visualizing smoothing in mousetester - Page 3 - Overclock.net - An Overclocking Community
Forum Jump: 

visualizing smoothing in mousetester

Reply
 
Thread Tools
post #21 of 53 (permalink) Old 07-20-2015, 04:17 AM - Thread Starter
lololol
 
qsxcv's Avatar
 
Join Date: Feb 2014
Posts: 4,288
Rep: 371 (Unique: 153)
if the spreading out of those blips is what smoothing is doing, it's definitely not as straightforward as averaging out data (like what the finalmouse mcu does). notice how the relative heights of the blips don't change that much from dpi to dpi.

too busy to check forums as regularly
pm me if i forget to respond
qsxcv is offline  
Sponsored Links
Advertisement
 
post #22 of 53 (permalink) Old 07-20-2015, 04:47 AM
New to Overclock.net
 
VolsAndJezuz's Avatar
 
Join Date: Nov 2014
Posts: 779
Rep: 95 (Unique: 53)
Implying that you would expect the the heights of the blips to be smaller at higher, more smoothed DPI, if the smoothing was a simple data averaging, because they would be averaged in with previous non-blip data? Maybe the TORQ X5's MCU disregards previous data upon framerate transition and starts the smoothing over fresh. That could explain the seemingly uniform relative blip heights in your graphs despite using straightforward averaging like FM.

sourceGL software
CS:GO and TF2 sensitivity - 1 @ 850DPI, 48.9cm per 360°
baby boy
(22 items)
CPU
Intel Core i7-4790K @ 4.8GHz - de-lidded (w/ ROCKIT COOL Copper IHS)
Motherboard
Gigabyte GA-Z97X-UD5H-BK
GPU
EVGA GTX 980 Ti Classified @ 1532MHz core, 4192MHz memory
RAM
OCZ Reaper HPC 2GB [x4] @ 2400MHz, 8-12-8-28 1T
Hard Drive
Samsung 850 PRO 256GB [x2]
Optical Drive
Lite-On DL-8A4SH
Power Supply
EVGA Supernova PS 1000W
Cooling
Cooler Master Nepton 140XL [x2] (w/ EK-Vardar F3-140ER fans)
Cooling
Thermal Grizzly Conductonaut liquid metal thermal compound
Cooling
Fujipoly ModRight Extreme thermal pads
Case
SilverStone Fortress FT05B
Operating System
Windows 7 Ultimate 64-bit (w/ NVIDIA Driver 399.24)
Monitor
BenQ XL2411Z 24" @ 144Hz
Keyboard
Das Keyboard Model S Ultimate (w/ MX Green modifiers and PBT keycaps)
Mouse
Logitech G900 - de-braided and de-battery'd (w/ Hotline Games Master feet)
Mousepad
Logitech G640 3mm
Audio
ASUS Xonar Essence STX II (w/ UNi modded driver)
Audio
beyerdynamic DT 880 PRO headphones (w/ AntLion ModMic 4.0)
Audio
Blue Sky eXo 160W 2.1 studio monitors
Other
Solarflare SFN6122F network adapter
Other
MikroTik RouterBoard RB751G-2HnD
Other
CyberPower 1350AVR UPS
▲ hide details ▲
VolsAndJezuz is offline  
post #23 of 53 (permalink) Old 07-23-2015, 08:04 PM - Thread Starter
lololol
 
qsxcv's Avatar
 
Join Date: Feb 2014
Posts: 4,288
Rep: 371 (Unique: 153)
Quote:
Maybe the TORQ X5's MCU disregards previous data upon framerate transition and starts the smoothing over fresh. That could explain the seemingly uniform relative blip heights in your graphs despite using straightforward averaging like FM.
maybe, but notice how the reported speed right before the 1st blip is 0.2m/s for low dpi, and ~0.1m/s or so for the highest dpi.

maybe something like this: smoothing works by keeping track of the motion data for some number of previous frames. as frame rate increases, this number increases as well in order to maintain a similar level of total smoothing. however it doesn't take into account that immediately after a framerate transition, the previous frames are still taken during the lower-framerate period and thus correspond to a longer amount of time.


here's a hypothetical example:
suppose the motion is a constant 6 counts per millisecond.
let the sensor's smoothing be configured like this:
at 2000fps, the smoothing is simply averaging the last 4 frames
at 6000fps, the smoothing is simply averaging the last 12 frames.

when the sensor is only at 2000fps, the motion data from each frame is always 3. if there is smoothing, well the average of 3's is still 3, so nothing happens. at 1000hz polling, we'd see a flat line of 6 in mousetester xcounts, since you get 2 frames per usb poll.

when the sensor is only at 6000fps, the motion data from each frame is always 1, and the average of the last 12 frames is 1. again, at 1000hz polling, we'd see a flat line of 6, since you get 6 frames per usb poll.

now what happens during a framerate transition?
the raw motion data would look like
<2000fps region> <6000fps region>
3, 3, 3, 3, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1
after the first 6000fps frame, the smoothing averages the last 12 frames, but doesnt take into account that the previous frames were 2000fps frames and thus correspond to larger amounts of motion. so it takes the average of the 1 and eleven 3's from those previous frames. the next frame takes the average of 2 1's and 10 3's, etc...

thus the smoothed data from the sensor looks like
<2000fps region> <6000fps region>
3, 3, 3, 3, 3, 2.83, 2.67, 2.5, 2.33, 2.17, 2, 1.83, 1.67, 1.5, 1.33, 1.17, 1, 1, 1, ...
| | | | |
usb polls come every two frames during the 2000fps region and every 6 frames during the 6000fps region.
so the mousetester data looks like (assume whoever coded the mcu is sane and didn't add mcu postprocessing/filtering tongue.gif)
6 6 6 14.5 8.5 6 6 6 6 6 6...

of course this is all ignoring the fact that actual sensors deal with integers not real numbers

edit: crap formatting is all messed up... quote this post to see how i intended for the numbers to line up

i think this is a very plausible explanation for the blips. and it also explains why the height of the blips don't change much. given this model, the height of the blips depends (mostly) on the ratio of the framerates during the transitions.

guess it's time to measure the torq x5 framerates again tongue.gif rip mousefeet

too busy to check forums as regularly
pm me if i forget to respond
qsxcv is offline  
Sponsored Links
Advertisement
 
post #24 of 53 (permalink) Old 07-23-2015, 09:16 PM - Thread Starter
lololol
 
qsxcv's Avatar
 
Join Date: Feb 2014
Posts: 4,288
Rep: 371 (Unique: 153)
btw i just abx'd 1550 vs 1600 dpi for the torq x5, 10 out of 10 times, by looking at jitter in mspaint diagonal lines when the mouse is lifted slightly:


3150 vs 3200 seems to be even more obvious... but i'm too lazy to do another abx test to make sure i'm not experiencing confirmation bias or something

i'm 99% sure this, and the plots of the blips, are due to smoothing.

too busy to check forums as regularly
pm me if i forget to respond
qsxcv is offline  
post #25 of 53 (permalink) Old 07-25-2015, 12:02 AM - Thread Starter
lololol
 
qsxcv's Avatar
 
Join Date: Feb 2014
Posts: 4,288
Rep: 371 (Unique: 153)
checked framerates of torq x5 by connecting oscilloscope to led

<0.2m/s: 3050
0.2-0.8m/s: 7100
>0.8m/s: can't tell, led is no longer pwm. estimation from mousetester dots: 12250

ratio of first transition is 2x.
ratio of 2nd transition is 1.73x

which more or less matches the relative heights of the blips.


anyone with one of the 3988 deathadders care to plot some medium speed (~1m/s max) swipes in mousetester?
adjust the data start/end so that it looks something like this:
mlF7Ao7.png

too busy to check forums as regularly
pm me if i forget to respond
qsxcv is offline  
post #26 of 53 (permalink) Old 07-25-2015, 01:51 AM
New to Overclock.net
 
Thunderbringer's Avatar
 
Join Date: Apr 2012
Posts: 360
Rep: 49 (Unique: 32)


Mouse is set at 1800 CPI & 1000 Hz.

Edit:

@OP
DA2013! (Click to show)


jaja, blabla
Thunderbringer is offline  
post #27 of 53 (permalink) Old 07-25-2015, 02:22 AM - Thread Starter
lololol
 
qsxcv's Avatar
 
Join Date: Feb 2014
Posts: 4,288
Rep: 371 (Unique: 153)
thanks. that looks really similar to my 1600dpi plot. which model of mouse is it exactly?

too busy to check forums as regularly
pm me if i forget to respond
qsxcv is offline  
post #28 of 53 (permalink) Old 07-25-2015, 02:28 AM
New to Overclock.net
 
Thunderbringer's Avatar
 
Join Date: Apr 2012
Posts: 360
Rep: 49 (Unique: 32)
Quote:
anyone with one of the 3988 deathadders care to plot some medium speed
tongue.gif Its a DA2013. Cool thread/idea btw! smile.gif

jaja, blabla
Thunderbringer is offline  
post #29 of 53 (permalink) Old 07-25-2015, 03:05 AM - Thread Starter
lololol
 
qsxcv's Avatar
 
Join Date: Feb 2014
Posts: 4,288
Rep: 371 (Unique: 153)
oh i thought there were other 3988 deathadders. and also forgot that da chroma has 3989... but iirc that's basically the same thing as 3988 except srom.
could you also plot at 800,1550,1600,3150,3200, and 6400dpi? (or just 1550 if you don't want to do all of them. i suspect it behaves exactly like the torq x5, unless synapse screws something up.)

too busy to check forums as regularly
pm me if i forget to respond
qsxcv is offline  
post #30 of 53 (permalink) Old 07-25-2015, 09:01 AM
New to Overclock.net
 
Thunderbringer's Avatar
 
Join Date: Apr 2012
Posts: 360
Rep: 49 (Unique: 32)
Unfortuantely no 50 CPI steps possible in Synapse driver:
Warning: Spoiler! (Click to show)
800 CPI

1600 CPI

1800 CPI again, this time with Synapse installed, post above with 1800 CPI was w/o Synapse installed

3200 CPI

6400 CPI

jaja, blabla
Thunderbringer is offline  
Reply

Quick Reply
Message:
Options

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in



Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off