Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Arri D-21 vs. Red

B

Brennan Klensch

Well-known member
Joined
Dec 17, 2009
Messages
143
Reaction score
0
Points
0
I'm not favoring one or the other here but would like to hear peoples' opinion on each camera and its quality. The only camera I've been impressed with as much as Red is the D-21, it had the most filmic quality in my eyes. ONly thing is, retail on that sucker is 170k..so...that's pretty much out. so that fact aside, let's hear your thoughts and facts
 
I've only seen tests, and the movie "RocknRolla" and the TV miniseries "Company Men" (shot on the D20), plus "Supernatural" on TV, which is shot on both the D21 and the RED.

The basic difference is that the ARRI D21 can send out either an uncompressed 2.8K RAW signal (for recording to a data recorder) or a 1080P 4:4:4 signal (for recording to an SRW1 deck). It has a 12-bit A/D just like the RED does.

In Log-C mode, the ARRI is a bit slower in speed/sensitivity than the RED. It also seems to have a little better dynamic range. The color science is optimized for conversion to film apparently.

Like all digital sensors, it prefers daylight-balanced lighting.

Obviously one of the main differences is the optical viewfinder on the ARRI.

Watching "Supernatural" in HD broadcast, I sometimes try to make guesses as to when they are using the ARRI vs. the RED, but it's pretty similar. I occasionally see that faint mottled "chunky & fuzzy" blue channel noise that is characteristic of the RED in underexposed tungsten lighting. I'm not sure why the blue noise with the RED looks that way, but I sometimes wonder if the REDCODE compression is somehow softening and fuzzing-out the blue noise, of if it's just an aspect of the debayering process. The noise in underexposed ARRI D20 images, as seen in the few night exterior scenes in "RocknRolla", is sharper and more pixellated, more of that "classic" noise pattern.

The lower sensitivity of the ARRI D21 is one reason I haven't used it yet, that and the expense of renting a separate recorder.
 
Last edited:
My thoughts... As far as owning a camera goes, I stopped playing keep up with the high end camera a long time ago. I found it far too expensive, and far too limiting.
Today's cutting edge is tomorrows old hat.
My suggestion... Be realistic with your camera needs. If you need to own a camera, then purchase one your client base will support. For example... The majority of the stuff I shoot is local TV commercials with budgets under $5K. It would be foolish for me to invest in a $170K camera when my client base would not support it. So I own a few lesser cameras that cover the needs of my client base at that level. When I do regional work with a higher budget needing a higher level of quality, I simply rent the appropriate camera for the job.
Filmmaking is really no different than shooting commercials.
You still have a client base, even if it is just you!
Sorry that I'm not debating the Red Vs Arri camera thing here, but these camera vs camera posts are a bit pointless, especially if you can't afford one.
If you like the look of the D-21, then do what other filmmakers do when they want to shoot on it, rent it if your budget will support it.

All the Best!
 
David Mullen ASC said:
I I'm not sure why the blue noise with the RED looks that way, but I sometimes wonder if the REDCODE compression is somehow softening and fuzzing-out the blue noise, of if it's just an aspect of the debayering process.
Click to expand...

We learned at REDUCATION that it indeed IS the bayer pattern in that there are 2 Green for every one 1 Red and 1 Blue and of course blue is the weakest.
EDIT: WE also learned that the new color science, is indeed adding some gaussian blur to the blue channel to enhance its performance in Tungsten.

David Mullen ASC said:
The noise in underexposed ARRI D20 images, as seen in the few night exterior scenes in "RocknRolla", is sharper and more pixellated, more of that "classic" noise pattern.

The lower sensitivity of the ARRI D21 is one reason I haven't used it yet, that and the expense of renting a separate recorder.
Click to expand...

I think whats important to remember here is the D21 sensor just came out and is being compared to RED's FIRST crack at a sensor in the MX (released 3 years ago now). I think the real comparison we are all waiting for is the D21 versus M-X Sensor in EPIC.

Also important to note that while ARRI started with a less compressed Wavelet Compression algorithm, EDIT: (Rumor has it, they may be releasing more compressed codecs in spring)
... (my guess is because the storage costs for uncompressed are still pretty prohibitive) while RED appears to be moving the opposite direction, i.e towards LESS compression.
Hence the higher REDCODE #'s coming down the pipe. (Yes, I understand those dont directly correlate to bitrate anymore) but still appears to me that as Graeme's math gets better
the sophistication on wavelet compression gets better.

ALL MY THEORIES, RUMORS I HEARD, like RED, are SUBJECT TO CHANGE ;)

All in all, what I find most flattering for RED, is that ARRI saw the value of Wavelet compression and RAW as the future.
 
Last edited:
Tim Whitcomb said:
We learned at REDUCATION that it indeed IS the bayer pattern in that there are 2 Green for every one 1 Red and 1 Blue and of course blue is the weakest.
EDIT: WE also learned that the new color science, is indeed adding some gaussian blur to the blue channel to enhance its performance in Tungsten.



I think whats important to remember here is the D21 sensor just came out and is being compared to RED's FIRST crack at a sensor in the MX (released 3 years ago now). I think the real comparison we are all waiting for is the D21 versus M-X Sensor in EPIC.

Also important to note that while ARRI started with a less compressed Wavelet Compression algorithm, they are releasing more compressed codecs in spring... (because the storage costs for uncompressed are prohibitive) while RED appears to be moving the opposite direction, i.e towards LESS compression.
Click to expand...

With Respect I think Epic with its M-X Sensor will be compared with Arri's New line of cameras with new sensors. Both should ship in 2010
 
Tim Whitcomb said:
We learned at REDUCATION that it indeed IS the bayer pattern in that there are 2 Green for every one 1 Red and 1 Blue and of course blue is the weakest.
EDIT: WE also learned that the new color science, is indeed adding some gaussian blur to the blue channel to enhance its performance in Tungsten.
Click to expand...

I think what David is referring to is the "clumpy" nature of the noise, especially in the blue channel, that is typical of Red images, particularly when underexposed. The cause of this could be the nature of the sensor itself, or more likely, the result of compression, which is done on a channel by channel basis and written to the R3d file, prior to any Bayer reconstruction (which is all a post process). Since Red has not released any information about or access to the RAW data prior to compression, and has not allowed any third parties to develop a demosaic algorithm, it is impossible for anyone outside Red to determine exactly why the noise looks like it does. But we can make educated guesses, as both David and I have.

I think whats important to remember here is the D21 sensor just came out and is being compared to RED's FIRST crack at a sensor in the MX (released 3 years ago now). I think the real comparison we are all waiting for is the D21 versus M-X Sensor in EPIC.
Click to expand...

The D21 has been around since early 2008, almost 2 years ago. And the D20, which used essentially the same sensor technology, was introduced in 2005, almost 5 years ago, and prior to any products from Red. And, as mentioned by Stephen Williams, the sensor technology to compare to Red's new sensor is the new technology in Arri's forthcoming digital camera line due in the spring. That sensor is quite different than that used in either the D20 or D21, and has features specifically designed to extend dynamic range, similar to what Red is trying to achieve. The early reports on that are very positive, much as the early reports on Red's efforts are very positive.

Also important to note that while ARRI started with a less compressed Wavelet Compression algorithm, they are releasing more compressed codecs in spring... (because the storage costs for uncompressed are prohibitive) while RED appears to be moving the opposite direction, i.e towards LESS compression.
Click to expand...

Arri has never used any compression, wavelet or otherwise. In fact, Arri has never even made a digital recording device. The output of the D20 and D21 has always been uncompressed, regardless of whether you're coming out as HD-SDI or Arri Raw (via T-Link on the same spigots). If you're referring to external recorders, like the Codex Portable (which uses JPEG2000 compression, much like that used by Redcode), that has nothing to do with anything Arri has done. The delivery of Arri Raw via T-Link is also uncompressed, and is recorded that way by all current T-Link partners (S.two, Codex, and Keisoku). Further, they have not, at least to my knowledge, ever publicly said anything about using compression, even on the new products, nor have they publicly stated exactly what the on board recording technology might consist of. All they've said is that there will be on-board recording for both video and Raw data, from either Arri, third parties, or both.
 
Tim Whitcomb said:
We learned at REDUCATION that it indeed IS the bayer pattern in that there are 2 Green for every one 1 Red and 1 Blue and of course blue is the weakest.
EDIT: WE also learned that the new color science, is indeed adding some gaussian blur to the blue channel to enhance its performance in Tungsten.



I think whats important to remember here is the D21 sensor just came out and is being compared to RED's FIRST crack at a sensor in the MX (released 3 years ago now). I think the real comparison we are all waiting for is the D21 versus M-X Sensor in EPIC.

Also important to note that while ARRI started with a less compressed Wavelet Compression algorithm, EDIT: (Rumor has it, they may be releasing more compressed codecs in spring)
... (my guess is because the storage costs for uncompressed are still pretty prohibitive) while RED appears to be moving the opposite direction, i.e towards LESS compression.
Hence the higher REDCODE #'s coming down the pipe. (Yes, I understand those dont directly correlate to bitrate anymore) but still appears to me that as Graeme's math gets better
the sophistication on wavelet compression gets better.

ALL MY THEORIES, RUMORS I HEARD, like RED, are SUBJECT TO CHANGE ;)

All in all, what I find most flattering for RED, is that ARRI saw the value of Wavelet compression and RAW as the future.
Click to expand...


Your info is GOLD Tim.
Thanks:)
 
Yes, Michael is correct, the sensor in the ARRI D21/D20 is several years old by now actually, most of the improvements have involved software and other tricks to get more out of the original sensor. Now I have no idea where the original RED ONE sensor came from or how long it existed before it was put into the RED ONE, but I suspect that it is newer than the ARRI D21/D20 sensor.

And Michael is also correct that the ARRI camera sends out an uncompressed signal. You can add compression if you want, as an HDCAM-SR deck would do (though mild). And some of the data recorders that record the RAW signal can add compression.

Both cameras use a Bayer CFA, so that doesn't explain why the blue noise looks different on the RED ONE camera. Like I said, the reasons might either be the debayering algorithm or the compression, I don't know. But the way that blue noise looks on the RED is unique to the RED, I get a different look to the noise with almost any other digital cine camera I've used or seen tests of. It's not a good or bad thing, it's just an characteristic.

The reason that ARRI first got their 1080P HD 4:4:4 output working on the ARRI D20 and only now have gotten around to getting 2.8K RAW to work is just a matter of (smartly) setting priorities: recording to HDCAM-SR using a portable SRW1 deck was a more common, easy to use, and cheaper approach a couple of years ago, and more supported by post houses, than recording RAW to data recorders was.

It's true that RED has made RAW workflows more and more feasible and commonplace, working ahead of ARRI in this regards, and RED should get a lot of credit for that.

But a 1080P workflow is still far more common today than RAW in moving images post-production, which is one reason why 1080P HD cameras are so popular for TV production. I think ARRI is offering RAW workflows just a the right time; two years ago, it would have made less sense compared to a 1080P HD workflow. But I think now is the right time. We all know that the future is an all-data shooting and post environment, free of video codecs and whatnot, and that HD video will only be one of many deliverable formats, but we are not quite there yet. RED has certainly pushed the industry ahead in that regards though.
 
Last edited:
All these advancements in digital image technology are becoming very exciting and will be much more exciting next year. Red, Arri, or anyone else.. Competition will give us something nicer ;-) We just need to try harder in order to keep up with new things.
 
M Most said:
I think what David is referring to is the "clumpy" nature of the noise, especially in the blue channel, that is typical of Red images, particularly when underexposed. The cause of this could be the nature of the sensor itself, or more likely, the result of compression, which is done on a channel by channel basis and written to the R3d file, prior to any Bayer reconstruction (which is all a post process). Since Red has not released any information about or access to the RAW data prior to compression, and has not allowed any third parties to develop a demosaic algorithm, it is impossible for anyone outside Red to determine exactly why the noise looks like it does. But we can make educated guesses, as both David and I have.

...
Click to expand...

I think another aspect that's being overlooked is the material used for the Bayer mask itself.

It's apparently quite difficult to create very precise absorption filters, e.g. permit blue to pass through but block all other spectrums. (You would think it would just be a very small CTB gel, but apparently it's not that simple or CTB isn't nearly as selective as we think it is.)

I've read silicon is actually very reactive to blue light, which make sense b/c blue light is very high energy and a sensor is a variation of a photovolatic cell. So I think we have to look at the sensor/mask combination when trying to evaluate why a sensor performs as it does.

I hope this information is accurate, as I'm not a scientist or chip designer (of course).

Oh, and Happy Holidays to EVERYONE.
 
Tim Whitcomb said:
All in all, what I find most flattering for RED, is that ARRI saw the value of Wavelet compression and RAW as the future.
Click to expand...

I'd think what's most flattering for RED is that people are comparing their 17K camera to another company's 170K camera and debating about which is better.

Do you think people often compare a Ford Taurus to a Ferrari and debate which is better?
 
Peter Moretti said:
I've read silicon is actually very reactive to blue light, which make sense b/c blue light is very high energy and a sensor is a variation of a photovolatic cell. So I think we have to look at the sensor/mask combination when trying to evaluate why a sensor performs as it does.
Click to expand...

Actually, it's very in-sensitive to blue light. And very sensitive to red light.

Graeme
 
You must log in or register to reply here.
Back
Top