dSLRs in general, some opinions please

[Paul Curtis]

Been thinking about upgrading my old canon dSLR and i'm interested in the video side of things but so far as i can tell there's no real dSLR out there that really isn't handicapped. This is how i see things at the moment but please correct me, that's why i'm here :)

(I have an EX1 but i'm looking at a dSLR for form factor and size and access to different lenses)

The canons have awful line artefacts because of the way the sensor is binned, which is a shame as a 7D would be ideal.

Not sure about the nikons in the same way but i don't think they're 1080p, but the rolling shutter is quite bad, worse than usual?

The GH1 and 4/3s suffer from poor compression but in someways offer the best features but a smaller sensor

The Pentax seems to tick an awful lot of boxes. Good compression ratios, only 720p though and only 30fps.

I have an LX3 which does really nice 720p in mjpeg, if that had manual control it would be very useful (tiny form factor and wide lens)

So there isn't really an obvious choice, no 1080p 24/25fps and any HDMI outs on the above don't output the full resolution.

Is that pretty much the state of things at the moment?

cheers,
paul

[Perrone Ford]

Very much depends on what you need to shoot. I also have an EX1 and am about to buy a 7D (or something similar) in 2010. It's nearly ideal for what I want. I'm hoping for a few tweaks in 2010 with it, but I'd gladly take it as it is.

But these are NOT video camera replacements. They are like film cameras that shoot on CF cards. If you think of them in that way, you'll be much more in line with what they do and don't do.

[Graham Hickling]

Just to clarify, the two Pentax options are:
K7: 1536 x 1024 (3:2 aspect) or 1280 x 720 (16:9) at 30fps
Kx: 1280 x 720 (16:9) at 24fps

The K7 will resize to 1080P quite well ... but as you say, its not 24fps.

[Paul Curtis]

Perrone,

Looking at the original 1080p download from

7D on Vimeo

for example, it's my humble opinion that the video output from the 7D is unacceptable for anything other than family snaps (which to be fair to canon is probably what it's for). The compression artefacts are incredible (i'm assuming this is direct from camera), the complete lack of colour tonality, people become just like watercolour paintings. This is something i see from most low bit rate AVC style compression, including the GH1.

In this case the little LX3 actually produces better video at 720p mjpeg, which can be up-rezzed and still look better than this 7D example.

So right now, there's little point in upgrading my dSLR as it still takes great stills!

the dSLRs seem so close, but they fail at the last stage - compression. Perhaps it's intentional while the likes of canon prepare next generation video cameras. Maybe 2010 will see some improvement.

I've yet to find full res examples of the Pentax. Thanks for the clarification Graham

Seeing websized example really makes everything look much better, difficult to judge from them

cheers
paul

[Perrone Ford]

Well in all fairness, you're looking at a 2Mbps recompression (Vimeo) of a 48Mbs capture from the 7D.

The 7D is fully capable of capturing footage that will equal or exceed my EX1 in "look". That is why I am buying one. The differences technically are there, and that cannot be ignored. But judging the cameras by what you see on Vimeo makes little sense.

[Graham Hickling]

Raw footage from the Pentax K7 can be downloaded here: Pentax K-7 Review: 21. Movie Mode: Digital Photography Review

And for the K-x here: Pentax K-x Review: 20. Movie Mode: Digital Photography Review

Note that the reviewers did nothing to limit shutter speed so some of the bright light clips look nasty in terms of motion - but IMHO the colors and tonality are acceptable.

[Paul Curtis]

Originally Posted by Perrone Ford

Well in all fairness, you're looking at a 2Mbps recompression (Vimeo) of a 48Mbs capture from the 7D.

The 7D is fully capable of capturing footage that will equal or exceed my EX1 in "look". That is why I am buying one. The differences technically are there, and that cannot be ignored. But judging the cameras by what you see on Vimeo makes little sense.

Am i? I downloaded the source movie (126MB Quicktime .mp4), i'm under the impression that's the original file vimeo use to compress with. Please tell me if i'm wrong because i would love a 7D. (When you register with vimeo you can download some of the original video files if the original poster allows for that)

I have not seen anything out of these cameras that can come close to an EX1, especially with regards to resolution. I think there are some more scientific tests out there with resolution charts that show how poor the 5D is and i suspect the 7 would be similar.

cheers
paul

[Perrone Ford]

Originally Posted by Paul Curtis

Am i? I downloaded the source movie (126MB Quicktime .mp4), i'm under the impression that's the original file vimeo use to compress with. Please tell me if i'm wrong because i would love a 7D. (When you register with vimeo you can download some of the original video files if the original poster allows for that)

Vimeo erases the source files after 1 week. All that is left is the conversion file they've done that is left for download. I've been a Vimeo user nearly since they started. Migrated there when Stage6 was winding down.

Originally Posted by Paul Curtis

I have not seen anything out of these cameras that can come close to an EX1, especially with regards to resolution. I think there are some more scientific tests out there with resolution charts that show how poor the 5D is and i suspect the 7 would be similar.

cheers
paul

I own an EX1 and am currently grading a film shot with a 5D and was on set doing the dailies with that 5D. Trust me, the images under many circumstances are nearly identical. The DSLR slaughters the EX1 in any scenario where the EX1 would need to come off -3 gain.

Yes, the resolution of the EX1 is better. And depending on the shot this will make a difference. I've seen the charts for the 5D and the 7D. I understand the moire patterns, chroma noise in the details, etc. Again, it's not perfect. But the Canon's are very, very good in how the look on many real world subjects. I've looked at the output on my 72" plasma, and on a 40ft screen. It's good enough.

[Paul Curtis]

Originally Posted by Graham Hickling

Raw footage from the Pentax K7 can be downloaded here: Pentax K-7 Review: 21. Movie Mode: Digital Photography Review

And for the K-x here: Pentax K-x Review: 20. Movie Mode: Digital Photography Review

Note that the reviewers did nothing to limit shutter speed so some of the bright light clips look nasty in terms of motion - but IMHO the colors and tonality are acceptable.

If you take a frame (say from the video 3) you see colour moireing over the grey areas indicating downsampling problems with the debayering. Also the compression is really quite harsh too, the people look very pixellated, the person with the red top looks like the colour sampling way lower than the resolution (even at 4:2:0).

I've included a crop example (the original is mjpeg so each frame is self contained and i saved it with jpg maximum to preserve it).

This is not 1500 x 1024 resolution in anyones book, a decent 720p has far more resolution than this.

I really hope i don't sound like i'm trolling here, but i see a nice active forum with people using dSLRs and it makes me wonder what i'm missing! These cameras produce wonder stills and have great sensors, the compression and downsampling from native resolution is killing them.

I'm going to put together a comparison image with the 7D, Pentax and the little LX3 to try and show what i'm seeing.

cheers
paul

[Paul Curtis]

Originally Posted by Perrone Ford

Vimeo erases the source files after 1 week. All that is left is the conversion file they've done that is left for download. I've been a Vimeo user nearly since they started. Migrated there when Stage6 was winding down.
Yes, the resolution of the EX1 is better. And depending on the shot this will make a difference. I've seen the charts for the 5D and the 7D. I understand the moire patterns, chroma noise in the details, etc. Again, it's not perfect. But the Canon's are very, very good in how the look on many real world subjects. I've looked at the output on my 72" plasma, and on a 40ft screen. It's good enough.

As i just said i'm not look to incite anyone but looking for real world examples.

Good news that the 126MB cannot be the original.

The canons are certainly excellent for shallow DOF shots - most of the detail is blurred which gets compressed very easily leaving more compression room for what's in focus. It's on detail - from the sea, leaves, grass, wide shots, clothing - that the real test is.

So, what i'd love to see if something from the 7D (or even 5D) that you'd consider a good example of what the cameras can do? Can you point me to any places with source footage?

You say you're grading the footage, do you have any examples showing the kind of range you're seeing.

Thanks Perrone, it would be really useful!

[i just want to add that i did find some source (women throwing a frisbee) and yes the compression is substantially better then the other examples i've found, although the source was seriously sharpened - i'd love to see more, it's quite difficult finding the original files!]

cheers
paul

[Graham Hickling]

For me, at least, a better comparison is my K-x ($500 on EBay) versus my Canon HF100 ($500 on EBay). The HF10 is sharper, useable for 'run and gun', can record for hours and is fairly resistant to moire. The K-x provides interchangeable lenses, far more DoF control (more 'filmic') and has better low-light capabilities.

For $1000 I now have both - life is good!

(Incidentally, that K7 clip does look bad - that subject material is poorly suited to the camera's abilities. Check out the K-x clips of the soldiers, and the fish .... they were OKish and encouraged me to buy the camera).

EDIT: Also, it's important to be aware of this: http://prolost.com/blog/2009/12/3/yo...elieve-me.html

[Perrone Ford]

Here is an ungraded snap from one of the scenes. This is straight out of the camera, handheld shot (well shoulder mount).

[Paul Curtis]

Perrone,

Yes, that looks great by comparison to what i've managed to find myself. That's a really positive image and suddenly i'm interested again.

This is from a 5D?

So bearing in mind you probably have a lot of footage from this where would you say the footage doesn't work? What are the problem areas from your experience?

cheers and many thanks for this
paul

[Perrone Ford]

Originally Posted by Paul Curtis

This is from a 5D?

Yes, untouched screengrab. Shoulder mount was a RedRock We were going to do it on the Fig Rig, but changed our minds at the last moment..

http://i666.photobucket.com/albums/v...Break_D197.jpg


http://i666.photobucket.com/albums/v...Break_D290.jpg

Originally Posted by Paul Curtis

So bearing in mind you probably have a lot of footage from this where would you say the footage doesn't work? What are the problem areas from your experience?

If both the camera and the subject are moving, you need to be careful. If they are moving in opposite directions, you're in trouble. The camera does not do well when it's moving quickly. Keep it on sticks, a dolly, a slider, etc., and you'll be fine.

[Paul Curtis]

>If both the camera and the subject are moving, you need to be careful.

Is that because of rolling shutter or compression artefacts or the way the sensor is subsampled (binned)?

In a locked off shot, for example the one you showed (i'm assuming it's locked off), how is the fine movement from frame to frame on the guys glasses? A concern would be that the aliasing due to the binning of the sensor would show up in subtle movements. Is that a problem too?

thanks
paul