AFM/MM Remake Flipper Delay


The question isn’t whether people can be 5ms accurate, but whether a 5ms variance in when the flipper energizes has a notable effect on shot trajectory. Anyone know roughly how fast a ball is moving down an inlane off a ramp? If you use that and interpolate over the flipper length (approximating the range of angles the flipper can shoot as being linearly along the flipper?) you’d be able to find how much of an angular change 5ms can cause

They post times a WPC game too and the delay is sub-millisecond

Of course there’s a transistor, but the thing is that those transistors are software controlled, whereas older games could have them wired directly, or at worst definitely via an interrupt.


Note to self; don’t try alley passes or live catches on WMS remakes.


His comments ‘I would guess…’ and his mistake in his initial testing don’t inspire a lot of confidence. He also mentioned that he didn’t try adjusting the power level for the flippers on the remake. Would be nice if someone a little more technical would follow up on his findings.

You mean a pre-WPC95 game? I’m not seeing that. Just WPC95 games. The older games with the Fliptronics 2 board have always felt the snappiest to me. WPC95 flippers are fine, but not as snappy as the older games IMO.


He posted his timings for his original AFM, and they were sub-ms. I’ve never done work with WPC-95 flippers, so I’m not sure how they’re wired, but on Fliptronic the flipper button signals go directly into the fliptronics board. Pre-fliptronic should have an even faster response time as there’s no logic circuitry involved at all, just the flipper switches directly sending high voltage power. Of course, there are other factors involved with that, so fliptronics may end up being snappier in the end. Either way, unless there’s a circuit running directly from the flipper switch to the transistor on the remake, it’s going through the CPU which is going to add some lag. If they don’t respond via an interrupt, but use polling, then there’s going to be even more lag, and if the signal actually gets sent into the emulator then that would definitely account for a variable 1-5ms lag


What was the mistake in his testing? As far as flipper flopping that’s how I remember that game when it was new. Also Theatre of magic, MM and Indy 500 had flopped flippers from factory.


I decided far too many people have speculated about what is an “acceptable” variation without doing any experiments, so here was mine.

I took a video of TWD with the ball taking the feed from the ramp on the right inlane down through the flipper and into the drain.

Then, I took a video editor and slowed it down by 0.3x three times. This means the final video was at 27/1000 speed.

Watching the video, it took 3 seconds to roll from the riot sweet spot to the barn sweet spot. This is about 81 ms.

At the distance of the entrance to the left ramp on TWD, there is about 5 pinball width of arc between those reference lines.

The “width” of arc that makes a shot is probably about 2/3 of a ball for the average shot ( maybe only 1/2 for AFM lock shot, but I’ll use 2/3 to be generous ).

81ms / 5 * (2 / 3) ~ 11ms

So, 11ms for the “width” of making a shot. The post on pinside claims variability in timing of button to flip of + or - 1.3ms.

So a variability of about 10% the width of a shot. Not a huge error, but definitely in the ballpark of noticeable. The people who are completely dismissive of the variability are wrong, IMO. The most accurate shooters could definitely suspect something is off.

Of course, the framerate on my phone isn’t great enough to have the smoothest source video, and different shots have different widths, etc etc etc. I’d be interested to see someone else try the same experiment to see if I’m way off or not.

Myth-busters rating: Plausible.

(Sadly, I did not find a reason to blow up my TWD at the end of the experiment like the myth-busters would)


End of the first post:

EDIT: Everything I originally said is correct. But it was pointed out that I actually measured the voltage difference between the power and hold coil. The latency issue remains the same as described here but I repeated the analysis using the ground and power leads of the coil. See post # 172. text

Not saying that his findings are wrong. Just would rather have them confirmed by a second party.


Josh shared my photo but I think it’s better to keep the discussion here than on my Facebook. Plus I posted that immediately after getting eliminated so I was a bit salty.

My main concern (which is probably something in my head) was all the trolling I was getting when I was complaining about the MMRs and AFMRs having flippers that felt weird. I’ve adjusted by completely changing my strategies on these games but I think it’s still important to make a distinction between these two games for tournaments while the games play noticeably differently.

I’m not going to pretend I understand the flipper. I took Electricity and Magnetism classes in college. I understand the idea of a magnetic field being created by induction. But this doesn’t affect how I actually think of a pinball machine. Pinball machines have a button that makes the flipper go. I let my subconscious do what it wants to resolve this and I don’t worry about it too much.

I also don’t understand how certain things work in pinball. I don’t understand Live Catches. Does anyone understand them? I think we have some sort of collective theory about a dampening effect at the end of the flipper stroke that transfers the balls momentum into ???

But here are the following things I’ve observed on both MMR and AFMR:

It’s nearly impossible to loop pass. Dropping the flipper is unpredictable. It happens at a random point after I let go of the button and there’s no feel for it.

Drop catching is weird. You can do it though.

Live catching is ok when the ball is mostly vertical and moving slowly on the playfield. It’s a little less crispy from an orbit feed, especially at speed. This may just be the kickouts.

Tap Passing is easy to do by accident.

Inlane shooting is way easy to shoot by accident.

A common drain on both of these games is an accidental Shatz (inlane pass) which travels across the playfield and out the outlane on the other side of the playfield with 0 chance to do anything.

All ramps are easier to hit on both games. Especially by accident.

All orbits are awkward. Shots rattle around

Lock shot on Medieval has no feel to it. It does not seem like a shot you make through skill. The ball doesn’t look like it can really make it’s way in there from a trap.

Lock shot on AFM wasn’t possible to hit when I was flipping as early as I thought was possible. (It’s possible that I subconsciously wasn’t letting myself flip early enough to compensate for some lag.)


Exactly. I was shaking my head all along browsing though the discussion on Pinside.

Added to this, the power to the coils is an oscillating 100/120 Hz signal. Having 10ms/8.3ms period times. Divide this by two is 5ms/4.15ms high-to-low time to take into account.

The seed to the speculative arguments and bashing of the remakes is the guys measurement of an original AFM (seemingly without any delay). Which must be wrong. Or some wierd stoke of coincidence. Maybe some other trigger point than what is assumed (signal ripple).

In any case, I would never make measurements on the coil as the only source of analysis of timing. I would start on control signal further upstream. Or maybe outside the driver with the coil replaced for a pull-up resistor.

For all we know, if the flippers do not feel consistant, or same as the original games, it may lie in properties of any of the electro/mechanical components, or control logic, of these designs. I see the driver is a FET type compared to a transistor on the original games.


I don’t know how the rise time of switched DC affects the magnetic flux of a coil. But it is guaranteed that, if the switching transistor turns on near the falling edge of the signal, the coil will take longer to fully energise than if the the transistor turns on near the rising edge.

As you say, that delay can be (presumably randomly) in the vicinity of 4-5 ms. Now add to that an up to 5 ms processing delay in getting the transistor to turn on in the first place, and we are looking at a worst case delay of 10 ms.

From having done lots of drumming, I know that a delay of 10 ms is easily perceivable, and not just by experts. Also, 10 ms is almost the total “width of a shot” reported by @johnnyfive.


BTW, there is a fairly easy way to measure all this empirically. Take an old AFM and a new AFMR. Mount a switch underneath the playfield that activates at start-of-stroke. Connect a circuit to detect switch closure at the flipper button. When the switch closes, run a timer to measure the delay between the switch closing and start-of-stroke.

Take a few hundred measurements on both machines under varying conditions of load. In particular, have some measurements that are run during multi-ball, when the interrupt load on the CPU is higher, there are more demands on the display and light effects, etc.

Now run an analysis on the figures for both machines to get best and worst case, median, average, and variance.


If the flip speed rise/fall is inconsistent, I’d say that’s more of a factor than just a difference - there are most likely differences from machine to machine regardless of original vs. remake. At least if it’s some kind of consistent, you can adjust correctly.


For whatever reason I’ve always found AFMR easier to play than MMR, and AFMR easier than a standard AFM. Things just feel… smoother to me.


Hmm, shouldn’t matter as both coils fire at the same time if it’s a standard williams setup.


The @johnnyfive experiment made me cringe on so many levels. It is a start in the right direction but this would be better if distances were not measured in pinball widths. [quote=“johnnyfive, post:26, topic:3212”]
At the distance of the entrance to the left ramp on TWD, there is about 5 pinball width of arc between those reference lines.

huh? pinball width of arc? is this an angle?


My trig is rusty, but it’s the equivalent of the distance from the flipper to the ramp entrance times sin(angle), I think?

In more concrete terms, if you took an arc from the barn shot to the riot shot roughly at the entrance of the ramp, the length of that arc is about 5 pinballs wide.


The beginning of this problem from the mechanical perspective starts with the acceleration of a pinball going down an inlane (or from rest on a flipper), how much distance is travelled in 5ms?

Then, measure where a flipper causes a pinball to go when at “x” then at “x+(5ms of distance)” and compare.


Later he said he tested a Hobbit and there was virtually no delay. The other Keith replied.

As proud as I am of the work I and others have done to get a great-feeling flipper system, it’s nigh impossible for me to believe that the response time is instantaneous.

First, a player has to hit a button. The I/O board has to read that button. The I/O board sends a usb packet to the cpu. The cpu/driver reads the packet. The game reads the packet from the driver. The game processes that fact and decides a flipper needs to be energized. The game sends data to the driver. The driver/CPU sends the USB packet to the I/O board. The I/O board reads the packet. The I/O board acts on the information to energize the coil.

In a best-case scenario, you have to be looking at an average of around .5ms from button to coil, because who knows when that data will be read and sent by the I/O board. I would have to think with all the different timings involved, there’d be 1-2ms between the player pushing a button and the coil starting to energize.

The original poster said he contacted CG and they said they will look into it. Will be interesting to see what they say.


Once again, it’s not whether or not delay exists. It’s whether the delay is inconsistent.


Just FYI from Duba posting over on Pinside . . .

"We spent the majority of yesterday looking into the concerns raised by twenty84 regarding differences in flipper responsiveness. I want to provide a brief update on what we have learned so far.

We repeated the test on two MMR and found the result to be similar to those of twenty84.

The results of the tests performed by twenty84 on the original AFM are not consistent with our understanding of how the original WPC95 hardware functions. It is our understanding that WPC95 polls the flipper switches every 4 milliseconds; therefore, we would expect the delay to range from almost zero to four milliseconds (in theory it can never be zero).
We then tested three separate original games and found delays of 0-4 milliseconds which is consistent with our understanding of the WPC95 system.

It would be greatly appreciated if twenty84 could retest the original AFM in the manner presented below:
Connect probe A (trigger) to SW1 (upper flipper) or SW2 (lower flipper) on the right flipper opto board, whichever is activated first in Switch Edges test. Connect ground to ground braid in cabinet. [Please note, the plastic flipper opto interrupter can warp over time. The first opto read will fire the coil]

Connect probe B to the middle terminal on the right flipper coil. You do not need to connect ground from probe B to anything.

Perhaps, others can perform the same tests on their original WPC95 games.

We are currently in the process of reviewing the original Williams source code to confirm our understanding of the frequency at which the system reads the flipper switches. Additionally, I intend to borrow another oscilloscope and validate our findings with the second piece of test equipment."