All the important (um) core functionality is there and solid - the main bit left to work on, really, is just doing a proper power-on selftest now that we have the hardware capabilities to. Beyond that, it's mainly a matter of implementing alternative UI behaviors that I or anyone else wants.
So into all the new stuff.
- Implementation of FlyShot protocol; integration with adjustable-speed flywheel drives
The protocol described in this post is implemented.
Support on the motor controller's end is required. This may become a compile time option, but in general, avoiding ESC reflashing and setting speed over the wire from the blaster manager's end is easier and more flexible, even if you program your blaster manager to set a fixed speed only.
- Implementation of speed feedback-based feed control (=closed-loop single-trigger control. CLOSE ALL THE LOOPS!!)
"Blind" feed delays/delay scheduling approaches to account for motor acceleration with no awareness of actual wheel speeds are so last decade and really needed to go, just as much as open-loop voltage command flywheel drives needed to go. In particular:
- Scheduled delays require tedious manual tuning to closely match the dynamics of a given flywheel/drive system and produce snappy shots, but consistent velocity, under a variety of conditions. Oftentimes they get just "Meh, good enough!"-ed to the generous side by a builder who can't be bothered squeaking every millisecond of latency and every fps out of the damn thing - including me.
- Scheduled delays cannot, obviously, react whatsoever to unanticipated variances. This could be anything from a change in DC bus voltage and sag due to batteries not being ideal and varying in voltage and IR with SOC and temperature, to a barely noticeable desync event during a spinup that slows it down slightly, to some debris stuck in a cage or cold grease in a bearing causing friction on a wheel, to an outright jam or other situation where a wheel fails to turn at all.
- Thus, delay settings must be conservative to cover the vast majority of these variances.
- Also, thus, scheduled delays will completely fail at their job in more severely unexpected situations - by feeding ammo when it would produce a crappy shot, or by feeding ammo when it is completely inappropriate to feed and causes or worsens a stoppage!
This not only greatly improves robustness against unexpected conditions, but renders the system 95% self-tuning, since the whole issue of predicting drive dynamics with math and/or multi-stage delays is fundamentally sidestepped. Plug in ANY motor, ANY controller tune, ANY flywheel inertia, ANY drag torque (cooling impellers and whatnot) - Doesn't matter one bit (well, as long as the drive is capable of reaching that speed at all). The feed delay optimizes itself on the fly for each spinup.
The main control logic is also extremely simple. Simpler than scheduled delays.
There is only one set of parameters left to adjust - a speed offset margin, which is a margin of error within which the speed is considered to have reached the setpoint. This accounts for achievable flywheel speed control loop performance, and it also provides a means of compensating for continuing wheel acceleration within the mechanical travel time of the bolt; while still reliably inhibiting feeding if the speed is out of range enough to produce a tangibly wonky shot, let alone is grossly unsafe to feed at. In this version, a linear (which is perhaps not the correct curve to apply, but whatever, works well enough) interpolation is provided between endpoints based on the current speed setpoint - for which the idea is to account for the greater torque and thus acceleration available from a real drive at lower speeds.
- Implementation of flywheel speed limit (tournament lock) with secured board-mounted potentiometer (i.e. requires tools to access)
This is self-explanatory.
- Implementation of user analog potentiometer and selector input devices
This is self-explanatory. See selective fire discussion below.
- Implementation of user speed configuration from minimum up to tournament lock setting
At boot time, if the trigger is not down, the blaster defaults to the tournament lock setting.
If the trigger is down, the blaster enters speed configuration mode. The flywheels spin continously for audible user feedback of the speed, while the analog knob varies speed between minimum speed and the tournament lock setting. Releasing the trigger enters normal operation at the current speed setpoint until the next power-up. This avoids accidental speed changes.
- Implementation of selective fire
Can you hear the crackling of ice forming? Because hell might be freezing over.
Yes, I put a selector on a T19.
This was motivated by a few factors:
- Popular demand. T19 has started to gain traction among locals as a platform - some of them are not full auto natives.
- Future flexibility of having the control itself - we aren't just limited to selecting burst modes! What if we want to select profiles that also contain speed, ROF, and offset margin ("snappiness level") combinations for various situations? But there is a lot more than that under software control. Maybe we could have a HvZ stealth mode that sets a low wheel speed, limits the voltage command of flywheel motors to hush the magnetostriction under high motor current on startup, sets a moderate ROF so the bolt doesn't go clack when it cycles, and sets 1/16 microstepping to quieten the bolt motor? Well that can be done now. If I have a need for something like that, I now have enough UI means available to turn it on and off. That's a way of justifying what I long saw as a superfluous control.
Straightforward - the logic is like a firearm's disconnector implemented in software and has the same behaviors as such a mechanical device. There is no paintball-style shot buffering/queuing mechanic, and I don't personally think there ought to be, although if any game organizer starts to push semi-auto-only rules I am totally open to implementing that along with 2 finger triggers!
Something to note is that I still have true full auto. Full auto is NOT a 99 or 999 shot burst (FDL, for instance). For each selector setting, there is an index in both a boolean isBursts and an integer bursts array. The first one is a mask for the disconnector and if that index is 0, the burst disconnecting logic is completely out of gear and it will fire truly ad infinitum. Does that practically matter? No, not at all.
Stock modes are full auto, 2 round burst, semi. I get a lot of demand and a lot of positive feedback on the 2 round burst.
- Implementation of live ROF adjustment
During normal operation, the analog knob sets ROF between the configured minimum and maximum at compile time.
ROF adjustment is linear, ROF adjustments take effect immediately at any time during the idle state (not firing), and ROF range endpoints are now configured in RPM rather than obtusely in microseconds per subcommutation. There is still a strict reliability limit to be set per-motor (88uS for the usual OSM 17HS16-2004S1). If you want to run it up to there, just set the maxROF a little beyond there.
ROF is a maximum bolt travel speed adjustment, and applies to all fire control modes at all times. Thus, it allows both changing the cyclic rate of fire, including for bursts, and changing the bolt speed and force (inversely related) - not that crappy ammo usually poses any problem in the field even at max ROF settings.