tsl1401 for filawinder

I have experimented with it, and also tried out the one posted on Thingiverse.   The resolution of the pixels puts the precision around .03mm.  To improve it, you need sub pixel sampling, or position the sensor and light source in a way that the filament casts a shadow larger than its diameter.  The biggest challenge however is the filament cannot be allowed to move in a way that changes its distance from the sensor.  Since you need to maintain accuracy at increments of .01mm, it is hard to do that in a way that still allows changes in diameter.  If the light is collimated the size of the shadow won't change much, but it changes enough to make the measurement untrustworthy at .01mm.

The best solution for that is to use two sensors that are perpendicular, like a laser micrometer.  Then movement away from one sensor will be detected as sideways movement on the other so changes in size due to movement can be factored into the calculation. 

I have been working on an alternative method using spring loaded bearings, a magnet and a hall sensor based on a design by Thomas Sanladerer - https://www.youmagine.com/designs/filam … ter-sensor

The sensor is sensitive enough for .01mm precision, and that can be improved by lengthening the lever so there is more movement of the magnet relative to changes in filament diameter.  Also I am using an Allegro A1302 hall sensor, and have since found there is an A1301 which is twice as sensitive which should improve the stability of the reading.

The code is much simpler than what is needed for a linear ccd.  All you need to do is measure a few drill bits to determine the relationship between the analog read and a known measurement for a few different diameters, and put them in a lookup table.  Tom simply lifted some thermistor code from Marlin to do this.  Reading out the exposure of the pixels in a linear array is more complicated, then you have to determine which ones represent the shadow and run some sub pixel sampling to detect where the edge probably is at the level of precision needed.

The biggest difference is price.  The cheapest I could find a TSL1401 is about $16.  A magnet and hall sensor cost $1.  From the standpoint of the Filawinder, the TSL1401 also needs a PCB so you can route the pins to a connector.  The legs of the hall sensor can plug straight into a wire housing at one end, with the other end of the wire plugged into the  headers I included for A7 on the Filawinder control board.

At the moment the only thing you can do with the sensor is read the diameter in the serial monitor.  Just showing the diameter can be done more simply with a thickness gauge mount like the one I posted on Thingiverse.  Eventually this could be used for graphing and logging on the PC if you have one you can leave connected to the winder.  The main goal however is to use this for closed-loop control of a puller to try and get at least .03 tolerance without needing to drop the filament into a loop.  I'd like to make this the front half of a puller, and use the rotation of the rollers to count the length of filament spooled as well.

It would be a bit of overkill for that application.    I considered it briefly, but the area it covers is very small so it would be easier  for the filament to accidentally move too far up or down to be detected.  If you are having trouble with the photocells and laser in the bright room, you might try using IR instead.