whoa - rumor of 4k/24p crack for 550d

Status
Not open for further replies.
Hi all,

Studying the reactions to this 4K hack announcement has provided a good opportunity to compare the quality of discourse on dvxuser with similar threads on other boards.

I have to save I am quite happy with with the level of critical thinking here and am quite happy to become a member because of it. I should have done it sooner.

Happy New Year, everyone!
 
With video mode, my guess is that the camera roughly turns every 3rd sensor line off, and lets all the others scan the image 24 times a second - when you think about, it seems that a couple different things could be built around this. For one, 2K isn't a higher resolution than 1080P essentially, it's just a bit wider. So if these cameras can record 1080P for 12 minutes, the honest truth is that it could handle recording 2K without a problem, with the biggest issue being overheating issues that might pop up or the buffer going out (maybe after 10 minutes?). 2K is not even as big as change as 720P vs 1080P is, yet it's a format that because of the aspect ratio, is very essential to modern filmmaking and Canon should have already implemented it into their newer DSLR's. Then the next step is 3k, which is where I'd suspect is where things start getting iffy. I'd suspect that 3k on a DSLR would start to give it problems, much like the extremely higher quality modes on the new T2i hack - I'm not sure if it could be done. And with 4k, I think it's just a bit out of the question, Red's new DSLR camera doesn't even do 4k, does it?
 
maranfilms said:
Im shocked people are still trying to believe this could be real. Did he mention anything about connecting a blower to the cam. Because thats what it would need to be able to keep it cool enough to shoot 8k. I cant even keep the damn thing from overheating at stock settings on a fairly hot day. Never mind 4k, 8k, and the funniest of all 16k. , I cant even type this without chuckling. it's just funny.
Click to expand...

Yeah, the 16k was SUPPOSED to be a joke in case you didn't follow. ::rolls eyes::

"4k" is the rumor, and it is based on the notion that the sensor can take that kind of resolution for stills (as someone already said, you can already take 4k video with this camera... as long as it is stop-motion- HA!).
Is it true? I'm going to have to say I don't believe anything until I see it. Sounds unlikely.

But at least it is based on something plausible as opposed to 8k, 16k, and the other numbers for exaggeration. Those are clearly jokes.
 
Snapper123 said:
With video mode, my guess is that the camera roughly turns every 3rd sensor line off, and lets all the others scan the image 24 times a second - when you think about, it seems that a couple different things could be built around this. For one, 2K isn't a higher resolution than 1080P essentially, it's just a bit wider. So if these cameras can record 1080P for 12 minutes, the honest truth is that it could handle recording 2K without a problem, with the biggest issue being overheating issues that might pop up or the buffer going out (maybe after 10 minutes?). 2K is not even as big as change as 720P vs 1080P is, yet it's a format that because of the aspect ratio, is very essential to modern filmmaking and Canon should have already implemented it into their newer DSLR's. Then the next step is 3k, which is where I'd suspect is where things start getting iffy. I'd suspect that 3k on a DSLR would start to give it problems, much like the extremely higher quality modes on the new T2i hack - I'm not sure if it could be done. And with 4k, I think it's just a bit out of the question, Red's new DSLR camera doesn't even do 4k, does it?
Click to expand...

My calculations is that it reads every third line so it skips 2 lines in 1080p mode. Also, in 720p mode, it skips even more lines and gets hotter faster. It seems the time between a lines read is the biggest issue for heat, not the resolution. So if you were shooting higher resolutions, sure, heat would build more because of the longer time the shift outs would be, but increasing the frame rate would be the biggest contributor. Not that I am saying any of it is possible, but just my observations of what the camera is doing. Also, the heat issue is a software issue as well. How do we know how far you can go without burning up the camera? Canon released a firmware update that relaxed it a bit on the 7D I know.
 
Exactly, Hunter. The T2i can shoot, what, like 3fps in continuous shooting mode at 4K-ish resolution? How the hell is it going to be ramped up to 24fps? Even the 7D can't manage more than 7fps (all of these are from the top of my head, so I may be off by a frame or so).
Click to expand...
I can't say that I know it's limits (T2i)....but my thought were that the FPS were the result of several things.

1. Mechanical requirements (not chip/data related). Shutter action etc.
2. Canon product/quality division. I don't think there is any question that all companies try to create "tiers" in their product lines...sometime it includes other factors (ie Adobe/NVidia video card support of Mercury playback)...sometimes it's LITERALLY not related to CAPABILITY...but other factors related to marketing and partnerships
3. .RAW vs. .JPG ....less data....how much less, I don't know. But I know that we're talking about much less.
4. Lessons learned from GH1/GH13....the marketed product (to be safe for consumer use and predictably work)....the ACTUAL ***LIMITS*** of ANY camera is NO WHERE NEAR what is enabled in the "out of the box" version. We simply had NO IDEA what the limits were and our PERCEPTION of those limits WAS WAY OFF. Many thought...24mbps was the AVCHD LIMIT(I'm aware it is the STANDARD's limit)....it couldn't be higher than that!! :) That ***LIMIT*** (on the GH13) ended up being 4 times the amount that. Not that anyone shoots AT THAT LIMIT either....but that's the point.

I'm not saying that it's true. I am saying that the fact that the T2i's out of the box photo limit is 3fps doesn't mean much about whether or not it can do 6 seconds with its sensor (sensor burnout risk and all). Remember...the D7000 can do 16 seconds RAW output. (6 fps) If you do the math.....jpg is 10%-50% size as RAW. This is 1/3 as long...1/2-1/10 the size....and it's NOT constrained, or consumer safe, or product protected, or sales division protected....he's also disabled Live View (during recording) to channel resources....so....I can't definitively say that it's NOT possible.

"It's NOT possible"....uh...no, I can't say that with a period. No problem that many countless others can.
 
Last edited:
Snapper123 said:
With video mode, my guess is that the camera roughly turns every 3rd sensor line off, and lets all the others scan the image 24 times a second - when you think about, it seems that a couple different things could be built around this. For one, 2K isn't a higher resolution than 1080P essentially, it's just a bit wider. So if these cameras can record 1080P for 12 minutes, the honest truth is that it could handle recording 2K without a problem, with the biggest issue being overheating issues that might pop up or the buffer going out (maybe after 10 minutes?). 2K is not even as big as change as 720P vs 1080P is, yet it's a format that because of the aspect ratio, is very essential to modern filmmaking and Canon should have already implemented it into their newer DSLR's. Then the next step is 3k, which is where I'd suspect is where things start getting iffy. I'd suspect that 3k on a DSLR would start to give it problems, much like the extremely higher quality modes on the new T2i hack - I'm not sure if it could be done. And with 4k, I think it's just a bit out of the question, Red's new DSLR camera doesn't even do 4k, does it?
Click to expand...


grimepoch said:
My calculations is that it reads every third line so it skips 2 lines in 1080p mode. Also, in 720p mode, it skips even more lines and gets hotter faster. It seems the time between a lines read is the biggest issue for heat, not the resolution. So if you were shooting higher resolutions, sure, heat would build more because of the longer time the shift outs would be, but increasing the frame rate would be the biggest contributor. Not that I am saying any of it is possible, but just my observations of what the camera is doing. Also, the heat issue is a software issue as well. How do we know how far you can go without burning up the camera? Canon released a firmware update that relaxed it a bit on the 7D I know.
Click to expand...


from what I've seen on my footage, the sensor works in an even weirder way

put your tripod and camera in front of a brick building, at that distance that makes moire completely kill your footage; now start recording, and do a pan

or just go on reading, because I've done it: even though you can't see anything clearly because of motion blur, the moire is still there

my guess is that:
* the microsensors take light samples many, many times a second (120Hz or more)
* then all that data is aggregated, and only a single reading per microsensor comes out of the sensor
* on top of that, the reading from most microsensors is discarded before encoding (two of every three lines are skipped; that information may not even be leaving the sensor)
* then the encoder has to process all that info

it totally beats me how these "hackers" could change that process; maybe underclocking the microsensors gives you the headroom you need to be able to read the information from all of them, without discarding lines and without burning your sensor; or maybe overclocking it allows you to take the information out of the sensor at a faster rate, while creating more heat and shortening the life of your camera; or maybe it is all just a hoax

I'll be very happy if all this is true; my guess is, it won't
 
NormanBates said:
from what I've seen on my footage, the sensor works in an even weirder way

put your tripod and camera in front of a brick building, at that distance that makes moire completely kill your footage; now start recording, and do a pan

or just go on reading, because I've done it: even though you can't see anything clearly because of motion blur, the moire is still there

my guess is that:
* the microsensors take light samples many, many times a second (120Hz or more)
* then all that data is aggregated, and only a single reading per microsensor comes out of the sensor
* on top of that, the reading from most microsensors is discarded before encoding (two of every three lines are skipped; that information may not even be leaving the sensor)
* then the encoder has to process all that info

it totally beats me how these "hackers" could change that process; maybe underclocking the microsensors gives you the headroom you need to be able to read the information from all of them, without discarding lines and without burning your sensor; or maybe overclocking it allows you to take the information out of the sensor at a faster rate, while creating more heat and shortening the life of your camera; or maybe it is all just a hoax

I'll be very happy if all this is true; my guess is, it won't
Click to expand...

I have panned at that moire level, what you see is the effects of the aliasing as you move, creating those crazy patterns. My opinion is that because there is a HUGE gap between the rows (you are throwing out 66% of the data) that your red/blue sampling goes to crap because now you get a sample every 6th row given the bayer pattern. With fine detail, it looks like the algorithms they use to interpolate the red/blue just don't do a good job. I think the GH1/2 do a better job because they obviously the pixel binning they are doing is preserving the color sampling more correctly, leaving the moire uncolored.

This makes sense to me because the moire I see is red/blue/purpleish which makes complete sense.
 
One thing to remember: they have only STARTED this project two months ago. IF this is true (I believe it is) then there has not been the time to say: "Oh, hey. This is for sure to work, on everyone's camera, without the chance of bricking. Also; it won;t shorted the battery life of the camera.'

New video (http://www.youtube.com/watch?v=L34Vw-I2Lzo&feature=feedu) says that 5, 000 people will get it, insinuating that A) 5000 people will get it at 4EST, or 5000 will get it for free, then it will go up in price. the problem? If he starts making people pay for it, and it bricks someones camera, they have a lawsuit on their hands.
 
the 3 fps still photo limit has absolutely nothing to do with it - that´s the limit of the SHUTTER speed of that particular shutter mechanism in the 550d - it would seem the biggest limitation is the buffer - hey but what´s waiting a few days with a cautious attitude going cost ? the truth is we have absolutely no, and maybe even canon designers have no idea what are the limitations of their sensor . I got the distinct impression from panasonic that they were flabbergasted at the gh13 crack . And like Pany was probably pushed to improve the gh2, this might have some real positive effects on future cameras, if there´s even a grain of truth in the rumor.
 
I don't believe the hype, but if there is something that is able to be downloaded...

Whomever gets it from here, send it around. Someone could host it on their web server for download.
 
kurth said:
the 3 fps still photo limit has absolutely nothing to do with it - that´s the limit of the SHUTTER speed of that particular shutter mechanism in the 550d - it would seem the biggest limitation is the buffer - hey but what´s waiting a few days with a cautious attitude going cost ? the truth is we have absolutely no, and maybe even canon designers have no idea what are the limitations of their sensor . I got the distinct impression from panasonic that they were flabbergasted at the gh13 crack . And like Pany was probably pushed to improve the gh2, this might have some real positive effects on future cameras, if there´s even a grain of truth in the rumor.
Click to expand...

I find it hard to believe the shutter is different in the 7D and the 550d. The difference is the 7D has two processors in it and the 550d one. That makes sense why the still framerate is higher. You are also right about the buffer, to me, that would be one of the big issues. When shooting 1080p with a slow card, that buffer doesn't last long if the card cannot keep up :)
 
The part I have a hard time with is the 7D hack, how they would be able to do it and yet the Magic Lantern project hasn't, and they've been at it for A LOT longer.
 
I find it hard to believe the shutter is different in the 7D and the 550d
Click to expand...
550d shutter specs from dpreview
• Focal-plane shutter
• 30 - 1/4000 sec (0.5 or 0.3 EV steps)
• Flash X-Sync: 1/200 sec
• Bulb

7d shutter specs also from dpreview
Focal-plane shutter
• 30 - 1/8000 sec
• 0.3 or 0.5 EV increments
• Flash X-Sync: 1/250 sec

I can´t find shutter life expectancy on the 550d but the 500d has 100,000 actuations and the 7d has 150,000

obviously they are two different mechanisms
 
kurth said:
550d shutter specs from dpreview
• Focal-plane shutter
• 30 - 1/4000 sec (0.5 or 0.3 EV steps)
• Flash X-Sync: 1/200 sec
• Bulb

7d shutter specs also from dpreview
Focal-plane shutter
• 30 - 1/8000 sec
• 0.3 or 0.5 EV increments
• Flash X-Sync: 1/250 sec

I can´t find shutter life expectancy on the 550d but the 500d has 100,000 actuations and the 7d has 150,000

obviously they are two different mechanisms
Click to expand...

Interesting, however, I still don't see the shutter as the limiting factor. We are talking about shooting a shutter at 1/48 for video mostly, and 1/3 or 1/7 for both cameras. That is FAR from the smallest 1/4000 speed it can go by several orders of magnitude. While we don't use the shutter for video, I just put that value in there because it is 100 times slower than the fastest speed on the t2i.
 
I actually like the idea of manual shutter activations - could it alleviate the skew by inverting it? Might be a nice look either way - a more film-like cadence.

The market for replacement shutters would skyrocket, as your 7Ds would die after shooting 104.16 minutes of 24P. What's a replacement shutter cost anyway, with labor?
 
You're still talking about 24 shutter activations per second, and the duration in milliseconds from open to closed to ready to activate again is the real question. Perhaps these tiny motors would overheat. Find out what the fastest commercially available mechanical shutter speed is, compare mechanisms, and there's your answer. Buffer, CPU, etc. are still a major issues, but let's take it one at a time.

Actual 4:2:2 or even 4:4:4 would be almost a gig a second. Some form of compression will obviously have to take place.
 
Last edited:
grimepoch said:
I have panned at that moire level, what you see is the effects of the aliasing as you move, creating those crazy patterns. My opinion is that because there is a HUGE gap between the rows (you are throwing out 66% of the data) that your red/blue sampling goes to crap because now you get a sample every 6th row given the bayer pattern. With fine detail, it looks like the algorithms they use to interpolate the red/blue just don't do a good job. I think the GH1/2 do a better job because they obviously the pixel binning they are doing is preserving the color sampling more correctly, leaving the moire uncolored.

This makes sense to me because the moire I see is red/blue/purpleish which makes complete sense.
Click to expand...

maybe it is that, but I still think it is because the microsensors take a lot of samples at 120Hz or more and then average them
my argument is that moire appears when there's fine detail, and there's no fine detail on a blurred shot

I'll have to test my theory with a vertical pan (the footage where I've seen this was a horizontal pan)
 
NormanBates said:
maybe it is that, but I still think it is because the microsensors take a lot of samples at 120Hz or more and then average them
my argument is that moire appears when there's fine detail, and there's no fine detail on a blurred shot

I'll have to test my theory with a vertical pan (the footage where I've seen this was a horizontal pan)
Click to expand...

I'm telling you, they do not multi-sample at 120Hz or whatever, you can't multisample and average if you are summing up to an exposure, and even if you could, that does not affect moire, it would help noise. Some CMOS designs acutally use multiple photosites under a single filter color to average out what is actually seen to minimize noise, but they don't multi sample in time, only space. The BIGGEST reason multisampling fails is due to motion. Once you start and stop the sensor, what it sees based on motion changes. I can assure you shooting at extremely fast shutter speeds will show you this.

Moire and aliasing are one in the same, as moire is just the result of aliasing of tightly packed patterns. It's a well known phenomenon. To reduce moire, one has to basically blur the image in simple terms to reduce the spatial frequency across the sensor. This is done using an OLPF. On cameras like the 550D, the OLPF is designed for the full sensor raster. However, given the density of green sites is twice that of blue and or red, they typically pick a value between them. This means some aliasing gets through, but is minized, and keeps from getting an overly soft image.

When we skip lines, we leave big holes where detail is lost. When that detail is smaller than the active area by this much, moire and aliasing get even worse.

Examining a bayer pattern CMOS sensor will reveal how they work, as this stuff is pretty straightforward.
 
Status
Not open for further replies.
Back
Top