Overclock.net › Forums › Components › Mice › Why exactly does MLT04 have such low tracking rates?
New Posts  All Forums:Forum Nav:

Why exactly does MLT04 have such low tracking rates?

post #1 of 8
Thread Starter 
Why does it have such low PCS and malfunction speed, even when having high FPS and lower than pixel-array cpi?
Is it a tradeoff of achieving such low LOD?

I read that angle snapping can increase tracking rates:
Quote:
Originally Posted by popups View Post

...It looks like the 3310 and 3366 have some angle snapping. That could be in an effort to increase the malfunction speed for marketing purposes. Older mice (like the Razer Diamondback) with angle snapping had very high malfunction speeds. The 3090 doesn't really rely on this (as far as I can tell).

How exactly does that work? Why does it increase tracking rate?
Shouldn't it just get rid of "other axis" coordinates, if "moving within a certain angle" (for lack of better words)?
For example (5; 1) becoming (5; 0).
MLT mice feel like they have the lowest angle snapping of all mice I tried..
As much as sensor position is tendentially more forward than other mice, even adjusting my grip on the DA and G400 I can't get that "freedom of movement feel".

Does the SoC design limit the micro-controller's capabilities in storing movement data?
Or better, is the specific controller in the MLT04 package not good for high tracking speed?

Is it because the sensor sends data as soon as it can determine movement, without buffering extra frames?
Or rather, is that actually happening?

As much as all these things can contribute to the low tracking speed, are there other reasons for it and which one counts the most?

@Skylit@popups@povohat
post #2 of 8
Internal buffer size limitations for one thing. Buffer overflow effectively means less or no fresh count data at all can be stored and sent to the host, i. e. the "negative acceleration" kicks in. Malfunction happens later with the WMO, so that has more to do with framerates and other things where the sensor simply doesn't pick up anything the MCU is firmware-wise set to be able to deal with. From an earlier post of mine in another thread (won't reference because there's outdated information in that post):
Quote:
This also goes hand in hand with the explanation why some mice experience an increase in tracking speed at higher polling rates. The internal buffer for count data in the mouse is emptied more frequently, so there is no buffer overflow which renders the mouse unable to store data, i. e. where it has to discard that data. Maybe mice like the WMO could benefit from bigger internal buffer memory sizes, although the fact that there is next to no increase in PCS achieved by going from 500Hz to 1kHz with the WMO indicates that buffer size is not the bottleneck for tracking speed beyond ~1.5m/s.

Edited by HAGGARD - 1/11/15 at 9:50am
post #3 of 8
I'm flattered you tagged me in this post. I wish I knew more about this stuff. I'm just an outsider like almost everyone here, and can only speculate based on the very limited amount of information available to the general public.

Regarding the low maximum tracking speed, we know that it can be increased by using a higher polling rate. Sujoy's mousescore article says you only need 250Hz to overcome the buffer bottleneck, but this only gets us up to ~1.5m/s. Why it cannot track linearly above 1.5m/s, I do not know and can only make poorly informed guesses about it.

Regarding angle snapping, I believe this is a side-effect of how the image correlation systems work. The direction and magnitude of mouse motion is determined by comparing a stored reference frame to the newly acquired frame. When the mouse has moved far enough, the newly acquired frames can no longer be correlated with the reference frame, and a new reference frame must be set (probably the last valid frame). Any difference between the old and new reference frame that was not considered different enough to register as a count will be lost. If the motion is quite fast, reference frames will be changing quite regularly and sub-count differences will be discarded more often. Maybe the low tracking speed and high frame rate of the MLT04 is actually the thing that is preventing this from happening. I imagine sensors with minimal angle snapping are resampling these images to higher resolutions before correlating, in order to discard less sub-count information (and allow higher CPI values).

Disclaimer: everything I wrote might be totally incorrect. I have based my understanding on this document, which may not be relevant to todays sensors. I hope somebody with more information about this stuff will correct me and offer a more accurate explanation.
post #4 of 8
Thread Starter 
^ Well I figured that since you recently added angle snapping to your driver, you must know something about how it works, at least more than me.. smile.gif
and I like how you write, I find it clear and easy to understand.
Interesting link, even if dated!
post #5 of 8
I don't have much expertise on the matter, but I used to think the reason MLT04 mice feel so "analog" to so many people was because of the fact that it doesn't have an adaptive framerate, but I've learned about so many other parts of mice since I started posting on these forums and I'm starting to think that a combination of MCU and very low angle snapping are the top contributors to that "rawness" MLT04s have.

I also think the MLT04's simplicity both helps and hurts it. The fact that the MCU doesn't have to deal with lots of features like adjustable DPI and RGB lighting probably means there isn't as much that could bog down the sensor's performance. But I can't help but think if there was slightly more power in the mouse, the perfect tracking would be a lot higher. Maybe they could even get the prediction lower than it is with firmware re-written for a more powerful controller. Not sure how exactly a better MCU would affect tracking overall, but I doubt it would have any negatives and would probably be able to get the max tracking speeds up. But it's probably a lot more complex than that, and I could be completely wrong, so feel free to correct me.

If it helps at all, this is a bit of info quoted by MaximilianKohler from another user:
Quote:
Originally Posted by MaximilianKohler View Post

"By djogedj h8m3 - Reply to #198
mlt04 has a simple architecture and DSP which doesn't allow any cpi scaling, therefore cause of simplistic algorithm which doesn't perform additional operations with frames it's behavior feels the most 'analog'"
Mars
(11 items)
 
  
CPUMotherboardGraphicsGraphics
Intel i7 4720HQ Lenovo Y50-70 Touch Intel(R) HD Graphics 4600 Nvidia GeForce GTX 860M 
RAMHard DriveOptical DriveOS
8GB DDR3L 1TB HDD + 8GB SSD Hybrid Drive External Windows 8.1 
KeyboardMouseMouse Pad
CoolerMaster QuickFire Rapid MX Blacks Logitech MX310 Razer Goliathus Control Large 
  hide details  
Reply
Mars
(11 items)
 
  
CPUMotherboardGraphicsGraphics
Intel i7 4720HQ Lenovo Y50-70 Touch Intel(R) HD Graphics 4600 Nvidia GeForce GTX 860M 
RAMHard DriveOptical DriveOS
8GB DDR3L 1TB HDD + 8GB SSD Hybrid Drive External Windows 8.1 
KeyboardMouseMouse Pad
CoolerMaster QuickFire Rapid MX Blacks Logitech MX310 Razer Goliathus Control Large 
  hide details  
Reply
post #6 of 8
Quote:
Originally Posted by mrvirtualboy View Post

MCU
 

 

MLT04 is SoC.

post #7 of 8
Quote:
Originally Posted by woll3 View Post

MLT04 is SoC.

I see. Are there any benefits to having everything integrated onto one chip versus having things separated?
Mars
(11 items)
 
  
CPUMotherboardGraphicsGraphics
Intel i7 4720HQ Lenovo Y50-70 Touch Intel(R) HD Graphics 4600 Nvidia GeForce GTX 860M 
RAMHard DriveOptical DriveOS
8GB DDR3L 1TB HDD + 8GB SSD Hybrid Drive External Windows 8.1 
KeyboardMouseMouse Pad
CoolerMaster QuickFire Rapid MX Blacks Logitech MX310 Razer Goliathus Control Large 
  hide details  
Reply
Mars
(11 items)
 
  
CPUMotherboardGraphicsGraphics
Intel i7 4720HQ Lenovo Y50-70 Touch Intel(R) HD Graphics 4600 Nvidia GeForce GTX 860M 
RAMHard DriveOptical DriveOS
8GB DDR3L 1TB HDD + 8GB SSD Hybrid Drive External Windows 8.1 
KeyboardMouseMouse Pad
CoolerMaster QuickFire Rapid MX Blacks Logitech MX310 Razer Goliathus Control Large 
  hide details  
Reply
post #8 of 8
Quote:
Originally Posted by mrvirtualboy View Post


I see. Are there any benefits to having everything integrated onto one chip versus having things separated?

 

The tl;dr version would be easier and cheaper, performance isnt affected in a direct way.

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Mice
Overclock.net › Forums › Components › Mice › Why exactly does MLT04 have such low tracking rates?