Sunday, April 29, 2018

Project T19 Part 10 Developmental Release LInk and Rundown. Plus Inconclusive Buscaps and Startup Gremlins

First things first: Google Drive link to the T19.101 developmental release part meshes and some assorted firmwares.

WARNING: Developmental release means developmental release. This is for the curious, those who walk on the very edge and want to be the first kid on the block to have one of these at all costs, and for the benefit of those who want to adapt or hack parts of my designs, mainly. Sort of like SimonK says but intenser: If it explodes, you get to keep all the shrapnel. No promises and no support here - yet. That is coming soon. I promise.

Actually, the models all came out relatively final on the first print, and assembled into that first build without many trims and adjustments. But these are the known bugs:

  • Breech - Lower screw bosses for the side covers come too close to the cover's lower edge and will need to be trimmed about a mm back parallel to the cover edge to clear.
  • Grip base - Needs a grind on the top front edge to clear the mag release spring perch reinforcement in the drive housing.
  • Cage (Hy-Con-GammaMajor_Main) - Insufficient clearance in the phase wire channels for the heatshrink on the phase wires coming out of the stator. Needs ends of the channels widened adjacent to motor mounting surface.
  • Stock - may need some supports and/or spaghetti cleanup after printing a surface over air.
Some roughnesses and improvement areas:
  • Trigger and mag release axial clearances are tight - hand fitting parts is required!
  • Magwell fit - I erred on the side of loose for drop-free and compatibility with a maximum tolerance range of mags, but this feels somewhat sloppy, so I will be adding 1 or more magwell models with tighter fit in both directions.
  • Cage groove filler clearances on Gamma Major are tight with the 9.5 wheel, and the lead-side fillers serve no concrete purpose (they are absent on all previous groove-filler Hy-Con cages which work flawlessly) so could be removed anyway. If you mess up your flywheel manufacturing and fitting and have too much runout, you will hit the cage.
  • Bolt clearances in the drive spacer rails are a tad loose and AK-like - this doesn't affect anything functionally though.
  • I need to get those 10mm and even larger gap flywheels up as an option.
  • Stock and stock base tube sockets need a bit too much grinding/sanding for 1" PVC fitment. Ideally, just removing layer change blips should do the job.
  • Grip frame install into the channel in the grip base is tight and needs hand fitting.
A word on PLA: I have no experience with and can't support PLA for T19 parts. Do that at your own risk. I use PETG only. Most parts should by all means work fine with PLA, but I haven't and may never test with it.

Words on my slice/print parameters:

All numbers consider my 0.4mm nozzle and 0.2mm layer height.

Most unspecified parts I use 6 top and bottom layers, 3 perimeters, and 30% hexagonal infill.

Flywheels: 4 bottoms and 3 tops, 3 perimeters, 20% hex - this is a possibly PETG-specific inertia minimization which is plenty strong and stiff.

Cage main section: 3 perimeters, 100% infill. Hy-Con cages are very sensitive to stiffness in the motor mount section. Also, 100% cages are distinctly quieter even if you were to find that solid isn't quite necessary for stiffness.

Grip frame and grip base: 10 tops and bottoms, 5 perimeters, 40% hex. Don't skimp on these, they are how a rather heavy blaster is handled.

Crank web: You may want some more perimeters on that as well for shaft bore, setscrew thread, and crank pin thread robustness if using something non-PETG.

Preliminary words on assembly:

3mm motor mount holes for wheels and pusher, 2mm flywheel bolt holes, all 3.6mm 6-32 fastener clearance holes and 3/16" 10-24 fastener clearance holes need drilling to final size to clean out.

2.7mm 6-32 tapped holes need drilling before tapping to avoid difficulty - I use ~3mm with PETG.

3.8mm 10-24 tapped holes should be tappable as printed. Be CAREFUL tapping the front rail hole in the cage! Don't dent the ID of the barrel with the tip of the tap. It comes CLOSE in there.

3mm shaft pilot bore in the flywheel center is a relic and now an overconstraint. Drill that out slightly oversize. The rotor OD locates the Gen3+ wheels, not that shaft pilot.

Watch out for screw lengths surrounding motors, both the stator base to cage and the flywheel to rotor flange bolts have the potential to hit and damage the windings if too long!

Alright, time to stick an image in here to break the textwall and shift gears while the clutch isn't biting for a sec:


I remain unsure of this upgrade.

By all textbook and EE means, upgrading the DC link cap on an inverter should ONLY be able to benefit proper inverter operation and should be a safeguard against voltage transients. You CAN'T have "too much" of it. The DC bus is supposed to act as a stiff voltage source in the design theory of a VSI. Adding a larger capacitance with a lower ESR to it near the switching devices only assists in making that closer to true.

However, zero-cross detection sensorless is a finicky beast, and lacking feedback current control, parasitic bus impedances ARE a variable that affects phase current and could affect the performance of sensorless. With this build, I have encountered less consistent motor startups than with my other Hy-Con gear, and dare I say, the controllers with the 1000uF Kemets were worst at occasionally "misfiring" (with some red desync warning LED flashes while SimonK powerskipped and attempted to reestablish what the rotor position was) on start. This would occasionally cause a slightly delayed spinup and a derp shot. Controllers with the 220uF Haicaps that I have tested so far are also not 100%, but are slightly better (I have tuned the SimonK build and feed delays both to a state I am happy with using for now and has prevented all derp darts).

Component tolerances (sense resistors, ...) can't be ruled out as why certain individual ESCs might be more apt to have sensorless glitchyness at low speed. But at the same time, one aspect that stands out is that Serial One's 14AWG harness, very high rated switch, and present 1.5Ah Graphene all result in lower DC bus impedance (both resistance and inductance) than the Model Pandora's long 16AWG battery wiring, 16AWG squids, cheap switch, and Monolith pack. Perhaps it's that provoking the trouble. It's not unprecedented for this to be a variable to worry about. With Ultracage, some posts give cases where simply switching to a high end battery can make a BLHeli drive that previously worked completely freak out in a FAR more epic way than my perfectionist griping is about.



So there are 2 directions to go from here, if that proves to be the cause and not just a dud Afro board or two.
  • The correct engineering one: tune SimonK to work on this hardware with a very stiff DC bus. Changing various startup mode and ZC detection parameters, using high side PWM, or adjusting duty schedules (I did have some success with cutting the startup duty back a tad) might do the trick on top of the timing that I have been dialing in. Then, we can run uprated caps and have less voltage spikyness on the bus (safer for the mosfets), and be more tolerant of a wide range of batteries.
  • The "stop fucking with it; you had the damn thing working like clockwork before" one - avoid the overkill caps, and avoid the overkill batteries.
I'm gonna keep chipping at 1, but I am also going to consider the validity of 2 by trying a single-P Monolith on the same drive parts and software as a comparison. I'm perfectly happy with the performance of the single-P on the Model Pandora, and it is safe. It may well be that practically it is better and gives more consistent startups to have a slightly non-ideal bus voltage stiffness and perhaps even slightly duller best-case accelerations if that's what makes the sensorless voodoo happiest. Because what it boils down to is what feed delays you can consistently get away with, and right now, the Graphene-equipped Serial One is doing slightly worse at that, not slightly better as I expected.

No comments:

Post a Comment