BlackMagic CEO Denies Existence of Fixed Pattern Noise in BMC4K

[Doctor Wu]

It all depends on how the MX chip is exposed. If one is underexposed in Tungsten, which has more red, than the blue channel is noisier. If there's enough light for good exposure than the blue light emitted by the tungsten source provides enough exposure to avoid noticeable noise in the blue channel. Or if the image is kept warm in the "yellow/reds" in post, than the blue color information is sufficiently supressed to avoid the appearance of noise. The MX chip is natively a "daylight balanced" chip like almost all chips so this is not unique to the RED MX at all. Given enough light however, there should be an acceptable level of noise whether the image is shot in tungsten or daylight, which is a given for any camera practically.

I've always shot with the RED Epic at 640 ASA for tungsten and 800 under daylight to compensate for lack of blue exposure under tungsten for a cleaner image, but in a pinch I've shot at 800ASA under tungsten without getting much added noise.

Epic hasn't been "noisy with tungsten" ever.

The original Red One was with the M sensor.

But this MX is clean under both tungsten and daylight sources.

 

[Doctor Wu]

Yes, most of the post houses can. Here in LA where you live, they've gotten smart and can charge UP TO $200 / hr to ingest your drives over USB3.0 or ESATA. It makes some sense - I agree there needs to a cost value to it - it's expensive technical labor and their machine time. It's wonderful that we have the ability to record in 16 bit 4K, lossless - amazing really compared to where we were just 5 years ago. But the post costs to finish from 4K DNG or raw files can be astronomical.

I
Most post houses should be equipped to handle 4K-5K by now, and shouldn't have much of an issue dealing with 4K DNG on those same boxes...

... average guy at home though? =P Well, that's a different story.

I neglected to post my feelings on the aliasing/moire/false detail part of this discussion: it's not good. And, it's completely worth calling attention to. While I don't think BMD cameras are difficult to work with at all provided you're using them as intended--built-out with batteries and accessories, the two largest setbacks are unfortunately very real setbacks.

- Moire
- Lack of 60P
.

 

[Kholi]

Yes, most of the post houses can. Here in LA where you live, they've gotten smart and can charge UP TO $200 / hr to ingest your drives over USB3.0 or ESATA. It makes some sense - I agree there needs to a cost value to it - it's expensive technical labor and their machine time. It's wonderful that we have the ability to record in 16 bit 4K, lossless - amazing really compared to where we were just 5 years ago. But the post costs to finish from 4K DNG or raw files can be astronomical.

Good point. It's like when Red Rocket Cards weren't readily available everywhere, the cost to run dailies were quite an added expense.

The tower's a business investment for sure, it's something I know can burn through 4K DNG RAW dailies at a 1:1 ratio the very least, but I'd likely charge a good 200/hr of footage to run them when it's all said and done.

 

[maarek]

I'm not arguing about your experience, I'm just curious if you meant that both looked about the same on charts (skintone swatches), or you had various kinds of skintones (humans) in your tests. It sounds like you were looking at actual skin, but just wanted to ask, anyway.

It was just one shot with little skin visible (actor had a facemask but hands were visible) There was also a color chart where there were skin hues. Of course it could be completely different with a different lighting situation where the Alexa could flex it's muscle more. Though the color chart makes me think that difference can't be that huge as the chart was quite close with both cameras. There were differences but not in the skin hues, mostly in saturated blues and reds.

Purpose of the shot was testing both on similar looking shots for a feature film. They're going with the Alexa.

I have no idea what filters they used in front of the lenses, if those could lead to noisy blues on the Epic. But it was daylight so how much could they mess it up?

 

[Kholi]

It was just one shot with little skin visible (actor had a facemask but hands were visible) There was also a color chart where there were skin hues. Of course it could be completely different with a different lighting situation where the Alexa could flex it's muscle more. Though the color chart makes me think that difference can't be that huge as the chart was quite close with both cameras. There were differences but not in the skin hues, mostly in saturated blues and reds.

Purpose of the shot was testing both on similar looking shots for a feature film. They're going with the Alexa.

I have no idea what filters they used in front of the lenses, if those could lead to noisy blues on the Epic. But it was daylight so how much could they mess it up?

Ah, I see. In that case, I personally wouldn't draw any conclusions, and it makes sense that you're seeing "similar". Skin and a Chart are vastly different. Charts can definitely look similar, the second subsurface scattering etc. enters the equation you begin to see the major gaps in color between formats. I'm not surprised that, if they can foot the bill for Alexa, they would go with that. =P

It depends on how the Epic's exposed, if it's under to try and match the DR of Alexa then it's likely to be noisy for sure.

 

[Robert Ruffo]

It all depends on how the MX chip is exposed. If one is underexposed in Tungsten, which has more red, than the blue channel is noisier. If there's enough light for good exposure than the blue light emitted by the tungsten source provides enough exposure to avoid noticeable noise in the blue channel. Or if the image is kept warm in the "yellow/reds" in post, than the blue color information is sufficiently supressed to avoid the appearance of noise. The MX chip is natively a "daylight balanced" chip like almost all chips so this is not unique to the RED MX at all. Given enough light however, there should be an acceptable level of noise whether the image is shot in tungsten or daylight, which is a given for any camera practically.

I've always shot with the RED Epic at 640 ASA for tungsten and 800 under daylight to compensate for lack of blue exposure under tungsten for a cleaner image, but in a pinch I've shot at 800ASA under tungsten without getting much added noise.

Yes, this is good advice. Teh MX sensor also has nice gentle fall-off into noise, and fairly forgiving (much more than BM 4K) falloff into white/burnout. Alexa is even better, but with less resolution, more weight, more expense.

 

[Robert Ruffo]

A phone camera film won Tribeca a couple of years back, but that doesn't mean most of us would choose to shoot a production on a phone.

If you rent cameras, as a lot of major market pros do, you can match the camera to the job by whatever criteria is relevant.

If you need or want to own one or more cameras the choice becomes far more complex. You have to balance practical utility for your type of shooting against budget limits and aesthetic preferences. These are often in conflict and the lower the budget, the greater the compromises one must make. There is no perfect camera.



Those of us that choose to own raw format cameras, Reds, BM's, or in my case a Digital Bolex, are driven much more by the aesthetic side of the equation I think. There is an extra level of skill and commitment involved to get the best out of any of these cameras and it takes knowing the camera and the post workflow intimately to make it work. But the upside of developing real skills and paying attention to critical details is that your clients pick you because of the aesthetic qualities of your work and not so much because you shoot any particular camera system.

Well said. Our clients (I work with my wife Amber) pick us not because we shoot with any particular platform (we own several, but rent many more, based on the needs of each project - Phantom, A7S, Sony shoulder mount ENG, you name it)) but partly because we are very picky about every last detail. That includes picking the right camera, the right lens, etc., very carefully. This does not get in the way of story. Quite the contrary, it enhances it rather than distracting from it with poor and incongruous visuals.

 

[Robert Ruffo]

Yeah it's both amazing and ridiculous that we're able to shoot 4K DNG lossless. Or 6K wavelet - whatever the camera. The storage requirements for a feature film are what's ridiculous, especially if you want to color grade a master off the original DNG's. For colorists and their technical staff, it's a big I/O problem.

5K Redcode si totally manageable without issue. You can edit it on Premiere on a laptop. Uncompressed is needly overkill of use only to nerds looking at numbers not actual image fidelity.

 

[Robert Ruffo]

Epic hasn't been "noisy with tungsten" ever.

The original Red One was with the M sensor.

But this MX is clean under both tungsten and daylight sources.

Well, it is a bit more so than Alexa, nothing tragic - you're right I wouldn't just slam it as noisy, like a condemnation, but for sure it is cleanER shooting with daylight sources than tungsten. Dragon is just plain noisy, and has seriously crushed blacks, and we skipped the upgrade until they sort that out, or release Dragon 2.0. Funny thing is Dragon is about the same noise-wise, under tungsten and daylight.

 

[Robert Ruffo]

Despite all the technological progress and ever cheaper cameras that shoot decent quality images, I still say a true entry level pro camera package, one that can shoot at least a decent 4:2:2 HD codec, with the minimum necessary lenses, media, and accessories is going to come in not much under $10k.

True, but ten years ago, for 1080p, that number was at least 5X that.

 

[Matt Gottshalk]

Well, it is a bit more so than Alexa, nothing tragic - you're right I wouldn't just slam it as noisy, like a condemnation, but for sure it is cleanER shooting with daylight sources than tungsten. Dragon is just plain noisy, and has seriously crushed blacks, and we skipped the upgrade until they sort that out, or release Dragon 2.0. Funny thing is Dragon is about the same noise-wise, under tungsten and daylight.

Agreed. I held off on upgrading to the Dragon as well.

 

[Postmaster]

5K Redcode si totally manageable without issue. You can edit it on Premiere on a laptop. Uncompressed is needly overkill of use only to nerds looking at numbers not actual image fidelity.

This is so wrong on on many levels.
Compared to uncompressed raw, Redcode feels like Prores.

 

[Kholi]

This is so wrong on on many levels.
Compared to uncompressed raw, Redcode feels like Prores.

=T Don't take the bait. It's an obvious insult and slur to compensate for the lack of a valid argument... Let it go...

 

[Robert Ruffo]

This is so wrong on on many levels.
Compared to uncompressed raw, Redcode feels like Prores.

It was good enough for several major studio films, but I suppose not good enough for you. Seriously I'll take famous directors and DPs word for it, along with my own careful testing.

 

[Robert Ruffo]

=T Don't take the bait. It's an obvious insult and slur to compensate for the lack of a valid argument... Let it go...

Not a slur, just a statement of fact. What you can't see as a difference, even under heavy grading, doesn't matter. Redcode 5:1 is almost perfect, 3:1 plain old perfect in terms of absent artifacts. Above 5:1 then yes you can be unlucky sometimes, but it depends on many factors.

 

[Kholi]

Not a slur, just a statement of fact. What you can't see as a difference, even under heavy grading, doesn't matter. Redcode 5:1 is almost perfect, 3:1 plain old perfect in terms of absent artifacts. Above 5:1 then yes you can be unlucky sometimes, but it depends on many factors.

Fact would imply that there's no difference between Lossy 5:1 data and mathematically Lossless ARRIRAW or CinemaDNG.

It seems like you're saying that it's a fact that there is no difference. And, if there is, that only "nerds" can appreciate that difference.

It's an opinion that it's unnecessary, or that you can't see it (you meaning yourself, as you really can't speak for anyone else), and that's all well and good. Fact versus opinion.


This isn't at you or about you, it's just an observation: It's incredibly awkward, for me at least, to actually acknowledge that someone used the word "nerd" today... not only because it's such a silly word, but media's also ruled by Geek Culture at this point in time... and well the word nerd? Like... what does that mean?

 

[Postmaster]

=T Don't take the bait. It's an obvious insult and slur to compensate for the lack of a valid argument... Let it go...

nah, I don't bite - I'm out - there is no hope.

 

[R Hudson]

 

[Robert Ruffo]

Fact would imply that there's no difference between Lossy 5:1 data and mathematically Lossless ARRIRAW or CinemaDNG.

It seems like you're saying that it's a fact that there is no difference. And, if there is, that only "nerds" can appreciate that difference.

It's an opinion that it's unnecessary, or that you can't see it (you meaning yourself, as you really can't speak for anyone else), and that's all well and good. Fact versus opinion.


This isn't at you or about you, it's just an observation: It's incredibly awkward, for me at least, to actually acknowledge that someone used the word "nerd" today... not only because it's such a silly word, but media's also ruled by Geek Culture at this point in time... and well the word nerd? Like... what does that mean?

I was a math wiz at school, and participated in science fairs. But, I acknowledge that cinematography is about the final, visual result, not mathematical theory. Nerd depends on context. here I am referring to tech nerds in cinematography, and pixel counting nerds. These always say for example that the lens that resolves the most pixels is "better", rather than looking at things from a human perceptual point of view, and picking the lens that suits the mood of their project best.

Endless testing by studios, top DPs, etc. all agree (not just me, by any means) that:

If you do not grade much (lift gamma gain mostly only) , Redcode 8:1 is fine. Many mainstream TV series are shot on Epics at that compression level.

If you do grade (secondaries, lost of curves, etc.) , then almost all the time 5:1 cannot be visually differentiated from uncompressed by human eyes, but you should not go higher, and 3:1 is ALWAYS identical to uncompressed for all practical purposes, including very heavy grading. Running footage through a color precision algorithm to mathematically compare to uncompressed, while amusing, is not a practical purpose, Being able to offload media cards 3X more expediently on set is a very practical purpose.

 

[Benjobe]

This thread has strayed pretty far from its original topic, which was the BlackMagic's CEO seemingly denying the existence of fixed pattern noise in the 4K Cinema Camera. Since that interview BlackMagic has unified the firmware for all its cameras and released numerous firmware updates to not only fix the problem of FPN in the 4K camera but also add more codecs, histogram, audio levels, recording time remaining, and other features.

I'll be the first to say it: I was wrong about BlackMagic's intention to improve its existing cameras through firmware updates, and I'm happy to admit it. It's a good time to own a BM camera!