The iPhone’s Positioning Sensors Were Never Good
Much is being made of Gizmodo’s tests showing that the positioning sensors in the iPhone 5s are off. Not just a little off, but off in a non-trivial way. The gyroscope read 3 degrees off, the compass 8 to 10 degrees off, and even the accelerometer seemed to be inaccurate.
There was only one problem with Gizmodo’s experiment: they only compared the iPhone 5s against the previous iPhone 5. That may have seemed reasonable at the time, but it assumes that the iPhone 5’s positioning sensors were accurate. Testing by TechHive in a variety of locations with an iPhone 4S, 5, 5c, and 5s now shows that the iPhone’s positioning sensors have never been any good.
While TechHive didn’t find anything wrong with the iPhone’s leveling capabilities, none of the iPhone compasses matched up — to themselves, to each other, or to an inexpensive Suunto A-10 recreation compass. Some were off as much as 20 degrees, and the worst deviation came in three different iPhone 5 units. TechHive also tested the compass of the Android-powered LG G2 smartphone and found that it was the closest to the Suunto, off by only 3 to 4 degrees. (The question this result raises is if the Suunto was itself accurate; a single cheap magnetic compass might not have been the best control.)
While Apple could, and should, make the iPhone positioning sensors more accurate and consistent, the moral of the story is to not rely on smartphone sensors for critical tasks. As our own Rich Mogull said during our staff discussion, “As a mountain rescue guy, digital compasses make me nervous. I have enough trouble keeping a physical compass calibrated and accurate. You walk out of an office building in a city near power lines, and no compass will be accurate. It’s just physics. Indoors? Not a chance.”
It’s also worth noting that despite the fuss surrounding this story, we’re not hearing from users about losing an orienteering race due to incorrect compass readings (iPhones wouldn’t be allowed anyway), having a woodworking project be tippy because of issues with the iPhone’s level, or even having trouble playing accelerometer-based games. In short, despite the proven problems, the iPhone’s positioning sensors still work sufficiently well for the uses that most people demand of them.
Even "physical" (analog?) compasses have issues. As a Surveyor who has on occasion used compasses with 6 inch long needles (nominally accurate to the quarter degree), I discovered any ferrous metal in close proximity to the compasses would cause inaccurancies. Even rocks with iron in them can pull the compass off. Reliable testing of compasses should be done under laboratory conditions.
I'm kindof curious as to the engineering problems of getting a magnetic heading within a device packed with devices that might be emanating electromagnetic fields..
The answer is to set up an observation place that is clean of interfering magnetic fields. If you can see the pole star or some well known landmarks from there you can check how a known compass behaves: rent one if you have to while buying local maps.
Then compare the performance of any iPhone that has been accused of misbehaving. Ships compasses were often calibrated by directly comparing a transit: Eg. between Alcatraz & one of the Golden Gate Towers — with the compass reading. Iron balls and bar magnets were bolted in place to compensate for the magnetism of the surrounding ship. Still must be for a lot of ships.
If you tell me that is not so anymore I shall stop going out on ships and probably aircraft, too ! !
The fuss over the iPhone 5s sensors is certainly justified for some of us. Granted, the iPhone should not be used as a compass for critical navigation tasks. Although it should be noted that some of the complaints posted to the web are based on poor testing methods and/or ignorance of how compasses are used.
More significant are the widespread complaints about the orientation sensors (accelerometer) in the iPhone 5s. Some apps rely on these sensors to make precise determinations of the phone's absolute attitude in space: inclinometers and other measuring tools, camera apps, astronomy apps, enhanced reality displays, flight simulators and other games. Users are discovering that these apps are producing significant errors when run on iPhone 5s, errors that they do not see when the apps are run on previous iPhone models. I posted some of my own test results in this photo gallery:
http://julianv.zenfolio.com/iphone5s
It's not immediately obvious to me why a cheap magnetic compass would be any less accurate than an expensive one: they're both just a magnet on a swivel.
When I got my first iPhone and tried to use the Pocket Universe star mapping app, I soon concluded that iPhone sensors gave rough approximations to the truth at best. Nothing has happened since to change my mind.
There's an interesting wireless iPhone peripheral called the NODE that's packed with sensors that might be more reliable: variableinc.com.
Little of it is obvious, Rob. The cheap compass may well have a cheap pivot that gets damaged easily and drags on the pivoting: causing random, erratic, misreadings.
Cars and buildings are full of magnetically influenced iron. I'm surprised that some readings were only 20º off. It becomes obvious with practice and experience.
Thanks for noting the NODE — I'll look for it.
"we’re not hearing from users about ... having trouble playing accelerometer-based games. In short, despite the proven problems, the iPhone’s positioning sensors still work sufficiently well for the uses that most people demand of them."
I would like to point you to the forums at macrumors.com. There are 79 pages of comments (1960 messages up til now) of people complaining about the problems the 5s is causing to their gaming and experience with astronomy apps, levels, clinometers, etc. Most of these are from people that still have an older phone and they can corroborate the difference. Their older phones are perfectly usable, within the expected range of a phone with this kind of app.
In my very limited, non-scientific tests with other older phones, every 5s including my own is off by 2 or more degrees, when compared to other 4s and 5, and to a real level (while the 4s and 5 are pretty accurate.)
Various app programmers have already come up with software solutions for their apps, although not a systemwide one- only Apple could do that- and there are high hopes that a coming iOS update could resolve this issue. Wait and see mode...
Thanks for the pointer - when I said that we're not hearing from people, I meant it. WE are not hearing from TidBITS readers about this - I hadn't fielded a single question or comment from a reader about the problem before we published the article.
That's only loosely indicative, of course, but still, it's not like I didn't get a lot of complaints about the visual aspects iOS 7, for instance.
I read five or ten pages in the MacRumors thread, and very few of the posts actually say that the person is having problems. It's a whole lot of people who are generally concerned, or who are verifying the experiments with their own devices, or reporting on their efforts to get replacements. Only 1 or 2 messages of all the ones I read said anything about actually experiencing a problem in normal usage.
Again, I'm not suggesting Apple shouldn't address this, just that it seems to be causing more online fuss than actual user problems.
Thanks for the response. The people reporting actual usage problems increased as days went by. I do not pretend you read all the pages, you are extremely busy keeping this great site!! :-) (which I've enjoy a lot, since the beginnings)
The point is mostly moot now, because 7.0.3 has resolved the issue for almost everyone, including me.
As per Apple's notes for 7.0.3: "Fixes an accelerometer calibration issue".
I saw that release note - so you're not seeing the same issues as you were before? That's great to hear.
My 5s had the problem. The level was clearly off. I hoped to use it for simple around the house tasks. After the 7.0.3 update it is much better, but still not perfect compared to a physical level. The analysis that the sensors never were very good and still aren't seems about right.
I've been using Pedometer+++ to track step count and I'm not convinced I can trust the accuracy. It reads the M7 data so any inaccuracy is the phone's fault, not the app's. I've had times the step count just seems too high, as if it's adding steps when I'm driving. This was observed after I applied the update, but not in a scientific way. I haven't cared enough yet to do a thorough controlled test, but I can't help but wonder if the phone's sensors just aren't reporting accurately.
As far as ships go, when I was a submarine Quartermaster 20 years ago (in the Navy Quartermasters do navigation) we used gyroscopes for direction, not a magnetic compass. Very sophisticated, very big gyroscopes.