Optical Mouse technology

Most optical mouses use a camera to track the movement of the surface under the mouse. I wanted to compare the features and technology used by all the optical mouses I can. Here are my results.

Optical Movement Sensing Methods

There are several optical methods used by mouses. Here is a short over-view of the main methods:

• Slit-wheel sensing (ball-mouses) The old ball-based mouse actually uses optical sensors to tell when the ball rolls. The ball turns two rollers connected to wheels with slits along the outside. As the wheel spins, it blocks the path of light between an infra-red (IR) emitter and two sensors. The sensors are built with one on-top of the other in a single package, so that the light strikes one then the other as the wheel turns. Thus, which sensor receives light first tells which direction the wheel is turning.
• Camera-based sensing (most optical mouses) The usual optical mouse has a sensor chip on the bottom that takes pictures of the surface as the mouse is moved. These pictures are compaired and the chip determines how many pixels the mouse was moved and in which direction.
• Laser interferometry (dual-laser mouses) There are a few mouses now on the market that use this new dual-laser technique. A laser shines light onto the surface and some of the reflected light bounces directly back into the laser, optically mixing with the light being emitted and partially cancelling-out the light. The movement of the surface changes the frequency of the reflected light in a phenomonon called a Doppler Shift. The difference between the frequency of the emitted light and the reflected light changes how-much the light cancels, and this power level is sensed at the emitter by a photo diode sensor. This power fluctuation correpsonds to how fast the surface is moving in-line with the sensor. The direction of movement (away from or toward the laser) is sensed by varying the frequency of the laser, and using two orthagonally-placed lasers senses the movement X and Y dimensions (and apparently using the info from both at once can even sense Z-axis movement too). A good description can be found in this Philips white paper.

The rest of this paper deals with camera-based sensing.

In regards to the algorithm used by the digital signal processor (DSP) in the sensor chip, Professor Berthold K.P. Horn of MIT emailed me this:

Enjoyed your site about mice.
Ever wonder what makes it all work?
Many people think it has something to do with correlation.
Actually, the underlying technology is "Optical Flow" (*) worked out in detail in http://csail.mit.edu/~bkph/papers/Optical_Flow_OPT.pdf
This algorithm was *not* patented, so that all could benefit.
regards, Berthold K.P. Horn

(*) Actually the "Determining Optical Flow" paper solves a harder problem, where the velocity need not be the same in different parts of the image. The simpler version for "Fixed Flow" (where all parts of the image move at the same speed) is described in  http://csail.mit.edu/~bkph/articles/Fixed_Flow.pdf (see eq. 24 and 25 for final formula).

• Quality factors
• Some more useful quality measurements
• The stats
• Images
• Related Links

Quality factors

The quality of an optical mouse is defined by how well it actually works. While some aspects (such as shape, size, weight, color, number of buttons, etc.) are personal preferences, we can actually compare the quality of other aspects, namely:

• Image sensor size (pixels)
• Resolution (counts/inch)
• Refresh rate (Hz or samples/sec)
• Image detail (lens purity, light color, etc.)

Image sensor size (pixels)

The optical sensor in the mouse is essentially a small digital gray scale camera. How big the image is helps determine how fast you can move the mouse and maintain accurate tracking. If the image processor in the mouse is able to handle all the data, bigger images are generally better.

Image sensor sizes vary from 16x16 pixels to 30x30 pixels.

Resolution (counts/inch)

The resolution of the mouse is determined by the optical properties of the focusing lens and the physical size of the image sensor. The mouse resolution is then further modified by the driver software on the computer, where the sensitivity of the mouse can be decreased by ignoring very small movements or increased by jumping more than one pixel on screen even though the mouse only moved one pixel on the desk.

The optical resolution of the mouse is usually given in CPI (counts per inch), but can also be given in pixels per inch or (somewhat erroneously) in DPI (dots per inch). Just keep in mind that this is how many optical sensor pixels per inch the camera sees, not how many screen pixels the pointer moves per inch of mouse movement.

The main advantage of higher resolutions is that the minimum physical distance needed to register as a mouse movement goes down. Usual mouse resolutions are 400 or 800 CPI.

Refresh rate (Hz or samples/sec)

Coupled with sensor size and resolution, how fast the camera takes pictures determines how far the mouse can be moved per second and not lose tracking.

Refresh rates are given in samples per second, Hertz, or (also somewhat erroneously) frames per second. As long as the image processor keeps up, faster refresh rates are better. Mouse refresh rates vary from 1500-7080 samples/sec.

Image detail (lens purity, light color, etc.)

The quality of the lens has an impact as blurry lenses distort and corrupt the image that the sensor sees, making it more difficult for the image processor. The light color can affect the contrast of the surface (red brings out details better than blue) and the image sensor is designed to respond best to a certain light wavelength as well.

For example, the Avago/Agilent ADNS-3080 optical mouse sensor has this response curve for a given wavelength of light (from page 14 of the sensor's data sheet available at the link above).

The visible light spectrum is made up of electromagnetic waves of wavelength from ~400 to ~750 nanometers (nm):

Wavelength Range (nm)	Color	LED's Available (nm)
(100-400)	Ultraviolet (UV)	(none)
400-450	Violet	"UV" 405
450-500	Blue	"Blue" 463, 470, 472
500-570	Green	"Green" 524-525
570-590	Yellow	"Yellow" 588-595
590-610	Orange	"Orange" 605
610-750	Red	"Red" 625-630, 660
(700-1000)	Infrared (IR)	"IR" 850-860, 880, 940-950

So, red light at 630nm works best (and is also very efficient power-wise, so that's what is used), near infrared at 850nm would work OK, IR at 945nm (like the LED's in most VCR-type remotes) wouldn't work very well. Blue LED's at 470nm would work better than 850nm near IR, so I'm not sure why more manufacturers don't use blue.

LED vs. Laser

Most optical mouses use a Light Emitting Diode (LED) as the illumination for the optical sensor. In September 2004, Logitech and Agilent (now Avago) released the first optical mouse to use a laser as the illumination source. The MX1000 Laser Cordless Mouse uses the same optical sensor as their high-end MX510 wired gaming mouse, but uses an 832-852nm Infrared laser instead of the LED (it also adds left-right scrolling with a tilt-wheel, like the latest incarnation of Microsoft's IntelliMouse Explorer: wired 4.0 or wireless 2.0). The Laser Technology Brief (pdf) does a good job explaining the upgrade, and even has nice images of what the optical sensor actually sees (shown at left).

Basically, because the laser is a uniform tight beam, it reflects more detail from minor surface defects and textures, even on what normal LED illumination shows as uniform-color smooth shiny surfaces (like white board), resulting in, "20x times the tracking power of optical." (That is, it can navigate on surfaces with 20x smaller surface features.) Logitech still warns that the "Laser still may not track on mirrored or clear surfaces, such as windowpane glass." But, at least one user claims that, "this mouse has had no issues with any surface I have tried it on, including a mirror."

Some more useful quality measurements

The quality factors listed above aren't the best way to compare optical mouses, as they must be taken together to get a meaningful representation of the over-all quality of an optical mouse system. Some more meaningful quality measurements:

• Image processing power (Mpixels/sec)
• Max speed (inches/sec)
• Max acceleration (g)

Image processing power (Mpixels/sec)

The total amount of image data processed per second goes up if you increase the image sensor size or the refresh rate. Mouses with the same amount of image data processed per second should work about the same, even if they have very different sensor sizes and refresh rates. A large image size with a slower refresh rate can do the same job as a smaller image size with a faster refresh rate.

To calculate the image processing power we multiply the total number of pixels in each image by the number of images per second. Some mouses process as little as 0.486 Mpixels/sec, up to 5.8 Mpixels/sec for the Logitech MX510.

Note: One might think that it would be better to have a fast refresh rate with a small image size than vice versa, as this would tend to actually update the position of the mouse pointer on screen faster. But, the update rate of the mouse pointer on screen is limited by the update rate of the USB system over which the updates are sent. So while a Microsoft IntelliMouse Explorer 3.0 takes pictures at 6000 images/sec and the IntelliMouse Explorer 2.0 takes pictures at 2000 images/sec, either way the mouse pointer on screen updates only 125 times per sec (the polling rate of USB). To test the screen update rate of your mouse, use mouserate.

Max speed (inches/sec)

This is the maximum speed at which the mouse can move and maintain tracking. In addition to resolution, sensor size, and refresh rate, this is also determined by the tracking algorithm used (how much overlap the image processor needs between each frame, etc.) Because we don't have access to info about the tracking algorithm, we can't calculate the max speed; the numbers listed below are as reported by the manufacturer.

If this figure is too low, your mouse will lose tracking at high speeds, resulting in random mouse movements. This is often encountered when trying to turn your character around quickly in a First Person Shooter (FPS) type game, such as Quake. One way to test if your mouse is prone to losing tracking is to move your mouse pointer to the edge of the screen and then quickly move the mouse further off-screen. If your mouse kept tracking properly, the mouse pointer will still be at the side of the screen (possibly moved up or down, but still at the edge). If your mouse lost tracking, the random interpreted movements will move your mouse away from the edge of the screen.

I can lose tracking quite easily on mouses with a max speed of 16 in/sec or lower. The newest mouses from Microsoft and Logitech have max speeds of 37 and 40 in/sec respectively, and I've not been able to get them to lose tracking.

Max acceleration (g)

This is how fast the mouse can change direction, given in units of 'g', the acceleration due to gravity on the Earth's surface. This also varies by image processor design and tracking algorithm, so the numbers listed below are as reported by the manufacturer.

I've seen no test to demonstrate when this is 'too low', so the numbers aren't all that useful.

The stats

Here's what I've found out about various optical mouse sensors (sorted by Mpixels/sec):

Notes:

1. The Razer Viper is a strange beast: Razer, makers of high-CPI ball-mouses, wanted to make a high-sensitivity optical mouse with the available Agilent sensor chips, so they used a tweaked-out optical lens with a standard ADNS-2051 sensor. The higher-magnification of the lens (which is officially out-of-specification according to the Agilent docs) means that the Viper has a higher resolution (1000 CPI instead of 800 CPI), but lower max speed and acceleration, making it more prone to problems if you move the mouse too fast. And since Razor used a sensor with a fairly small original max speed (normally 12 inch/sec), many users complained of poor tracking of the mouse. Thus, the Razer Viper's homepage warns that, "For those of you who play and compete at low sensitivity settings [and thus move the mouse fast], we do not recommend the Viper mouse." This wouldn't have been a problem if Razor could have used a sensor with a higher max speed, but the sensors with max speeds >30 inch/sec were made exclusively for Microsoft (by STMicroelectronics) or exclusively for Logitech (by Agilent).
2. Since the MX310 and the MX510 both use the same sensor chip (S2020), but claim different Mpixels/sec, presumably the S2020 has a variable refresh rate.
3. The PixArt PAN101 is the only mouse sensor I've come across that I know for a fact is NOT made by/in agreement with Avago/Agilent. I don't have many specs on it, but it is pretty cheap and you'll find it in your $4.99 cheepo optical mouses (the one I looked at was a free-after-rebate iConcepts-branded PS2-only mouse from OfficeMax). Surprisingly, the mouse works pretty well. Avago (a spin-off of Agilent which was a spin-off of HP), has lots of patents on optical mouse sensors, and is currently suing PixArt for infringement.
4. (New Jan 2005) ELAN Microelectronics is now being sued by Avago/Agilent for patent infringement as well.
5. (New Apr 2005) The Razer Diamondback was the first 1600 CPI optical mouse. For the first 6 months of its release, it was the only 1600 CPI optical mouse. In April 2005, Avago/Agilent released their ADNS-3080 1600 CPI optical mouse sensor to other manufacturers, and Logitech released their "MX518 Gaming-Grade Optical Mouse", the second 1600 CPI optical mouse on the market. Since the Diamondback was out before the ADNS-3080 was finalized, the ADNS-3080 may have fixed some bugs that were in the version of the sensor in the Diamondback.
6. (New Apr 2005) The Razer Diamondback Plasma Limited Edition uses an Infrared LED (or IRED actually) for illuminating the sensor, which is outside the visible light range. Blue LEDs are used for lighting effects.

Images

Here's some images of several opened-up mouses and microscope images of the sensor chips from a few. If you have a dead optical mouse I'd be willing to pay shipping to gut it and put it's picture here.

Click images for huge versions. WARNING: I do mean huge (most are 1280x960 or 2048x1536).

Comparison of sensor systems

Philips Laser Sensors sent me this image showing the total size of various mouse sensor systems. Since Philips' twin-eye laser sensor has the sense/illumination system (the two lasers) and the lens all in one package, it makes for a very small system:

Avago/Agilent ADNS-2610, ADNS-2051, ADNS-3060 sensors:

Thanks to Avago/Agilent for these samples. Note that close-ups are all oriented with front of the mouse up.

iConcepts/HorngTech HTM-62WT:

Uses the PixArt PAN101 sensor.

I-Rocks IR-7300:

Thanks to Stan Horn for these pictures. Uses the ELAN OM02 sensor.

Logitech MX310:

Logitech MX510:

I have one, but haven't taken the pictures yet. Uses the same sensor as the MX310.

Logitech MX1000 Laser Optical Mouse:

Microsoft IntelliMouse Optical 1.1A:

I have the pictures, but haven't organized them yet. Uses the same sensor as the IntelliMouse Explorer 4.0.

Microsoft IntelliMouse Explorer 4.0 with tilt-wheel:

Microsoft Wireless Optical Mouse 2.0 with tilt-wheel:

Hacking around the Avago/Agilent ADNS-2051:

Using the 2051's pixel dump mode, I managed to snag a few images. I've attached one of the more interesting ones. It is of two threads of burlap material crossing. Images from the ADNS-2051 are 16x16 with 6 bit pixel depth. I multiplied the pixels by four to increase the contrast and use the full 256 grayscale offered by RGB bitmaps.

I got some sample chips (ADNS-3060 and ADNS-3080) from Avago/Agilent, and I hope to use them for high speed, low resolution microscopy of metallurgical events out in the mechanical engineering department of the University of Kentucky. I've been using the ADNS-2051 as a
learning platform, but I'm gearing up to use the 3080 soon. I'm waiting on some inverting level shifters from an ebay purchase.

The other attached picture is of the bench where I have the setup operating. Hopefully I'll have some mouse movement video soon - hopefully up to 500 or 1000 fps.

Feel free to post the pixel dump of the ADNS-2051 to your website if you wish.

Thanks,
Sasha

Related Links

Inside images of an Apple Pro Mouse (old version that uses an Avago/Agilent HDNS-2000) I especially liked this image of the CCD (Charge Coupled Device) image sensor.

Inside images of a newer Apple Pro Mouse (uses Avago/Agilent ADNS-2051)

Opened-up Apple Might Mouse (uses Avago/Agilent ADNS-2051)

Inside images of the Razer Viper (uses an Avago/Agilent ADNS-2051 with out-of-spec lens)

How to get an MX1000 working in Linux (with some of my stats)

uberOptions mod for Logitech SetPoint (Enables editing of all the buttons on Logitech SetPoint-controlled keyboards and mouses, by replacing the XML device settings files. Adds lots of features. Created and maintained by me.)

Optical Mouse Sensor Manufacturers

• Avago Technologies (formerly Agilent Technologies' semiconductor group, part of HP before that.)
  • Avago Chip Docs
• AtLab Inc.(.com)
  • AtLab Optical Mouse Sensors
• Elan Microelectronics Corp.
• Philips Laser Sensors
  • twin-eye laser sensor
• PixArt Imaging Inc.
  • PixArt Mouse Sensors
• STMicroelectronics (Makes the clear sensors used by Microsoft.)
  • Press release about all-in-one square sensor
• Unity Opto Technology Co., Ltd. (not updated since February 9, 2003)
  • Mouse shown with Unity Opto Sensor
  • Unity Opto Sensor (page with no info)

If you have questions, comments, optical mouses you'd like to donate, or more info to include in this page, email Richard L. Owens (remove NOSPAM).

-Richard L. Owens

This page visited times since Mon 25-Apr-2005 11 PM UTC (6K+ visitors before that).

A no-frills page whipped-up in Crimson Editor and Nvu.