This is such brilliant stuff, thanks for doing this testing. I'm probably crazy, but this is actually very reassuring for how I'm using my pile of BN-180 and BN-220 units, because being within 5m is more than adequate for my uses, but knowing that the M8P would be worth it for position hold navigation is exactly what I'd want to run.
I love how much detailed info you're able to cram into this, and yet still make it so easy to follow and understand! lol I've used several of these GPS modules for my flying over the years and never really noticed much in the way of accuracy or performance differences, other than some of them lock much quicker than others at start up. lol That said, this makes it clear to me that uBlox is probably the way to go, especially the M8N seems to be a good performer. I don't think I've seen a module targeted for FPV with the M8P, or maybe I wasn't paying attention if I did. ;) Excellent job!
Yeah for flying planes I think any of them will be fine. I use the BN220 almost exclusively on mine and haven't noticed any shortcomings even though it didn't look that great in this test. The u-blox don't cost much more though, so if you do want a bit more accuracy that's the way to go. The M8P is made for RTK usage so it's helpful for surveying or aerial mapping photography. It's not really a peer of the other ones, I just happened to have and thought it might make an interesting comparison.
Brilliant exploring and visuals and great data set and explanations. Pretty sure the best I’ve seen. I wouldn’t say mistakes made, just the raw process of learning exposed. It’s courageous of you to put out videos that are impossible to edit once published. (youtube is an interesting and somewhat scary medium to share original content) At 13:35 you say separating into “two groups”, I see 3 distinct groups. The heat-map plots at 16:44 and 15:30 really highlight the accuracy of data collected from individual units. • Group1 { MSP, M8N, 7M } … most consistent position data • Group2 { BN180, BN880, BN880Q } • Group3 { BN220, 6M } … least consistent position data A merged overlay for each of the 3 groups would really show similarity of GPS units in each group, and contrast the 3 groups (or levels of accuracy, consistency). For “true GPS” reference position, you might want to do weighted average of the 3 groups, something like a 3, 2, 1 weight, or 2, 1, 1 weighting. This weighted average “true GPS” reference position may better highlight the inaccuracies of individual units. Weighted average would help reduce the “smudging” effects you mention at 28:57 from the less accurate GPS units. BTW: At 27:50 … Are you sure it’s the 7M that’s off, not the 6M? Like you said, I’m expecting the 6M to be one doesn’t belong in this group (Group1 as I noted above). See 7M vs 6M at 14:40 … that seems to contradict the plot(s) at 27:50. I’m a bit sad as just received 3x BN220 units that I ordered 3-4 weeks ago. Have to thank you in advance, as your results will save me some future head aches. Also thinking to set up a similar experiment to test my 3 BN220 modules for my region. Likely a winter project, as I’ll just use the units for now, having better insights. Many thanks for sharing your deep dive into GPS data.
Glad my comment was useful! Your comparison is now by far the best I’ve seen, loving all the visualizations. But again the poor performance of the BN-880 is surprising to me along with the decent performance of the 7M. Maybe my 7M is a fake and just a 6M? The decent performance of the BN-880Q just makes it more confusing. I guess I’ll have to look closer at my past data again. And yes, my username is a Monty Python reference as a result of my name being Tim Chant and thus I’m often called the 'electronics enchanting wizard'.
A very comprehensive and well explained video. Thanks for efforts of finding breaks in the weather. Being in Melbourne (AUST) I understand your pain this time of year. Have you thought of doing the test against an RTK setup to show the advantages of an RTK? I realise that this is quite expensive, although not as bad as it used to be. Just a thought as it would provide a good control signal. Thanks for your videos.
I actually had an RTK setup provided to me a few years ago, that's what the M8P is part of. But there's not really any point making a heatmap of that because it would barely move at all. It might make a good demonstration of how it can return exactly to the same location, but I kinda already did that here: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-n8PUyOtiGKo.html Still, I'll consider dusting it off and using it if I do any more tests.
19:37 - Dang! You could use the yellow/red dots to (seems like a matched pair in terms of resolution, sample-rate, lag, etc.) to also determine the absolute heading of a thing without relying on compasses
Couple of observations. You can do a contour histogram of position. Or if you are just looking at one of height, lat, long, you can do a histogram which shows the spread and any offsets nicely. ie. A vertical alignment of histograms, then the x-axis as a consistent range between all. The other interesting thing would be to check to see if there is any correlation between the different devices when it comes to errors. Is there a relationship where positions tend to move in the same direction. Otherwise interesting.
If you could add support for non UBX mode GPS modules, this could be a real series. There are so many different GPS chipsets out there so it would be very interesting to see a comparison. The other problem is trying to have many UARTs available to simultaneously test multiple modules. Thanks!
3 года назад
Right now I'm watching a lot of your older video series again, like the Farm Rover or the ardupilot experiments. I am right to say that if you did this again, would you get significantly better results after discovering this accuracy problem? Thank you very much for sharing your experiences!
One may think about the wondering or rotation of the sats in space while looking from really far away on to a point on earth?? then the ones that are wondering more probably have lock down to the farthest sats.... I do however am impressed by the M8N I mean unless that particular antenna is locked down in to the closest possible satellites "Very much directly in top?? but no because the test give you a whole 10 hours meaning that GPS antenna changed satellites at least 130 times... great work, probably not a subject for every one !!
I wonder if different gps devices use different Kalman filters. I think some units try to detect if they are stationary or moving and use different filters based on that. So, while moving, there can more or less different results. Switching to 10Hz should improve accuracy, especially when moving. (Most modules calculate and filter position at the output Hz rate. More Hz means more data and seems to give better accuracy to points at least when moving). I think at least some older Ublox modules had different modes for stationary, moving (with low change in elevation like car) and flying (where there can be a lot of height change).
I wonder if I could ask you for some help? I have two F2s Flight controllers both of which lost their GPS pucks in crashes last year. I've tried all over but cannot get replacements unless I get one's similar that need various changes making which I don't understand about! Can I email you about this please? Best regards Nick.
Your test data and visualization are impressive! I did have one thought to pass on. I noticed you seem to have taped some of the modules down. I'm no RF engineer, but I've always wondered about the effect that dielectric plastic coverings might have on the ceramic antennas used in these GPS modules. Would the extra capacitance the tape adds shift the center frequency of these antennas enough to matter? If so, the GPS antennas which come with tight fitting cases would presumably need be tuned to account for this effect when they are manufactured. I've always added heatshrink to bare GPS modules and hoped it wouldn't matter. Your test methods might actually be sensitive enough to answer this question in a practical way by testing identical modules with and without cases or plastic tape, then switching and testing again to control for sample differences.
I don't think the tape (or heatshrink) would change the results even slightly. If anything, when it was raining there was a chance that a raindrop might sit on the antenna and perhaps have more of an effect. These antennas are often housed in hard plastic casings, as are just about any 'rubber ducky' or whip style antenna.
Just need to mention that almost all the modules you use have a ceramic antenna patch that is usually connected to a single chip amplifier directly on the feed into the board. This basically makes "active antenna" moot.
Great and comprehensive comparison! Would be awesome to see a comparison between the newer Beitian models (BK with M9 and BE with M10) and maybe also the more expensive Matek SAM-M10Q. Did you get to use any of those?
I have a BN-452 that uses M9, with quite large patch. I found it to be slightly better than the M8P. It's the only module I've had that was able to get a fix from cold, indoors, right out of the box and never having seen any NZ sky. ru-vid.comUgkxDidLT7TTL7J_DlTj59Cnrux9vBUTediO
Thanks, didn't see the post. Just ordered a discounted BK-880Q as my very first receiver, will have to see how it goes. I'm curious what FW version it has and if it's updatable.
After all your experiments and your new experiences, which module do you recommend buying? (precision wise, non RTK) Have you tried to use two or more modules and average them?
For most usages you might as well get the cheaper BN220, especially for a plane which will not benefit from super precise numbers anyway. For a ground vehicle I would probably want a bit larger antenna patch, maybe M8N is a good choice. If you don't have a single use case in mind the M8N is probably the best all rounder, but there's not much in it. Just avoid the BN180 because it doesn't save any config settings on power-off, which can be really annoying.
Hello, Nrf and neo6m gps (tinygps library) work, but Nrf and neom8n gps (tinygps+ library) the code does not fail and loads but does not work. What do you think might be the reason. I would be very happy if you could help me. I wish you healthy days.
Super interesting thanks! Can you do a comparison with an M8P utilising RTK? Your previous video on RTK showed an improvement but it'd be great to see this detailed analysis.
Ah I didn't see this video ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-n8PUyOtiGKo.html only saw the landing comparisons, which had the extra error from the flight controllers landing accuracy.
I don't think there would be any point because the RTK will not move more than a centimeter or so, which is basically the limit of the 1/10^7 degree resolution. It might be useful to use it as the 'true' value I suppose.
It'd be interesting to see C/N0 and pseudorange performance metrics between the different modules. Seeing pseudorange variation for the same PRN/Frequency between the various modules would be a good indicator of the tracking loop performance for the various modules. Some of performance issues on the lower performing modules could be from bad measurements for satellites (multipath, bad Iono/tropo, etc). Checking for multipath in the measurements would be interesting. There should be a few chips of variation between the multipath signals and the regular signals (on the same PRN different Freq), accompanied by a lower than average C/N0 if they're present.
AFAIK, only very expensive «professional» modules provide needed information. For example, products based on uBlox 9M «standard» module costs about 10 times less than ones based on 9P «high precision + RTK» modules (I can not google retail prices for modules itself from mobile). AFAIK, these modules differs only in firmware and additional pads, chipset are physically the same, but prices are very different.
@@blacklion79 Hmm I'm not as familiar with off the shelf modules, but I see in the Ublox-8 protocol specification they have the UBX-RXM-RAWX messge that contians the raw GMSS measurememts per PRN. Though I'm not sure if its in any of the earlier products since i haven't checked their specs, so probably couldn't do the comparison.
@@fureter6986 Problem is, is this message enabled on given firmware or not. I didn't study 8 an 9 series data sheets, but 6 and 7 series had this and other «high precision» messages enabled only for top-tier variants, which was much more expensive and virtually unobtainable in my country anyway.
@@fureter6986 These «professional» features are used for market segmentation, unfortunately. You could google receivers which supports data output needed for RTK post-processing, for example - they cost more than complete Garmin GPS navigators, where Garmin unit has screen, firmware, battery compartment, IP65+ protection, routing features and all these bells and whistles, and RTK-ready receiver is simple PCB in black plastic box with USB connector. And I'm speaking about receiver only, which needs PC for real RTK processing, not full-featured autonomous RTK solution! Ok, to be honest, I was at this market 2 years ago, maybe something changed in these 2 years...
The weather has been solidly awful for months and nanny state says I can't leave home to buy things. I was making a big CNC router and although smaller parts can arrive by mail I'm currently stuck waiting until local businesses are allowed to work again for some things (laser cut aluminium plate, bolts, steel tubing). The cost of the chip I need for sbusmixer increased from $2 to $32 and no stock to be found so I lost motivation to get that finished any time soon. On a rare non-rainy day I was able to test ardupilot and FPV on homeslice8, but the video is kinda bumpy and lighting is not good, also prop in the view is more annoying to watch than I expected. but fwiw here are some clips: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-qtMzMgLvR-Y.html Perhaps I could dig up some old footage and make a video somehow...
Fascinating stuff! I've learned a lot from your video's so thank-you for making them. A couple of things spring to mind though...did you really get the 6M to output data at 10Hz? I thought it was only capable of 5Hz max. I'm not doubting you but would like to know how you did that. The other thing about the 6M...isn't it GPS only? I don't know about the others but to be fair it does have a disadvantage at only getting GPS data, which kind of means it didn't really do too bad considering. I'm probably wrong on both accounts but whatever....thanks again for the fantastic info, a real gold mine.
Yeah I forgot to mention that. I had always thought the 6M could only do 5Hz but it seems to be doing 10Hz. I thought perhaps it gives messages at 10Hz but the values only change at 5Hz, or maybe the modules got mixed up, but they are all doing unique 10Hz values at some points, eg. www.iforce2d.net/tmp/6mDoes10Hz.jpg It's a Neo-6M-0-001, the data sheet says only 5Hz so I don't know what's going on.
@@iforce2d Fantastic! Thanks for the reply, great stuff! One other quick question if I may....do you level shift the RX pins on the modules? I think I've knackered one of my units by not doing so when sending it commands, just curious. Thanks again.
These modules will typically output 3.3v signals. I'm not using any shifting here since the STM32 is also 3.3v. In the past I have used 5v signals from arduino and not had any problems. The datasheet says max 3.6v for the GPS unit itself, so I guess it depends on whether the module it's mounted on includes any protection.
Now I know why my fixedwing always change circle point in loiter mode using 6m. But everytime my wireless PIR doorbell triggered it will cause my gps loses all satellites.
Good, but I miss to compare same with RTK fixed station, that should be on another level.. But need the long range wifi or other way to transmit this information to the modules tested.
From what I've seen the RTK just doesn't move at all, so not much point 'testing' it, unless maybe to use it as the 'true' value. ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-n8PUyOtiGKo.html But it's still only as true as the surveyed-in value.
@iforce2d I'm trying to configure my BN-180. I'm fiddling with the u-center program. Do you know if the SBAS option has any effect on the position accuracy? I live in Europe so I tried the EGNOS system but I don't see any difference. Did you try the SBAS functionality?
SBAS should be off for my settings, there is no useful system here and in fact sometimes it seems to briefly pick up the one from Japan which actually makes things worse. Keep in mind the BN180 has not onboard flash to record any of your u-center settings, it must be configured on every power up. They can be extremely frustrating and tbh I would be inclined to use something else.
@@iforce2d Yes, my SW configures the BN180 at start-up. I have fiddled with the U-center program and found out that in the default configuration the SBAS is already active. It scans many frequencies, but apparently (as I understand it) it doesn't pick up any satellite. Where I live I should be able to pick up the EGNOS satellites. And I also found out that in the default GNSS configuration, there are at least three constellations active. I wonder if the accuracy would be better if I chose only one constellation, GPS for example. Do you have any idea about that?
your weather is just like england, thanks for the correction video, If you can find the fence post on google earth can you tell which gps was closest ?
@@iforce2d didn't realise that, that's no good then. So I suppose the most accurate one would be the one that returns home closest to its start location.
Ah, a good ol' quantization bug. Now you know the pattern for the future at least (if that's a comfort) :P The modules have different accuracies at different refresh frequencies, you should probably see different patterns at 1Hz than at 10Hz.
This is why we tend to use an offset when working with geo data, the points you're using are seldom far apart, so that way you can have high precision, even when just using floats. Also, converting between projections (wish there was just one to rule them all - *sigh*) you loose precision, even when using doubles, since there's so many intermediate calculations.
You should try the ATGM336H module, I've seen some good written reviews about it, but there is so little documentation about it, it can track the Chinese constellation too, some people claiming that they sometimes caught 20 satellites with it.
I have one of those actually. I bought it to use on the high-altitude balloon since it's so small, but the lack of UBX output made it not so convenient. The sbusmixer thing is only planned to support UBX protocol.
@@iforce2d I guess for my purposes it would be fine then, I just need the UART communication. After some deep digging I found some commands you can send to it to change some parameters. But the module is so cute, I couldn't resist
@@iforce2d Is UBX supported by any non-uBlox-based devices? Are these BNxxx uBlox-based? I think, NEMEA is must to have in such «product». Yes, it is old and messy protocol, but it is universal one.
All the modules in the video output UBX. Yeah I guess you're right, supporting NMEA would open up access to just about every GPS out there. It's just so much less efficient compared to UBX.
hi chris, I'm still using your iforce2d editor, which is great but you're website has no valid ssl certificate. Maybe you can fix that. Thank you, great content and great editor.
The weather has been solidly awful for months and nanny state says I can't leave home to buy things. I was making a big CNC router and although smaller parts can arrive by mail I'm currently stuck waiting until local businesses are allowed to work again for some things (laser cut aluminium plate, bolts, steel tubing). The cost of the chip I need for sbusmixer increased from $2 to $32 and no stock to be found so I lost motivation to get that finished any time soon. On a rare non-rainy day I was able to test ardupilot and FPV on homeslice8, but the video is kinda bumpy and lighting is not good, also prop in the view is more annoying to watch than I expected. but fwiw here are some clips: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-qtMzMgLvR-Y.html Perhaps I could dig up some old footage and make a video somehow...
Floating point precision is a rake that every programmer steps on.. don`t get me started on FP equality comparison and denormals, those are equally "fun"
This is great, but I'm now wondering if you have two identical modules, and put them head to head. Will the error be similar or not? Also you might know this but there was an experiment where precision was increased using 2 GPS devices, and one was stationary. All the errors the stationary one had, were subtracted from the moving one. As far as I remember clouds do affect the GPS location, and satellites do move fast, using the stationary GPS you negate most errors, but you need to be in the proximity of the ground station. It's nice that the modules gained accuracy while moving, but bad if you want to have a ground station. I've noticed not all drawn lines ended up in the same point. But this is something you might be interested in testing. BTW, did all the GPS modules agree on the average location?
I might do a test with multiple BN220s since I have many of them. Yes, the 'experiment' you mention with a stationary and roving unit might be what is called RTK correction, eg. ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-n8PUyOtiGKo.html No, the average location for each module is shown as the purple cross in the heatmap images.
@@iforce2d So you've already done this. Great video on RTK and... RTK seams extremely nice, but expensive. If the error on BN220s is consistent over time, then using a simple "Differential GPS" (or substraction) to find the error this could be send to a drone or something. This could become a product, that would just connect to the remote control and send error to the receiver. I have zero hands on experience with this stuff, but if there are remotes that can connect with secondary slave remote, then a GPS module would be no problem. The idea is not to use extra radios to send data. Great stuff. Thanks.
But....would the paths naturally be off since the satellites themselves are in .......well, not in the same position(s)? To me, this is somewhat of a TDOA measuring the sat/ gps signals, where they'd been captured at ....different times/ of arrival? Now, if there'd been 1 gps satellite and all the gps modules linked to this 1 sat..... then, this would provide the comparison desired.... with the set up as is in the video, there are way too many..."moving parts".... In which case I demand that a video series be created where sending up a satellite of your own to perform this test and having the modules link to it so that the data can be without so many variables, documented and presented to us, the viewers- kidding- but, that'd be cool....and I know i'd watch that!
Well, all the modules are using mostly the same satellites. For example if one module is using 9 sats and another is using 17 sats, 9 of those 17 are probably the same 9 that the first one is using. The fact that some modules can use more sats than others (either because they support multiple constellations or simply because the antenna is better), is in itself part of measuring the quality of result. Typically about 6 or more satellites are required to get any reasonable fix, so you wouldn't get far with just one satellite, especially when it's on the other side of the Earth :)
