cool to see DKF has branched out from just DK. I check in here and there to read news but never took the time to sign up.
figured I better chime in on the topic above.....I should probably rewrite that guidebook stuff, it has been a work in progress and it could use a final rewrite. sorry for the confusion.
Basically with Williams games that use blitter (not Defender or Stargate), the one that really suffers bad emulation is Robotron because it is so wicked fast with the movements and projectiles.
So here is the scoop. Sean Riddle, the hobby mamedev, started reprogramming MAME's williams code around v14x. He would make coding changes after doing benchtesting on his real boardset and or JROK pcb, and I'd play test it or compare to my real machine that I owned at the time.
by MAME 148-153 he had recoded everything that he could control in the mame world. and in general it is perfectly coded for accurate emulation. BUT it leaves THIS topic he couldn't code for, only hedge.
A USB interface doesn't take in nearly as many control inputs as a real Williams control interface board. So the CPU isn't quite as busy in mame as a real machine. So the only solution is to hedge a bit and tie up the CPU a fraction to account for the control inputs that simply don't exist in mame with a USB interface.
So the version of mame that I "tuned" at the guidebook will match the gameplay experience of a real arcade machine, which even Jeff Harrist has played and confirmed is undetectably accurate with his real machine. A quick mame download will also be at williamsplayersunite under the robo extras section by later tonight or tomorrow TRY OUT THE NEW TIE-DIE ROMSET, TOO.
long story short if you play old mame like v106, you will perceive that 153 is close to an arcade experience in comparison.
but if you get really good at robo and become aware of nuances you'll find that 153 is a titch harder than a real machine. noticeable in the wave 40-60 range on tank and brain stages.
MAME at mamedev is as good as it can get by pure coding so it'll forever be a bit harder than real machine, unless you hedge it with the williams.c adjustments I outlined.
Joust is a slight bit harder too but Joust is a slower game so it isn't as detectable, like you'd have to be in the wave 30-60 range to notice it and you'd have to be experienced enough to be aware of the variations.
PS-ArcadeSD is outstanding for vertical games some of which you mention above. But it is the worst out there for Williams emulation. 19-1 pcb does it better than arcadeSD. Clay has the timing info from Sean, but as noted above the pcb seems to have lost steam and i expect to never see him rewrite the Williams emulation. Zoo Keeper is a huge plus on ArcadeSD, though.