Phyphox is amazing: there is a wide variety of different experiments available using all sorts of sensors found in phones. The iPhone accelerometer seems to saturate at roughly 13g, easily achievable by pulling your phone very fast in an increasingly narrow arc towards yourself. What were you thinking how I found out?
(Fun fact: this is another project that Sebastian Staacks of http://there.oughta.be worked(s?) on. You might remember his wooden Gameboy shell, the Gameboy interceptor to screenshare the Gameboy screen to a PC or his bullettime video booth.)
> iPhone accelerometer seems to saturate at roughly 13g
That surprises me. As soon as a sensor saturates, you can no longer use it for dead reckoning (integrating the acceleration values to know position).
13g is way below what you'd expect when tapping the phone against a solid object for example - that could easily reach 1000g briefly.
I was under the impression that a big part of iphone's superior GPS performance was down to the clever use of dead reckoning and re-calculating past datapoints in a way android manufacturers rarely do.
I don’t think the iPhone does naive dead reckoning like that, as the sensitivity of the needed inputs is so high for it be reasonable in a consumer handheld. The cost of components would be way to high, then there’s the issues of power consumption, because you would need to poll all your sensors at a very high rate to get enough accurate data to perform naive dead reckoning accurately.
Instead I think the iPhone cheat a bunch, and use a mixture of step counting, magneto and gyro inputs to perform dead reckoning. The iPhone uses periods when you GPS is in use anyway to calibrate your stride length at different speeds, which when mixed with magneto data to estimate a course direction, would allow an iPhone semi-accurately perform dead reckoning, without needing high resolution, wide scale, accelerometer input.
In short, the iPhone cheats. It’s dead reckoning systems take advantage human bio-mechanics to simplify the problem of dead reckoning, at the cost of building a dead reckoning system that only works effectively in a handheld device that semi-permanently lives on your person. A perfectly reasonable assumption for a phone, but a terrible solution for a more generic dead reckoning system.
Sure, but if an entire human person is accelerated at 1000g for a period of time that's long enough to lead to a significant change in their location...
If my math is right, assuming an iphone deforms linearly on impact with 1000g deceleration, it would only give way 0,7mm when dropped from 1m height onto concrete. that seems too low a deformation? I would expect it to whobble quite a bit in slow-mo.
(Fun fact: this is another project that Sebastian Staacks of http://there.oughta.be worked(s?) on. You might remember his wooden Gameboy shell, the Gameboy interceptor to screenshare the Gameboy screen to a PC or his bullettime video booth.)