This is part of my main Garmin 945 Review and specifically looks at many aspects of GPS accuracy or, more correctly, GNSS accuracy.
I’m looking at FITNESS FOR PURPOSE but what purpose do you want a GPS track for? If it’s a pretty post-run or post-ride/run track of where you’ve been that’s ‘about right’ then most sports watches will be fine for you. If you want to use GPS to determine your walking and cycling speed then you’ll probably be alright too. It’s just that when you are running GPS has a tricky job of dealing with the environment and your swinging arm to give you a correct pace and distance. Most sports devices are not fit for the purpose of giving instant pace but may be alright with your current lap pace. Think about it….a running watch can’t really tell you how fast you are running. Hmmm.
They probably also misestimate distance by 0.5% or more in tricky, urban conditions.
I’m also looking at the accuracy of using the American GPS satellites and a little bit at augmenting that with the potentially superior European GALILEO satellites and we’ll leave the Russian’s GLONASS system out of it. There are other regional satellite systems too.
I’ve completed 2.5x 10 mile runs on my standard GPS course and following a set methodology with my normal usage in running and other sports. That means you can compare the 945 with the best GPS watches from yesteryear which most of you will know as the Polar V800 and Suunto AMBIT 3.
Note: I’ve only had 2 OW swims so far. One had an elemnt of ‘user error’ to contend with 😉 I’ll update that section in June probably (ping me below in the comments if I haven’t done that)
FYI: Garmin also uses the accelerometer in the watch as an input to their pace/distance algorithms.
GPS Accuracy – GLONASS, GALILEO and GNSS Functions
It seems that the new Sony GNSS chips used by Garmin from 2019 onwards can only use the E1 frequency (Galileo) and the L1 frequency (GPS). My understanding is that the accuracy of Galileo with a single frequency is similar to the accuracy of GPS with a single frequency. Thus you can only expect to get +/-5m accuracy BUT enabling GALILEO increases the number of visible satellites and hence GALILEO increases the probability of +/-5m accuracy being achieved.
The massive improvement in GNSS accuracy is most likely to come from dual-frequency Galileo (frequency E1 and frequency E5a). Put simply, if you run down the middle of the road today with your 945 then it’s perfectly acceptable for the watch to say you are running on the pavement on either side – that’s true of EVERY OTHER sports watch.
Credit: @Mirko, and others
Of course, the reality of any given situation is more nuanced and complicated than simply what piece of tech is on your wrist…try this link to see what else can go wrong.
As of 23 May 2019 my recommendation for you is to just use GPS but consider enabling GALILEO+hope where there are tall buildings and trees. @DCR notes that Garmin’s focus appears to be GALILEO right now, which is a roundabout admission from Garmin that it doesn’t currently work as well as it should.
It’s UNLIKELY that the 945’s current GNSS technology will ever deliver accuracy & precision in larger cities from a satellite signal alone. Buy a footpod like STRYD if you want more accurate running pace and distance measurement
I would contend that: GPS+GALILEO generally seems to make no overall difference or perhaps some slight, positive difference – but then sometimes it’s notably off.
GPS + GALILEO
Fiest up is cycling in the suburbs.
This was generally fine and I’ll highlight some exceptions. Here the 945 gets a nice track through the tunnel where there is no satellite reception (!) but to the left, it is too inaccurate, straying 10m from the true location.
In more open conditions or near these low houses (right of image) the 945 is good. However, on the left and nearer to larger buildings, the 945 soon becomes more than 5m inaccurate like the other 2 devices.
Around some low trees and at low speed the 945 is good
Although these navigations around a circle/roundabout look bad, I guess all are just about in an acceptable 5m off from where I really was…
Same here
Neither the 945 nor the Bryton are that great here in relatively benign conditions
Summary Cycling with GPS+GALILEO – it looks alright. It could/should be a bit better. I don’t have too many data points and am quite happy if others say they have better results. It seems pointless me doing too much with Galileo, as dcr says, Garmin are doing most of their tinkering with Galileo so whatever I find will soon be superceded by new Garmin firmware
Next up we have some GPS+Galileo-based running in the park
Here it’s not too bad and you can see that some nearby trees throw the Forerunner 245 off-course, whilst the 945 gives the smoother of the other two tracks – compared to the budget Ticwatch S2.
Then, later, this is pretty good in easy conditions (245 is not good)
And this is pretty good in easy conditions on part of the Bushy Park parkrun course
With some trees and slightly taller buildings, we get this not-so-great performance from each watch. Very strange.
Yet on a different day in similar conditions, different results were obtained and the 945 was quite good.
I put the 945+GALILEO through my formal test, albeit on the wrong arm (so the results don’t count) and it performed with a score of 69%. Compare this to the 935 which got 71% (MediaTek chip+GALILEO). Yet the 935 scored higher with GPS+GLONASS (79%+73%) and GPS-only (73%).
Summary Running with GPS+GALILEO – it seems passable. Generally in good conditions the results are good enough, it’s just that GPS+GALILEO handles tricky reception conditions variably – sometimes good and sometimes not-so-good. I don’t have too many data points and am quite happy if others say they have better results. It seems pointless me doing too much with Galileo, as dcr says, Garmin are doing most of their tinkering with Galileo so whatever I find will change soon.
GPS- Only Accuracy
My testing with GPS-only has been more extensive with well over a few hundred miles of testing. I’ll do more over time and add bits in here as needed as the 945 will probably be one of my main ‘reference watches’ and so will be used quite a lot.
First to Riding
This was on a ride to Brighton on Ditchling Beacon which, despite awesome scenery, is one of South England’s more unpleasant rides when going uphill. The Bryton strayed a bit here under the trees on a gently winding road but both the 530 and 945 were good.
Then on to one of South England’s more pleasant and famous cycling hills, Box Hill. A very pleasantly smooth road to put the power down on for a few minutes. The Bryton struggled here quite a bit and the 530 struggled as well to find the road. The 945 was not perfect but certainly ‘quite good’ (ignore the funny circles at the top of the hill, I was waiting for someone)
Yet on a different day, there are different results even though the devices are the same. The Bryton was still off-track and, whilst the other two were generally good, it was the 945 that this time strayed 10m from the road (top, centre of image)
From the delights of Virginia Water in the previous image and, earlier, two of Englands pretty hills we move to Shepperton High Street, not too far from the OW swimming mini-mecca of Shepperton Lake. The Edge 530 is pretty solid here but the 945 drifts at the bottom, although to be fair it’s probably just about 5m from where it should be.
Running with GPS
Off to a flying start as I pass the Pheasantry Cafe in Bushy Park. I would have popped in for a coffee but the 945 was metaphorically on fire, putting down a good track under some trees. For a novelty, I even tried the 530 as a hand-held (blue)
I might use this bridge over a train line a little more as it seems to stress some devices even when there are clear skies. Here the 945 did well and the 245, errr, didn’t.
Then for some hill reps in Richmond Park close to where, today, Olympian Stuart Hayes was knocked off his bike by a deer. It all happens where I live (and best wishes to super-nice-guy Stuart for a speedy recovery). This image is still at quite a high level and all the 3 devices have moments of madness as well as moments of glory. the 245 was noticeably the worst and I ‘d say the 945 was reasonable-to-good in these fairly challenging conditions.
Then, just when you thought you knew how to predict the performances, you get this where the 245 nails a fairly hard section with very high brick walls and tree cover and the 945 wants to go swimming.
Further on, in the same run, the 245 again nails a hard looped section under a bridge whereas the 945 doesn’t look so good but actually performs reasonably well if the difficulty is taken into account and if the tendency for ‘going for a swim’ ignored. #Sigh
Finally, my proper running test with the 945 returned an ‘incomplete‘. The watch froze after about 11km (second time ever with a watch on the test). The test is physically non-trivial for me to fit in with my training so it is back in the queue for the 945. Luckily for the 945 it will probably get a few more re-tests as well and I will probably focus those on GPS+GALILEO for reasons outlined earlier (ie as Garmin improve the algorithm)
Open Water – GPS
To be added June 2019
Running Pace – GPS
It’s easy enough for anyone to wear a spare watch, go for a run and then look at the maps afterwards. I try to ‘test’ one watch at a time and do endeavour to use it to periodically check my instant running pace – in fact, just as you would do if it were you only watch 😉
The accelerometer/algorithms on the 945*DO* seem to give a generally good and stable instant pace when running in reasonably straight lines and in reasonably good ‘open sky’ conditions. But at other times there are notable discrepancies of 20-30 secs/km at any one time.
ULTRA-TRAC Mode
This is a mode that has nothing to do with accuracy and everything to do with battery-saving. The 945 has a great battery life and 99.5% of you will never need this mode. If you use it by accident it will give a ‘vague impression’ of roughly where you have been and you wouldn’t really have any idea of the exact route you took unless there was only one possible path within, say, 50m.
It’s only JUST better than nothing, IMHO.
Summary
Despite quite a lot of time with the 945, I don’t feel in a position to more definitive that GPS is better than GPS+GLONASS but that would be where I ‘d place a bet right now.
As of 3Jul2019 I would now say that the GPS-only performance on the 945 is on par with the 935. ie quite good but not excellent.
I understand that Garmin ARE STILL WORKING TO improve 945+GALILEO accuracy but I do NOT expect it will be a massive improvement and I DO NOT think it will end up being as good as the GPS-only V800/AMBIT3. Let’s wait and see…I hope I’m wrong.
Once again the sad, tentative conclusion is that if you want a pretty picture of where you are going then all is good. If you want accurate running paces and accurate distances then you need a footpod. I think I’ve said that before, somewhere.
Return to the main Garmin 945 Review
Interesting article, cheers.
ps. the images don’t all load in Chrome, but do in IE, all be it quite slowly. (just for info)
thank you Will
I have a CDN problem apprently. on it now
Which footpod do you recommend for the 945? The Garmin one? I only want accurate current pace
STRYD
https://the5krunner.com/2019/01/16/stryd-review-2019/
Is the current restriction to a single band for Galileo a hardware issue or firmware issue i.e., could it be enabled in a future firmware update?
hardware
I believe it is impossible on the current Sony chip
More than what tracks look like, the killer metric to me is laps, do you remember how the 945 behaved compared to the 935 in that respect ? The F5 I briefly had was all over the place (no consistency, sometimes it beeped early, sometimes late)compared to my old FR610 and FR235, unlike the FR935.
I did a run today with the old FR610 and the FR945 with gps+galileo. Maybe today for the first time the FR610 has a better track. The FR945 is quite good in openfield, but near buildings and near mountains it seems to struggle a little more than ghe FR610. Maybe we really need the dual frequency gps to improve the performance of the watches near mountains and near buildings.
https://www.mygpsfiles.com/app/#nem2h5Py orange FR610 blue FR945
Were you monitoring for when Laps (say 1km) buzz on each watch ? That’s really the best way I’ve found to check for accuracy, at least the one I need as a runner.
Next week I’ll do a run in a cycle path with km mark to check the accuracy in the way you describe. About the run of today: for the first few km the FR610 and the FR945 beeped about in the exact moment. Then the FR610 beeped about two – three seconds earlier and this difference was growing slowly with the distance. I think that the FR945 is more accurate because usually the FR610 is always a bit long. Overall a good performance, because I was always happy with accuracy of the FR610.
A friend of mine who was running with me had a FR735 and his watch beeped some seconds later than the FR945, and the difference was growing with the distance.
My impression is that the more accurate of the three was the FR945 or the FR735.
The FR945 was with GPS+Galileo and 3D speed and 3D distance on. The FR610 and FR735 were with only GPS satellites.
About instant pace, for the most part of the run the FR945 agreed with instant pace of my Runscribe pod. But I experienced the same of TFK, in some part of the runs instant pace of the FR945 was off about 20-30 seconds for 200-300 meters and then came back to the exact instant pace.
At the end, a good performance of the FR945 but in difficult moments (near buildings or mountains that reflect the signal) we really need dual frequency to avoid multipath errors. Unfortunately using more satellite constellations seem not so helpful. Probably it’s better to use only one constellation (gps or galileo) but with dual frequency.
I have also to say that today I was not looking so much at instant pace, because I was looking at my CIQ Runscribe field on the FR945 and I was monitoring my flight ratio imbalance, and I was not thinking to much at instant pace. Only after the run I looked at the tracks on Garmin Connect and I was a bit surprised because in the other two runs the FR945 was much better (for example in the cycle path it was exactly on the path, today often is some meters off). Maybe today was more difficult for gps due to the weather (I think that the name is tropospheric errors? I don’t know).
Till the end of the run, the FR610 and FR945 beeped quite near in time. The FR735 of my friend beeped a bit more seconds later than the FR945.
On the other hand, in difficult moments for gps (for example building) the FR945 has the accelerometer that helps to keep an accurate distance. The FR610 is all alone, without the help of the accelerometer, and in bad moments for gps has bigger errors than the FR945 in recording the distance.
Thanks for all the additional info on your testing. That sounds really good for the 1km Laps. My FR610 and FR235 were pretty close but the F5 was all over the place, very inconsistent, disaster for a runner. I sent it back and got the FR935 that was consistent with the FR235 in the way you describe from your test.
The FR935 has an accelerometer too but I wasn’t aware that it used it to correct GPS distance, is that new on the FR945 ?
Maybe I dreamed about it because I don’t find it any more, but I believe that I read that in the Sony chipset the algorithm uses gps signals and accelerometer data to give the data to the processor of the watch. The accelerometer is inside the Sony chipset itself, so the firmware of the sony chipset take in account both gps and accelerometer data and there is a good integration.
Just speculation, but I believe that in the FR935 the accelerometer was outside the Mediatek chipset and maybe was the processor of the watch that was mixing gps data with accelerometer data with his firmware. It’s difficult to find information about this, but for the users it’s only important to verify the watch performance.
At the end, probably there is a good performance of the mix accelerometer+gps in the FR945 maybe thanks to the Sony chipset..
The FR610 doesn’t have an accelerometer inside, for example without a footpod it can’t give a cadence.
About the FR935 I can’t help because I didn’t have one. I had a FR35 and running with it I had the feeling that it was using a mix of gps and accelerometer data (because sometimes it buzzed earlier, sometimes later), but it’s just speculation.
hmm not sure.
the garmin is supposed to use accelerometer data + GPS to modify pace. If that is in the sony chip i do not know.
one strategy that garmin et al use to save power is to have multifunction chips, so its very likely the accelerometer is combined with “somthing else”.
my guess would be that it is NOT in the Sony chip.
I found the article: “https://www.sony.net/SonyInfo/News/Press/201302/13-022E/index.html”
Extract:”2. Equipped with Sensor Fusion functionality, realizing seamless and high-precision position measurement for both indoor and outdoor location
The Sensor Fusion feature combines information received from multiple smartphone sensors such as accelerometers, magnetic sensors and gyro sensors to perform arithmetic processing that achieves highly accurate measurement of positional information. In conventional systems, the host CPU is required to host each specific sensor driver. However, the new models are equipped with a sensor control function that enables mobile product manufacturers to easily incorporate Sensor Fusion functionality, as a sensor driver is no longer required for the host CPU.”
There is also a beautiful picture that explains the new architecture.
I didn’t understand very much of this article, but probably it’s a good thing 🙂 .
lots of bits that were all separate are now in one power-saving unit.
looks like you were probably right. tho it does not necessarily mean that garmin will use all those sensors or maybe they can be individually controlled…IDK
thanks for the info, interesting
yes i OFTEN do that. problem is when one goes out of sync for a one-off event (like handling a tnnel). but it is a good sign if they all beep together.interestingly 245+945 often beep together but 945 track is usually noticably better
Bike GPS short accuracy test of 2 km in easy track for GPS (cycle path in openfield) with the watch fixed to handlebar. Perfect track in Garmin Connect and watch beeped exactly at km marks.
Also my FR610 with bike gave perfect tracks.
Next week I will try a short test running in the same cycle path to see the effect of the swing of the arm while running.
even the cheaper watches have a good attempt at open sky conditions
that’s why i have my test methodology and route that takes into account different GPS hazards. it’s both how the watch handles the hazard and then how it recovers afterwards.
I did today a test run in the same bicycle path with km marks. In open field the FR610 and and the FR945 beeped at the exact km mark and were very consistent between each other (I was looking at the distance data field in the two watches). In the path there is a small pedestrian underpass that goes under a road: when I went under it, the FR945 lost 20 meters. Then I went ahead and this 20 meters difference between the two watches remained constant.
Then I did a U inversion and I came back: under the same underpass the FR945 lost again other 20 meters. Later I looked at the tracks in Garmin Connect and both watches have good graphs, so I don’t know why the FR945 lost so many meters. In theory, the FR945 should perform better in the pedestrian underpass because the accelerometer helps him. In reality the FR610 remained aligned with km marks on the bike path, even if it didn’t have the help of the accelerometer help. The tunnel under the road was very short, about 10 meters, with a small downhill and a small uphill.
It seems to me that the FR945 is very good in openfield. In difficult condition, it goes easy in trouble. The FR945 was in GPS+Galileo and 3 D speed and 3D distance on. I don’t know if Garmin will improve the performance of FR945 in difficult conditions with firmware updates. Probably we really need dual frequency to help in difficult conditions.
Overall, another good performance of the FR945, and with the exception of the tunnel at least as good as the FR610 and beeped about in the same time.
I had the FR610 in the left wrist and the FR945 in the right wrist.
yep.
it is possible to interrogate the FIT file to see what sensor was used each second. well, at least I was looking at 945 with hrm and surprised at how often it seemed to revert back to oHR (battery on my HRM-TRI was low)
maybe same thing is possible with distance from gps or accelerometer? i might look one day
These are the tracks http://www.mygpsfiles.com/app/#pe1XyK6S
the tunnel is the top of the track where the bicycle path goes under the road
but would the 945 sans galileo have been the same?
I think so, because from time 20:18 to time 20:50 I run in the middle of buildings and you can see that the FR945 has the same deviation from road as the FR610. The second constellation of satellites didn’t help the FR945 , because the building are too near to each other (the road is just 4 meters wide in that point). Another problem is the mountain on the right that block the sight of satellites in the east side). For about 30 seconds the two watchs were without any GNSS signal. I believe that was the accelerometer inside that helped the FR945 to keep a good instant pace and a good recording of distance even without a satellite signal. The Galileo satellites could give no help because I believe that for some moments the FR945 and the FR610 lost sight of all satellites. The accelerometer is build inside the Sony chipset (it’s Sony in his website that declares this), and maybe for this reason the accelerometer does a good job, because maybe the sony chipset gives with his firmware a position fix using both gps signals and accelerometer data and without the need of any input of data from the other sensors of the watch.
The rest of the run was in open field and I believe that the American GPS satellites were enough to give a good position fix.
But in the future I will continue to use the combination GPS+Galileo, because in some other difficult moments the presence of more satellites can help, and I believe that in the future it will become the best option, so I will not use American GPS alone in the next runs.
I thought that DCR had written that Garmin had focused on GPS+GLONASS and recommended using that setup? I have seen good results with GPS+GLONASS on a Marq (better than my 5x)
I have written that the 945 is not that great, nor the 935 before it.
I don’t see the “mind-bogglingly good results” other than in clear skies, when other devices can do the same thing.
here is what I do https://the5krunner.com/2016/11/05/test-route-for-gps-devices/ which broadly has always lined up with what fellrnr used to do arrving at similar conclusions by a totally different method. I have zero links with Garmin.
So personally would you just go with GPS only, or suggest using one of the other constellations?
for what purpose?
yes I would use GPS-only myself. That’s what I do.
Maybe if I lived in a city I would use Galileo/GLONASS
Best pace & distance consistancy in open & tree cover.
I typically have 9-14 available GPS satellites on a typical day.
Glonass gets 9-12 more.
I couldn’t find how many this particular chipset can “sync” up to at once. I know 4 gives you the ready although I always allow 3-4 minute lock time during warmup.
it aint how many you’ve got it’s what you do with them (or some similar quote)
https://the5krunner.com/2016/11/05/test-route-for-gps-devices/
I’m just looking for best pace and distance consistency on a regular basis and was curious since I just got mine. Living in Georgia GPS has the most visible always and I used to lean that way with my 935 but every watch is different especially when they change the hardware. This one has so much more going on in the background compared to the 935.
STRYD gives best pace and distance
But without the stryd?
You’re obviously avoiding the question. Guess I’ll just have to run my own testing.
as i said: https://the5krunner.com/2016/11/05/test-route-for-gps-devices/ this has all the results of my formal tests, including the breakdown to easy/med/hard gps conditions
I thought only my 945 isn’t accurate, but it seems like when you zoom in, the GPS are often 10+ meters off trail/road. It’s pretty disappointing. I hope they keep working on the accuracy.
yes, well spotted.
I specifically look for inaccuracy above 5m. that would then be a no-no unless there were mitigating circumstances eg next to a building
it’s pretty bizarre that accuracy is not more highly sought after by garmin…at least it seems that way. I’m sure they would deliver super-accuracy if they were able (probably would come down to a hardware re-design with better and bigger aerials…then there’s less space for the battery….
GPS tracking, this is exactly why I ‘m still using the V800. New watches are cramming in so many superfluous features at the expense of accurate gps tracking and distance.
Why should we be paying hundreds of pounds for watches that can’t match older watches in the essential metrics for runners. It’s crazy that if you want accuracy you have to add another £200 for a footpod to correct pace and distance.
Until a company actually makes a real running watch again, by dumping the stupid features that aren’t needed for serious running, then the V800 is staying on my wrist.
A powerful gps chipset with basic metrics is all you need!!!
i tend to agree
I have aked squillions of times “How many running watches can tell you how fast you are running now”….answer” none, at least not reliably”
unfortunately, people buy because of all the features they’ll never use, so i can’t see anything changing any time soon. (Actually I have some vague hopes for Polar’s next iteration)
wow… I hope that come off, i’ll keep an eye out for any news from Polar.
Purely from a cost point of view, I’d still like to se a running version of a watch. Paying full price for a multi-sport watch when you aren’t going to use virtually any the features seems wasteful.
Sunnto did it with the Ambit 3, so I don’t see why it can’t be done again…. and if someone does that, why not put a better gps chipset in while their at it.
i think you might see a decently-featured running-only watch very soon (or at leat focussed on that). maybe two.
but i don’t hold up hopes for V800 levels of gps for either.
I really have given up hoping for that! #stryd
Mmmm… so disappointing.
Aw well, may buy another V800 for backup.
I dug my V800 out last week as it happens. Managed a few OWS swims. GPS wasn’t too great there (quite bad in fact). What are your experiences of swimming accuracy with the V800?
I only run, short road races and trail running before I got bad knees earlier this year. Once I fix that problem I’ll be back on the trails. So swimming isn’t an issue for me.
That’s why I was hoping for… well, really just an updated V800, I’ve had Garmin and Suunto, just too many faults with them, sold them after a few months.
Admittedly, I like to see a tidy gps track, some aren’t bothered, but for the the best option (and cost) is a second-hand V800 and the Stryd footpod. Great running gps tracks and accurate metrics for around £300.
It’s just missing a few new training features I’ve liked on the newer watches I’ve tried, especially the HR ring on the Suunto I really liked that, even though I know my HR and pace zones, I was impressed with that. Shame you can’t see the screen while running and it couldn’t calibrate Stryd.
What is that ominous running-focused sports watch you speak of? I’m done with my VV, and my 935 is getting long in the tooth but still rocking (battery lasting for like 10hrs of on-off GPS, and OHR sensor is cracked beyond recognition but is still spot on!)
I use a Stryd but won’t pay for a 945 if GPS is worse than my 935 – that Stryd won’t work forever.
July…
Useless