iPhone 5s accelerometer/gyroscope/compass values incorrect

[richard371]

The thing I find interesting is the drifting on units that are close to 0 after a few days. My understanding it the gyro/accelerometer chip are the same used in the previous iPhones. The only new chip is the m7 which only collects and processes info from the sensors and is not an actual sensor.

I'm staring to wonder if there is some sort of built in auto calibration that's way off and causes these results.


If it's hardware I can't imagine what could be causing these changes. Maybe they used some sort of new pcb material or solder that's warping a bit internally with heat etc.

I wonder if the units that are sifting are reset do they go back to 0?

 

[WilliamG]

The thing I find interesting is the drifting on units that are close to 0 after a few days. My understanding it the gyro/accelerometer chip are the same used in the previous iPhones. The only new chip is the m7 which only collects and processes info from the sensors and is not an actual sensor.

I'm staring to wonder if there is some sort of built in auto calibration that's way off and causes these results.


If it's hardware I can't imagine what could be causing these changes. Maybe they used some sort of new pcb material or solder that's warping a bit internally with heat etc.

I wonder if the units that are sifting are reset do they go back to 0?

Nope. They don't. And to be honest this "drifting" business seems a little fishy. I'll believe it when I see it...

My replacement from Apple should be here tomorrow, so we'll see what happens... -6 degrees was starting to drive me insane.

 

[Sodner]

It seems like the calibration process is off to me. You need rotate the phone so the red ball rolls around but it never seems to completely "fill in" the circle before the compass comes up.

Anyone else see that?

 

[gs28]

http://www.insideactivitytracking.c...ate-accelerometer-and-gyroscope-measurements/


"With the iPhone 5s, Apple changed producers for its accelerometer component. According to teardowns by iFixit and EDN, the iPhone 5 used the LIS331DLH 3-axis accelerometer and the L3G4200DH gyroscope, which are both made by STMicroelectronics. According to a Chipworks teardown, the iPhone 5s (see the M7 teardown) uses a gyroscope component from STMicroelectronics, but Apple switched to Bosch Sensortec for the accelerometer, using the BMA220 3-axis accelerometer. According to the companies’ websites, the old accelerometer was capable of higher data output than the new accelerometer (16 bit output compared to 6 bit digital resolution). The Bosch accelerometer also has more typical output noise and a higher Zero-g Offset than the STMicroelectronics component–two aspects also contributing to the lower level of accuracy in the iPhone 5s. This new accelerometer might be a source of users’ issues measuring levels."

 

[WilliamG]

It seems like the calibration process is off to me. You need rotate the phone so the red ball rolls around but it never seems to completely "fill in" the circle before the compass comes up.

Anyone else see that?

Yes, if you are slow. It is possible to fill the circle in if you do it quickly.

http://www.insideactivitytracking.c...ate-accelerometer-and-gyroscope-measurements/


"With the iPhone 5s, Apple changed producers for its accelerometer component. According to teardowns by iFixit and EDN, the iPhone 5 used the LIS331DLH 3-axis accelerometer and the L3G4200DH gyroscope, which are both made by STMicroelectronics. According to a Chipworks teardown, the iPhone 5s (see the M7 teardown) uses a gyroscope component from STMicroelectronics, but Apple switched to Bosch Sensortec for the accelerometer, using the BMA220 3-axis accelerometer. According to the companies’ websites, the old accelerometer was capable of higher data output than the new accelerometer (16 bit output compared to 6 bit digital resolution). The Bosch accelerometer also has more typical output noise and a higher Zero-g Offset than the STMicroelectronics component–two aspects also contributing to the lower level of accuracy in the iPhone 5s. This new accelerometer might be a source of users’ issues measuring levels."

Glad this is getting picked up more and more. Cheers for the link!

 

[jpcarro]

Nope. They don't. And to be honest this "drifting" business seems a little fishy. I'll believe it when I see it...

I agree, the drifting seems fishy and I waited to report mine for that reason. After seeing others posting about it, I posted my drifting as well.

It's only a 1 degree drift, it went from -1° to -2°. I would have hung on to my 5s with -1°, but now I want it fixed/replaced.

 

[WilliamG]

I agree, the drifting seems fishy and I waited to report mine for that reason. After seeing others posting about it, I posted my drifting as well.

It's only a 1 degree drift, it went from -1° to -2°. I would have hung on to my 5s with -1°, but now I want it fixed/replaced.

My friend's new 5s read -1 when I got it with him last week. We tested it immediately. I'll call him later and have him check it.

 

[Sodner]

http://www.insideactivitytracking.c...ate-accelerometer-and-gyroscope-measurements/


"With the iPhone 5s, Apple changed producers for its accelerometer component. According to teardowns by iFixit and EDN, the iPhone 5 used the LIS331DLH 3-axis accelerometer and the L3G4200DH gyroscope, which are both made by STMicroelectronics. According to a Chipworks teardown, the iPhone 5s (see the M7 teardown) uses a gyroscope component from STMicroelectronics, but Apple switched to Bosch Sensortec for the accelerometer, using the BMA220 3-axis accelerometer. According to the companies’ websites, the old accelerometer was capable of higher data output than the new accelerometer (16 bit output compared to 6 bit digital resolution). The Bosch accelerometer also has more typical output noise and a higher Zero-g Offset than the STMicroelectronics component–two aspects also contributing to the lower level of accuracy in the iPhone 5s. This new accelerometer might be a source of users’ issues measuring levels."

Interesting insight to a potential cause for the level issue but why does my 5s not know where north is?

 

[Ceebee1980]

Apple switched to a less-accurate accelerometer? That kinda pisses me off. Sure, it may not be the exact cause of this issue, and the difference in accuracy may be negligible by a factor of 10 or 1000 or whatever, but decreased accuracy doesn't seem like progress.

 

[madKIR]

http://www.insideactivitytracking.c...ate-accelerometer-and-gyroscope-measurements/


"With the iPhone 5s, Apple changed producers for its accelerometer component. According to teardowns by iFixit and EDN, the iPhone 5 used the LIS331DLH 3-axis accelerometer and the L3G4200DH gyroscope, which are both made by STMicroelectronics. According to a Chipworks teardown, the iPhone 5s (see the M7 teardown) uses a gyroscope component from STMicroelectronics, but Apple switched to Bosch Sensortec for the accelerometer, using the BMA220 3-axis accelerometer. According to the companies’ websites, the old accelerometer was capable of higher data output than the new accelerometer (16 bit output compared to 6 bit digital resolution). The Bosch accelerometer also has more typical output noise and a higher Zero-g Offset than the STMicroelectronics component–two aspects also contributing to the lower level of accuracy in the iPhone 5s. This new accelerometer might be a source of users’ issues measuring levels."

Now we all can kiss Tim's ass for degrading the quality of the iPhone!
First they degraded the vibration motor and noise canceling equipment in the 5, now this!
It's bad news! It means the fix might not arrive until next year.
Can you imagine how many of those chips are being shipped to the manufacturing facilities now?

 

[richard371]

I'm thinking the gyro may be calibrated based on the accelerometer as they both seem to be off by the same degree.
If the accelerometer is correct then the gyro will be. The accelerometer could be throwing off the compass calibration too.

It's a shame apple downgraded on a part that only cost a few dollars and now the phone performance in this respect is degraded.

 

[MLAM]

If the 5c doesn't have this issue then does it still use the old chip as per found in the now discontinued iPhone 5?

I was under the impression that the 5c and 5s would share as many components possible, if that is the case then it's a little surprising the 5c doesn't have this issue.

 

[benzn]

I just wanted to chime in that if you run some math on the data provided in the data-sheets of the old parts you get accuracy which is very close to what we're seeing.

The zero-g offset of the old device: +/-20mg (milli-earth gravities)
The zero-g offset of the new device: +/- 95mg

Now the angle of inclination of the phone is the arctan of any pair of axis, many of us have measured the inclination of the phone laying flat, in which case we're measuring the gravity from along the axis projected out along the back of the phone and the gravity in the direction out from the top, bottom, left, or right (it doesn't really matter which).

In the best case scenario, when flat, we read 1g-force out the back and zero out whichever side we're using. If you take atan(0/1), you get 0 degrees, meaning the device is flat.

In the worst cases, assuming the above points from the datasheet (if I'm understanding them to mean what I think they mean).

Old device worst case: (0 + 20mg) and (1g - 20mg)
New device worst case (0 + 95mg) and (1g - 95mg)

Old worst case inclination: atan(20/980) = 1.16 degrees
New worst case inclination: atan(95/905) = 5.99 degrees

So I think we've found our culprit. This is pretty disappointing, and clearly not going to be fixed by a software update without further calibration. I guess the best thing we can do is hope that Apple switches back to the older chip.

 

[gs28]

So what you are saying is that the new sensor it's just not accurate enough and there is nothing that can be done via a software calibration?

 

[benzn]

Not quite, I'd imagine there is some calibration that can be done, but it would require user intervention (like the compass) and I think it's unlikely that Apple would ship a software update that would require 5S users to selectively calibrate.

 

[H3m]

I just wanted to chime in that if you run some math on the data provided in the data-sheets of the old parts you get accuracy which is very close to what we're seeing.

The zero-g offset of the old device: +/-20mg (milli-earth gravities)
The zero-g offset of the new device: +/- 95mg

Now the angle of inclination of the phone is the arctan of any pair of axis, many of us have measured the inclination of the phone laying flat, in which case we're measuring the gravity from along the axis projected out along the back of the phone and the gravity in the direction out from the top, bottom, left, or right (it doesn't really matter which).

In the best case scenario, when flat, we read 1g-force out the back and zero out whichever side we're using. If you take atan(0/1), you get 0 degrees, meaning the device is flat.

In the worst cases, assuming the above points from the datasheet (if I'm understanding them to mean what I think they mean).

Old device worst case: (0 + 20mg) and (1g - 20mg)
New device worst case (0 + 95mg) and (1g - 95mg)

Old worst case inclination: atan(20/980) = 1.16 degrees
New worst case inclination: atan(95/905) = 5.99 degrees

So I think we've found our culprit. This is pretty disappointing, and clearly not going to be fixed by a software update without further calibration. I guess the best thing we can do is hope that Apple switches back to the older chip.

If this really is the meaning of zero-g offset I think, we should see a lot more phones with a larger deviation than 6°. In this case the +/- would mean that 66.6% of the devices show a deviation of up to 6° and 33.3% should have an even larger deviation. I can't imagine that Apple would use such a sensor...

 

[U4icLab]

We all know there are more important things to worry about, but hey, we like our gadgets and we expect them to work as advertised. They're not cheap. What is frustrating to me is that they boast about precision and being "the best." They throw a new processor + co inside, talk about gaming, and emphasize motion in the keynote. We expect bugs and mishaps with technology, but with such a strong, loyal customer base, I would think it would be nice for Apple to at least bring a little bit of light or acknowledgement on the issue. Being in the dark is only adding fuel to the fire. They can fix and tweak all the other bugs/settings, but this thing is hitting us smack in the face.

 

[Ceebee1980]

I can't imagine this is the actual case, though. If Apple switched component-makers, there'd have to be some process by which to gauge the old vs new accuracy.

Also, I highly doubt that Bosch would sell a component that has the described level of inaccuracy. It wouldn't be worth marketing, let alone building in the first place if it's off by 6 degrees.

 

[gCloud]

I just wanted to chime in that if you run some math on the data provided in the data-sheets of the old parts you get accuracy which is very close to what we're seeing.

The zero-g offset of the old device: +/-20mg (milli-earth gravities)
The zero-g offset of the new device: +/- 95mg

Now the angle of inclination of the phone is the arctan of any pair of axis, many of us have measured the inclination of the phone laying flat, in which case we're measuring the gravity from along the axis projected out along the back of the phone and the gravity in the direction out from the top, bottom, left, or right (it doesn't really matter which).

In the best case scenario, when flat, we read 1g-force out the back and zero out whichever side we're using. If you take atan(0/1), you get 0 degrees, meaning the device is flat.

In the worst cases, assuming the above points from the datasheet (if I'm understanding them to mean what I think they mean).

Old device worst case: (0 + 20mg) and (1g - 20mg)
New device worst case (0 + 95mg) and (1g - 95mg)

Old worst case inclination: atan(20/980) = 1.16 degrees
New worst case inclination: atan(95/905) = 5.99 degrees

So I think we've found our culprit. This is pretty disappointing, and clearly not going to be fixed by a software update without further calibration. I guess the best thing we can do is hope that Apple switches back to the older chip.

Great post
I cant see them switching back anytime soon with millions of devices sold and millions being made daily. They will probably issue a software update of sorts to massage the readings and leave it at that, this means u can never trust the information on the phone that relates gyro/acclero.
I think this baby is going back before return period expires. Bought the largest capacity thinking i'll keep for a while, might look at smallest 5c or a 5 if i can get and wait till next year, this is probably the worst of the gates

 

[Marco123]

http://www.insideactivitytracking.c...ate-accelerometer-and-gyroscope-measurements/


"With the iPhone 5s, Apple changed producers for its accelerometer component. According to teardowns by iFixit and EDN, the iPhone 5 used the LIS331DLH 3-axis accelerometer and the L3G4200DH gyroscope, which are both made by STMicroelectronics. According to a Chipworks teardown, the iPhone 5s (see the M7 teardown) uses a gyroscope component from STMicroelectronics, but Apple switched to Bosch Sensortec for the accelerometer, using the BMA220 3-axis accelerometer. According to the companies’ websites, the old accelerometer was capable of higher data output than the new accelerometer (16 bit output compared to 6 bit digital resolution). The Bosch accelerometer also has more typical output noise and a higher Zero-g Offset than the STMicroelectronics component–two aspects also contributing to the lower level of accuracy in the iPhone 5s. This new accelerometer might be a source of users’ issues measuring levels."

Thanks for this, apple picked cheap but retails at higher. This is what happens when all you care about is the profit margin.
Apple will bury their head in the sand over this and I will be going back to ip5.
Can't wait to see how they try to fix this mess.

 

[benzn]

If this really is the meaning of zero-g offset...

Yeah, you're right.

One potential source of error in my calculations is that it may be such that there is less error in the presence of gravity (which to me makes intuitive sense), in which case (assuming it's zero), we get simply:

atan(20/1000) = 1.15 degrees
atan(95/1000) = 5.43 degrees

There have been reports of errors of 6 deg or more, but they are far and few. I highly doubt that the error follows a normal distribution (no reason why it should), so that would explain why we don't see errors like 10 degrees. There could be a bunch near the 5.5 degree threshold so that the notation of the tolerances being one standard deviation away still hold, but we can't really know.

Also, I highly doubt that Bosch would sell a component that has inaccuracy. It wouldn't be worth marketing, let alone taking to clients like Apple.

Every component has inaccuracy, even the old sensor. For whatever reason, Apple didn't appear to consider the accuracy in this case.

 

[Megakazbek]

Every component has inaccuracy, even the old sensor. For whatever reason, Apple didn't appear to consider the accuracy in this case.

But what exactly they DID consider, since it now breaks the primary functionality of even Apple's own standard iOS 7 apps? What's the point of having sensor at all if it doesn't do what it's supposed to do?

 

[jpcarro]

Here's what -2° looks like in the level app and the Star Walk app. Measured from a known level surface. Not horrible, but not good. Can't imagine tolerating any worse.

If a cheaper component with a broader inaccuracy range is the culprit, this is disappointing.

 

[RiccardoA2]

The range

Yep, but the new sensor have a largest range up to +- 16 g.

... and the g-offset is not the sensitivity, i think.

Apple //c

 

[benzn]

But what exactly they DID consider, since it now breaks the primary functionality of even Apple's own standard iOS 7 apps? What's the point of having sensor at all if it doesn't do what it's supposed to do?

Of course we can't really know. The two guesses I have are a) cost, and b) something related to power consumption. According to the datasheets, the new chip has nominally less power draw < 10uA vs 10uA (I have no idea how much "<" actually means), but it could function in a way that allows for easier interoperation with the M7 (it can run on a 1.6V supply instead of a 2.3V like the previous chip). I'd have to spend some time reading the datasheets more thoroughly to get a better idea, but I don't have the time today, unfortunately.