Just an update, I did stop motor from full speed (above 50km/h) to full stop by full reverse powering very often now, and have never seen a problem with that.
From physics lessons long ago at school I know that the distance s for constant acceleration dependent of time t is:
s = 0.5 * a * t^2
Regardless at which speed I start full reverse power braking, it is a constant negative acceleration. So I assumed that the slow down curve has to be somewhat quadratic.
In order to investigate that with my slow android camera (gives 15fps) I did draw North, South, East and West, doubled and doubled again to get 16 directions in Motor Test Station. After having done that I realized that what I did is well known:
Then I used a video for up to now record run with 14.37m/s or 51.7km/h, where I made the Arduino do a full stop by reverse powering for 1800ms. This is the animated video I created from it (whith blue lines from center nail passing the front lit LED). In order to know when braking happens, I lighted LEDs on pins 13, 1 and 0 in addition to the always lit power LED. This gives thick LED arcs easily distinguishable from thin line of just power LED on. In order to be able to see something the animated .gif is slowed down by factor of 10:
Next I determined roughly the angles of the blue auxiliary lines and determined the distance in meters to the point of full stop. Finally I played with coefficients of quadratic function in order to fit the measured points well and came up with 2.8 and -0.3 (you can view image in new tab to see it at a better resolution):
This shows pretty well that in fact the braking curve is quadratic, the linear term may be caused by friction. For completeness these are the 28 frame pictures being basis for the animated .gif. They are screenshots from video viewed by rowvid.com in order to know the time each frame happens at hundredth of second resolution:
https://stamm-wilbrandt.de/en/forum/mts.brake.frames.zip
Yes, more than 8m braking distance is a lot and a no go for robot running in house (at least in my house), but speed is 51.7km/h and I don't think robot moving autonomously will ever come into that speed range.
There is the problem of knowing how long to reverse power at a given speed to get to a full stop and do not start moving into the opposite direction. This may be solved by using a mouse sensor for detemining whether motion still happens. Since speed of both robot motors may be different (eg. when in curve), having two such sensors vertically and "looking" on each wheel's motion should allow to do a perfect full stop:
Last, but not least, the slow android camera is fine for recording normal stuff. For more details slow motion camera is useful, and I am happy that my younger son bought a LG G Flex 2 android phone with 120fps.
I did not find anything easy allowing to single frame step through a 120fps slowmo back and forth, so I googled and based on what I found did my own 120fps viewer (do "View Page source" for details). It shows a short 2.8 second video of Motor Test Station run without Arduino, just with a finger tap, with 341 frames, and does fps snap-in time:
http://stamm-wilbrandt.de/videos/120fps.html
Hermann.