Don Haines said:jrista said:
I knew exactly what you meant.... but I didn't consider the faster frame rate with the smaller files... Even a Rebel can read an 18M sensor at 60 times per second, and any "big megapixel" camera that supports video should have no problem with being able to read the sensor 60 times per second.
I think that what slows down frame rate is the shutter speed, the time needed to create the files, and mostly the ability to dump the files out to storage. I agree, smaller number of pixels to be used in the image gives faster processing and less time to write, and that gives you more frames/second.
As an interesting aside, I have a p/s with a 16M sensor, it can shoot video at 240 frames per second so that implies that time required to read the sensor is not important... As you shrink down the recorded size of the image the frame rate goes up...
11.5fps/17 images (16M)
60.3fps/60 images (3M)
If Canon put out a DSLR where you could put it into a 10Mpixel crop mode and fire off a burst at 30 or 40 frames per second there would be a lot of interested bird photographers
I am talking about cropped stills, not cropped video. (Personally, I could really care less about video in my DSLR...it's convenient for some uses, but I really use my DSLR for photography.) Also, keep in mind, read occurs at the front end of the pipeline...compression, such as video compression or compression of photos into RAW files/JPEG, occurs at the back end of the pipeline. The readout rate requires that the front end speed, the data being pulled off the sensor and shipped into the DSP, support the full RAW data size of the full sensor at it's native bit depth, masked pixels and any error correcting or other intrinsic overhead included.
As for readout rate, they aren't reading the entire sensor at 60fps. Video reads are different than full frame stills reads. You only need to read two megapixels for full HD video. If they actually WERE reading the full frame at 60fps, that would mean the data throughput rate was TWO GIGABYTES PER SECOND. The DIGIC5+ is only capable of 250 MEGABYTES per second each (and the 1D X needs 480MB throughput to support 14fps, hence the use of dual digic.) Conversely, at 2 megapixels, the total throughput rate for 1080p readout at 60fps is 218MB/s, and is very likely achieved via some kind of basic hardware binning at best, and row skipping at worst.
Your little P&S is using binned readout or something like that to achieve 240fps, and even then, is it a progressive readout, or interleaved readout? If interleaved, the true readout rate would be 120fps, very likely for a mere 1mp worth of data. To actually read a full 16mp worth of data at 240fps (even at a mere 12 bit), you would need over 6BG/s throughput (that would be faster than SATA 3!)
What I think most people are referring to is a true native stills photography read, but with native cropping. Assuming we get DIGIC7+, capable of 7x DIGIC5+ performance. That would allow a 1750MB/s (1.75GB/s) raw data readout rate. That would allow around 19fps @ 46mp (assuming masked border pixels and some additional overhead). It would allow 37.7mp APS-H reads at ~33fps, and 18.1mp APS-C reads at ~51fps. For all that seven times more processing power than a single DIGIC5+, it still doesn't get you over the 60fps hump...even with hardware cropping.
Upvote
0
