Optimal ISO for Alexa

What is the optimal ASA for the Alexa for minimum noise? They say the native ASA is 800. I don't need near that speed.

I know the 5D has some ASAs that have less noise than lower ASAs.

Does it work like that with the Alexa? If 800 is the best, I could always ND it down.

Also, should I use an IR with this camera?

Thanks for any help.

Marty Marty Mullin DP

my experience is that the 800ASA gives you nice, but significant, subtle noise, at least when recording ProRes. Looks a bit like a fine grain, not ugly, but.. Working with 400 gives much better results. VFX guys don't like the 800ASA noise.

The Tiffen T1s have performed fabulous on one show for me, stunning skin tones, while performing not ideal on another (green renderings within skin tones). Since I didn't dit the camera on that one however, that might have been a totally different issue. But I read that the T1s can have such effects, hence the mentioning. A little research revealed people talking pleased about the Schneider Platinums.

Art did some testing for his Article 'IR cheat sheet', which maybe helpful, possibly Art has some updated info: <http://bit.ly/V0GJim> .

Cheers,

Rocco Schult dit

I would agree that the 400ISO setting gives a cleaner result, although it depends on how one likes to hold a lightmeter. Either way you are not giving up much in the way of exposure range. The Alexa can capture more than 14 stops in Raw or Log, at 800 that translates to about 7.5 stops over and at 400 it's about 6.5. You have a lot of room.

Above ND.9 you should use IR protection. You'll need to re-white balance for it, but you can save the setting as a custom color temp. for easy and quick switches.

Mitch Gross Applications Specialist AbelCine NY

I agree but...

You have to understand where your exposure split is.

At 400 ISO you have less headroom than you do at 800 ISO

So, it depends where you need to place your range.

You have 14 stops and you can split it however you want.

Arri are very clear on this.

If you think you get better results thinking 400 ISO then fine, equally if you like 1600 ISO then fine.

Whats recorded doesn't alter...

Cheers

Geoff Boyle FBKS Cinematographer EU Based

At ISO 800 you get roughly +7/-7 stops above/below 18% gray. At ISO 400 you get +6/-8 with noticeably less noise.

I shoot it at ISO 400 because I'm not a big noise fan. And, as Rocco said, VFX houses don't like ISO 800. They much prefer 400. Many do rate it at 800 with great results, but they often have better noise reduction tools on hand than I do.

I've shot at ISO 800 at night and it worked very well. I still like 400 better as it's cleaner, and less noise means better detail resolution in shadows.

If you're shooting LogC you won't be able to use the full bit bucket at ISO 400 as the maximum signal caps out at or slightly below 100 IRE, but 800 only goes to 104 IRE. You have to go to ISO 1600 to actually get a highlight up to 109 IRE in Log C. You don't lose much until you get into the lower ISOs... I think 200 puts max LogC signal into the 80s.

Whenever I've asked manufacturers how they determine their "native" ISOs I get an answer that basically boils down to how much noise someone in QC is willing to stomach. Sometimes it's quite a lot. For some curious reason several cameras have "native" ISOs that put the same number of stops above and below middle gray, but surely that's just an accident.

You should use either Tiffen IRNDs or Schneider Platinum IRNDs. Regular NDs will definitely cause color shifts in far red/near IR reflective objects.

-----------------------

Art Adams | DP San Francisco Bay Area

What they mean by that is that 800 gives the best compromise between highlight handling at the top and noise at the bottom. The available bits are distributed more or less evenly from darkest to brightest.

As you decrease the ASA, you assign more bits to the bottom, the low end gets quieter, but you have fewer bits at the top, and so need to control your highlights better.

--Bob

Bob Kertesz BlueScreen LLC Hollywood, California

Each sensor has a point of maximum exposure at which photons completely fill the capacity of the photosites (Full Well Capacity). Each sensor has a point of minimum exposure at which photons are so few that it is no longer possible to distinguish signal photons from noise.

If you reduce exposure by one stop, the wells which were full before become half-full. The wells which before were readable as one stop above noise are also reduced by half and are now just noise. As a result, the dynamic range of the sensor is reduced by one stop.

You can amplify the sensor signal using analog gain, digital gain or a combination of the two to move the highlights back up one stop. But at the low end you are just amplifying noise.

You can attempt to regain the lost stop by applying noise reduction at the low end, but if that is really possible (without error), you could also have done it in the first place and thereby increased dynamic range of the initial exposure. You have still lost a stop of dynamic range.

So I must take issue with those (not Art) who claim their cameras achieve the same dynamic range at different exposure (ISO) settings. Each sensor has a dynamic range sweet spot. Exposures above or below that point decrease dynamic range (increases noise).

Cheers.

Noel Sterrett Admit One Pictures

I don't mind having a little less headroom. But what is the effect on noise?

And I will most likely be shooting Rec709, so does the "what's recorded doesn't change" still hold if I'm not shooting RAW?

Marty

Marty Mullin DP

Here's straight from ARRI's website, and I have found this to be pretty accurate.

EI 160 (+5.0 / -9.0) EI 200 (+5.3 / -8.7) EI 400 (+6.3 / -7.7) EI 800 (+7.4 / -6.6) EI 1600 (+8.4 / -5.6) EI 3200 (+9.4 / -4.6)

Between 400 and 800 you basically flip the levels of over and under range. If ARRI were to call the "correct" sensitivity the point where the under & over ranges were equal, I'd say it would land at something very close to 640 ISO.

Just think of this as the same as film. How many of us would take the exposure rating from the manufacturer and then choose to over-expose the stop anywhere from 1/3 to a full stop in order to get the look we prefer? Same deal.

In shooting REC709 at 400 ISO, you will tighten up the blacks some (less noise). What's recorded does change, and it is clear on the monitor.In REC709 it is a WYSIWYG camera like any other video camera, albeit with a great look with a fat signal that has more room for post adjustment than others.

Mitch Gross Applications Specialist AbelCIne NY

As most people have commented, I. R. & also more to the point Far Red polluting light waves from 680nm to 750nm can or will come into play. This is down to many contributing factors. The SPD of your light sources and the absorptive / reflective capabilities of fibres, material and dyes will affect the outcome.

In essence you have a moving target here, with a range of filtering options to address the issues over a wide range of camera sensors. All of the Far Red / I.R. Cutting filters will require an amount of post work, though we advise that any colour balance you can perform in camera with the aid of a grey and colour scale will enable you and your post team to obtain a quicker grade. So a reference with-out filtration is also helpful.

If you would like to review my over White Paper of this issue here is the link.

http://www.thelondonfiltercompany.com/filters/pdf/Far_Red_Pollution_White_P aper.pdf

So when you have factored in ISO and all its latitude and noise considerations, then worked out what ND to use. Another question you might like to ask yourself is will adding diffusion make any difference. Or then what diffusion should you use to take into account all of these considerations in conjunction with your lens package to get the look you are after.

Carey Duffy Technical Consultant MPTV Camera Filters Tiffen International Ltd

Speaking from a VFX standpoint, it's not always the noise that is a problem. Not having enough DR can be a problem too. It depends on the context. Denoise/renoise tools are pretty good, so usually the noise is fairly straightforward to deal with, and if handled correctly, can help sell the integration. In most cases, people should be denoising before pulling keys to get the best results (not to say that everyone does this).

I think the exceptional DR on Alexa makes it more forgiving than, say, Red Epic, which yields footage with very little headroom when shooting with a low ISO. But in either case, if you know nothing you're capturing needs the upper part of the range, bringing down the ISO to reduce noise seems to make a lot of sense, if you don't specifically want the look of the noise.

The thing I've always been curious about is why, from an engineering standpoint, the digital cinema cameras can't treat ISO the same way stills cameras do. Is there some sort of analog gain applied in a stills camera that lets it capture raw images with different ISOs without affecting the dynamic range? There are cameras out there now shooting ISO 128000, etc, which seems like it would completely wipe out the range in the shadows if the stills cameras worked like digital cinema cameras. Being locked into a fixed relationship between noise and dynamic range seems rather limiting.

Andy Jones Director of VFX MassMarket Venice, CA