How much grain structure should there be in the Helium - Epic W?

[Michael Hoffman]

Sorry to sound a bit green - but I'm running into a serious issue that has effected my work.

I was an early adapter to the Helium sensor - but; consistently - no matter how many blackshades; etc - I keep getting exceptionally grainy footage - close to my old Varicam gained to at at least 6DB - sometimes 12DB.

I've owned an MX - and cut/graded many Dragon shows - but I'm at a loss here.

I black shade before every shoot - and have tried to "Shoot to the right" - even dangerously.

I've tried the low light OLPF. When I render with the Noise Reduction on High in the RAW settings - I see absolutely no difference.

When I first reached out - I heard it was a "Color Science" issue.

I just updated to the IPP2 in camera - and shot at 1280 ISO as recommended by RED on their videos.

If I develop in RED LOG - I can see there is TONS of exposure there (VS the 3G10 that is used in camera).

I understand I should see a little noise - etc - but my AC - when zooming in for focus (who works lots of Alexa and Weapon shows) - asked during the shoot - if we should be seeing that much noise on the monitor.

Again - when I dump the footage - I've got a grain structure throughout - no matter how I develop. Considering how bright this all is - I'm kind of at a loss.

I'd love to apply a "Cool Look - or LUT" to some jobs - but when there's that much grain - it gets rough.

It's generally the blacks and shadowed areas that are moving off the charts - but blues are super noisy too.

The last shoot I did was a super bright trade show - I used a Wescott light gelled to match the greenish 4400ish white balance in the buildings for interviews.

Compared to the C300 footage - this is almost a no go - without heavy Temporal Noise reduction.

Hoping I am just missing something here. How I'm developing - etc?

I generally use Resolve. I can dig through and see if I have some smaller RED RAW files to upload if anyone wants to check it out. Hoping it's user error on my end somehow?

Any help would be appreciated. Thanks!

 

[Jaimy Soares]

What are your settings? Do you shoot 8K or do you window down to 4K or less?

What are the Compression settings with that resolution?

And last can you post a r3d. snapshot?

Thanks!

 

[Jaimy Soares]

Also are you running the latest beta firmware or the ipp2 preview firmware? Check in camera at the camera info tab which one it is.

 

[Michael Hoffman]

I always shoot 8K - unless overcranked.

I did just use the latest firmware beta. Yes.

The problem was the same before I upgraded. I had waited a LONG time to do so.

I generally was shooting 7:1 - but shot this trade show 10:1 - and I see no difference.

I can export and attach a still from resolve - processed in Wide RGB and RED LOG Film.

I think you'll see the grain all over the image before any general corrections. Especially the blues and blacks.

This is exporting a grab of the ungraded image - in Resolve.

https://wetransfer.com/downloads/ec...f8992f5211bf1f3f7de7d4ea20170426122632/1eae7b

Thanks!

- Mike

 

[Björn Benckert]

Upload / link to a r3d snap or it never happen.


Are you at 50ISO or 12400ISO when you develop, that makes difference. What your AC sees at 1:1 has some sort of relevance but I doubt that your AC ever in his life before has seen a 8k frame that up close before. Sure it should not be more grainy than any other 1:1 pixel punch in as helium pixels are simply very good. But I would not take an AC´s word for it. They are used to look at on set monitors and judging from whats showing on the the touch screen is a difficult one.

Also sure the way of transcoding makes difference. But normally it´s not really what usually is at fault, nor is black shading. Exposure and light fill is usually what the problem, especially when working in unbalanced light. For example, expose to the right is a good first step but then if your dark areas are still far to the left they are still underexposed so naturally you dont have a clean image if your left end is still out of the scope.

To me the helium sensor really need to be pushed to get some sort of character, if exposed well its like a picture on glas with no visible noise, especially when viewed in HD or even UHD. Something, that to me, is so clean it gets boring. but sure everyone has their perception of acceptable or desired noise levels, I try to film with some noise in camera and I find my self often adding noise in post.

 

[Jaimy Soares]

With the new beta firmware the iso settings are now redone. So 640 was the old 1280. and 800 now was 1600.

It doesnt'look right but you also said you where in redlog film. You need to be in log3G10.

Can you please make a .r3d snapshot? You can do that with redcine x. So I can import that here to take a closer look.

 

[Michael Hoffman]

Sure. Thanks!

 

[Michael Hoffman]

I have a bunch of others - that were lit in a more "Studio" environment; at 800iso - before the firmware updates - with the same - if not worse results.

Apologize for the green cast - convention center lighting. LOL. We did our best to gel our LED to match their funky ones.

I can grab those shots at my office later - but my concern - is the overall noise in the walls; blacks - and especially those blues.

Appreciate you looking at this!

https://wetransfer.com/downloads/0b...b77d4185940b048d6ad643b620170426123914/876d77

 

[Björn Benckert]

So from looking at the png I can only see that you exposed for his white shirt and the white TV in the background the rest of the shot does not really hold any high levels of exposure, especially not the black carpets etc. So yes even if the convention is bright, that does not simply make things easier, as your dark areas then get more pronounced and you kind of have to let go of your highlights or get poor exposure in your lowlights. HDRx is your friend in these cases simply expose for the background and then use the B track with lower exposure for his shirt and the TV.

Simply put, your set is not optimally lit. But helium has more tools than most and could tackle it better than possibly any camera, still or motion.


Here is an image where I just set the exposure to the shirt / not clipping it and gave it a normal curve, then parts of the back becomes rather dark but to me that looks rather noise free. So another way to do the shot would be to offer the guy another shirt....

writefile75.000001 by Björn Benckert, on Flickr

 

[Michael Hoffman]

That background wall is noisy - even if you grade it down and crush the blacks.

So is the blue shirt - etc. Super grain structure.

I'm at a loss here - but the C300 handled this much much better.

I assumed there was more dynamic range than just crushing the blacks - no?

I've shot conventions for years - and this is the 1st time I've ever run into this much problematic stuff.

We exposed to the peak of clipping on those shirts - via the waveform.

 

[Jaimy Soares]

First thing I see is you are really under exposed. The blacks are gone.


When I use iso 800 and balance a bit I can say that it looks the same as my camera. As of the noise.



But main fact is that the light was not well balanced around the persons you where shooting. A canon c300 is more suitable for such envoirements.

 

[Michael Hoffman]

BTW - we followed the lowlightt/noise warning on the camera.

Never came up more than 1/10th during the shoot.

If this is underexposed - then I have no idea how to shoot helium.

Monitor showed 50 on faces - 5 above black - and detail everywhere.

 

[Jaimy Soares]

I gave it a go in resolve. You can get all the info out no problem.

 

[Michael Hoffman]

Curious - what settings did you use to develop that?

Thanks!!!!

I guess I can chalk this all up to experience - despite having an experienced AC with me.

Question - how far right should I go for exposure on this camera on the IPP2 - Medium Contrast - setting?

We were shooting faces around 50 on the waveform of our external monitor.

Further?

I had issues with highlight retention before upgrading to this firmware on some controlled shoots - so we had read it was best to err on the side of caution.

Especially when dealing with LED TV's etc in these environments.

Thanks!

 

[Jaimy Soares]

Camera raw settings.

Iso 1280
denoise high.
color temp 3600
tint -21

I just used 2 nodes. One with the ipp2 lut. medium contrast roloff 4.
And one where I used the gain to 1.24
gamma 0.05
and lift -4.

exposing is tricky no right or wrong. Testing is key and I never trust monitors I trust my lightmeter, that worked wonders for me when switching camera's.

And you can add a bit of noise reduction too if you want around 20%

 

[Cole Sprague]

If your goal posts are lighting up at all then you are clipping blacks and you are going to see noise in the shadows.

50 on faces is pretty dark for white people. I try to keep it around 70 for people with a light complexion.

BTW - we followed the lowlightt/noise warning on the camera.

Never came up more than 1/10th during the shoot.

If this is underexposed - then I have no idea how to shoot helium.

Monitor showed 50 on faces - 5 above black - and detail everywhere.

 

[Björn Benckert]

Curious - what settings did you use to develop that?

Thanks!!!!

I guess I can chalk this all up to experience - despite having an experienced AC with me.

Question - how far right should I go for exposure on this camera on the IPP2 - Medium Contrast - setting?

We were shooting faces around 50 on the waveform of our external monitor.

Further?

I had issues with highlight retention before upgrading to this firmware on some controlled shoots - so we had read it was best to err on the side of caution.

Especially when dealing with LED TV's etc in these environments.

Thanks!

The IPP2 has nothing to do with it, the medium contrast has nothing to with it. Your external monitor has nothing to do with it, the wave form on your external monitor has absolutely nothing to do with it.

All that is relevant is how much / little light you give the sensor. Goal posts has a lot to do with it, so does the RGB stoplights and the Gioscope.

To be sure not to clip none of your stoplights should be lit, if you expose right under that then you avoid clipping. But you also need to see what is causing it to clip, is it the bright white sign or a lamp in the background or is it something relevant. Sure its nice to not clip the spotlights in the background or the white sign but as you are not in a controlled light environment possibly at times it´s worth clipping such feature to get good exposure on something more relevant. But again with red you can shoot double exposure with HDRx so as I see it there is never a reason to clip, or better put, there is always good to cover such clipped highlights with a HDRx pass as it´s possible. 20stops of DR is no problem to create for a scene like that as not much has relevant movement and you can really bring up the exposure like 2 stops over clip of your A track and let your b track be 3 under or such. then you got more range than a C300 would even dream of.

But should be very possible to get that scene well exposed without HDRX as long as your highlights are just under clip so open up until the stoplights goes on and then close back down just a fraction of a stop and that should beat all other cameras to date in terms of DR.

Also it´s important to compare fairly. Sure if you zoom in to the 8k frame and look at a pixel level, possibly it has more noise at the same exposure as your C300 but when compared at the same mastering resolution and the helium frame is downscaled to match your C300 frame, then it should be cleaner.


You say your lowlight goal post was lit to about 1/10 that means that 1 out of 10 of your pixels saw nothing / could not register a thing except for black noise. 1/10 might sound a little but the question is how far the other nine where from staring out in the dark as the only bright part of the image is the white sign the rest is quite far down there so simply when you expose to the right but not clip most of your image is still short of exposure.

And to me these images does not look to shabby. I assume they will be HD in the end and at that resolution they look quite glossy if you dont push up the exposure in the darks too much, no? And if you want brighter / cleaner then I would just clip the spotlight above the Philips sign that hit right into camera, To me the dynamic rounding in IPP2 would handle that just fine and you would get more exposure in the suits, carpet etc.

 

[Mike Finegan]

With the new beta firmware the iso settings are now redone. So 640 was the old 1280. and 800 now was 1600.

It doesnt'look right but you also said you where in redlog film. You need to be in log3G10.

Can you please make a .r3d snapshot? You can do that with redcine x. So I can import that here to take a closer look.

Are you sure about that? I found 1280 a little noisy - and for that now to be equivalent to 640 sounds wrong - so what I now need to shoot at iso400 to get the old iso1000?! (that seemed to be best for me not 1280 as Phil and others liked best)

 

[Michael Hoffman]

Thanks guys!

Thanks guys!

Appreciate all the input.

I think I need to get used to the idea of using the goalposts VS just monitoring exposure VIA external waveform.

I'll have to do more tests.

Those darn conventions are tricky because the overhead lights are such an ugly color to begin with.

I think I may have been exposing too far to the left to begin with.

On my Panasonics etc - I've always tried to keep faces in the 40-50 category.

Best!

- Mike

 

[Mike Finegan]

I always saw noise in EpicW footage and thought ti was related to the black shade issues. Just manually black shaded but have not done a 'balanced contrast' test to see if I can get that clear as glass mentioned above!
On the other hand if you post to 4K I don't think you are going to be complaining about grain - that is the magic of RED 8K - finally nyquist criteria and so 4K should be golden