z 10.37 1.4 6
-35 2008-01-26T06:25:19.348Z 10.11 1.4 6
-35 2008-01-26T06:25:20.348Z 10.23 1.4 6
-35 2008-01-26T06:25:21.348Z 10.43 1.4 6
-35 2008-01-26T06:25:22.347Z 10.42 1.4 6
-35 2008-01-26T06:25:23.347Z 10.65 1.4 6
-35 2008-01-26T06:25:24.347Z 10.65 1.4 6
-35 2008-01-26T06:25:25.347Z 10.94 1.4 6
-35 2008-01-26T06:25:26.347Z 11.15 1.4 6

You lost me at -35 2008-01-26T06:24:00.815Z 12.93 1.4 6

Select to expand quote

seanhogan said..
You lost me at -35 2008-01-26T06:24:00.815Z 12.93 1.4 6

Not that hard Sean. It's just the data in the GPX file

Select to expand quote

decrepit said..
Some gps units output both doppler and trackpoints, they make different speed graphs and different tracks, so it's obvious there's both. Other units only output one set of data, (trackpoints). The analysis software we use, has a check box, for "doppler" and "doppler if available". If this is set to doppler, any file that doesn't have a doppler output will give no results.
So we don't have to reverse engineer anything, The output itself, tells whether it's doppler or not.

No, it's not really that easy. I have tested at least one GPS unit that gave speed numbers that clearly differed from the trackpoint-derived speeds. Checking and unchecking the "Doppler" box in the analysis software gave different results, so it seemed safe to assume the unit gave doppler speeds. However, some tests had large artifacts that caused enough suspicion to double check with the chip manufacturer, who then confirmed that the chip was not calculating doppler speeds. Most likely, the speeds where processed with Kalman (or similar) filters that were not used for the track points.

Select to expand quote

sailquik said..
... it is very clear in the raw data if Doppler speed is recorded. There are specific fields in the data structure that correspond to the various outputs.

No, it is absolutely not. There are only two easy cases:
1. The GPS files only records positional information, and does not have a separate entry for speed.
2. There is a separate speed entry, but the speed there is identical to the position-based speed (evidenced over a sufficiently large number of points).
In these cases, we know that the device does not record doppler speed.

But the reverse conclusion is not necessary true. If a device file gives a speed that differs from the positional speed, that only means that this speed was not calculated directly from the positional data. For example, any simple filter applied to the speed will cause a deviation, and speed data are often filtered. The use of doppler to calculate speeds is only one possible reason for discrepancies.

Select to expand quote

segler said..
If we had to depend on ds/dt for speeds, with s being ?10 meters most of the time, our speeds would be noisy all over the place.

I'm trying to understand what you are saying here. I thinkwhat shows up as (question mark) 10 is supposed to be (plus/minus) 10 meters, which is based on typical positional errors of 5 meters. This is missing a critical piece of information about the positional accuracy: the error margins describe theabsoluteaccuracy of the position, not theaccuracyrelativeto preceding points. Formostpoints, therelativepositional error is a lot smaller than the absolute error. For these, the positional speed is very close to the doppler speed,as is the noise. Only occasionally, the track points "jump", causing a spike in the speed graph. There are a couple of examples in Mike's graph above, or look at this track:


Even with ds/dt, the speeds are pretty close to the correct speeds most of the time.

Select to expand quote

segler said..
Since my old Legend, and later Legend HC, a couple different HCx units, the Oregons, the Montanas, and my Timex all report speeds that plot out with noise not too different from that of my GW-60, their speed points, even without moving averages, are definitely not ds/dt. Doppler has to be their method.

Again, just by looking at a gpx file, and looking at the number, I don't know how you can conclude whether it is doppler-derived or ds/dt-derived or somehow-otherwise-derived. You have to look at firmware code.

Now you are really confusing me. First you say that since the noise on your data does look similar to the GW-60, so they must all use doppler.

But then you say you can't just look at the numbers in a file and "conclude whether it is doppler-derived"? That seems to be a direct contradiction.

"For well over a decade, most GNSS receivers use a combination of position deltas and Doppler shift to determine speed."
The issue I had seen with a "non-doppler" GPS before was that spikes in the positional speeds "bled through" into the speeds, albeit much reduced. I just looked at a whole bunch of GT-31 and GW-60 files, and did not find a single example of that happening. Here is a typical comparison of the positional speeds (top) and doppler speeds (bottom) from a GW-60:


In general, the positional speeds are very similar, and that includes the noise. The obvious big difference is a spike after a crash - probably when the GPS lost reception briefly, and then re-acquired satellites and come up with a brand new position. That spike is completely absent in the doppler data, as would be expected for pure doppler-based speed.

Select to expand quote

boardsurfr said..

sailquik said..
... it is very clear in the raw data if Doppler speed is recorded. There are specific fields in the data structure that correspond to the various outputs.

No, it is absolutely not. There are only two easy cases:
1. The GPS files only records positional information, and does not have a separate entry for speed.
2. There is a separate speed entry, but the speed there is identical to the position-based speed (evidenced over a sufficiently large number of points).
In these cases, we know that the device does not record doppler speed.

But the reverse conclusion is not necessary true. If a device file gives a speed that differs from the positional speed, that only means that this speed was not calculated directly from the positional data. For example, any simple filter applied to the speed will cause a deviation, and speed data are often filtered. The use of doppler to calculate speeds is only one possible reason for discrepancies.

Well, yes, I said as much in an earlier post where I gave an example of a GPS watch that appears to output Doppler speed, but does not.

I should have used the word, 'usually'.

Both of the speed graph example given by Boardsurfr and decrepit show excellent examples of what we called 'spikes' in the Locational data speed calculations. We rarely see a similar type of obvious 'spike' in Doppler data errors, ( but I am sure Boardsurfer can find and example ) which is why other data to validate Doppler speeds, like 'Doppler speed error' is so important.