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

[Michael Hoffman]

OK. So - question about 4K then - because the grain structure in the blacks is still pretty evident.

Should I set Resolve to 8K - then render out in 4K....

Or - should I have Resolve at 4K - and render out to 4K?

I've been doing the bottom.

I have a Dual D700 Mac - so it'll do the 8K to 4K bounce - but I didn't know if it added increased quality - better downsample - etc.

Thanks!

- Mike

 

[Homer Dulu]

Hi Michael,

I'll take a look at the R3D when I get home, but just wondering, how does the noise look when you transcode down to 4K? Also how does it look if you do a really light noise reduction pass just to get rid of chroma noise (leaving the luma noise and still retaining most of your detail), and then scale down to 4K? The grain might look pretty big when looking at 100% at 8K, but when you scale down to 4K or even 1080p, the the perceived amount of grain gets reduced as the grain size also gets smaller.

Also, regarding what you see on the monitor, if you have focus peaking turned on, it will get noisy in the shadows because of the added contrast.

One thing that might help is checking the RAW view when determining exposure as well.

Btw, with the new firmware, 1280 is equivalent to 2500-ish ISO on the previous versions of the firmware so if you were exposing based on that ISO, yes, you will be seeing some grain at that point.

The previous advice in other threads holds true - in low-light situations, use a lower ISO to protect your shadows, for well-lit situations, use a higher ISO to protect your highlights. It was confusing to me, as the way you would use ISO in a Red is backwards compared to say, your C300 or DSLR, but I found it helps a lot.

Hope this helps!

 

[Homer Dulu]

OK. So - question about 4K then - because the grain structure in the blacks is still pretty evident.

Should I set Resolve to 8K - then render out in 4K....

Or - should I have Resolve at 4K - and render out to 4K?

I've been doing the bottom.

I have a Dual D700 Mac - so it'll do the 8K to 4K bounce - but I didn't know if it added increased quality - better downsample - etc.

Thanks!

- Mike

On my Hackintosh, I do both. In Premiere and FCP X, I have either an 8K or 4K timeline and render out into 4K. I don't really notice a difference but then again I don't have eagle eyes haha.

Maybe try an 8K timeline, and as I mentioned above, do a light NR pass, then render out to 4K.

 

[Michael Hoffman]

Thanks!!!

I will try both.

Interesting to hear on switching around ISO's.

It's the opposite of what I'm used to - even with an ND - usually I'll shoot outdoors 800 Max - then maybe boost to 1250 indoors if I need extra light - and try to light to crush the shadows.

On the RED video - they said to expose all to 1250 with that new LUT - so this must be a learning curve.

Weird thing is - I'll have to find the R3d in a bit - before this build - we lit a product video - and the blacks were super grainy at 800 ISO - well lit with an LED/HMI mix.

It was the top of a black desk - so - even if I crushed it - it was moving around pretty crazy. Like grain on these super 16mm film transfers I'm working on (I was shocked - as I was used to seeing lattitude in the grades on RED - especially when people soft clip the blacks and crush back down).

I had thought it was some kind of user error (at the time - we shot 800 ISO - then developed RED LOG/DRAGON 2 I believe).

Will try some DNR.

 

[Björn Benckert]

OK. So - question about 4K then - because the grain structure in the blacks is still pretty evident.

Should I set Resolve to 8K - then render out in 4K....

Or - should I have Resolve at 4K - and render out to 4K?

I've been doing the bottom.

I have a Dual D700 Mac - so it'll do the 8K to 4K bounce - but I didn't know if it added increased quality - better downsample - etc.

Thanks!

- Mike

No difference when you render it out is what counts.

 

[Brendan H. Banks]

Thanks!!!

I will try both.

Interesting to hear on switching around ISO's.

It's the opposite of what I'm used to - even with an ND - usually I'll shoot outdoors 800 Max - then maybe boost to 1250 indoors if I need extra light - and try to light to crush the shadows.

On the RED video - they said to expose all to 1250 with that new LUT - so this must be a learning curve.

Weird thing is - I'll have to find the R3d in a bit - before this build - we lit a product video - and the blacks were super grainy at 800 ISO - well lit with an LED/HMI mix.

It was the top of a black desk - so - even if I crushed it - it was moving around pretty crazy. Like grain on these super 16mm film transfers I'm working on (I was shocked - as I was used to seeing lattitude in the grades on RED - especially when people soft clip the blacks and crush back down).

I had thought it was some kind of user error (at the time - we shot 800 ISO - then developed RED LOG/DRAGON 2 I believe).

Will try some DNR.

Any idea what compression/resolution combo you were at? Noticing a lot of folks with weird grain on Helium. I've had some issues myself but I thought they were self-contained or specific to certain moments.

 

[Michael Hoffman]

The shoot that we did for a headphone manufacturer - before this - was at 7:1 - 800 ISO.

This shoot - being a convention center with little camera movement (mainly lockdowns and slow pans) - I assumed 10:1 was OK.

Some shots are 8:1 - but exhibit the same issues I'm seeing.

I'm noticing - some shots are exposed to the right - where I wasn't contending with LED screens and Shirts - and at a better white balance - as I was exposed for windows that were exposing a hall - and I still see the grain structure pretty heavily on mid-exposed items (signs - walls) - even on a 1080P Mac Cinema display (yep - old school here).

It's also pretty noticeable on the Flanders Scientific I'm grading on - VIA SDI.

Weird.

I had thought it was a blackshade issue - but I did the camera the night before - in my bathroom - at 73 degrees - lights off - censor covered - TWICE.

The convention center was 72 degrees.

 

[Homer Dulu]

Thanks!!!

I will try both.

Interesting to hear on switching around ISO's.

It's the opposite of what I'm used to - even with an ND - usually I'll shoot outdoors 800 Max - then maybe boost to 1250 indoors if I need extra light - and try to light to crush the shadows.

On the RED video - they said to expose all to 1250 with that new LUT - so this must be a learning curve.

Weird thing is - I'll have to find the R3d in a bit - before this build - we lit a product video - and the blacks were super grainy at 800 ISO - well lit with an LED/HMI mix.

It was the top of a black desk - so - even if I crushed it - it was moving around pretty crazy. Like grain on these super 16mm film transfers I'm working on (I was shocked - as I was used to seeing lattitude in the grades on RED - especially when people soft clip the blacks and crush back down).

I had thought it was some kind of user error (at the time - we shot 800 ISO - then developed RED LOG/DRAGON 2 I believe).

Will try some DNR.

Interesting - I'd like to see that R3D.

One thing I noticed with my Helium is that I was shooting a scene with party LEDs - when I changed my colour temp down to 3200K (as opposed to 5600K), the blue channel clipped much earlier and is noisier than the other channels, especially after I applied the IPP2 LUT. I think I read somewhere that Red sensors work their best around the 5000-5600K mark.

 

[Brendan H. Banks]

The shoot that we did for a headphone manufacturer - before this - was at 7:1 - 800 ISO.

This shoot - being a convention center with little camera movement (mainly lockdowns and slow pans) - I assumed 10:1 was OK.

Some shots are 8:1 - but exhibit the same issues I'm seeing.

I'm noticing - some shots are exposed to the right - where I wasn't contending with LED screens and Shirts - and at a better white balance - as I was exposed for windows that were exposing a hall - and I still see the grain structure pretty heavily on mid-exposed items (signs - walls) - even on a 1080P Mac Cinema display (yep - old school here).

It's also pretty noticeable on the Flanders Scientific I'm grading on - VIA SDI.

Weird.

I had thought it was a blackshade issue - but I did the camera the night before - in my bathroom - at 73 degrees - lights off - censor covered - TWICE.

The convention center was 72 degrees.

Hmmm. Auto or Manual blackshade? How did you cover the sensor? I've been hearing a lot of people with blackshading issues when they only use a lenscap and not a bodycap with no lens attached. Not saying this is the case with your setup, but I know some people found that that solved the issue.

This is a very interesting situation. Perhaps grab a few R3Ds from various scenarios and export a log file and send in a ticket. Could help yourself and others!

 

[Michael Hoffman]

Manual blackshade - body cap on - no lens.

It forced me to do one after the last update erased all my priors.

In a dark bathroom - lights out.

I'll have to dig through and find some sample R3D's.

Hope some haven't been deleted - but I have the new stuff - which - sadly - isn't ideal lighting.

 

[Michael Hoffman]

Hey, guys.

After MUCH digging through - something is definitely WAAAAY off here - either with me - or the Epic W.

I found 3 BTS stills of my camera - and I have the same shot in an R3D.

I'm uploading all.

The camera claimed I wasn't even CLOSE to clipping the whites on a prominent white shirt - with peaks going to 60 - and no warning on the LED monitor - YET - all detail is GONE upon import on that shirt in the foreground.

So - I was exposing to the right as much as humanly possible - even though the camera said I had tons of head room.

How do I compensate for this?

Again - was monitoring/shooting in the IPP2 build - medium contrast - 1280 ISO as recommended.

3 photos of the camera shooting are uploaded here - and an R3D of the actual shot I got.

I'm finding it impossible to monitor exposure.

My waveforms in Resolve are totally different - running it through the same process.

Despite being super grainy; I'm glad I exposed to (what I thought was) the left.

https://wetransfer.com/downloads/e5...4fe99a3f7be87cf76770a38020170427024019/ecace5

 

[Graeme Nattress]

I'm confused by " 1280 ISO as recommended." - we only ever recommend "starting points" for ISO setting. If you're shooting the IPP2 beta with the new ISO calibration, that starting point would be ISO640 - 800, and you'd modify that starting point based on a number of factors: your personal noise tolerance, nature of the scene (high key, low key, low light, balanced) and scene dynamic range (do you need to see into the shadows, do you need to hold detail in really strong highlights). We can't recommend an ISO because all those factors matter, and you are there and we are not, so what we recommend is the starting point where you should be, as the camera has a range of ISO settings, they all could be valid depending on the factors above.

I'm travelling and not easily able to look at files, but if you can put up an R3D still I can look at code values in the raw data and see where your exposure is in the range and perhaps offer guidance. One thing to learn in camera are the eposure tools as they can really help. The "stop lights" can tell you if you're clipping in the raw data, and the "raw exposure" overlay will show you which areas of the image are clipping, or mid grey, or in shadow (i.e. could be noisy).

Graeme

 

[Homer Dulu]

Hey, guys.

After MUCH digging through - something is definitely WAAAAY off here - either with me - or the Epic W.

I found 3 BTS stills of my camera - and I have the same shot in an R3D.

I'm uploading all.

The camera claimed I wasn't even CLOSE to clipping the whites on a prominent white shirt - with peaks going to 60 - and no warning on the LED monitor - YET - all detail is GONE upon import on that shirt in the foreground.

So - I was exposing to the right as much as humanly possible - even though the camera said I had tons of head room.

How do I compensate for this?

Again - was monitoring/shooting in the IPP2 build - medium contrast - 1280 ISO as recommended.

3 photos of the camera shooting are uploaded here - and an R3D of the actual shot I got.

I'm finding it impossible to monitor exposure.

My waveforms in Resolve are totally different - running it through the same process.

Despite being super grainy; I'm glad I exposed to (what I thought was) the left.

https://wetransfer.com/downloads/e5...4fe99a3f7be87cf76770a38020170427024019/ecace5

Hi Michael, I opened these files in the latest RCX Beta Build 50 and you're correct - exposure is within the range - you're not crushing the blacks, nor are you blowing out the highlights. Looking at it in RWG/3G10 view, the details in the shadows are definitely still there.

I changed the ISO to 640 (equivalent to ISO 1280 in previous firmware builds) and I don't see any significant noise at all at 100%. And basically disappears at 50%. Especially in a video downscaled to 4K or 1080p, you would barely see any noise. I did start seeing noise at 1280 ISO, but then again, that is equivalent to 2500 in the old firmware and that was the point I would also start seeing noise in my Epic-W, especially if the scene wasn't well-lit.

However, the scene itself seems to be just dim. At 640 ISO, the histogram distribution is very concentrated in the bottom 50% of the range, so even though the exposure is within limits I feel it's still a bit underexposed. Especially the skin tones.

Edit: I saw the JPGs you attached - I noticed that you may be underexposing, because the left hand goalpost is lit up about 1/4 of the way up, even though you're at ISO 1280 (remember, ISO 2500 equivalent). I believe that's why it looks noisy in post. I think you would've been golden in terms of noise if you shot at 400-640 ISO and exposed more to the right, where you wouldn't have that left goalpost lighting up.

 

[Michael Hoffman]

Graeme;

I was following RED's own tutorial on the IPP2 Build - https://www.youtube.com/watch?v=GyG456rKSws - which states they've boosted the ISO to 1280. If I swing down to 800 - it is SUPER DARK because the LOG 310 is crushing back the image.

Homer - I know that shot of my camera showed a goalpost signaling noisy blacks on the left - I put up those stills to show - if you play with that particular R3D - the whites in the shirt in the foreground are clipped/blown - but the right goalpoast never gave me a warning - and my histogram was all to the middle left - so I'm having a hard time of following the rules of "Being in between the goalposts".

I generally shot with the goalpost about 10 percent up on the left - because if we opened the lens up anymore - things were clipped all over.

On wider shots that were well exposed daylight - we shot in between the goalposts - but would lose white highlights (despite seeing no warnings on the right goalposts - and the histogram saying otherwise) - so we had to stop down to compensate to our external monitor's waveform.

So - I should be processing these 1280 rolls that are in the IPP2 build at 640?

Thanks!

Graeme - I can get some R3D snapshots later.

My other question is - when processing the same clip - with the same settings in resolve (1280 - Wide Gammut RGB/3G10 - IPP2 Medium Contrast) - you can see a totally different vectorscope and waveform than I had on set - and see that guy's shirt is clipped from the get go.

Thanks!

 

[Graeme Nattress]

That's not what Nate says though, and the video doesn't say to use ISO1280. That just happens to be what the camera is set at for that particular shot of the user interface. In the past when people have asked for a good starting point ISO I have suggested to them to use ISO1280-ISO1600 (with the old calibration), but always as a suggested starting point.

To go further, I'll have to see the R3D, and then I can get specific. Setting correct ISO and exposure is not always an easy thing and some scenes can present greater challenges than others. Scenes where you have zero control over the lighting can be most challenging of all, and it's looks like you were in that kind of situation. Couple that with IPP2 being new and that's why I want to help with advise to get the best out of the cameras.

With regards to protecting highlights - the stop lights are the quickest way to see if you're too hot on your exposure. If they're clear you may have the odd clipped specular highlight in the scene, but no areas will be clipping. If you switch on the raw exposure view you'll quickly see any areas clipping indicated in bright red. If that's all good you're not clipping in the raw data no matter what any external monitor is saying. Now of course, it could look clipped, but that's a bit different. Under IPP2 there's good control on what to do. It could be that you have a very high key shot and setting the highlight roll-off to "very soft" will allow you to see all the detail in the highlights.

In terms of processing IPP2 clips the different ISO calibration is all handled automatically behind the scenes so there's nothing different you need to do.

For Resolve - I've done testing there and found excellent results. If the image is totally different my guess is that Resolve is not decoding as RWG/Log3G10. Quickest way to check is to make sure that before you apply the LUT the image looks properly log like. Again, I'll be able to help more once I can look at R3D and when I get back to the office next week after this week's NAB meetings and travel.

Graeme

 

[Homer Dulu]

Graeme;

I was following RED's own tutorial on the IPP2 Build - https://www.youtube.com/watch?v=GyG456rKSws - which states they've boosted the ISO to 1280. If I swing down to 800 - it is SUPER DARK because the LOG 310 is crushing back the image.

Homer - I know that shot of my camera showed a goalpost signaling noisy blacks on the left - I put up those stills to show - if you play with that particular R3D - the whites in the shirt in the foreground are clipped/blown - but the right goalpoast never gave me a warning - and my histogram was all to the middle left - so I'm having a hard time of following the rules of "Being in between the goalposts".

I generally shot with the goalpost about 10 percent up on the left - because if we opened the lens up anymore - things were clipped all over.

On wider shots that were well exposed daylight - we shot in between the goalposts - but would lose white highlights (despite seeing no warnings on the right goalposts - and the histogram saying otherwise) - so we had to stop down to compensate to our external monitor's waveform.

So - I should be processing these 1280 rolls that are in the IPP2 build at 640?

Thanks!

Graeme - I can get some R3D snapshots later.

My other question is - when processing the same clip - with the same settings in resolve (1280 - Wide Gammut RGB/3G10 - IPP2 Medium Contrast) - you can see a totally different vectorscope and waveform than I had on set - and see that guy's shirt is clipped from the get go.

Thanks!

Hi Michael,

If you do swing down to 800 and it's super dark, unfortunately you have to do what you did back in the days of film - give it more light or open up your aperture haha.

Regarding the clipped whites in the foreground - are you talking about the guy with his back to the camera? I need to go back and check it again. I didn't notice any clipping in RCX and I thought the loss of detail was due to him being out of focus.

Did you post up the well exposed daylight ones? I might have missed that post.

I'll check it out tonight, I'd like to help you get to the bottom of this too!

 

[Daniel Reed]

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

A: With low compression ratios, as much as you want (or none).


Also, almost always in my experience w/ (any camera), shoot at the absolute lowest ISO you can and boost (or curve up) in post as needed.

Frankly, higher ISO base values look ugly to me - this amateur high ISO craze helps me out a lot, to mean, I get double takes on many of my (low ISO) shots.

 

[Michael Hoffman]

I will find those RED RAW files.

Sorry. I'm working. Still have to cut all this stuff.

On the ISO front - 1280 doesn't seem to be "high" with the new IPP2 lut running over the 3G10 space. In fact - it was darker than a neighboring cam at 800 ISO.

I was under the assumption - that I would seriously lose highlight dynamic range by stopping down to low ISO's - which - aside from being dark - would really effect me in run and gun/BTS situations like I'm shooting - especially when contending with LED's; white shirts - etc.

I was told on my previous RED model - that shooting outdoors at 200 ISO - etc - was a no no. That's why I always shot at 800 - and ND'd down with my Firecrest - as the other way seriously clamps the highlights - lowering the dynamic range.

Most had rated this camera at 1280 to 1600 - which; I assumed was a good tradeoff for grain/noise VS dynamic range.

- Mike

 

[Brendan H. Banks]

On the ISO front - 1280 doesn't seem to be "high" with the new IPP2 lut running over the 3G10 space. In fact - it was darker than a neighboring cam at 800 ISO.

Are you using the new Helium Calibration or the old? Seems like you're using the old calibration which would make your use of 1280 more like a 640 in terms of what people normally know and understand. I don't think that's crazy high. This new calibration thing though is throwing a lot of folks. I'm excited that the values will be more in line with a more standard rating system (light meters, other cameras, etc), but it does beg the question about how much noise is in the images. I was shooting 1280 on STH and there was some noise but it was OK. Now that that number is effectively 640, I'm a little surprised at the amount of noise. I'll probably live in the 400-800iso range under the new calibration for sure.

 

[Michael Hoffman]

I updated to the Beta Firmware with the IPP2 lut - and followed this video for setup: https://www.youtube.com/watch?v=GyG456rKSws

I then did a custom blackshade.

Graeme is being very nice - and when he has time - will help point me in the right direction.

I am wondering if I'll have to do like you though - and for certain jobs - shoot it more in line with 800 - which is pretty dark on comparable cameras.

- Mike