Does Melee have input timing jitter?

[CoreyJL]

So for a while I've been building my own Smashbox-esque controller. Over time it has developed into a "Cheater-box" for fun.

I am currently working on a perfect wavedash macro, and that concept can be applied to multishining as well.

I have an older build of the controller that is a Frankenstein padhack of a GameCube controller. I used a Teensy 3.2 to interface with it. So essentially, I am using a microcontroller to "push" the buttons electronically for me with precise timings.

I am having some very interesting findings though, perhaps some of the experts can provide an explanation:

If I space out a jump input and a trigger input to exactly what it should be for a perfect wavedash for the respective character, it will only perfect wavedash about a third of the time, maybe less (I didn't do any sort of statistical analysis). Sometimes the airdodge would happen early and get eaten, and sometimes it would be a frame late. I believe on extremely rare occasions it could also be two frames late.

Is this a known issue, or has anyone tested this? I suppose my code and application could be at fault, but I am fairly certain that's not the case.

 

[Dolla Pills]

I've had similar issues working with macros on a controller modded with an Arduino that I was messing around with. For applications like adding or subtracting from the analog stick values it seemed to work okay, but anything that relied on pushing a button press on a strict timing was inconsistent at best

I do not know what the issue is though, maybe it's the board receiving an interrupt at a critical moment or maybe something else, I honestly have no clue

 

[CoreyJL]

I originally thought it was a problem like that. I'm using the NicoHood Nintendo library for my new build so I thought it had something to do with that, but when I use a regular official GameCube controller, the problem is still there.

Even running super simple code that simply jumps and then airdodges at a specific interval seems to give the same result. I honestly think that the game's input polling is really messed up. I haven't tested on an actual GameCube yet though, only Dolphin. I wonder if the same issue is on the regular GameCube.

If this really is a problem with the game, it is interesting to me that no one has noticed. I suppose it is hard to notice a +/- 1 frame jitter though. Maybe all these years people have just blamed themselves for missing a wavedash when really it was just bad luck.

 

[Dolla Pills]

I've been testing it on a GC/Wii myself. As far as I can tell it's almost certainly more of an issue with the micro controllers based off how inconsistent it is vs how consistent normal play is

 

[CoreyJL]

I really hope so. I would hate for it to be a problem with the game itself since it is such an amazing game.

I just really wonder what's going wrong. I'm using timing in microseconds, and none of my code is using interrupts as far as I know. It is hard for me to believe that a microcontroller would provide a time unit in microseconds that is up to ~16 milliseconds off in either direction.

 

[CoreyJL]

Upon further testing, I believe I have confirmed the problem is with the microcontroller. I tried it out in Project M and the problem is still there.

I really wonder if there is a way to overcome this problem. I'm not even sure what the cause is. If anyone has any advice I would appreciate it, although I know this is sort of a unique problem.

 

[Rococo]

Melee does have polling drift which causes a couple dead frames that don't allow inputs every 600-900 frames. That's obviously not significant enough to account for the errors that you are finding, but it could mess up a wavedash once in a blue moon even if you configured your microcontroller correctly.