T3i kills the t2i and 60d sales?

[M. Gilden]

Not entirely true.

The problem has been easily overcome by all camcorder manufacturers (Canon included) by seamlessly continuing the clip in a new file once the 4gb limit is reached - then re-joining the clips later with the bundled importing software. Canon chose not to implement that ability in this camera, therefore it is a self imposed limit.

It is, however, a logical and intentional self-imposed limit. The reason they didn't choose to implement it is probably because (a) it would attract a higher tax in some regions if it can record video continuously, and (b) because the overheating problem would become more apparent if people were able to record continuously and (c) they didn't think anybody would use a still camera to do a video cameras job.

That's one way of looking at it.
But the truth is, even as you admit, those other dedicated camcorders that worked their way around this problem have had to do exactly that: work around the problem. The problem still exists in those situations, however the manufacturers accepted that the limitation was not acceptable for a dedicated video recorder, and had to design a solution.

What is that solution?
Well, I may be wrong about this, but the way I understand it is that the video is held in an extended buffer before being sent to the encoder which eventually writes the file to the card. If the encoder receives an EOF error (end of file), the buffer begins to store a few extra seconds of data until the next file is ready and the encoder can begin working the video out of the buffer again.
This extended buffer system can also help accommodate slower memory cards, as more data can be held in the buffer before it has to give up and stop. Both are considerations that must be made for a consumer camcorder.
But NOT for a DSLR.

Of all your options (a, b, and c), I'd say C is the only likely real reason- they didn't expect this to be such a serious feature that required the work-around that other cameras had to implement. An extended buffer requires a lot of high-speed RAM, and since it is not the primary function of the camera, would add unnecessary cost and engineering.
As for reasons A and B...
a) If you flip the resolution down to standard def, you'll see it can record nearly a half hour of footage per clip without any hacking or tweaking (29:59). I've heard that some countries require the video camera tax if it records more than 30 minutes, so that might well be an imposed limit. But that limit can be breached via Magic Lantern, I believe. I think I've heard of people reaching 45 minutes with a lower bitrate, but then we are still hit with the original 4GB limit. Either way, 12 minutes is more of an arithmetical limit since the camera will record much longer than that out of the box in standard def.

b) Again, perhaps on the 7D. But the other cameras don't heat up nearly as often, and many users on these forums have run their cameras for hours without ever seeing an overheat warning. I shot almost an hour and a half the other night, and aside from the camera being slightly warm to the touch around the battery cavity, there wasn't any warning of overheating.

Still, you could look at it as "They could have worked around it but they didn't" as a self imposed limit. But I disagree with that point of view. That's like comparing my car without a sunroof to the luxury model that has the sunroof as a standard luxury feature... according to your logic, one could claim that Nissan actually removed the sunroof in my car purposely to not impose the "luxury" status on the vehicle.

 

[andythefilmer]

I know the DSLR's have buffers, esp. for photography, but it seems like ML should implement a feature like the one described, where it uses the buffer for a few seconds while it begins a new file upon reaching the 4GB limit. From what I've heard, the Lumix's can record longer. Do they do start new files like the way described? What about high-end card systems like P2? What about high-end SD card cameras like the HMC150? Sorry for the questions, but I am just curious how these differ.

 

[Samuel H]

ML, at least for the 550D, has that feature, but it's not completely seamless: a few seconds are lost with the file change

given ML can't access the encoder at all, probably this is the best it can do (though I'm continually surprised by the improvements on ML, so who knows)

 

[Kin Lau]

What is that solution?
Well, I may be wrong about this, but the way I understand it is that the video is held in an extended buffer before being sent to the encoder which eventually writes the file to the card. If the encoder receives an EOF error (end of file), the buffer begins to store a few extra seconds of data until the next file is ready and the encoder can begin working the video out of the buffer again.
This extended buffer system can also help accommodate slower memory cards, as more data can be held in the buffer before it has to give up and stop. Both are considerations that must be made for a consumer camcorder.
But NOT for a DSLR.

Of all your options (a, b, and c), I'd say C is the only likely real reason- they didn't expect this to be such a serious feature that required the work-around that other cameras had to implement. An extended buffer requires a lot of high-speed RAM, and since it is not the primary function of the camera, would add unnecessary cost and engineering.

The DSLR's already have more high-speed RAM than your typical camcorder - buffering RAW files is way more intensive than HD video. My 7D can buffer 15 18MP RAW files which are about 20-25megs each, and that can be written out in about 30 secs. That's at least a 300meg buffer that gets filled in about 2 seconds - the 7D is 8fps. If I switch to JPEG instead of RAW, the buffer goes _way_ up, I was able to crank out about 200 shots @ 8fps continuous before I got tired of it. The camera was able to capture 8 18MP images per second, compress and write them out to a CF card w/o skipping a beat for close to a minute. Even my T3i can do 4fps 18MP jpegs continuously to a Class 6 16gig card for longer than I care to keep pressing the shutter button - I stopped around 50.

As for the tax reason, the most obvious example would be Panasonic. The GH1 non-EU version does not have a 30 min limit, the EU version does, and it's one reason only - taxes.

 

[M. Gilden]

As for the tax reason, the most obvious example would be Panasonic. The GH1 non-EU version does not have a 30 min limit, the EU version does, and it's one reason only - taxes.

Yes- and I already address the 30 minute limit. That's a lot longer than 12.
Heck, I even already pointed out that if you turn down the resolution to standard, even a Canon out of the box will record for 29 minutes and 59 seconds before reaching a REAL self imposed limit (which, like the GH1, is easily hacked to remove). I have no doubt in my mind that the 30 min limit is due to taxation.
However, we're discussing it stopping long before it reaches that limit. In fact, the greatest proof that it isn't considered a self-imposed limit is that it IS NOT EXACTLY 12 MINUTES. In fact, it fluctuates depending on the shot. If there is very little motion and a solid black background, you can record a few extra minutes before the compression routine hits that magic 4GB file size and quits.

The DSLR's already have more high-speed RAM than your typical camcorder - buffering RAW files is way more intensive than HD video. My 7D can buffer 15 18MP RAW files which are about 20-25megs each, and that can be written out in about 30 secs. That's at least a 300meg buffer that gets filled in about 2 seconds - the 7D is 8fps. If I switch to JPEG instead of RAW, the buffer goes _way_ up, I was able to crank out about 200 shots @ 8fps continuous before I got tired of it. The camera was able to capture 8 18MP images per second, compress and write them out to a CF card w/o skipping a beat for close to a minute. Even my T3i can do 4fps 18MP jpegs continuously to a Class 6 16gig card for longer than I care to keep pressing the shutter button - I stopped around 50.

I hear what you are saying, but I don't think your math is accurate. Before the video images are sent to the compressor and saved to the card, the buffer has to store all the uncompressed data. Uncompressed video is EXTREMELY large, and at 24p or even 30p, can easily take up more space than 15 RAW stills, even at the lower resolution that video requires. Let's take 1080p at 24fps for a moment:
1920 x 1080/24p 4:2:0 Data rate: 746Mbps 328GB per hour of video
That means each second of HD video requires almost 100MB to buffer before going to the compressor to be encoded into the MUCH more compact h.264 format (328 per hour / 60 = 5.4GB per minute, /60 = 93.29MB per second, on average). If your buffer is really only 300mb, that barely gives you 3 seconds to save the file and its metadata before starting the next one.
And, as you pointed out yourself, that buffer could sometimes take as much as 30 seconds to write out to SD card, which would surely mean the camera has to give up.

Therefore, while the sensors, optics, and other parts of a camcorder are most likely lower quality, I don't think that a DSLR requires more RAM than a camcorder at all- especially one that writes in the computationally expensive h.264.

Can they get around the limitation? Yes, its possible. But I'm not entirely certain our cameras have the hardware to do it seamlessly at 1080p.

 

[Kin Lau]

I hear what you are saying, but I don't think your math is accurate. Before the video images are sent to the compressor and saved to the card, the buffer has to store all the uncompressed data. Uncompressed video is EXTREMELY large, and at 24p or even 30p, can easily take up more space than 15 RAW stills, even at the lower resolution that video requires. Let's take 1080p at 24fps for a moment:
1920 x 1080/24p 4:2:0 Data rate: 746Mbps 328GB per hour of video
That means each second of HD video requires almost 100MB to buffer before going to the compressor to be encoded into the MUCH more compact h.264 format (328 per hour / 60 = 5.4GB per minute, /60 = 93.29MB per second, on average). If your buffer is really only 300mb, that barely gives you 3 seconds to save the file and its metadata before starting the next one.
And, as you pointed out yourself, that buffer could sometimes take as much as 30 seconds to write out to SD card, which would surely mean the camera has to give up.

Therefore, while the sensors, optics, and other parts of a camcorder are most likely lower quality, I don't think that a DSLR requires more RAM than a camcorder at all- especially one that writes in the computationally expensive h.264.

15 shot buffer is just the buffer waiting to be written out... that's a 18mp image with a 14bit A/D and that 25megs is _after_ compression (lossless like LZW).

Lets do some simpler math. 1920x1080 = approx 2mp, 24 x 2mp = 48mp/sec, 18mp x 8 = 144mp/sec. Guess which system has to push more data? Video is h.264 compression, stills is jpeg compression - a raw file actually is raw with lossless compression _and_ a jpg embedded. At jpeg only stills, the 7D can actually do that continuously - the bottleneck is the memory card, the buffer fills up because the camera is pushing too much data. That's compressing 144mp/sec, and jpeg compression gets the file sizes small enough to overcome the bottleneck

Even my little T3i is 18mp x 4 = 72mp/sec and it can write jpgs to the SDHC card continuously for as long as I care to keep the shutter pressed. The only bottleneck again is the memory card, not the camera.

 

[wilcal]

The only bottleneck again is the memory card, not the camera.

Just for yuks I've tried a 4GB SanDisk Ultra II
Class 4 SD chip in my T3i. Either this is a very
good Class 4 chip or the data rates arn't as
high as advertised. It seems to work just fine.
No missed frames or pauses.

 

[Fritz bakon]

http://www.youtube.com/watch?v=tO0ZvYoV57o

T2i Color correction Video

 

[maranfilms]

Fritz, I think you can push that footage a bit more to get better blacks and more color from it. Although it is artistic choice, it seems to me there's a bit more room left before it falls apart.

 

[deltoidjohn]

Still, you could look at it as "They could have worked around it but they didn't" as a self imposed limit. But I disagree with that point of view. That's like comparing my car without a sunroof to the luxury model that has the sunroof as a standard luxury feature... according to your logic, one could claim that Nissan actually removed the sunroof in my car purposely to not impose the "luxury" status on the vehicle.

Actually, by my logic one would argue that the cheaper model does not have a sunroof because it would add to the production costs.

I know the DSLR's have buffers, esp. for photography, but it seems like ML should implement a feature like the one described, where it uses the buffer for a few seconds while it begins a new file upon reaching the 4GB limit. From what I've heard, the Lumix's can record longer. Do they do start new files like the way described? What about high-end card systems like P2? What about high-end SD card cameras like the HMC150? Sorry for the questions, but I am just curious how these differ.

All the current card based camcorders use FAT32 formatting which means they have a 4GB individual file limit. However, other codec's and cameras are much smarter than the DSLR's and are able to seamlessly start recording a new file as soon as the first file reaches 4GB. They also have the ability to stitch the files back together when they are imported to a computer. That's why AVCHD has a more complex folder structure than the DSLR's. The additional files/folders contain all the necassary information to reassemble the clips into one continuous file.

I have no doubt in my mind that the 30 min limit is due to taxation.
However, we're discussing it stopping long before it reaches that limit. In fact, the greatest proof that it isn't considered a self-imposed limit is that it IS NOT EXACTLY 12 MINUTES. In fact, it fluctuates depending on the shot. If there is very little motion and a solid black background, you can record a few extra minutes before the compression routine hits that magic 4GB file size and quits.

Yes, exactly my point (even though you were disagreeing with my point). It is a combination of factors (such as those I listed at (a), (b) and (c)) that leads to the 4GB limit remaining on the DSLR's. They did not bother to overcome the 4GB limit, even though they know how, because it would still be limited to 29:59 for Tax reasons. So why waste time, money and effort to implement a feature that is not going to be all that much more useful anyway?

I don't hear many people crying out "give us 30 minutes of uninterupted recording!" People want unlimited recording. Either you don't need long record times (narrative, action sports, music videos) or you need very long record times (live shows, weddings, events, speeches). If they overcame the 4GB limit and offered us 30 minutes of recording, those people who need long recording times would still not be satisfied (take a look at the European GH1/2 for proof of that fact). And then everybody would also be asking questions like "why are my 30 minute clips split up into 3 seperate 4GB files" or "I recorded a 30 minute speech and copied it to my PC from my card then formatted my card but the clip I copied only has half the speech, where is the rest of the 30 minute clip?"

 

[W00DPUSHING]

My personal thought is that from watching Nike SB's very own Jason Hernandez behind the lens.... well he's what made me definitely go for the t2i/550d. He makes t2i with ML and a Tokina 10-17 fish eye look amazing... For the money you can grab a t2i nowadays due to the release of the t3i then watch footage from Jason's t2i and you'll definitely wanna stick with the t2i for the money.

http://vimeo.com/12883374?ab

the t2i footy is in that video somewhere... check' it !