DRAGON vs EPIC MX Noise - NOT GOOD

[Martin Stevens]

After This thread I'm very disappointed about the Dragon.

I think you should be disappointed with this thread and not the Dragon, even though
this is a great thread, but can be misleading at times, and maybe you have been a bit mislead.



The Dragon is fantastic and after a few more firmware tweaks will be even better.

 

[Matt Ryan]

here is an example of same clips but both RG3 Dragon on the right. The dragon has slightly more detail but I hesitate to call it 3 more stops. both were at 800 ISO exposed for an skylight lit interior (I was testing exterior highlight roll-off among other things when I found the noise)

Open both in RCX, set them both to Redlogfilm, then roll them back to 250ISO and check out the difference. Dragon should have much more info than MX and you should see a difference. Feel free to post the R3D grab from each.

 

[Terry VerHaar]

ISO and ASA were introduced as industry standards for film light sensitivity.....so while you can obviously use iso (along with shutter and f/t stops) to adjust your exposure, every sensor still has a certain sensitivity that really does not change.....just like film has one, and still can be pushed and pulled....the metadata is just information for the post (or raw) software.....
certain hardware (ND filter, OLPF, ....) obviously changes the amount, quality, color, path,....of the light hitting the sensor so it obviously changes what the sensor sees (and can record)

when dragon was announced the detail that struck me was the claim of 16.5 stops DR from 250 -2000 iso....there was quite a bit of discussion about it....
now it turns out that maybe one can squeeze that DR out across that entire range but at what cost?
the hardware (camera, sensor, OLPF) still provides a pretty narrow optimum range (in terms of DR and noise) and the software can only get so much out of it...

i guess the idea of the user replaceable OLPF is interesting and in reality the only solution to provide the best file across an entire range of ISOs....i would be ok with it....

A major difference in the RED world and with raw is that no electronic gain is EVER applied to the sensor in capturing the image. In the past, if you wanted to shoot in low light, you often flipped a switch on the camera adding up to 18db of "gain" in some cases. That was electronic gain and introduced noise into the signal. Since RED captures raw (i.e. it captures EVERYTHING the sensor "sees") ISO is a less useful concept in some ways. Changing ISO does NOT inherently change what is being captured and it can be changed at any time with no change to the underlying (raw) image. The value in ISO seems, to me, to be more in understanding the approximate "base" sensitivity of the sensor, in a term that is relevant.

I have heard others say that changing ISO for capture DOES change the image captured but I have never seen any convincing argument as to why that would be the case, nor have I personally experienced this. YMMV

What is interesting about this discussion is that it begs the question of the capture of dynamic range vs. being able to view it. In the hypothetical extreme, if you could capture 30 stops of DR but only show 8 on your monitor, it definitely gives you room for creative expression in post but, ultimately forces the dilemma of how to "interpret" what is captured. If there is noise in the last 1/2 stop on the low end or clipping in that 30th stop, it still means you have a ton with which you can work. :undecided:

But I digress.

 

[paul schefz]

Not being one to hold every statement to a standard of perfection, I'm assuming with the old OLPF that RED did indeed provide 16.5 stops DR from 250-2000 iso. If not, why didn't everyone scream "you cheated me!" when everything was on the old OLPF? But I agree with your concern about hitting that target with the new OLPF... that is, at what cost?

I really don't want to lose anything in the high lights.

the problem is that even with the old OLPF the 2000 iso come with a certain noise penalty.....and obviously there were issues with the OLPF otherwise it would not have been changed....

someone just posted that they are now disappointed with the dragon....not sure that is fair....they should be disappointed with the claims of what dragon can be....
it still provides 6k resolution, great DR (with old and new OLPF IMO) and is clean but not at the iso promised.....
and in reality all this is getting into an area that is very subjective.....and with delivery nowadays going to the web (or even on TV) i doubt a well shot piece of work would be THAT much different from epic (if shot on dragon)....epic is still great and the air on top gets thinner and thinner.....

the issue for me is that this is simply a tool....i hate to bring up other companies here but sony is putting out sensors left and right, the new A7s shoots in darkness....a full size (5K) sensor with noise at 25 000iso comparable to epic at 800.....15+ stops DR....yes it is in a consumer body and the video is more or less forgettable.....but if sony decides to figure out compression and puts any of these in one of their many existing pro bodies....
as much as i love my red but the competition does not sleep and this dragon was announced a while ago, i am still not able to upgrade and it still does not seem to be finished......

 

[Ivan McCullough]

A major difference in the RED world and with raw is that no electronic gain is EVER applied to the sensor in capturing the image. In the past, if you wanted to shoot in low light, you often flipped a switch on the camera adding up to 18db of "gain" in some cases. That was electronic gain and introduced noise into the signal. Since RED captures raw (i.e. it captures EVERYTHING the sensor "sees") ISO is a less useful concept in some ways. Changing ISO does NOT inherently change what is being captured and it can be changed at any time with no change to the underlying (raw) image. The value in ISO seems, to me, to be more in understanding the approximate "base" sensitivity of the sensor, in a term that is relevant.
.......clipped.

Whatever way you cut it, wherever and whenever you apply the 'gain' electronic or digital you amplify and bring out noise that is there.

I don't see the inability to do this in camera as a feature as much as Jim trumpeted it as one. For me it was always a potential option ignored.

Alexa applies a version of this (or so it is said in a round about way blah blah blah) and it doesn't hurt the image making ability of the camera, quite the opposite, it is probably its best feature. (this works with RAW on the Alexa too in spite of assertions that it is incompatible with RAW)

For me Dragon is a great leap (from what I can see so far without my own tests) its just a shame that the Jim marketing machine of big promises is still having to be managed by RED.

 

[paul schefz]

A major difference in the RED world and with raw is that no electronic gain is EVER applied to the sensor in capturing the image. In the past, if you wanted to shoot in low light, you often flipped a switch on the camera adding up to 18db of "gain" in some cases. That was electronic gain and introduced noise into the signal. Since RED captures raw (i.e. it captures EVERYTHING the sensor "sees") ISO is a less useful concept in some ways. Changing ISO does NOT inherently change what is being captured and it can be changed at any time with no change to the underlying (raw) image. The value in ISO seems, to me, to be more in understanding the approximate "base" sensitivity of the sensor, in a term that is relevant.

I have heard others say that changing ISO for capture DOES change the image captured but I have never seen any convincing argument as to why that would be the case, nor have I personally experienced this. YMMV

What is interesting about this discussion is that it begs the question of the capture of dynamic range vs. being able to view it. In the hypothetical extreme, if you could capture 30 stops of DR but only show 8 on your monitor, it definitely gives you room for creative expression in post but, ultimately forces the dilemma of how to "interpret" what is captured. If there is noise in the last 1/2 stop on the low end or clipping in that 30th stop, it still means you have a ton with which you can work. :undecided:

But I digress.

every still/dslr sensor (and that is pretty much what the red sensors are) have a set/optimum ISO...they usually are rated at 100 or 200 and pushing or pulling is nothing but software magic (incl heavy noise reduction and serious loss of DR the further one moves that iso dial) a good way to get some info on this is DXOmark and their sensor tests.....the first fun fact is usually the sensors "real" iso rating and the next is that with each turn of the iso dial the DR goes down....so a 5DII with 11 stops DR (at 100) and beautiful clean iso 3200 only has 8 stops at that setting...that is how every sensor works.....
sony has their new A7s with a FF 12?mpix sensor 15+ stops DR and pretty amazing 25 000+iso...big pixels always help and there is obviously a lot going on internally and i am positive that 15+ stops DR turns to less then 10 pretty quick (before you hit the 10 000 iso) but still pretty amazing and cool....pushing the limits....but they also can't really create magic.....yet...

 

[Elsie N]

the problem is that even with the old OLPF the 2000 iso come with a certain noise penalty.....and obviously there were issues with the OLPF otherwise it would not have been changed....

someone just posted that they are now disappointed with the dragon....not sure that is fair....they should be disappointed with the claims of what dragon can be....
it still provides 6k resolution, great DR (with old and new OLPF IMO) and is clean but not at the iso promised.....
and in reality all this is getting into an area that is very subjective.....and with delivery nowadays going to the web (or even on TV) i doubt a well shot piece of work would be THAT much different from epic (if shot on dragon)....epic is still great and the air on top gets thinner and thinner.....

the issue for me is that this is simply a tool....i hate to bring up other companies here but sony is putting out sensors left and right, the new A7s shoots in darkness....a full size (5K) sensor with noise at 25 000iso comparable to epic at 800.....15+ stops DR....yes it is in a consumer body and the video is more or less forgettable.....but if sony decides to figure out compression and puts any of these in one of their many existing pro bodies....
as much as i love my red but the competition does not sleep and this dragon was announced a while ago, i am still not able to upgrade and it still does not seem to be finished......

This very reasonable post very well lays out your position and, may I say, helps me to better understand your arguments.

Still, I do not share your fears of RED Digital Cinema being overtaken by the competition (including any and all manufacturers whose products are adored because of their customers' ingrained perceptions and willingness to spend more for less '-) RED listens (to reasonable matters, presented reasonably) and then delivers.

And contrary to what you are doing, that is, holding off on a Dragon sensor upgrade, I have just this day notified my bomb squad rep that I am ready to upgrade my second camera. I haven't been holding off doing that because of the pandemonium set off by this thread... for me this thread has been only valuable in reading who has remained unconcerned by this noise characteristic or who keep on shooting and sharing their methods while RED addresses the complaints.

My sole reason for delaying my upgrade has been a financial one. And I'll happily take it in the same condition as my first camera became after the OLPF change to the new version.

 

[Nate Clapp]

Open both in RCX, set them both to Redlogfilm, then roll them back to 250ISO and check out the difference. Dragon should have much more info than MX and you should see a difference. Feel free to post the R3D grab from each.

here is a screen grab of each at 250 ISO REDLOGFILM. crazy magenta in the clipped sky of Epic MX. otherwise I don't see a huge difference. Exact same stop (to the extent two RPPs can be set for the same stop!)

 

[Elsie N]

Whatever way you cut it, wherever and whenever you apply the 'gain' electronic or digital you amplify and bring out noise that is there.

I don't see the inability to do this in camera as a feature as much as Jim trumpeted it as one. For me it was always a potential option ignored.

Alexa applies a version of this (or so it is said in a round about way blah blah blah) and it doesn't hurt the image making ability of the camera, quite the opposite, it is probably its best feature. (this works with RAW on the Alexa too in spite of assertions that it is incompatible with RAW)

For me Dragon is a great leap (from what I can see so far without my own tests) its just a shame that the Jim marketing machine of big promises is still having to be managed by RED.

OBJECTION YOUR HONOR: Argumentative

YOUR HONOR: Sustained.

 

[Blair S. Paulsen]

Reality check.

Reality check.

Virtually every "pro" camera out there does exponentially more noise reduction internally than any RED camera. A perfect example is the Sony F55 which uses extensive on board DSP to achieve roughly 14 stops of DR. IMO the image looks "airbrushed", with subtle textures lost along with the noise. When viewed on a 1080 monitor, even the Dolby, the image looks pretty darn good as the downscale process combined with the resolution limits of the monitor dovetail nicely. The "price" of this approach is a more "artificial" looking result with no guts - AND - no way to change that personality.

IMO, noise reduction in developing/processing (RCX, Scratch, Resolve) that didn't require significantly more resources (read time) would be ideal. As a practical matter, a new calibration that trades a little bit of high frequency detail for less noise seems like the right move. Thanks to the Dragon's (roughly) 18 megapixels (or even the 9 John Galt might posit) I really doubt even a "de-tuned" decode would be too soft. As others have noted, the reality of such adjustments are several times more complex and interdependent than such broad terms can adequately describe.

Cheers - #19

 

[Anthony Marino]

Fwiw I'm following this thread closely and this is exactly why my next camera is a RED. Input from the big dog, ultra creatives sharing their knowledge. They'll get it right, I'm not worried, thanks.

 

[Matt Ryan]

here is a screen grab of each at 250 ISO REDLOGFILM. crazy magenta in the clipped sky of Epic MX. otherwise I don't see a huge difference. Exact same stop (to the extent two RPPs can be set for the same stop!)

But then you're not getting the full range of the Dragon. When you shot both, did you notice you could open up on Dragon more before the sky clipped? If you shot both at the same stop you're not pushing Dragon to its limits, therefore you will not see much of a difference.

Next time, shoot both again, expose for MX and open up until it begins to clip, note the t stop. Do the same for Dragon and open up until clipping, note the t stop. Dragon should be able to open up a couple stops more. Then do this Redlogfilm test and see how the Dragon highlights come back while MX is gone... You could even take them down to 50ISO and investigate further.

 

[Mark Wilkinson]

Nate, can you share trims of those r3ds? (did you already and I missed it?)

Thanks -

 

[Gunleik Groven]

here is a screen grab of each at 250 ISO REDLOGFILM. crazy magenta in the clipped sky of Epic MX. otherwise I don't see a huge difference. Exact same stop (to the extent two RPPs can be set for the same stop!)

If you look at the wooden bars and the sky, you see that the Dragon holds them, while the epic loses them. By how many stops? Not possible to say from a comparison like this, but Phil, John and I have seen kinda the same tolerance vs the Epic in the highlights

If you start grading these two, this is not a worstcase for Epic color-wise. It would be more interesting to compare to shots with a bit richer colorpalette, highlight blowouts on both and shadow-info.

But the tendency you see in these two images is practically not "little"

 

[Elsie N]

Fwiw I'm following this thread closely and this is exactly why my next camera is a RED. Input from the big dog, ultra creatives sharing their knowledge. They'll get it right, I'm not worried, thanks.

Interestingly, your reasons were my own back in early 2007. Someone came to the forum with concerns that sounded outrageous, but none-the-less were concerns. Some wanted to ban him from speaking and I spoke up in his defense saying I liked the fact that he was able to put voice to his feelings. He was called out by those with more knowledge and more sense, which I agreed with. His misstatements were handled by the members and kept the FUD from spreading. I stated at the time that was why I decided to put down a deposit (and a month or so later, a second one) for a RED One camera.

I hope your ride will be as gratifying as mine.

 

[Jan Reiff]

magenta in clipped highlights at MX can be handled in RCX with the DRX slider ... Dragon doesn´t have this "issue" - my early tests with my Dragon and my MX 2 month ago were significant, Dragon holds highlight much better -

 

[Tim Morten]

I see a lot of assumptions (on both sides) about how big or small an issue this will be. Probably best to not be reactionary: I've found that speculation is not the most accurate predictor of the future. Now that the community helped provide examples, let's give Red time to work their magic..

 

[paul schefz]

And contrary to what you are doing, that is, holding off on a Dragon sensor upgrade,

i am not holding off.....still waiting for that email.....it just does not add to the "joy" of waiting if what you are waiting for might not be all that was promised....as of right now i am still going ahead with the upgrade once i get my notice....and i can't really think of anything that might change that....

 

[David Battistella]

If you look at the wooden bars and the sky, you see that the Dragon holds them, while the epic loses them. By how many stops? Not possible to say from a comparison like this, but Phil, John and I have seen kinda the same tolerance vs the Epic in the highlights

If you start grading these two, this is not a worstcase for Epic color-wise. It would be more interesting to compare to shots with a bit richer colorpalette, highlight blowouts on both and shadow-info.

But the tendency you see in these two images is practically not "little"

the big subjective and creative question is, "what are we looking at?" As in, is that sky the focal point of the frame and is it fine to blow out and how does that affect the overall image?

pixel peeping is great, but we have to step back. To me highlights are as important as what we dig out of the shadows and dragon can be excellent on both ends. We've seen that. Now, I think if it falls right in the middle of old and new OLPF it won't be a compromise and we will have all that DR on both extremes to go after.

Both those are the first parts of the image to be graded out generally speaking. It's nice to have the dr though.

Battistella

 

[Gunleik Groven]

Agreed David. 100%

point of my comment is that the poster saifd there wasn't much tech difference.

IMHO, even though this is not the best example to display the differences, they can clearly be seen