iPhone 5s accelerometer/gyroscope/compass values incorrect

[Ceebee1980]

Good math, but poor analysis.

Let me restate the exact same thing, more strongly: There's no way in hell that Apple or Bosch would work with a sensor with the kind of error being alleged. +/- 5% is not a valid margin of error for anything in the modern technical world.

Unless there's an abnormality with the batch of sensors, I suspect software or interference from other component(s).

 

[WilliamG]

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.

This is what my -5 to -6 looks like, resting on a flat surface.

 

[benzn]

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

+/- 16G is only really useful for detecting shocks/when the device is dropped. When you put the device in this mode you get very low resolution readings (around 0.5G I think) which is useless for orientation detection.

And you're right it's not, but sensitivity doesn't help with bias (thus it isn't causing this problem). But while we're at it, the sensitivity in the 2G mode is 62.5mg in the new vs 1mg in the old (the old uses a lot more bits to represent its readings)

----------

Good math, but poor analysis.

Let me restate the exact same thing, more strongly: There's no way in hell that Apple or Bosch would work with a sensor with the kind of error being alleged. +/- 5% is not a valid margin of error for anything in the modern technical world.

Unless there's an abnormality with the batch of sensors, I suspect software or interference from other component(s).

I share your sentiment, but this isn't alleged, this is straight from the datasheet that Bosch wrote (http://ae-bst.resource.bosch.com/media/products/dokumente/bma220/BST-BMA220-DS003-08.pdf). I suppose there could be the possibility that the part number was misreported, but I'd welcome any corrections anyone has.

 

[richard371]

I'm trying to find a teardown that shows what accelerometer the 5c is using but not having much luck. Ifixit doesn't seem to show it.

 

[Ceebee1980]

I think the math may be flawed, then, or your interpretation of "range".

Any chip that's accurate enough for recommended use in gaming couldn't be that acceptably inaccurate. Sorry.

It's still a good find (the sensor pdf), but just seems like a more complex issue. If it was the sensor, then the first internal tester who tested Infinity Blade 3 on this thing would've called Tim Cook and rang the emergency bell.

 

[highlightshadow]

There is zero reason for Apple to use a system that could result in such a large margin of error .... even cheap modern gyros are accurate to less than that.

 

[gCloud]

There is zero reason for Apple to use a system that could result in such a large margin of error .... even cheap modern gyros are accurate to less than that.

Reason or not it looks like the did as we are all seeing the issue on the 5s and not the 5c, 5, and 4s

 

[highlightshadow]

The thing is accelerometers and gyros are relative devices

Still think its the interface between M7 & the gyro chip - not storing the values .... its not like the gyro isn't accurate relative to its own plane... it is

I can adjust the angle of device and get an accurate reading (if you click on the compass page it ignores the stored 'zero' and uses relative position ... its accurate

I think its a case of the software cannot calibrate a level and hold it .... i'm pretty certain this is software ....

Same principal in RC Heli self-level systems that use a combination of accelerometers and gyros to know which way is 'up'... you have to TEACH it that true centre then all movement is done as an offset of that.
I don't think its the accuracy

 

[Ceebee1980]

Reason or not it looks like the did as we are all seeing the issue on the 5s and not the 5c, 5, and 4s

This is a fanboy and hipster forum now, I guess. Thinking that this not being an issue on other devices shows absolutely zero understanding for software builds based on devices.

It could easily, EASILY be a build-specific glitch.

Do you really think Apple "ran out of time" when it came to testing these devices in the field? They've been showing up for a year or more, not to mention the fact that every stakeholder has a lot to lose if the build was defective. Don't you really think that they test several of these per hundred that come off the assembly line?

In addition, I'd imagine there were thousands of these devices in the field with pre-approved software testers and devs.

Most likely, this is akin to an OS memory-leak. Something in the software is allowing some play in a number that's supposed to be constant, or a known and acceptable bit of interference is not being accounted for in the latest OS build that came loaded on the devices, or the orientation process at time of manufacture was askew. I highly doubt that as well.

Even at my most cynical, I'd have to imagine that Apple did so much testing that there would've been five hundred new beating victims at Foxconn if the manufacturing process had a mistake in it.

 

[RiccardoA2]

+/- 16G is only really useful for detecting shocks/when the device is dropped. When you put the device in this mode you get very low resolution readings (around 0.5G I think) which is useless for orientation detection.

And you're right it's not, but sensitivity doesn't help with bias (thus it isn't causing this problem). But while we're at it, the sensitivity in the 2G mode is 62.5mg in the new vs 1mg in the old (the old uses a lot more bits to represent its readings)

This is the reasons:

Bosh Ultra Small package 2mm X 2mm X 0.98 mm
ST 3mm X 3mm X 1 mm

And consumption:

Bosh 1.6 V to 3.6 V
ST 2.16 V to 3.6 V

--------------------
Apple //c

 

[walmartmartyr]

so how do you explain the 5s's that do not have any issue with this problem?

 

[gCloud]

This is a fanboy and hipster forum now, I guess. Thinking that this not being an issue on other devices shows absolutely zero understanding for software builds based on devices.

It could easily, EASILY be a build-specific glitch.

Do you really think Apple "ran out of time" when it came to testing these devices in the field? They've been showing up for a year or more, not to mention the fact that every stakeholder has a lot to lose if the build was defective. Don't you really think that they test several of these per hundred that come off the assembly line?

In addition, I'd imagine there were thousands of these devices in the field with pre-approved software testers and devs.

Most likely, this is akin to an OS memory-leak. Something in the software is allowing some play in a number that's supposed to be constant, or a known and acceptable bit of interference is not being accounted for in the latest OS build that came loaded on the devices, or the orientation process at time of manufacture was askew. I highly doubt that as well.

Even at my most cynical, I'd have to imagine that Apple did so much testing that there would've been five hundred new beating victims at Foxconn if the manufacturing process had a mistake in it.

Fanboy indeed great testing if its showing up on every 5s device in every apple store.

 

[benzn]

Ah, good catch! That means they reduced 9mm^2 to 4mm^2 and freed up 5mm^2 of space (which is probably needed by the M7 and associated chips). (FYI, voltage isn't quite the same as power consumption, but a higher voltage typically results in higher power consumption, and there's a reasonable chance both chips are being run on the same voltage anyway (say 3.3V) simply because they already have a supply line at an acceptable voltage being used somewhere else)

 

[richard371]

If this is the case I don't see apple going back to the old chip due to the size.

Ah, good catch! That means they reduced 9mm^2 to 4mm^2 and freed up 5mm^2 of space (which is probably needed by the M7 and associated chips). (FYI, voltage isn't quite the same as power consumption, but a higher voltage typically results in higher power consumption, and there's a reasonable chance both chips are being run on the same voltage anyway (say 3.3V) simply because they already have a supply line at an acceptable voltage being used somewhere else)

 

[gCloud]

so how do you explain the 5s's that do not have any issue with this problem?

Are u sure they exist? I looked at half a dozen in store, 3 mates have the issue, others have looked at thier local store all had em replacments had em, people who reported zero one day reported drift the day after.

 

[Ceebee1980]

Fanboy indeed great testing if its showing up on every 5s device in every apple store.

What do you think is more likely, if it's showing up in EVERY 5s:

a) software bug in final-build of brand new OS?

b) most valuable company on earth, led by maniacally-dedicated supply-chain guru Tim Cook, forgot to check the production line?

 

[gCloud]

What do you think is more likely, if it's showing up in EVERY 5s:

a) software bug in final-build of brand new OS?

b) most valuable company on earth, led by maniacally-dedicated supply-chain guru Tim Cook, forgot to check the production line?

In this instance probably b

I'd love to be wrong, i want to keep the phone.

 

[benzn]

If this is the case I don't see apple going back to the old chip due to the size.

Yeah, they could theoretically buy a lot more than they need and just pick ones that meet lower error thresholds, or negotiate with Bosch to have Bosch tighten their verification standards (for those who don't know, a very large number of chips are often thrown away as part of normal production when they don't meet the required error thresholds, the percent that make it past verification is known as the yield). Sucky situation to be in either way.

 

[Sodner]

Can I ask why no one outside this thread cares?

Honestly are all the reputable news agencies afraid of Apples backlash (think Gizmodo) if they go public with this?

 

[walmartmartyr]

90% don't know their phone can even do this, hell I didn't till I came here.

I have 3 5s, delivered today, each color. I'll check them each for skew

----------

Are u sure they exist? I looked at half a dozen in store, 3 mates have the issue, others have looked at thier local store all had em replacments had em, people who reported zero one day reported drift the day after.

well 9% are at zero and 18% show one so.....

 

[gCloud]

90% don't know their phone can even do this, hell I didn't till I came here.

I have 3 5s, delivered today, each color. I'll check them each for skew

Yes, please report findings and production week.

 

[i-aamir]

QUOTE - walmartmartyr

I have 3 5s, delivered today, each color. I'll check them each for skew

----------

Please post your findings when you get a chance. I have the following info :

Nice Name: iPhone 5s
Family name:
Group1: iPhone
Group2:
Generation:
CPU speed: 1.3GHz
Screen size: 4 inch
Screen resolution: 1136x640 pixels
Colour: Gold
Production year: 2013
Production week: 39 (October)
Model introduced: 2013
Capacity: 16GB
Memory - flavour: xx
Factory: C3 (China)

And am -4 degrees off at every angle!

 

[WolfSnap]

Nice Name: iPhone 5s
Family name:
Group1: iPhone
Group2:
Generation:
CPU speed: 1.3GHz
Screen size: 4 inch
Screen resolution: 1136x640 pixels
Colour: Gold
Production year: 2013
Production week: 38 (September)
Model introduced: 2013
Capacity: 64GB
Memory - flavour: xx
Factory: C3 (China)

Off by -4º

 

[BMcCoy]

Week 39, factory C3... Perfect 0 degrees.

 

[schizbomb]

My boss' 5C is off by -1 but I guess that could be anything. Regardless a 5, 2-4S and a 4 show 0 on our desks. My 5S is between -2 and -4 off and my boss' 5C is -1 off.