Bug TPA Flipper Bug Confirmed

[Matt McIrvin]

Presumably there's a sampling rate in the physics engine, if the ball motion somehow exceeds the speed of that, it could pass through a solid object before the engine has acknowledged the collision.

While I don't know how the actual physics engine works, what I imagine is that it'd detect the collision if the unperturbed path of the object would have intersected the surface since the last time step. That, in itself, wouldn't be prone to frame-rate problems.

But it wouldn't be efficient to check the path against every single surface on the playfield in every time step, so for a given surface, there's probably some kind of bounding box it has to be inside to even get checked. And if it were to fly entirely through that in between time steps, you'd have a problem.

 

[Huffster]

Fantastic work superballs. Nice video capture and editing of the slowed down version to really show what's happening. It should be very useful to the dev team. Too bad you didn't capture at 60fps, but this still show the issue well.

It does seem logical that the sample rate of the physics may be too slow. I know that in Forza Motorsport 3 & 4 for the xbox 360 platform the physics engine runs at 360fps, while the display runs at 60fps.

 

[Sean DonCarlos]

It does seem logical that the sample rate of the physics may be too slow. I know that in Forza Motorsport 3 & 4 for the xbox 360 platform the physics engine runs at 360fps, while the display runs at 60fps.

Running a physics engine at 360 Hz seems excessive, unless the calculations have been considerably simplified and/or the number of calculable objects artificially limited. Forza probably does not have to run its calcs to quite the same degree of accuracy, either - you can afford to be several centimeters off in detecting collision in a racing simulation, while 1 millimeter might be the difference between a successful save and a SDTM drain. (Note also that pinballs and racecars are capable of moving at similar speeds!)

I agree that upping TPA's physics sampling rate to 120 Hz - assuming it's not already there - might help fix collision issues in general. Cirqus Voltaire seems to be made of Swiss cheese, and I managed to forcibly eject a ball through the back wall of TOTAN the other day. But playing with the physics engine is a delicate business, and you'd introduce more fragmentation of TPA's code structure, as the mobile devices almost certainly aren't going to be able to use a 120-Hz physics engine without experiencing a severe performance hit on the display side.

Then there's all the arcana you can get into with bounding box optimizations and so forth, but that's Bobby King's place to hold court, not mine.

 

[Huffster]

I agree 360Hz may be too high. I was just making a point that other games do have their physics running faster than the games visual frame rate. BTW: the following forum post (another forum) says the new Forza Horizon game physics will also run at 360Hz. LINK But, that game will have a visual frame rate of only 30fps.

 

[RKade8583]

I KNEW IT WASN'T JUST ME! I remember posting this in BH waaay back when and I couldn't provide a vid. Now maybe this can finally be fixed.

 

[Dumahim]

Forza needs the high sampling rate not for collisions, but for physics. Realistic physics in racing games is very difficult. So it needs the high sampling to model suspension travel, tire deformation, grip levels, and how heat and road irregularities affect it all.

I suspect it’s somewhat low in TPA, which is probably why it seems like I can never hit certain shots from repeatable angles. Like a ball kickout that hits the flipper the same way every time, in TPA it seems like there’s a limited number of possible hit locations.

 

[Sean DonCarlos]

Forza needs the high sampling rate not for collisions, but for physics.

I bet if I played Forza, it would need the high sampling rate for collisions!

 

[Dumahim]

You'd have to keep up with the other cars in order to worry about hitting them. Many of the tracks have a lot of runoff area to avoid hitting walls.

 

[Huffster]

Forza needs the high sampling rate not for collisions, but for physics.

Collision detection in TPA is a part of its physics. Hope for a FarSight employee, most likely Mike, to chime in on what they've found out and if they've been able to duplicate in the studio and their theory to what's happening with thier physics/framework as seen in the video.

 

[Gord Lacey]

I think this is more of a Bobby question. I'll forward the thread to him and see if he can chime in.

 

[Bobby King]

This is something we've been working on (and improving) since the original Gottlieb collection.

In this video, the ball did not "collide" with the top side of the tip of flipper and slightly deflected down off of the back side. Had the ball been halfway through the flipper (below the upper collision and above the back side), the ball would have behaved strangely (moving it down fast or up at a weird angle). You guys have probably seen this behavior. Thankfully, it's pretty rare.

The tip (midway through moving up) is the point at which the flipper collision is moving the fastest so this is where the ball is most likely to miss colliding. Our game typically runs at 60 fps which is fast enough for displaying graphics, but not fast enough for our physics collisions to work perfectly - so our flipper physics collisions step through what happens between frames. This is a setting. However, increasing the steps to where they need to be to prevent the ball from ever passing through the flipper tips hurts the framerate on devices that are CPU bound (as this is a CPU intensive calculation).

There is also setting for the speed of the flipper.

Ideally, we are aiming to increase the number of steps (physics calculations between frames by optimizing this code) and speed up the flippers OR completely rewrite the way deflection angle hits off the flipper are calculated.

 

[Dumahim]

Are there consequences to allowing faster physics for systems that can handle it, (hopefully) such as consoles?

 

[bossyman15]

I never seen this. WHY HAVE I NOT SEEN THIS!?

 

[superballs]

Bobby,

Thank you for the insight on why this is happening,
I don't recall ever seeing this happen on any of the mobile devices. I've played a bit on the Samsung Galaxy SIII and extensively on the iPhone 4 (basically played MM from iOS launch to XBOX launch). If this is in fact happening on the mobile platforms I'm hoping someone will chime in and correct me on that.

Is it possible to change how often collision detection is run on the flippers on the fly? If so, could it be set higher if the ball is below a certain level on the playfield? Something close as in wherever would be the theoretical maximum distance the ball could make to to the flippers from in 1/60th of a second?

I know squat about programming and physics routines, but I would hope a solution like that could have minimal impact on the graphics engine, even maybe tie it to ball speed.

While we're at it, would it be possible to work on the flipper mechanics in order to make drop catches, flip catches and cradle seperation a bit more feasable? I find them quite difficult in TPA and they are some of the few things I'm actually very good at IRL.

Thanks in advance and I hope I can be of some assistance.

 

[Pod]

I only play on iphone4 and while I've never seen the extreme case in the first post I do get it to go through the tips of the flippers all the time.
Its actually starting to really annoy me. I think in a typical 45 min session (my tube journey) It'll happen to me 2 or 3 times.
It'll be exacerbated because recently I'm trying to tick off some of the harder wizard goals and so I'm doing lots of flipper transfers and ball control.
Most of the time I see it is when doing tip to tip saves when the ball is heading for the drain at a narrow angle.

If it matters I've recently been playing a lot of Funhouse and Monster Bash but I've also seen it on the other tables.

Pod

 

[FurVid]

Yeah, I wouldn't say this is rare at all. Happens to me pretty much every time I play for more than 30 minutes or so. Seems to have been there on all versions, all tables, and all platforms (I've tried all but android). If people aren't seeing this, I expect it's because they just aren't flipping the second flipper in these cases. After 3+ decades of play, it's instinctive for me at this point to flip both flippers in sequence when the ball comes down like that - otherwise, you run the risk of the ball shooting up the opposite inlane and then down the outlane. This almost never happened the PHOF games, at least the later ones that I played more of. There were some weird physics sometimes, like having the ball get momentarily stuck inside a flipper before exiting (often through the bottom), but they were very rare. This new problem is so common I'd say it's basically the norm - seems like I'll lose the ball in these situations more often than I save it. I'd rather not try to train myself to stop flipping in these cases, so looking forward to this being fixed, or at least reduced.

In addition to adding more steps when moving the flippers, I wonder if it'd help to change the sequence so that the new flipper position is in place before the old one is removed. Then instead of there being an instant in which the flipper vanishes, there'd instead be an instant in which the same flipper exists twice, making it much less likely the ball would ignore both.

 

[an ox]

I've kinda come to rely on a particular manifestation of this glitch, when the ball drains from down the hole in BH, if it passes underneath (actually above as you're viewing it) the left flipper you can generally flip it back into play through the flipper.

I'll miss that little advantage when it's fixed, but by then I'll be so chuffed that the gate behaviour is back to normal that I won't mind that much. Because by then the gate behaviour will be back to normal. yes. It will.

 

[SKILL_SHOT]

I experienced another flipper issue twice today at first I wasnt sure but the second time was obvious! On CV for PS3 I trapped a ball with left flipper and had a second ball approaching left flipper I shot the ball as they made contact 1 ball goes up the other goes through the flipper draining.

 

[Gordon]

self bug on my second game Monster Bash

 

[Crush3d_Turtle]

I can also confirm that this bug occurs on the PS Vita as it cost me the Battle for the Kingdom on MM last night. I had two balls trapped on the left flipper and I shot the first one up the table with a quick flip and it got caught behind the right troll as I had hoped giving me single ball play without that 10 second timer. The second ball was still trapped and I tried to do a quick post transfer to get the ball over to the right flipper. I hit the button and the ball jumped up and hit the post and came straight down through the left flipper and drained.