Lightwars - Low light video GH1 versus 5D

[Ben_B]

Great video Ben.

OK first issue I found versus 5D today in my final tests...

Here is how my GH1 sunset came out versus the 5D tonight :violin:

The velvet red / dark orange tones are a pale greenish yellow in the GH1 shot, even though the cameras were at identical exposures and white balance (6500k which is quite warm). I wasn't able to solve this colour issue with the GH1 no matter what.

**GH1**

**5D**

I think the GH1 looks better here...

 

[commanderspike]

Okay, now the next issue.

Magic hour sunshine just before the sun sets, on a lot of detailed foilage and fencing. The colour and detail just isn't there in the GH1 shot. If you watch the 1920x1080 footage on a large plasma screen (with decent colour) the difference between the GH1's footage and the 5D's shot is quite painful, especially for a avid GH1 like me).

Frames resized to 1280 horizontal, so the difference is much less noticeable here than it is on a proper screen or projector. Bear that in mind. LCDs cannot do the 5D's colour justice like a plasma can.

**5D**

**GH1**

Out of the 12 or 13 different scenes I shot, the 5D lost in one due to moire and the GH1 lost in all the others. It wasn't a good showing guys. Maybe this thread should be best moved to the 5D forum, because they'd like this over there. I feel like I'm doing you all a bit of disservice, but I can't change the results. Damn. :violin:

 

[ajamils]

Interesting stuff but one thing that GH1 wins every time is the value for money. With same quality lens as GH1 Kit....5D will cost twice as much (if not more) which makes it out of reach for most enthusiasts/newbies (which makes up a big chunk of GH1 users) that jumped on GH1 without any prior interest in DLSR.

For people who'll use it to make money can definitely justify the price but again most of us are just using cam for personal use.

 

[commanderspike]

I totally agree. That's why I got the GH1 first and not the 5D. *thumbs up*

But I realised that to get a Leica 25mm F1.4 or even the 20mm F1.7 pancake was £600 or £300 out of the price difference between the 5D and GH1 if I needed a fast portrait lens (40 or 50mm equiv.) on the GH1. At first I was relying on a c-mount lens for that purpose but they're also now too expensive (after everybody caught on to them on eBay). And if you want a fast wide, say 20,24 or 28mm equiv. like the Sigma F1.8s on the GH1 then you're out of luck. Only will full frame do you get such low light performance and shallow DOF on wide lenses.

The GH1 reminds me of a Replica from Bladerunner. It's capable of beauty. But the 5D is a human being

 

[commanderspike]

Next part:

Here's a tree in twilight sun, the leaves were giving off a fantastic glow of colour, from shades of orange like in the red brick wall, to the warm sunlit yellow of autumn. Many of the leaves were still fresh and green from summer. You can guess which camera did the better job of recording all that:

GH1... erm, green? Orange? Ooops. Maybe it forgot. :violin: But I still love it.

 

[commanderspike]

I think the GH1 looks better here...

Of course, it's always going to subjective to a certain point. But my own eyes perceived that sunset as being blood red. It really was beautiful. The GH1 rendered it greenish yellow in some places whist with the 5D that wasn't the case.

I've found it hard to reproduce what I saw on the plasma from the same footage, by doing frame grabs on my Macbook. There is something wrong with the gamma in the 5D clips on my Mac, especially in Quicktime though not so much in FCP 7. It does look a bit too dark and the colours aren't allowed to sing. If anyone wants me to upload the footage to Vimeo for downloading and displaying on a 1080p plasma let me know. I really do think that's the only way to view the differences accurately, and not on a LCD or by resizing static frames. It's hard to judge colour and detail that way.

 

[philiplipetz]

Next part:

Here's a tree in twilight sun, the leaves were giving off a fantastic glow of colour, from shades of orange like in the red brick wall, to the warm sunlit yellow of autumn. Many of the leaves were still fresh and green from summer. You can guess which camera did the better job of recording all that:

GH1... erm, green? Orange? Ooops. Maybe it forgot. :violin: But I still love it.

Looks like GH1 is over exposed and more detail in dark areas and 5d under exposed. I prefer the GH1 in all shots. The sunset GH1 does have a slight green cast, but that can be corrected easily in post, while the 5d seems to have striated colors and so not give a pleasant gradient to my eyes, and not so easy to correct in post

 

[Park Edwards]

so far the contrast on the mk2 has been way too much. something that's been noticeable since it first hit the streets. dunno if it's something that can be changed or it's just it's codec. i really like how the gh1's yellow pops in the last comparison.

 

[sdhurley]

"The GH1 reminds me of a Replica from Bladerunner."

It's Replicant. Your mistaken use of Replica, though ironic and funny doesn't refer to Blade Runner.

 

[commanderspike]

so far the contrast on the mk2 has been way too much. something that's been noticeable since it first hit the streets. dunno if it's something that can be changed or it's just it's codec. i really like how the gh1's yellow pops in the last comparison.

That would be great, but the tree wasn't actually yellow.

 

[Park Edwards]

the the mk2 is messed up too. i definitely see yellow in that one as well. but mayhaps it's not meant to be yellow. i still like how it pops.

 

[commanderspike]

Well thats ok. Nothing wrong with liking it, as long as you know that there were really strong greens and oranges in that shot that simply didn't make it on the GH1 shot, and are mostly hidden on the 5D shot in my frame grab for some strange reason only known to LCD monitor manufacturers! On the plasma screen it is more obvious which camera got it right.

Detail also. Even on static shots, the 5D holds onto it a lot better. Again this isn't noticable when a 1920x1080p moving image is made into a single frame and resized to 1280x720 for the forum.

If anyone reading this is forming their view of the GH1 versus 5D image quality based on the thread, I think it'll be a good idea to get down to a retailer and plug both cameras into a HD plasma if possible. I think that is the only way you can judge how big the gap is between the two. Trust me it is big. Honestly!

For whatever reason, a computer screen hides a lot of the 5D's advantage and covers up a lot of the GH1's short comings.

*Both great cameras however*

 

[PappasArts]

"commanderspike"

You shouldn't shoot vibrant ( setting 6 ) mode as you posted to that being the setting you are using/guessing on your Japan language GH1. I made a comment/response about that a few posts back; you seemed to have overlooked my response to you. I suggest standard ( 1 ) or smooth ( 4 ). Your results will be wayy better. My tests show Vibrant mode contrast is cranked up ( creating a dark less exposed looking image ), and color goes little mush. The tonal latitude is crushed as well ( aka worse DR than 1 or 4 setting ). Your choice to do it as you see fit; however considering your doing these tests, why not do it as right as possible. Just some helpful insight to share!


Michael Pappas
http://www.pbase.com/Arrfilms
Arrfilms@hotmail.com
http://www.PappasArts.com

Well thats ok. Nothing wrong with liking it, as long as you know that there were really strong greens and oranges in that shot that simply didn't make it on the GH1 shot, and are mostly hidden on the 5D shot in my frame grab for some strange reason only known to LCD monitor manufacturers! On the plasma screen it is more obvious which camera got it right.

Detail also. Even on static shots, the 5D holds onto it a lot better. Again this isn't noticable when a 1920x1080p moving image is made into a single frame and resized to 1280x720 for the forum.

If anyone reading this is forming their view of the GH1 versus 5D image quality based on the thread, I think it'll be a good idea to get down to a retailer and plug both cameras into a HD plasma if possible. I think that is the only way you can judge how big the gap is between the two. Trust me it is big. Honestly!

For whatever reason, a computer screen hides a lot of the 5D's advantage and covers up a lot of the GH1's short comings.

*Both great cameras however*

 

[Martti Ekstrand]

I'm mostly concerned with the fixed pattern noise in those bricks - I think the brick maker should do a firmware upgrade to fix that. How can they ship such bad looking bricks to a poor unsuspecting home owner? Shame on them.

Looks like GH1 is over exposed and more detail in dark areas and 5d under exposed. I prefer the GH1 in all shots.

The GH1 is also sharper all through the depth of field. Look at the leaves in front of the tree trunk, there you can sense the serrated edges and texture in GH1 picture but they are soft on the 5D. Could be the lens though.

That would be great, but the tree wasn't actually yellow.

I think you should spend some time learning to colour correct. There's more data to play with in the GH1 picture even after your jpeg compression. Here's a one and a half minute Potatochop hack. Notice I could push up a little more detail in the near blacks that are just gone in the 5D picture. Not saying it's needed on this particular image, just showing that the info is there.

Then HD plasmas are just plain hideous - if that's your standard then god help you. I wouldn't judge image quality from anything on those. Especially not by plugging the cameras straight into them, then you start to compare their playback electronics more than anything else. In general you are much too sloppy in your testing methods to come out with guns blazing stating opinions as hard facts like you do.
.

 

[philiplipetz]

I have always had two responses to 5/7d footage - initially WOW and then SOMETHING IS WRONG. Of course there are the issues Barry revealed with his resolution chart thread, but I think there is something else. The 5d sunset pic looks to my eye, on recompressed web pics, as if there is a subtle color banding. The 5d seems to initially give an impression of more color detail but the method of downsizing from a full 35mm frame to 1080 may make the algorithm compromise on the color values as it averages to value to that can be expressed in a single line. I feel, but do not know how to test, that this results in subtle color banding when a color gradient that extends over a large area (you will not see this if the color gradient is so small that it does not cover a significant number of lines on the 1080) and the algorithm makes the same compromise repeatedly until it reaches a magic value when it can make a different choice. So it makes sudden transitions that initially seem to reveal detail but are really artificial. I think this is what we see in the 5d sunset pic.

I am a pilot, we look to clouds, and the detail on the feathering of the trailing edges of the clouds in the upper line of clouds between the towers is not correct in the 5d pic. Each feathering region looks almost like a separate cloud, when that is not what happens. A single formation has feathers pulled from it as is depicted in the GH1 shot. I think this is a prime example of the limitations of the 5d codec and the way it may use color banding to create false detail.

Why might there theoretically be more post processing potential in the GH1? It has a smaller sensor so there is less downsizing to produce a 1080 image, and so less of what I call color banding. Whatever it is, the 5d seems to take the Canon tendency for a processed look and magnify it. Two passes, not one, through the DIGIC color amplifier. At this very very primitive stage in hDSLR evolution is there an advantage to less binning, less processing? Look at what Martti Ekstrand did with the leaves pic, it shows the great potential in post processing for GH1. People talk about RAW video as the holy grail but Marrti showed us that the GH1 had enough headroom for excellent post. At least in this image, we will have to wait to see if this is a generalized rule.

I was number one on my local dealer's 7d wait list and got the camera and it is now back on his shelf. This thread is leading me to the GH1 as an interium solution, but I will wait until the last minute before I shoot my next big project, and then choose the best for me. Soon we will have the Fuji m4/3 (I love and use the Fuji sensor and it will have distinct advantages for downsizing images to 1080 since it has a different sensor arrangement that reduces what I call 5d color banding but is not as good with highlights - at least so far), the revised EP2, a rumored mirror-less SONY, the E5 Olympus high end, and the DIGIC V processor cameras, the expansion of video to the entire EOS line, and possibility offerings from Nikon and RED.

It is threads like this, and the work of people like Barry and CommanderSpike that will teach us how to evaluate hDSLRs. Obviously, they present new challenges that were not so prevalent in pro video cams of the past. All of this obsessive pixel peeking is a necessary stage to develop the needed vocabulary. I do not think that is represents a move away from actually making films but a transitory step to build a foundation to use these new tools properly. They are not like the existing pro videocams and so we have to learn new ways to use and evaluate them. These threads that would be neurotic if applied to technologies we totally understood, but we do not know enough to film around the limitations of a technology that has not been thoroughly field tested.

 

[GMC]

Well thats ok. Nothing wrong with liking it, as long as you know that there were really strong greens and oranges in that shot that simply didn't make it on the GH1 shot, and are mostly hidden on the 5D shot in my frame grab for some strange reason only known to LCD monitor manufacturers! On the plasma screen it is more obvious which camera got it right.

Detail also. Even on static shots, the 5D holds onto it a lot better. Again this isn't noticable when a 1920x1080p moving image is made into a single frame and resized to 1280x720 for the forum.

If anyone reading this is forming their view of the GH1 versus 5D image quality based on the thread, I think it'll be a good idea to get down to a retailer and plug both cameras into a HD plasma if possible. I think that is the only way you can judge how big the gap is between the two. Trust me it is big. Honestly!

For whatever reason, a computer screen hides a lot of the 5D's advantage and covers up a lot of the GH1's short comings.

*Both great cameras however*

Dear commanderspike, I can only very strongly agree with Martti that in no way you should compare footage by direct playback via HDMI. When I was doing that for the first couple of times I was extremely disappointed by the "yellowish" image I was getting. I only realized a little later that my Panasonic Plasma was identifying my GH1 as an NTSC model, and changed the colour space to NTSC with some strange color settings.

The same footage (1080p24, smooth film mode) played back after pulldown removal via HDMI by my ACER computer (and the plasma set to 24p) gives very nice and correct colours.

Maybe even the GH1 is messing up the colors during playback via HDMI. In any case, with my Plasma the colours are pretty wrong by direct playback from the camera.

Maybe this helps.

best regards

 

[commanderspike]

Cheers. I'll take on board all of the above advice and include it in the next test, it's very constructive.

 

[commanderspike]

The GH1 is also sharper all through the depth of field. Look at the leaves in front of the tree trunk, there you can sense the serrated edges and texture in GH1 picture but they are soft on the 5D. Could be the lens though.

Hi Martti . It's not the lens. Both Contax Zeiss, the GH1 has a Distagon 28mm known as one of the sharpest on earth, and the other is a Zeiss 50mm F1.7 stopped down to F2.8.

It may be DOF. The focus plain is thinner on the 5D.

It may also be my Macbook or your screen clipping the colour latitute of the 5D clip, because...

Then HD plasmas are just plain hideous - if that's your standard then god help you. I wouldn't judge image quality from anything on those. Especially not by plugging the cameras straight into them, then you start to compare their playback electronics more than anything else. In general you are much too sloppy in your testing methods to come out with guns blazing stating opinions as hard facts like you do.

... a good HD plasma gives better colour and contrast and a wider colour gamut than even the best computer LCD monitor. I'm sorry, but that is a irrefutable industry fact. LCDs are only just now - in 2009 - beginning to catch up. I have 4 screens here and BBC Planet Earth on Blu Ray looks by far at it's best and most natural on the plasma screen. If that's not a benchmark for a screen I don't know what is, and I am sure the 5D and GH1 don't demand half as much from the screen as that kind of footage does in terms of colour and detail.

I think you should spend some time learning to colour correct. There's more data to play with in the GH1 picture even after your jpeg compression. Here's a one and a half minute Potatochop hack. Notice I could push up a little more detail in the near blacks that are just gone in the 5D picture. Not saying it's needed on this particular image, just showing that the info is there.

Again, colour correcting a compressed frame in JPEG format on a LCD screen is no substitute to viewing the full HD feed direct from the camera on a 50 inch HD plasma. I made sure colour and shadow detail were at their maximum on the GH1. The 5D was actually toned down in-camera. This was after experimenting with all the different in camera settings. I don't think it's sloppy testing.

Regarding the darkness of the 5D's shadow detail, that's because for some reason the footage looks darker after being pumped through my Mac. Especially in Quicktime. Less so in FCP. But the Mac does something strange to the 5D's H.264 footage's gamma. I'll try and solve this issue in future.

Damn I hate to be the doom monger here. If I had known how the test would have turned out I'd have started this thread in the 5D forum. It is not my intention to be negative against the GH1, as a GH1 owner myself who badly wants to keep just one camera (i.e.... the cheaper one!!). But I have to tell it as I find it and the 5D is making a very strong case against the GH1. Maybe my testing IS flawed? I hope so.

 

[commanderspike]

Dear commanderspike, I can only very strongly agree with Martti that in no way you should compare footage by direct playback via HDMI. When I was doing that for the first couple of times I was extremely disappointed by the "yellowish" image I was getting. I only realized a little later that my Panasonic Plasma was identifying my GH1 as an NTSC model, and changed the colour space to NTSC with some strange color settings.

The same footage (1080p24, smooth film mode) played back after pulldown removal via HDMI by my ACER computer (and the plasma set to 24p) gives very nice and correct colours.

My GH1 is a NTSC model but the TV output can be set to PAL in-camera. I think what your say is still plausible but I am sceptical because the colours from the GH1 footage do not improve on my Mac, rather the 5D footage seems worse. This is something I do need to look into though!!

 

[commanderspike]

"commanderspike"

You shouldn't shoot vibrant ( setting 6 ) mode as you posted to that being the setting you are using/guessing on your Japan language GH1.

I will try various modes next time, but one thing I should mention was that this WAS an optimised test based on prior experiments to see which mode maximised the GH1's picture quality. It wasn't vibrant straight out of the box, rather vibrant with contrast -2 and colour +2, yet the 5D still beat it on colour and contrast (when viewing the footage back on the 50 inch plasma), that strange JPEG frame grab Macbook gamma issue aside (which I am working on to correct).