Canon T3i

[Video Head]

Has anyone here ventured into the new Canon T3i for full HD video yet? If so, what good and bad are you encountering? I am looking to pick up one of these as a light weight still camera and would be pleased if the video is as good as Canon says and could use it for double duty.

Thanks.

VH

[racer-x]

I own a t2i, which is the same thing minus the swivel LCD. It does shoot good video if you know what you're doing. It also has it's quirks. I have mine hacked via Magic Lantern firmware which adds lot's of cool new features.

If you go this route, get ready to spend lot's of money on lenses, unless you already have a good lens collection...........

[jagabo]

Not specifically about the T3i but an interesting comparison of film and DSLR video:
http://www.zacuto.com/shootout

Watch out for rolling shutter problems:
http://www.youtube.com/watch?v=UBN469gRXlg

[racer-x]

Here is a short sample clip I shot the other day to use as stock footage. It was compressed about 50%

[jagabo]

Here is a short sample clip I shot the other day to use as stock footage. It was compressed about 50%

You mean you took the raw video from the camera and further compressed it by 50 percent? Or did you use a "50 percent" setting in the camara when you shot it? Just curious.

[festmaster]

I have the T2i and have not read any great reviews of the T3i. Most seem to say you should save $100 and get the T2i or spend $200 more and get the 60D. That said, the HD movie capability is great and you can use any of them as dual duty devices. The one caveat is that the SD card is FAT32, which means there is a 4GB size cap per file. On these systems, that is only 20 minutes per shot.

[intracube]

Not specifically about the T3i but an interesting comparison of film and DSLR video:
http://www.zacuto.com/shootout

That's an interesting link. I've seen a few similar comparisons recently but missed that one. Here's another comparison - between a Canon 7D and an Arri Alexa:
http://vimeo.com/17627567

Going back to the Zacuto link and looking at 'Webisode 1', there's issues with peak white level throughout most of the video - the film examples are mostly stuck at ~235, while many of the digital camera examples are far lower:

the left and middle images are unaltered, the right has had the highlights lifted.

Initially I thought it was related to my graphics card and a 235 vs 255 peak white issue. But some of the footage at the end of the video reaches 255:

image is as it appears on my screen.

It's hard to draw conclusions with variations like this.

I would've liked to see them under expose the digital cameras to capture more of the light filament, then colour correct with an s-shape curve to bring the shadows/mid-tones back up. This would reduce some of the ugly digital clipping (but might then show up noise or compression artefacts in the shadows).

[jagabo]

Going back to the Zacuto link and looking at 'Webisode 1', there's issues with peak white level throughout most of the video - the film examples are mostly stuck at ~235, while many of the digital camera examples are far lower:

You should download the videos and look at the levels before conversion to RGB for display. On your computer monitor whites should be at 255, blacks at 0.

But yes, the peak levels for some of the digital clips are much lower than Y=235:


The brights on the other shot are overblown but clamped to Y=235:

[lordsmurf]

I have the T2i and have not read any great reviews of the T3i. Most seem to say you should save $100 and get the T2i or spend $200 more and get the 60D.

It's the same camera, aside from a few features.
There was a recent post here: Which camera should I buy, Canon T2i or Canon T3i ?

Generally speaking, SLR video is crap. It's either high compressed H.264, and therefore unfit for editing. Or it's MJPEG, and stuck at 5 minute shots. So you'd better be shooting short scenes. Those claims that "well they used one to shoot scenes on XYZ TV show" are misleading, because the cameras have been loaded down with about $10K+ worth of cinema gear. When the rig is set, you can't even see the SLR under all the other gear.

SLR video is for a photographer (or simple camera owner, in many cases, regardless of whether they think themselves a "photographer") that wants to "play videographer". Honestly, 99.99% of the people who even have to ask about SLR video capabilities fall into this category automatically. That's fine -- I fall into that category. I've never been a video shooter, I work solely in post.

Everything has to be tripod mounted.

[racer-x]

Jagaboo wrote:

You mean you took the raw video from the camera and further compressed it by 50 percent? Or did you use a "50 percent" setting in the camara when you shot it? Just curious.

I took the raw footage and ran it through Ipod4me (for a quick, simple encode to mp4) and got about 50% compression. The original footage is 47 mb/s .MOV.

The main advantages to shooting video with DSLRs is that it is relatively easy to achieve a nice shallow dept of field because of the large sensor. It can also get extremely good low light footage when used with fast lenses like an F1.4 or F1.8. I have a 50 mm F1.8 that I can practically shot in dark with when I set ISO to around 3200 - 12800. The noise level is surprisingly low due to in camera denoising.

[jagabo]

I took the raw footage and ran it through Ipod4me (for a quick, simple encode to mp4) and got about 50% compression. The original footage is 47 mb/s .MOV.

Thanks for the info.

The main advantages to shooting video with DSLRs is that it is relatively easy to achieve a nice shallow dept of field... It can also get extremely good low light footage

Exactly.

[festmaster]

[QUOTE=lordsmurf;2069226][QUOTE=festmaster;2069073

Generally speaking, SLR video is crap. It's either high compressed H.264, and therefore unfit for editing. Or it's MJPEG, and stuck at 5 minute shots. So you'd better be shooting short scenes. Those claims that "well they used one to shoot scenes on XYZ TV show" are misleading, because the cameras have been loaded down with about $10K+ worth of cinema gear. When the rig is set, you can't even see the SLR under all the other gear.

SLR video is for a photographer (or simple camera owner, in many cases, regardless of whether they think themselves a "photographer") that wants to "play videographer". Honestly, 99.99% of the people who even have to ask about SLR video capabilities fall into this category automatically. That's fine -- I fall into that category. I've never been a video shooter, I work solely in post.

Everything has to be tripod mounted.[/QUOTE]

True that those cameras record in MP4, but they do so at 44.9 Mbps, which is by no means highly compressed. The current versions of all of the standard video editing apps have no problem with it. As for playing videographer, I think that applies to anyone coming to this site looking for video camera reviews. The question really is whether to get a dedicated consumer video camera in addition to a DSLR or getting a DSLR that does video, too. I think the latter is clearly the way to go not only for cost savings and needing to carry only one gadget, but the vast amount of lenses available for DSLR's.

Agree that tripod or monopod is a must for steady shots.

[intracube]

You should download the videos and look at the levels before conversion to RGB for display.

I can't find an easy way to dump the stream using Linux. DownloadHelper doesn't work, and tools suggested in other theads are Windows based. Any suggestions?

On your computer monitor whites should be at 255, blacks at 0.

I'm still having trouble figuring out what my computer is doing when displaying video files. When I've collated some more info I'll start another thread, rather than take this one off topic.

The brights on the other shot are overblown but clamped to Y=235:

[Attachment 6303 - Click to enlarge]

*Hmm, I can't seem to make that attachment visible any more.

Are you saying that the peak Y values in the raw video for that specific shot (attachment 6303) are actually 235, and my computer is expanding the range to 255? If so, it's odd as earlier on in the video the Fuji 8573 example seems to be clipping at Y=235 - and in that instance it's not being brought up to Y=255.

Am I making sense?

[poisondeathray]

You should download the videos and look at the levels before conversion to RGB for display.

I can't find an easy way to dump the stream using Linux. DownloadHelper doesn't work, and tools suggested in other theads are Windows based. Any suggestions?

You can try rtmpdump which can run on linux

On your computer monitor whites should be at 255, blacks at 0.

I'm still having trouble figuring out what my computer is doing when displaying video files. When I've collated some more info I'll start another thread, rather than take this one off topic.

The brights on the other shot are overblown but clamped to Y=235:

[Attachment 6303 - Click to enlarge]

*Hmm, I can't seem to make that attachment visible any more.

Here is a mirror

Are you saying that the peak Y values in the raw video for that specific shot (attachment 6303) are actually 235, and my computer is expanding the range to 255? If so, it's odd as earlier on in the video the Fuji 8573 example seems to be clipping at Y=235 - and in that instance it's not being brought up to Y=255.

Am I making sense?

That is a luminance waveform, it measures Y' in Y'CbCr , not RGB

That hard white line at the top means values > Y' 235 are clamped .

When you view anything on a monitor, it has been converted to RGB . The video itself is Y'CbCr. On PC monitors, black is usually at RGB 0,0,0 , white at RGB 255,255,255 - he's referring to RGB values, NOT Y' when he says that.

Note these are RGB values for R,G,B respectively, not Y' values in Y'CbCr . I think the confusion probably lies there. What usually happens is Y'235 is "white", but gets converted to RGB 255,255,255 for display on your PC. This is what he means by "On your computer monitor whites should be at 255, blacks at 0. "

Jagabo posted some test videos with greyscale patterns for monitor calibration. I don't have the links handy, but they are useful in determining what is going on. I'm sure he'll post the links
EDIT: here it is , and the link in that post
https://forum.videohelp.com/threads/313178-A-problem-with-dynamic-ranges?p=1965119&view...=1#post1965119


The choice of decoder used makes a differece on how the natively recorded Y'CbCr h.264 MOV's are decoded from the 7D, T2i, T3i etc... The native Quicktime decoder will clamp values and decode in standard range Y' 16-235 , but libavformat based decoders (such as ffmpeg ) will decode full range (Y' 0-255). There are significant gamma differences and significant quality differences between the QT decoder and other decoders as well when decoding this DSLR footage , but enough hijacking. If you want more info specific to the DSLR processing , just ask

[Video Head]

Thanks for the great replies, information and samples.

My original post was a little vague. I am looking for a light weight field camera that does high quality stills. If it can handle HD video reasonably well it will be a good all round solution for my application. This camera will be taken mountain and rock climbing frequently so it's weight and lens flexibility will be important. For HD video it will be used to produce sport coaching material and allow the armchair enthusiast to experience what athletes do without all the effort. As a result it will be mounted on bicycles, rowing shells, etc.

I have been involved in video production since tape and toasters. I used to do promo/demo still and motion work for musicians, actors and models. I have been away from production for some years now and I am just getting back into it, specializing in sports training videos. I have cleaned out all of my old cameras and production equipment as well as upgrading my computers to handle a move to full digital. Naturally, I still have all of my tripods, lights, umbrellas, screens, etc. Now I am aquiring new digital cameras suited to the application. If I require a 3 chipper for some of my production I will just rent a Sony, Canon or Ikegami as needed. A 5D is already on the shopping list and I do have a collection of EF lenses and access to more. I have read a couple of comparisons on the T2i and T3i and have heard the same comments about the difference between the two models. The swivel LCD would be usefull in my application so I will go with the T3i.

All of these replies have brought up another question though: The T3i captures in .mov format. What are you using to edit with on the PC or are you using Mac?

[poisondeathray]

You can edit with Vegas 10 pro, premiere pro cs5, or edius . You need decent "horsepower" to edit the footage (hardware). You didn't list yours in the profile

Vegas 10 has improved DSLR support and h.264 decoding , as does edius. Premiere pro CS5 has an optimized engine "mercury playback engine" that leverages Nvidia CUDA cards for accelerated effects . Editing is much smoother with these programs . Older software titles have much more difficulty editing the footage, because they are not optimized as well for DSLR and h.264 decoding

For HD video it will be used to produce sport coaching material and allow the armchair enthusiast to experience what athletes do without all the effort. As a result it will be mounted on bicycles, rowing shells, etc.

Beware of rolling shutter cmos artifacts (partial exposure) if you are mounting the camera for sport footage . It will look very "jello-y" and wobbly . Jagabo's earlier video only showed 1 problem: linear skew. This can actually be treated in post production fairly well with filters. A more difficult problem in real mounting situations is from vibrations. These videos are from a different cameras, but I'm sure you get the idea . The "jello" from CMOS partial exposures cannot be treated effectively with filters.
http://www.youtube.com/watch?v=UEaDrS-yzIE
http://www.youtube.com/watch?v=fSWDP1Tsj1Y

[intracube]

Jagabo posted some test videos with greyscale patterns for monitor calibration. I don't have the links handy, but they are useful in determining what is going on. I'm sure he'll post the links
EDIT: here it is , and the link in that post
https://forum.videohelp.com/threads/313178-A-problem-with-dynamic-ranges?p=1965119&view...=1#post1965119

Follow-up questions here:
https://forum.videohelp.com/threads/333661-YCbCr-16-235-to-RGB-0-255-and-PAL-NTSC-diffe...-%287-5-IRE%29

Video Head - sorry for cluttering your thread with my questions

[Video Head]

poisondeathray: Thanks. I will probably go with Premiere. I am going to buy the T3i and start testing. I hope to not be in a totally hostile environment (for the camera) at all times, and it's true roll is a light weight still with good lens choices. A POV camcorder is also on my shopping list and it will probably do the worst of the worst. The technical side of this project does have some serious challenges, but if it was easy everyone would be doing it...

intracube: NP, your posts made a good read.

[jagabo]

You should download the videos and look at the levels before conversion to RGB for display.

I can't find an easy way to dump the stream using Linux. DownloadHelper doesn't work, and tools suggested in other theads are Windows based. Any suggestions?

Odd. DownloadHelper worked for me. I started playing the video and it appeared in the DH pulldown.

[intracube]

Odd. DownloadHelper worked for me. I started playing the video and it appeared in the DH pulldown.

You're right. What I did wrong was pause the video before clicking on the DownloadHelper menu, which screwed things up. Leaving the video playing made the .flv file appear on the menu. I've just updated to the latest version of DH, and that particular bug has been fixed.

Back to the luma issue. Looking at the raw video with mplayer and forcing it to not convert the levels from 16-235 to 0-255 seems to confirm what I was thinking - that some of the shots go above Y=235.

The image I posted of the arch & vines was a bad example - I can now see that shot stays within spec.

The images on the left have pixels that go above Y=235 (but not below Y=16), the images on the right seem OK:


On the topic of using DSLRs to shoot video - I'm really excited by what's now possible. In addition to the shallow depth of field and low light capabilities, using DSLRs opens the door to using a wide range of specialist lenses; super wide angle rectilinear/tilt shift/telephoto/macro/etc - without the need for clunky adapters that affect the image quality.

Philip Bloom's short film at the end of the first Zacuto video looks gorgeous. The DOF, the lack of noise, lens artefacts (chromatic aberrations/purple fringing etc) and skilled camera work all give the video a silky smooth quality. Great stuff.

[poisondeathray]

IMO "achilles heel" of the canon DSLRs is aliasing and moire. I've shot with 5Dmk2, 7D, T2i . It will pop up unexpectedly at any time and ruin your shot. But they are very capable cameras if you are selective in what you shoot and plan ahead . Just "google" this topic or look at any DLSR board. It's the #1 complaint by far, even more so than CMOS jello.

[racer-x]

poisondeathray
Re: Canon T3i

IMO "achilles heel" of the canon DSLRs is aliasing and moire. I've shot with 5Dmk2, 7D, T2i . It will pop up unexpectedly at any time and ruin your shot. But they are very capable cameras if you are selective in what you shoot and plan ahead . Just "google" this topic or look at any DLSR board. It's the #1 complaint by far, even more so than CMOS jello.

Never seen the "aliasing" issues, but I believe the "moire" issues have to do with the encoding settings Canon uses. It uses AVC with a Baseline profile and no CABAC encoding. The stock camera doesn't allow bitrate changes, but Magic Lantern firmware hack will let you set encoding bitrate from 6 Mb/s - 120 Mb/s. However, no cards available can keep up with that. The best I could do so far was 60 Mb/s........

Here are the stock camera encoded MOV specs:


General
Complete name : E:\Pictures\2011\Myrtle Beach\Sunrise\MVI_0127.MOV
Format : MPEG-4
Format profile : QuickTime
Codec ID : qt
File size : 187 MiB
Duration : 32s 932ms
Overall bit rate : 47.7 Mbps
Encoded date : UTC 2011-02-20 06:53:26
Tagged date : UTC 2011-02-20 06:53:26

Video
ID : 1
Format : AVC
Format/Info : Advanced Video Codec
Format profile : Baseline@L5.0
Format settings, CABAC : No
Format settings, ReFrames : 1 frame
Codec ID : avc1
Codec ID/Info : Advanced Video Coding
Duration : 32s 932ms
Bit rate mode : Variable
Bit rate : 46.2 Mbps
Width : 1 920 pixels
Height : 1 080 pixels
Original height : 1 088 pixels
Display aspect ratio : 16:9
Frame rate mode : Constant
Frame rate : 29.970 fps
Color space : YUV
Chroma subsampling : 4:2:0
Bit depth : 8 bits
Scan type : Progressive
Bits/(Pixel*Frame) : 0.743
Stream size : 181 MiB (97%)
Language : English
Encoded date : UTC 2011-02-20 06:53:26
Tagged date : UTC 2011-02-20 06:53:26
Color primaries : BT.709-5, BT.1361, IEC 61966-2-4, SMPTE RP177
Transfer characteristics : BT.709-5, BT.1361
Matrix coefficients : BT.601-6 525, BT.1358 525, BT.1700 NTSC, SMPTE 170M

Audio
ID : 2
Format : PCM
Format settings, Endianness : Little
Format settings, Sign : Signed
Codec ID : sowt
Duration : 32s 932ms
Bit rate mode : Constant
Bit rate : 1 536 Kbps
Channel(s) : 2 channels
Sampling rate : 48.0 KHz
Bit depth : 16 bits
Stream size : 6.03 MiB (3%)
Language : English
Encoded date : UTC 2011-02-20 06:53:26
Tagged date : UTC 2011-02-20 06:53:26

[poisondeathray]

poisondeathray
Re: Canon T3i

IMO "achilles heel" of the canon DSLRs is aliasing and moire. I've shot with 5Dmk2, 7D, T2i . It will pop up unexpectedly at any time and ruin your shot. But they are very capable cameras if you are selective in what you shoot and plan ahead . Just "google" this topic or look at any DLSR board. It's the #1 complaint by far, even more so than CMOS jello.

Never seen the "aliasing" issues, but I believe the "moire" issues have to do with the encoding settings Canon uses. It uses AVC with a Baseline profile and no CABAC encoding. The stock camera doesn't allow bitrate changes, but Magic Lantern firmware hack will let you set encoding bitrate from 6 Mb/s - 120 Mb/s. However, no cards available can keep up with that. The best I could do so far was 60 Mb/s........

It's not an encoding or bitrate issue ; it's a sensor sampling issue and processing (lack of OLPF) . A higher bitrate will just make the aliasing and moire more visible where the encoder would have dropped detail

There are a gazillion articles explaining why, but this article sums it up nicely
http://www.dvxuser.com/articles/article.php/20

I've experimented with the dozens of workaournd , techniques (defocusing), filters (caprock), post production techniques (chroma blurring), it's very disconcerting

You have to learn what you can and cannot shoot. Certain objects (like fences, brick walls, patterned shirts, tiles on a roof) cannot be shot at all. A seemingly "clean" situation will fall apart at certain distances . For example (this is from a 5D)
http://www.youtube.com/watch?v=Lgw90AYbaYM


Moire issues are way more visible to the average viewer, but aliasing is there as well. They go hand in hand. GH1,GH2 are way less prone, but suffer from other issues , and not as good for stills

I 100% agree with what you said earlier: Main advantage is shallow DoF for a low price . Some shallow DoF adapters for "conventional" cameras cost more than the body of a 7D !

[jagabo]

The picture with the resolution test pattern is showing the result of a simple bilinear (maybe simple bicubic) resize.



Here's their "good" image with VirtualDub's Resize filter in Simple Bilinear mode:



Click on the image to see it full size. You'd think Canon could use a better resizing algorithm like Precise Bilinear.

[racer-x]

Wow.....thanks for that info guys. That aliasing deformed propeller really drives home the point. I've only come across aliasing in one of my videos so far, but it's good to learn about this for when I get some nasty aliasing in the future.

[poisondeathray]

You'd think Canon could use a better resizing algorithm like Precise Bilinear.

It's a CPU/processing "horsepower" issue

The huge sensor is designed for stills, not video. In burst mode the fastest you get in most DSLRs is what? 4-6FPS ? 1920x1080 is a tiny fraction of the size of your 20+ Megapixel still photo.

The processing isn't fast enough to implement a proper resize algorithm at video framerates in realtime. The 7D overheats as it is now. So in order to achieve "video" framerates like 24fps, 30fps , they subsample the sensor. This is done by either pixel binning or line skipping - this is one of the reasons why you get aliasing and jaggies

If you had the "horsepower" to be able to take 24, 30, 60, even 120 fps off the full sensor.... guess what... you would essentially have a Red Epic/ Red MX/ Red One. But with a higher pricetag of course

[jagabo]

The backwards/forwards spinning of the prop happens with film to. You see it all the time with wagon wheels in old westerns. You can see it the attached clip just before the end of the shot. It's just a matter of the exposure time and frame rate. The bending and detaching of the prop blades is just another manifestation of the jello effect.

[racer-x]

The bending and detaching of the prop blades is just another manifestation of the jello effect.

Yeah, I would say the Prop blade distortion is caused by "Rolling Shutter". All CMOS cameras that I know of would suffer from this. It would be a cool experiment to shoot that prop with different shutter speeds. That would give different variations of deformity.

I mostly use my t2i to shoot stock footage and stills anyway. I have an HV20 and an HG20 that I use for most of my video work.

[poisondeathray]

CMOS cameras with a global shutter wouldn't suffer from the bendy prop effect or related jello & skew effects - but these are very rare and expensive

These are great cameras for the price . It's just a tool - and every tool has pros / cons / limitations and a learning curve in using it use ideally

There is some incredible work by talented folks shot by Canon DSLRs . They often get requested specifically at production companies and by artists for music videos for the "look". It's a very popular item.

Canon has recycled the Digic 4 processor the past few years, people were expecting a new improved model by now - especially with all the complaints. But the primary market is photography, not video. These are designed for photos, and video is an afterthought . Panasonic has improved their processing; there is much less moire and aliasing visible in the GH2 for example. But GH2 has other issues and isn't as good a stills camera
ttp://vimeo.com/20565849