Electronics, Lighting, & Gadgets - Edge 305 speed fluctuation

I'm still pretty new to my Edge 305, and got this for a short 22mile loop this morning

http://i321.photobucket.com/albums/nn384/RDW5980/edgePinehurst.jpg

I noticed while I was riding that my speed was fluctuating way more than made sense at several points (e..g mile 5 to 8) . I've been used to riding with a Cateye wired Strada cadence before this and am accustomed to steady speed readout, however this ride seemed to have my speed jumping around 1-2mph at a time, when I know that my cadence was steady.

So, I'm wondering, does the Edge ever try to get speed from GPS? (It looks like it starts at a time when tree cover was more significant). Secondly, anyone ever had any problem with the 305 loosing contact with the speed/cadence unit, or any problems with electromagnetic interference? I'm a bit puzzled, and hope I can fix it.

The speed/cadence sensor can have some issues. They are usually caused by low battery and wheel magnet not close enough to the sensor arm. I would check both and post back. It might help to overlay your cadence just to verify that cadence was not fluctuating.

1. Are you using a wheel-speed sensor with the 305?
2. What program are you using to generate the graph?

I suspect the jagged graph may be more about the data smoothing in the display program, than the device that recorded the data.

I agree with Shimagnolo, it looks like it could be a smoothing issue. Assuming you are using the Garmin Training Center, you can smooth your data out by hitting the button circled in red on the screen shot below. The two graphs are of the same ride, but the one on the bottom has been smoothed.

http://i58.photobucket.com/albums/g251/freemanmatt/smoothing.jpg

If you're not in GTC, if you can export your speed vs time data, just set up a moving average and smooth it yourself in Excel.

As I understand it, the Edge always gets speed data from the GPS unless the GPS is turned off (like on a trainer indoors) or the satellite coverage is poor.

Thanks for the input guys. I've not had internet access to reply until now. In answer to the points:

I generated this graph using motionbased.

I don't think that this is a smoothing issue, since I noticed the issue real time on the unit itself - if your speed jumps 1 mph or more either way without a significant change in your cadence this doesn't make sense. I also didn't "see" an equivalent change in the cadence readout on the screen.

I've tried to reposition the speed sensor (yes, I'm using the speed/cadence monitor) to see if this recurrs. I was curious if others have seen the same issue or not.

Cadence detected by sensor once or twice a second at 60 to 120 rpm versus a cluster of satellites some 23000 miles away every 100ms or so with a position accuracy without tricks of say 6 to 12 feet. Hmmm, I wonder which will show more variablility?

torgrot

Cadence detected by sensor once or twice a second at 60 to 120 rpm versus a cluster of satellites some 23000 miles away every 100ms or so with a position accuracy without tricks of say 6 to 12 feet. Hmmm, I wonder which will show more variablility?

torgrot


As I understand it, the Edge always gets speed data from the GPS unless the GPS is turned off (like on a trainer indoors) or the satellite coverage is poor.

That second point is interesting - the start of the route is relatively open, with good satellite coverage. Where I started seeing this variability (Canyon, CA for anyone who knows the area) is gentle rolling terrain leading to a climb, but is quite heavily wooded. So if this was a GPS derived speed measurement that would make sense. I thought that you set the sensors in the setup menu?

As I understand it, the Edge always gets speed data from the GPS unless the GPS is turned off (like on a trainer indoors) or the satellite coverage is poor.

What I have read somewhere, (maybe in that loooong 705 thread), was that is there is a cadence/wheel sensor, once under way GPS is initially used to calibrate output from it, then the speed is taken only from the sensor for the duration of the trip.

I had an experience where the speed reading kept jumping between two different speeds, one of which was twice the other. I finally figured out the the sensor arm had been jarred so it was right on the border line of where the magnet could be detected. Apparently, ever time it missed the magnet, I would get the low reading, then the next time it detected it on two revolutions in a row, it would read correctly again.

So I *know* it does not get speed from GPS if the cadence sensor is detected.

So I *know* it does not get speed from GPS if the cadence sensor is detected.

Well, no, you don't. If the Edge 705 uses the GPS position data where possible, and falls back to the cadence sensor if necessary, and you were in a bad-GPS-reception situation, you would have seen the same results.

I don't *know* the answer, either, but it's easy to notice fluctuations in reported speed on a flat, level, ride, when the cadence does NOT change. If the speed came solely from the wheel sensor, then - when in a given gear and not coasting - the two would have to move together and maintain a fixed ratio. If the speed varies by, say, more than 1% but a cadence of around 100 isn't changing, then the two are being calculated independently.

Due to GPS position errors, measurements of instantaneous speed will fluctuate while average speed will remain (more) correct, if the speed is being measured by GPS.

What I have read somewhere, (maybe in that loooong 705 thread), was that is there is a cadence/wheel sensor, once under way GPS is initially used to calibrate output from it, then the speed is taken only from the sensor for the duration of the trip.

I had an experience where the speed reading kept jumping between two different speeds, one of which was twice the other. I finally figured out the the sensor arm had been jarred so it was right on the border line of where the magnet could be detected. Apparently, ever time it missed the magnet, I would get the low reading, then the next time it detected it on two revolutions in a row, it would read correctly again.

So I *know* it does not get speed from GPS if the cadence sensor is detected.

I thought we were talking about the 305, not the 705. They could do things differently. This is from page 76 of the Edge 305 User Manual "The speed data from the GSC10 is only recorded and used for distance calculations when GPS signal weak or GPS is turned off." I guess I interpreted that to mean the speed was only taken from the wheel sensor in those cases, but it doesn't really say that. I know my Edge will display speed even when I'm just walking with it and the bike is sitting still.

Sorry - good point, and it was a typo on my part (I have both the 305 and 705 in our house). Your quote from the manual seems to agree with what I was describing; I have no reason to think the two units would work differently.

Yes, the 305 and 705 will both display "speed" when sitting still and within range of a wheel sensor. Again, due to GPS position errors, you appear to be moving because your calculated GPS position changes. That's a great way to demonstrate how the GPS speed measurement is good in the long run (all those little speed twitches would, if signed properly, average to 0) but not as an instantaneous measurement.

I repositioned my sensor and took this from a ride yesterday afternoon up Diablo. Looks better to me, but then again, I'm sure that GPS coverage was also good. I'll likely need to wait for another opportunity to do the original loop.

http://i321.photobucket.com/albums/nn384/RDW5980/052109DiabloNG.jpg

The speed numbers seem tighter to me (looking at the climb), and were more reflective of my ride here. I did notice the cadence numbers showing some odd values at the start of the descent, when I was coasting, but figured that could be me backpedalling/repositioning the cranks as I set up to go through corners and probably swept the magnet past the sensor.

These replies are interesting though - I'm sensing that despite how common the 305's are, they are still something of a black box.

My concern is that if your speed can vary by as much as 1mph in either direction depending upon intermittent GPS coverage, that's going to make any kind of ride to ride comparison difficult.

I thought we were talking about the 305, not the 705. They could do things differently. This is from page 76 of the Edge 305 User Manual "The speed data from the GSC10 is only recorded and used for distance calculations when GPS signal weak or GPS is turned off." I guess I interpreted that to mean the speed was only taken from the wheel sensor in those cases, but it doesn't really say that. I know my Edge will display speed even when I'm just walking with it and the bike is sitting still.

That exact phrase is on page 40 of the 705 manual.
I have heard before, (but never seen a source cited), that the 305 and 705 do things differently.
My experience with a dislodged GSC 10 make me skeptical of that line in the manual.
Also, the way it is done could change in a firmware upgrade, w/o documentation being updated.

Love that elevation graph!

There's certainly something else going on here. Under relatively steady conditions, the magnitude of the GPS position error will fluctuate within a relatively constant range. You seem to be showing a larger absolute fluctuation at higher speeds.

It's important to remember that most consumer GPS units do quite a bit of smoothing and filtering before giving you a result. Most consumers prefer information that's predictable rather than correct :) You're unlikely to go from 20 mph to 0 mph then back to 20 mph in 3 seconds, and if you went through an underpass and completely lost sateliite reception in those 3 seconds, your GPS would probably look like nothing happened (even without a wheel sensor to back it up).

I wonder if the data that's recorded and downloaded from the Edge is different from the data that's displayed? I could see the display being smoothed but the data kept in original format. I'll bet your displayed speed on the descent didn't fluctuate that much.

There's certainly something else going on here. Under relatively steady conditions, the magnitude of the GPS position error will fluctuate within a relatively constant range. You seem to be showing a larger absolute fluctuation at higher speeds.


Actually the higher speed fluctuation is probably accurate, the descent down Diablo is full of curves so there are lots of times when I'm scrubbing off speed before going into a corner - I think that's what you're seeing in the graph. I highly recommend the ride!

My suspicion looking back at the first graph is that the GPS coverage was intermittant during miles 5-8, and this meant that the speed was constantly being readjusted as I moved through a clear patch w/o trees. This showed up real time on the 305 unit, and is probably why my mph readings seemed all over the place despite having a sensible cadence. I guess that's one of the drawbacks of the system. Otherwise I'm loving it.

Here's my final update. Did the following ride today:

http://i321.photobucket.com/albums/nn384/RDW5980/2009_05_23_PeninsulaRide/052309ElevationSpeed.jpg

I started off without the 305 picking up a signal for the GPC10. So the first hill (Old La Honda in Woodside) was ridden entirely on GPS - so speed very variable as the signal came and went through the trees.

By the time I got past OLH my cadence reappeared, and by the time I got to the last big climb (Tunitas Creek) the same situation was occuring. In fact i watched as I was climbing with a cadence of 70rpm, and watched my speed vary from 3.8mph to 6.8mph in the space of a second.

I guess this will just be a regular occurence when riding through tree cover here unless anyone has any smart fixes.

Thanks all for your replies.

Well riding the Silver Comet trail in Georgia, you go through a tunnel and my 305 loses satellite reception completely. How do I know, the 305 tells you. It flashes a box saying it has lost satellite reception. As I recall it stopped displaying the speed until I got out of the tunnel or it displayed a zero. I have never had it do that while riding through a forested area. I think they do a very good job of locking on and not giving up. It is just from Garmin's own web site that the accuracy is worse than I stated above, more like 15 meters.

How accurate is GPS?

Today's GPS receivers are extremely accurate, thanks to their parallel multi-channel design. Garmin's 12 parallel channel receivers are quick to lock onto satellites when first turned on and they maintain strong locks, even in dense foliage or urban settings with tall buildings. Certain atmospheric factors and other sources of error can affect the accuracy of GPS receivers. Garmin® GPS receivers are accurate to within 15 meters on average.

http://www8.garmin.com/aboutGPS/

torgrot

Garmin® GPS receivers are accurate to within 15 meters on average

Right, which means half the time the error's more than 15 meters :)

The chief cause of the wild fluctuations in GPS speed is not "loss of GPS coverage", strictly speaking, but switching between different GPS satellites (and solutions).

At any time your GPS will use signals from four satellites to calculate your position. A "twelve-channel" receiver doesn't use twelve satellites at once - it still uses only four. However, it can keep track of 12 at one time, and it can dynamically select the best four (based on geometry and signal strength) at any instant to calculate your position.

Let's say there are five satellites being tracked, all with excellent signal quality, and for which any four will produce comparable geometry. And let's say you're riding by a row of trees on your left, which intermittently block reception for one of those five satellites.

You ride along with a GPS position determined by signals from satellites 1, 2, 3, and 4 for a while. You have a 10-meter error in your position based on the conditions experienced by those signals at that moment. Then you pass the trees, and signal 4 is briefly blocked. No problem, your GPS switches to satellite 5, and calculates a NEW position based on 1, 2, 3, and 5. That new position has a DIFFERENT 10-meter error; remember, you can always have a 15-meter error and still jump around by 30 meters every second.

When you hit a break in the trees, you go back to your 1, 2, 3, 4 solution, then back to 1, 2, 3, 5 when you hit the next tree. The result is that you ping-pong back between multiple GPS positions rapidly. Each is equally valid, with comparable error, but different. This causes short, instantaneous jumps in apparent speed.

Remember that when the GPS marketing guys selected among several different ways to measure position error, they picked the one that gave the smallest number, not the most useful one! CEP (or Circular Error Probable) of 15 meters is still a pretty big number, and means that 98% of the time you'll be within 36 meters of the right place.

I resurrected a thread in the road cycling forum about elevation estimates, based upon my observation of how some parts of yesterdays ride were plotted out by MB. I suspect that this relates to Ed's point above - compare the road that I took with the route that MB claims that I rode. Given that both MB and GTC gave me higher elevation gain for this ride, I think it comes from them trying to measure elevation points from a topo map, rather than the value I got from the Edge 305 unit:

http://i321.photobucket.com/albums/nn384/RDW5980/2009_05_23_PeninsulaRide/Tunitas_GPS.jpg

Actually, one of the techniques used to filter GPS data to meet user expectations is - in automotive GPS units - to strongly presume that you're actually driving on the road!