357 replies to this topic

[paulstevens]

One last measurement for the night.  I hooked the old scope up to the left flipper on a Black Hole to measure the time between flipper energizing to EOS switch opening. 

 

After about 15uS (microseconds, or 0.015mS) the EOS switch is opened.

 

Measuring directly across the coil itself, I measured the pulse width of when there is high voltage across the flipper coil to when the high voltage is cut off - EOS switch opened.  I can see switch bounce, but clearly and consistently the EOS is definitely opened by 15uS and the hold coil is engaged. 

 

The VPX F11 measurements in AfM, Fathom and Bad Cats have EOS completing at 30mS best case, compared to 15uS measured across a real coil, a factor of 2000 difference.

[Drybonz]

When I first got a 144hz monitor and placed it side by side with my old computer monitor the difference was amazing and I decided I could never go back to a 60hz monitor.  That's why I mentioned, above, if you have access to one somehow that you can compare... even if it's not a monitor you plan to use in your cab or whatever... if you get a chance to compare side by side it's worth it.

[Ben Logan]

Drybonz -- If I press f11 while playing a table and see that I'm only getting 75 fps, what happens if I engage Vsync on a 144hz monitor? Lots of stutter?

From reading your posts, I'm thinking that you generally leave vsync engaged. Does your graphics card consistently put out more than 144 fps?

Fps varies with my GTX750ti. Not sure if I should make the jump to a 144hz monitor without upgrading my graphics card first. And, like a lot of us... I'm pretty broke!

Thanks in advance for your response. Totally appreciate you for sharing so much of your experience.

[toxie]

you will get at least a bit smoother gameplay then (and slightly less latency), but of course only if you can drive the 144Hz you will get the full benefit of both smoother ball movement and less overall latency.

[Drybonz]

Hey Ben... Yeah, I always have vsync on for pinball, pc games or anything else like that.  I never pay attention to fps.  I tend to just obsess over the number if I start looking at it.  As long as everything is visibly smooth to me, I'm happy.  If there is a problem, I may look at the fps just as a benchmark, but if I can play a table or game without noticing any stutter or slowdown I'm happy.  I'm not saying people should operate this way, but this is how I do it. 

 

As far as being broke, I am right there with you... lol.  A monitor is a sizable purchase for me, so I make sure I read up and do research before I make a purchase.  Also, a monitor that works for my hybrid desktop setup isn't going to be something that would work for people with a large cabinet, but there is a whole spectrum of quality available for either.

 

*edit*  Also, toxie is right, but most cards that are pushing VPX tables at full speed should be fine... but if your system/card is struggling, picking up a better monitor is not going to fix it.

Edited by Drybonz, 28 November 2016 - 09:07 AM.

[paulstevens]

Sorry - I entered a post with an Excel table, but the formatting got lost. I could not delete the post outright, so I'm leaving this.  Sorry for the goof.

Edited by paulstevens, 29 November 2016 - 06:27 AM.

[paulstevens]

Long post - sorry.  Hopefully it is somewhat interesting to JimmyFingers (and others) as one of his posts got me thinking about this.

Looking at the PAPA flipper skills videos, I decided to try measuring some timings while my daughter tried flicking the flipper button to simulate a tap pass.  The goal was for her to press the button as lightly as possible while still registering the slightest of flipper movement.  This was done on a WPC-95 Monster Bash.

I only have a 2 channel scope, so I monitored the flipper button opto and the flipper coil solenoid drive signal. I could measure the following:
Button press to solenoid drive
Button press to button release
Button press to solenoid drive release

 

My forum ignorance is keeping me from figuring out how to format this better, so my apologies for the table below.  Hopefully you can make sense of the numbers

 

WPC-95 Monster Bash                                 
Button press to Flipper Energized      1.6           2.0   0.3   3.5     4.1    3.3     0.5         3.8    2.3    1.6    3.9   0.5   1.8   2.6    1.1      2.9         0.4
Button press to Button Release          2.2           8.8  10.3  10.4   9.0    11.9    0.9       12.9   22.9  11.6  13.6  2.6   2.4  10.1  12.5    10.3       1.8
Button press to Flipper de-energized 5.3/7.5* 10.1  12.4  15.6  11.8   15.1  4.6/8.5* 16.1   26.5  13.9  16.1  4.2   5.8  11.0  13.6  11.1/12.3*  4.4/7.1*
all times in milliseconds

 

My observations:

- as before, WPC introduces different "lags" between button press and when it fires the flipper.  It would seem to depend on when WPC samples the button press and then the CPU delays internally before the solenoid drive is activated
 

- My daughter has a softer touch than I do.  She could do a button press and release in under 1mS.  This is the time from when the button press interrupts the opto on the flipper opto board to when the button is released, and the opto path is no longer interrupted by the button mechanism.

- the bottom row shows that WPC delays are evident on de-energizing the flipper as well.  It depends on when WPC samples the button released, and then some CPU delay later WPC turns off the flipper solenoid drive signal.
 

- the "hard" flip case (22.9mS button press) was still not a full-on power flip but hard enough to smack the coil stop pretty loudly.
 

- in some instances, instead of the flipper de-energizing completely, it went to around half power (35V or so) for a few milliseconds before de-energizing completely.  I don't know why this happens.

For me, the take aways are:

- flipper button presses for tricks, tap passes, etc, have to be very short.  As such, for a true simulation, the button press has to be sampled at as high a rate as possible (1mS = 1000Hz, or faster) to detect the press and, very shortly after, the release.

- flipper button press to coil drive is even shorter - again suggesting a faster sample rate of the button is better.

I have not looked at WPC-95 button press to EOS activation yet, but it's not going to be slower than what is in this table.  It's only going to be faster.

If there is something else anyone wants measured, I can try to oblige.

I hope this stuff is useful.  If not I won't clutter the thread.

[toxie]

Interesting stuff for sure! One piece of the puzzle.. Thanks!

[toxie]

just for future reference: this thread is also interesting in the whole context: http://www.vpforums....showtopic=36421

[paulstevens]

A couple more measurements

 

WPC-95 Monster Bash, Right Flipper

 

Time from coil-energized to EOS active (upswing):  19mS-24mS 

(19mS when I see EOS first closure, 24mS when it is done bouncing and solidly closed.)

 

Time from when coil DE-energizes to EOS open (flipper released, downswing)   89mS

[Iain1986]

Howdy

Sorry I don't mean to drag up old posts but I was pretty new to VP around January and I've seen this thread linked to a fair bit with regards to the investigations that went into flipper/input lag.

It looks like things just came to a fairly abrupt stop around Dec last year....I wondered if there is anything I should be doing with regards to my VPX setup, or any script changes people do? Are people all using this "invisible flippers" thing and having flippers instantly react instead of waiting for the ROM? Or was that all kinda abandoned as just an experiment?

I'm on latest VPX and Pinmame so I assume the tweaks that went into 10.2 are present, and maybe the work that looked into getting pinmame to process solenoid flippers faster?

-Iain

[toxie]

All the (realistic/possible) VPX and VPM enhancements discussed here actually ended up in the christmas releases.. Then there were some further minor adjustments in the june releases (VPX 10.3 and VPM 2.9)..

 

As for the script changes (like not waiting for the VPM roundtrip, etc): These are up to the table authors to implement, or if you're up for the challenge, you need to modify the table scripts yourself.. :/

 

Then, all the system configuration things like increase USB polling rate and buying/using a 120Hz monitor are of course also still valid and will always provide less flipper lag..

[vogliadicane]

^^ btw does anybody know about a 43 inch 4k Monitor with 120 Hz? Didn't find something yet...

[toxie]

i think you cannot have all three at once yet.. :/

[BorgDog]

^^ btw does anybody know about a 43 inch 4k Monitor with 120 Hz? Didn't find something yet...

 

would be nice, I'd prefer a 42, or a 41 or a 40, to fit in a standard body cab heck I'd even down size to a 39 and have more bezels showing for a 4k 120 Hz monitor.  It also seems that anything they have that is getting close they keep putting a curve on it which I don't think would work to well (maybe it would, don't know)

[vogliadicane]

yes, true, don't think curved would look good.

 

@toxie you mean also not with a GTX 1080 Ti (or whatever is the mightiest graphic card atm)?

[Iain1986]

@toxie

Thanks for that, was just interested to know what came of it all. I'm on the latest of everything and actually I think I did notice an improvement a while back when I upgraded. Either way I don't feel like I have flipper lag issues so everything working great from my POV.

[mattbini]

Wow this is a long thread with some great ideas. Just wanted to add something that helped improve my lag massivly so that it was not noticable. I have an old q9650 processor max i can get for my old pc. I updated my gpu from gtx750ti to gtx1050ti. I found when i overclocked the gpu i could get more fps and have everything switched on to max but when i played the slight lag occured and it was even worse in future pinball. When i turned off the overclocking the lag went away. So my conclusion is that much as the graphics card can paint a very good picture, the cpu is incharge of puting everything together and if you give it too much to do it slows down a little causing lag, but still managing to look great and give higher fps. I want real feel and no stuttering so vsync has to be on for me in vpx and fp as otherwise i am only getting around 100fps on my set up on the high power tables and i feel you need over 200 fps to be able to get stuuter free ball without vcync
In sumary What works best for me with a low end cpu is the following and i dont get any noticable lag.

1. Do not overclock GPU.
2. Turn off table reflections in options menu, this seems to take a lot of energy from cpu.
3. Have vscyn on 1 and table in exclusive full screen. (Vsync is apower drain but a must if your getting fps below 200 for me).
4. Enable 4x brute anti lasing (another power drain but makes tables look real so imortant for me).

Hope this my help other guys with less powerfull cpu too.