Morier: “In optical sensing, [resolution is] completely different than your camera. Your camera has a resolution of the number of pixels times the number of pixels that the pixel array has. The optical sensor is not that. The resolution is not the number of pixels. It’s the size of the pixel on the table. So you take only one pixel and you image it, because there is a lens and optical system, and you [determine how] big is this pixel on the table. Then you [determine] how many can you put in one inch? And this is the resolution. So if your pixel is 30 micron big in the matrix, how many times can you put 30 microns into one inch? This is the resolution. [In this example] it’s about 840 DPI.”
If a CMOS sensor array uses pixels 30 micron big, the resolution works out to about 840 DPI or CPI, the number of counts the mouse makes with one inch of movement. But here’s where it gets more complicated: a mouse with 8400 DPI doesn’t necessarily have pixels that are 10 times smaller, like you’d expect. Why? Because sensor DPI increases often come from subdividing those pixels into smaller and smaller increments. This is also where higher DPIs can be bad news.
Morier: “The physical resolution is a single pixel [but] the sensor system is able to see less than one pixel by doing image processing. It’s able to understand fractions of pixels. If the system is pretty strong, it’s able to discriminate, say, eight times less than one pixel. So you cut a pixel into eight slices...This mini slice of eight times less than 30 micron, you put it on the table, and how many can I put on one inch? I can put a lot, but the original resolution has not changed. The original resolution is still the native physical pixel.
“But the processed resolution, or the one that has been created by the sensor algorithm, is this very thin slice. And then you can imagine, you can put a lot of them. But you gain nothing in terms of accuracy with this. It’s just being faster in terms of, you need to move only that [smaller distance] to see one count. But a pixel is not very big, 30 micron, if you cut this by eight, you are talking about something that is very small. And this very small item is sensitive ... much more sensitive than the big 30 micron system.”
Every mouse sensor picks up a certain amount of valid signal and a certain amount of noise, called the noise floor. Now, imagine you have an entire 30 microns of data to work with: it’s going to be (relatively) easy to filter out the noise from the signal. Now imagine, by comparison, having only one eighth the amount of data. The more you subdivide your pixel, the closer your signal and noise floor are to one another, and the harder it becomes to discriminate valid data from noise. When they become too hard to discriminate, the mouse sensor starts to report noise, which results in inaccurate movements.
Morier: “That is the reason why increasing DPI is very dangerous if you don’t understand, in terms of sensor design, what is the basic capability of the sensor. If you just spot for very high number of DPI, and your design thinking is on that only, you will fail. At the end you will have a system that is very poor. It is doing what you call spurious motion, that means you do nothing, you put the mouse on the table, and the cursor is just floating away, it’s picking the noise and creating counts. So this is the problem if you design it wrong. The right approach is to design it right for the low resolution, to make it strong and robust, and then see how much you can cut the slices in, but not to spot first for the high resolution, and then break everything in terms of design.”
Many gaming mice, especially laser-illuminated gaming mice, are actually using sensors that have been around for years. An upgrade from a 2013 edition of a mouse to a 2015 edition could be using the same exact sensor, but offer a higher DPI count by subdividing the original resolution. “When you start to subdivide pixels you reach a noise floor, and you end up getting to a point where, with older generations of sensors, the consistency of the images that you’re processing is a little bit suspect,” says Chris Pate.
That’s bad. Real bad. Granted, it doesn’t mean that a mouse using an older, DPI-boosted sensor is going to perform badly at every setting. At lower DPIs, it’s going to perform just as well as an older model, since that lower DPI is within (or closer to) the original design parameters. But crank the DPI all the way up to the max, and you may start to see spurious motion, resolution error versus speed, cursor ripple, and other issues. So when you see a new mouse come out, advertising a higher DPI setting: be cautious. It’s not necessarily a good thing.