Also, keep the temperature cool:
When filament starts, it is thick because there is no weight to "draw" it out, as the loop forms, the weight increases so as the "draw" weight increases and then forms thinner filament which in turn decreases "draw" weight and causes it to thicken again.

You will end up with an oscillating thickness and lower temperature will help to minimize it.

I think that a "L" configuration would work better than a "U" but I have not tried it yet.

My 3mm set up:
My "U" config has 31" between winder and extruder, filament entry and exit at the same height, and a 20.5 inch drop to the laser
the PTFE tube starts at the bottom of the guide which is solidly mounted and has no flex and is looped in the direction of the extruder.

Also thought of separating the motor power from the heater circuit as I noticed change in pitch when heater turn on/off.

So i finally got time to play with this setup again.
I got the new motor driver that displays voltage and amperage.
The amperage is fluctuating about .2 amps at the most.

How much fluctuation do other people see???

The fluctuation kinda sorta matches the fluctuation in diameter, buts its hard to really tell.
The nozzle does trace a small circle, its not much, maybe a few millimeters...

Without completely replacing the nozzle/barrel/mounting hardware, and replacing the wood body with a real metal body, i doubt i can get things aligned better...

Any other thoughts out there?

FYI, i did end up allowing the filawinder to stop, but its reaction is much much slower, which definitly helps it consistently it pull, and not start/stop so much and abruptly.

Here is video of the lastest setup.
https://youtu.be/HMHZV-eDhL4
Running 17V, i got out 700 grams, in maybe 7 hours, of colored PA 746.
Diameter is around +/- 0.5, good enough, that im about to give up on making improvements.

Tonight i grabbed a spool of Taulman bridge, and another spool of purchased filament i had. Total variation was probably +/- 0.3

0.3? Sure about that?

Tonight i grabbed a spool of Taulman bridge, and another spool of purchased filament i had.
After initially being annoyed both rolls were at 1.7 nominal... instead of 1.75, I ran some through the filawinder thickness gauge setup. Probably 2-3 minutes worth. The general deviation matched what I as hand measuring and didn't do the oscillation. Total variation was probably +/- 0.3 and it kinda slowly meandered around, not large swings like I see when i extrude.

What I am most curious to see, is video of extrusion through other real time diameter monitoring setups (or copies of mine). I know there is a light based thickness sensor.
Really want to know if this oscillation is inherent to the vertical setup in general, or if its specific to my machine in some way.

Maybe their is something odd about the power getting to the motor, or the motor itself causing slight cyclical changes in motor speed.... I can easily monitor voltage... Should probably get the new motor controller board so i can see the current also... Any change in barrel pressure I would think would have a corresponding voltage/current change.

I will mention, now that I think about it.
Back when i was trying to fight PLA (kinda given up that fight for now.) I do remember seeing the barrel kinda walk around a bit, i think it happened when i was using too low a temp. It did stall once, but i was right their and stopped it. The nozzel end probably does still wonder a bit... Probably means my motor to barrel setup is completely straight, although I had thought it was kinda tolerant of misalignment since the coupling isn't rally rigid. Maybe it indicates something is bent... I will rig up a pointer or something so its easy to see how far the actual hole is drifting around while extruding. Wouldn't think it to matter much initially, but maybe whatever is causing the drift also causes a cyclical change in barrel pressure... And if barrel pressure is oscillating, that would explain diameter oscillation. All that said, the motor sounds consistently annoying, but bearably loud.

Im out of town next week for work, so its going to be a couple weeks till i get back to this.

Edit: Totally blanked, I already do have a voltage read out hooked up to the output of the PWM Board version 1. Voltage is stable. Maybe thats expected though, and i have to look at current.

I had run a bit of store bought filament through and seem to remember the measurement being more stable. However, I never ran bought filament through and actually winding it on the filawinder.....

Will try that.

By comparing your measurement setup to what most people do (calipers), its a bit apples to oranges. Additionally, store bough filament is generally rated at +/-0.05mm or +/-0.1mm, so you're right in the ballpark.

As for improving variation, reducing temperature and adding fan cooling should reduce variation, but the 2.1mm nozzle will be too large for that. 180C is recommended for ABS if filament diameter tolerance is the highest priority. Most folks have a little less drop in the loop, not sure if you've experimented with that while at the 33" apart.

So i spent the weekend playing with the filastruder.
I was able to get things better, but still not where i think it should be.
Filawinder is now about 33 inches to the right of the extruder (cannot go any further, out of room)
Loop goes down about 30", temperature is 184.
Diameter average is about right.
Played with the firmware on the filawinder, setting min speed from 25 to 60 (which is a bit above stall speed). Also made the maximum increment of speed change 6 instead of 25. (I think i will reduce this a bit further to 4). In general, the filawinder seems to be running about as smooth as possible, introducing minimal oscillation in the filament loop. I did remove the PTFE tube loop, and now just have a straight section of tube, up to my dial caliper measurement setup, which seems to provide enough tension for the spool.
Also adjusted the guide angle step from 1.5 to 1.9. I had put in a linkage arrangement on the quide to limit overall range from +/-180 to about half that, while reducing the stress on the motor. Adjusting this angle definitely makes the spool, spool up better. Also tweaked the location of the hall transducer, i think it was bent back a bit too far from the magnet and not guite reading reliable.

Filament height sensor is slightly offset to the filawinder side.

Turned the fan off, so their is no wind affecting the filament.
At 184, my diameter is fine without it. (Although maybe i need coolant to minimize the unsteadyness of the diameter, see below)

I did manage to put 3 guides that limit movement of the filament about 6" below the nozzle. Somewhat surprised that appears to be working out just fine.

Here is a video mid way through the weekend.
https://youtu.be/wmB0wgmJpFk

So in the end, I am just barely achieving +/- 0.05 mm Every once  in a blue moon it maybe goes +/- 0.06 mm.
I can live with that, as it is usable.
However, what really surprises me is the overall behavior.
When I hear +/0.05mm, I would think 90% of the time its between +/-0.03, and maybe spends 5% of the time at the full +/-0.05mm.
What i see, is it goes back and forth between the extreme pretty fast. Like if its at +0.05, then it will likely end up at -0.05 within 6-20 seconds, and then it swings back the other way. Sometimes it will hunt around the midpoint for a bit, but I would say its swinging between the extremes at least 50% of the time.

What behaviors do others see? You really have to have a thickness gauge on it to track the diameter over time.

My thickness gauge holder does have the centerline of the pulley 0.5mm off center of the in/out hole, to provide a bit of tension. I was thinking maybe this offset causes changes in diameter to be exagurated in the gauge due to the angle of flex of the filament, but in general, the thickness gauge match my hand reading with a caliper...

Picture of guides.

What speeds are you seeing currently, and what were you seeing at 180C?