The Master Pedestal mystery

[kidzrevil]

I haven't had time to do any scientific tests of my own but I hear some filmmakers get good results by adjusting the master pedestal on compressed 8bit cameras like the GH4 and NX1

1) some say you get more dynamic range. This is false and speaking with visioncolor via fb they have advised against it. They as well as others say you lose data in the 0-255 range and it can introduce banding

2) some say it's merely to get the camera compression to prioritize the mids over the deep shadows. 

Im interested in #2. Has anyone experienced better internal compression when raising the master pedestal ? 

[Lintelfilm]

I haven't had time to do any scientific tests of my own but I hear some filmmakers get good results by adjusting the master pedestal on compressed 8bit cameras like the GH4 and NX1

1) some say you get more dynamic range. This is false and speaking with visioncolor via fb they have advised against it. They as well as others say you lose data in the 0-255 range and it can introduce banding

2) some say it's merely to get the camera compression to prioritize the mids over the deep shadows. 

Im interested in #2. Has anyone experienced better internal compression when raising the master pedestal ? 

#2 is interesting. I hadn't thought of it like that before. I've always pretty much stayed away from the MP. 

[kidzrevil]

#2 is interesting. I hadn't thought of it like that before. I've always pretty much stayed away from the MP. 

same here. I started looking into it after I found out technicolor cinestyle is 16-255 and when I heard someone say the data in the 16 and under range of 0-255 is useless. I figure maybe if that range isn't recorded more date would be recorded elsewhere and that can alleviate h.264 & h.265 macro blocking issues 

[M Carter]

My experience with master ped over the years is: it's not more light. It's not like upping your shadow fill ratios.

But - with some compression schemes and some camera-tweaked looks/profiles… it can help to lift it a touch. It can be handy to bring two different camera a little closer look-wise on set. IN MODERATION...

You just really, really have to run tests. If there's time on set, run a take and slate it with the new master ped setting so you can see it on a real-world clip.

It's really the kind of thing that's too weird and specific to expect a useful forum answer. You have to see it on your gear with your style.

[Lintelfilm]

same here. I started looking into it after I found out technicolor cinestyle is 16-255 and when I heard someone say the data in the 16 and under range of 0-255 is useless. I figure maybe if that range isn't recorded more date would be recorded elsewhere and that can alleviate h.264 & h.265 macro blocking issues 

Why not just change the Luminance Level to 16-255 in the menu?!

[kidzrevil]

Why not just change the Luminance Level to 16-255 in the menu?!

maybe that will work, hopefully it doesn't cause a disaster in my NLE

My experience with master ped over the years is: it's not more light. It's not like upping your shadow fill ratios.

But - with some compression schemes and some camera-tweaked looks/profiles… it can help to lift it a touch. It can be handy to bring two different camera a little closer look-wise on set. IN MODERATION...

You just really, really have to run tests. If there's time on set, run a take and slate it with the new master ped setting so you can see it on a real-world clip.

It's really the kind of thing that's too weird and specific to expect a useful forum answer. You have to see it on your gear with your style.

any experience with it on the nx1 ? Im not using it to get a look on set, im using it to fight macroblocking in the shadows

[Guest Ebrahim Saadawi]

The idea of Log Curve (Raising the lower curve (MP) and/or lowering the upper curve) being a loss of information due to the tones being recorded in a smaller portion of the 8bit shades, has proven to me, to be just not true. I've never had any introduced artefacts with using an in-camera LOG curve and de-Logging to see (lost colour tones) compared to standard curves. Even on the Leica's ridiculous LOG curve that recorded the data in basically a strip on the waveform, it didn't cause any issues with colour tones or banding. While on the other hand, some camera just have banding more than others and I've found out that profiles don't introduce nor eliminates them, just hides them or show them. The NX1 has banding, the A7s has banding, GH4 less so, It's just that I've never tested and saw the number of colour bands increase or decrease with using a LOG gamma or not, they're identical in each camera. I think since they're all 8bit camera, the banding performance variation is based on softeware reduction/noise reduction/colour processing, for example the C300 shows very very little banding while the 5D shows heaving banding. It's also due to the imaging sensor and how it's debayered and downscaled to the final image. 

 

In short, I've done a lot of heavy tests on Log vs Non log on 8bit cameras (Canon, Nikon, GH4, A7s, C100, C300), and banding was never a variation which is supposed to be the biggest one. It remains the same.(However there are reports on the Sont A7sII S-Log3 to increase banding) 

Using a Log curve I wouldn't worry about banding, worry about other things. V-Log for example introduces very bad shadow magenta blocks, S-Log on the original a7s crushes colour science, Canon Log and Nikon Flat, Cinestyle, Arri LogC, Blackmagic LOG come with no downfalls. 

By the way, Technicolour scientists along with Canon officially declared (while creating the Cinestyle LOG gamma for the Canon 5D MKII) that raising the black point up in the range is to make the codec reduce compression in those areas, as the normal profile, when the codec sees black it compresses this area far more than the higher tones (how H.264 works) and thus raising the black point up is for allocating more data (lower compression artefacts) in the shadow areas. 

Does it work? No. Tested if it reduces shadow compression and no it doesn't. I haven't checked with the Master Pedastel GH4 setting this trick, perhaps it will work and create less shadow compression by tricking the codec into thinking they're mid-tones. 

Worse a try side by side magnifying the shadows (after raising the normal one on post to the MP level) and see if one has less lower compression artefacts. 

[kidzrevil]

The idea of Log Curve (Raising the lower curve (MP) and/or lowering the upper curve) being a loss of information due to the tones being recorded in a smaller portion of the 8bit shades, has proven to me, to be just not true. I've never had any introduced artefacts with using an in-camera LOG curve and de-Logging to see (lost colour tones) compared to standard curves. Even on the Leica's ridiculous LOG curve that recorded the data in basically a strip on the waveform, it didn't cause any issues with colour tones or banding. While on the other hand, some camera just have banding more than others and I've found out that profiles don't introduce nor eliminates them, just hides them or show them. The NX1 has banding, the A7s has banding, GH4 less so, It's just that I've never tested and saw the number of colour bands increase or decrease with using a LOG gamma or not, they're identical in each camera. I think since they're all 8bit camera, the banding performance variation is based on softeware reduction/noise reduction/colour processing, for example the C300 shows very very little banding while the 5D shows heaving banding. It's also due to the imaging sensor and how it's debayered and downscaled to the final image. 

 

In short, I've done a lot of heavy tests on Log vs Non log on 8bit cameras (Canon, Nikon, GH4, A7s, C100, C300), and banding was never a variation which is supposed to be the biggest one. It remains the same.(However there are reports on the Sont A7sII S-Log3 to increase banding) 

Using a Log curve I wouldn't worry about banding, worry about other things. V-Log for example introduces very bad shadow magenta blocks, S-Log on the original a7s crushes colour science, Canon Log and Nikon Flat, Cinestyle, Arri LogC, Blackmagic LOG come with no downfalls. 

By the way, Technicolour scientists along with Canon officially declared (while creating the Cinestyle LOG gamma for the Canon 5D MKII) that raising the black point up in the range is to make the codec reduce compression in those areas, as the normal profile, when the codec sees black it compresses this area far more than the higher tones (how H.264 works) and thus raising the black point up is for allocating more data (lower compression artefacts) in the shadow areas. 

Does it work? No. Tested if it reduces shadow compression and no it doesn't. I haven't checked with the Master Pedastel GH4 setting this trick, perhaps it will work and create less shadow compression by tricking the codec into thinking they're mid-tones. 

Worse a try side by side magnifying the shadows (after raising the normal one on post to the MP level) and see if one has less lower compression artefacts. 

sorry for the confusion. My concern wasn't with LOG profiles. I refuse to use them and in the event I have to I never alter the curve. The last part of your statement was very interesting regarding Technicolor Cinestyle and their reasoning behind raising the black point. I feel like h.265 has even heavier compression than the h.264 codec resulting in  macroblocking in the shadows. I will take a page out of technicolors book and try raising the pedestal to trick the camera into prioritizing other areas in the image other than deep shadows

[Guest Ebrahim Saadawi]

LOG profiles are not just what the companies call them . Raising the master pedastel level and lowering a highlight curve in the GH4's settings is in itself a log profile, just different and tweaked to your own liking. So when I say the word LOG above I am talking generally any contrast-lowering settings that occue in-camera before compression. 

I'd be interested to ser your MP test, seeing if raising it lowers the shadow compression artefacts like technicolour intended to do (but never practically achieved in real world images really it was the same)

[kidzrevil]

LOG profiles are not just what the companies call them . Raising the master pedastel level and lowering a highlight curve in the GH4's settings is in itself a log profile, just different and tweaked to your own liking. So when I say the word LOG above I am talking generally any contrast-lowering settings that occue in-camera before compression. 

I'd be interested to ser your MP test, seeing if raising it lowers the shadow compression artefacts like technicolour intended to do (but never practically achieved in real world images really it was the same)

ooohhh I definitely have to look into that. I thought log was short for logarithmic. I know cameras like the NX1 shoot a linear rec.709 profile which is supposed to be way different in terms of dynamic range and how the bits are assigned to each stop of encoded dynamic range.

http://www.qvolabs.com/Digital_Images_ColorSpace_Log_vs_Linear.html

anyway so far so good. i seem to be recording more of the visible dynamic range by raising the pedestal. Nothing is clipped on either side of the histogram

[Guest Ebrahim Saadawi]

Your understanding of Log is correct kidzrevil. But each log curve is so different in different companies that the term has become very loose. Aside from that which is of no value in real world (terms), back to the original subject, I am going to test with you how Master Padestal affects the shadow compression. I'll do it on a GH4 so an NX1 test would be great 

I aleays expose to the right, meaning I lift the shadows up as high as possible (with exposure not curves) and indeed the codec allocates much better compression to these areas as they appear as mid tones, while recording it with black shadows it turns to complete compression elimination (of this non-useful area according to the codec), so it might actually explain also why ETTR yields much better images in noise and compression artefacts. Testing Master Padestal could be valuable 

[mercer]

Your understanding of Log is correct kidzrevil. But each log curve is so different in different companies that the term has become very loose. Aside from that which is of no value in real world (terms), back to the original subject, I am going to test with you how Master Padestal affects the shadow compression. I'll do it on a GH4 so an NX1 test would be great 

I aleays expose to the right, meaning I lift the shadows up as high as possible (with exposure not curves) and indeed the codec allocates much better compression to these areas as they appear as mid tones, while recording it with black shadows it turns to complete compression elimination (of this non-useful area according to the codec), so it might actually explain also why ETTR yields much better images in noise and compression artefacts. Testing Master Padestal could be valuable 

Ebrahim, speaking of the gh4 and ETTR, I would ETTR with my NX500 and got pretty good results, but with my G7, I am having issues ETTR. I use the histogram and I expose just right of center (proper exposure) but I am getting crazy clipped highlights. Any thoughts? In fact, with the G7, it seems a slight underexposure works better. For full disclosure, I am only using the aperture ring, ISO and a variable ND filter to control exposure... Should I +/- exposure compensation?

Also, do you use zebras... If so... What do you have them set at?

Sorry about the OT.

[Guest Ebrahim Saadawi]

Mercer, with the g7, when you have a clipped higlight, a single millimeter, then you ARE already exceeding ETTR, and you need to underexpose until no highlights are clipped, then you are in correct ETTR. 

That's a common misunderstanding of ETTR. 

ETTR technique is exposong to the right to the point just before clipping, it doesn't mean getting a bright exposure at all times, sometimes at ETTR you will have a pretty dark image because there's a small very bright area in your image, this means the dynamic range of the scene is just too high. In this situation, ETTR would be upping exposure until the point just before that bright small area is clipped, and leaving the rest of the image dark. In this scenario this IS correct ETTR. You're just doing your best by upping exposure (ETTR) to help the shadows bring up, and the best you could do (upping exposure/ettr) is not enough to get clean shadows, then your camera needs more dynamic range. In this case if the shadow areas are more important than the small bright area, you can give up and blow out the unimportant highlight, then up the shadows as high as possible, getting the cleanest image in the important area. It's a compromise you'll have to do at many times when you have a camera with a lower dynamic range. 

[Mattias Burling]

Ebrahim, speaking of the gh4 and ETTR, I would ETTR with my NX500 and got pretty good results, but with my G7, I am having issues ETTR. I use the histogram and I expose just right of center (proper exposure) but I am getting crazy clipped highlights. Any thoughts? In fact, with the G7, it seems a slight underexposure works better. For full disclosure, I am only using the aperture ring, ISO and a variable ND filter to control exposure... Should I +/- exposure compensation?

Also, do you use zebras... If so... What do you have them set at?

Sorry about the OT.

To ettr the g7. Set the Zebras to 100% and avoid them on any highlight of importance.

[kidzrevil]

Your understanding of Log is correct kidzrevil. But each log curve is so different in different companies that the term has become very loose. Aside from that which is of no value in real world (terms), back to the original subject, I am going to test with you how Master Padestal affects the shadow compression. I'll do it on a GH4 so an NX1 test would be great 

I aleays expose to the right, meaning I lift the shadows up as high as possible (with exposure not curves) and indeed the codec allocates much better compression to these areas as they appear as mid tones, while recording it with black shadows it turns to complete compression elimination (of this non-useful area according to the codec), so it might actually explain also why ETTR yields much better images in noise and compression artefacts. Testing Master Padestal could be valuable 

this makes a lot of sense now that I think about it ! When the gamma c and dr curves for the nx1 we announced the samsung documentation states :

"Two Gamma modes now provided for more control of aesthetic look of footage. Both gamma modes are designed for a more cinematic look.

• C Gamma mode is a preset logarithmic color setting designed to deliver higher overall contrast.
• D Gamma is a preset logarithmic color setting designed to capture a wider tonal range for increased dynamic range."

Your theory on ettr seems about right. On the nx1 i wont say that I am gaining dr but shadows on this camera has a habit of

looking clipped but when you take the files into post production you will see there were alot of detail hidden in the shadows.

When I raise the pedestal those details are no longer swamped in shadows. I will combine this with your ettr theory and expose as far to the right as possible without clipping the important highlights such as in the skin tone regions. 

 

I am using a custom picture profile by the way. My only adjustments have been to lower in camera sharpening and slightly reduce saturation. I have made no adjustments to contrast since I am not aiming to create a "flat" profile, I am aiming to merely mimic the technicolor cinestyle theory of moving the shadow bits towards the 16-255 range in the available 0-255 luma to counter the compression artifacts in the shadows

[mercer]

To ettr the g7. Set the Zebras to 100% and avoid them on any highlight of importance.

Thanks Mattias, I've been using a mixture of the zebras and histogram, but the zebras seemed overly sensitive, so I have been paying more attention to the histogram. I think I have my zebras set at 95, I'll have to check. With other cameras, I only used zebras and never paid much attention to the histogram... I think I'll go back to that. 

Thanks again!!!

[M Carter]

So far, with the NX1 - the Gamma DR gives a nice shadow lift - it's still kind of a shock to see just how much the image changes when you pop out of video mode to set white balance and then go back in. 

I'm only a couple weeks in with it, and I've been keeping master ped at about +2; so far, the footage comes into the NLE a little flat in the shadows, and bumping up the blacks makes it all good and very pretty. This will generally be my camera for controlled light situations, where I'm used to adding some overall fill if I want some lift in the shadows.

The bigger thing I see is in very contrasty situations - the color gets very saturated (though color balance stays good, other than cyans and blues really seem to pop). I try to dial the saturation down in those situations - I can de-sat in post, but I assume that not asking the codec to compress all that extra color may be a good thing. My gut feeling is that this sensor excels when you can rein things in a bit.

As far as ETTR - I dunno. I did some tests and wasn't happy with the highest values in the skin. VS exposing to the right, I'd say light to the left: get some more fill going and push the left side of the histo towards the middle a bit, and make sure you're well shy of clipping facial highlights. So far, I've been happy getting skin where I want it and then bumping up maybe a 3rd of a stop. I'm much more comfy when the face looks proper but bright on my monitor.

The other key to a really lovely look on this camera is glass. I don't own the S zooms, but my go-to glass for years has been these:

Nikkor 28-70 2.8; Nikkor 85mm 1.8 (just a greek god of an interview lens, generally use at 2.8 to 5.6); Old push-pull 80-200 2.8 (LUSCIOUS at the long end!!! And you can get a beater for $350 or so); Series E 100mm 2.8. The NX sensor is loving these lenses. I have a wide Nikkor zoom but rarely use it for video. The Samsung 16-50 kit zoom is really a nice lens wide open - stop it down much and there is some freaky sharpness stuff going on.

[Mars]

 

The other key to a really lovely look on this camera is glass. I don't own the S zooms, but my go-to glass for years has been these:

Nikkor 28-70 2.8; Nikkor 85mm 1.8 (just a greek god of an interview lens, generally use at 2.8 to 5.6); Old push-pull 80-200 2.8 (LUSCIOUS at the long end!!! And you can get a beater for $350 or so); Series E 100mm 2.8. The NX sensor is loving these lenses. I have a wide Nikkor zoom but rarely use it for video. The Samsung 16-50 kit zoom is really a nice lens wide open - stop it down much and there is some freaky sharpness stuff going on.

My fav lens on Nikon cameras at the longer end has always been the old AF-D 180mm ED. Not to heavy, really really nice rendering, gains a bit when stopping down to f4. Highly recommended. Have also the push-pull 80-200 you mention and agree with your assessment - used it recently a bit less though, especially since the 70-200/4 VR found its way into my bag. But on the NX1 should be certainly a good choice if you need the flexibility.

[kidzrevil]

So far, with the NX1 - the Gamma DR gives a nice shadow lift - it's still kind of a shock to see just how much the image changes when you pop out of video mode to set white balance and then go back in. 

I'm only a couple weeks in with it, and I've been keeping master ped at about +2; so far, the footage comes into the NLE a little flat in the shadows, and bumping up the blacks makes it all good and very pretty. This will generally be my camera for controlled light situations, where I'm used to adding some overall fill if I want some lift in the shadows.

The bigger thing I see is in very contrasty situations - the color gets very saturated (though color balance stays good, other than cyans and blues really seem to pop). I try to dial the saturation down in those situations - I can de-sat in post, but I assume that not asking the codec to compress all that extra color may be a good thing. My gut feeling is that this sensor excels when you can rein things in a bit.

As far as ETTR - I dunno. I did some tests and wasn't happy with the highest values in the skin. VS exposing to the right, I'd say light to the left: get some more fill going and push the left side of the histo towards the middle a bit, and make sure you're well shy of clipping facial highlights. So far, I've been happy getting skin where I want it and then bumping up maybe a 3rd of a stop. I'm much more comfy when the face looks proper but bright on my monitor.

The other key to a really lovely look on this camera is glass. I don't own the S zooms, but my go-to glass for years has been these:

Nikkor 28-70 2.8; Nikkor 85mm 1.8 (just a greek god of an interview lens, generally use at 2.8 to 5.6); Old push-pull 80-200 2.8 (LUSCIOUS at the long end!!! And you can get a beater for $350 or so); Series E 100mm 2.8. The NX sensor is loving these lenses. I have a wide Nikkor zoom but rarely use it for video. The Samsung 16-50 kit zoom is really a nice lens wide open - stop it down much and there is some freaky sharpness stuff going on.

Im playing with the standard profile with master pedestal +15. The next one i'll try is gamma c. Both profiles have saturation turned down with no contrast adjustment. Gamma DR is great but im looking for an alternative to get close to a straight out of camera look with profiles not intended for grading. I have a theory that h.265 is so compressed that ideally you want to get even closer to the final look than h.264 . 

I was going to use my nikkor with a tiffen diffusion filter because another thing I need to tackle is the in camera sharpening but i'll save that for another test. One thing at a time you know ? For now im using some m42 lenses seeing if that gives me more of a "film" look. My nikkors and canon fd L have more of a modern feel 

[mercer]

Mercer, with the g7, when you have a clipped higlight, a single millimeter, then you ARE already exceeding ETTR, and you need to underexpose until no highlights are clipped, then you are in correct ETTR. 

That's a common misunderstanding of ETTR. 

ETTR technique is exposong to the right to the point just before clipping, it doesn't mean getting a bright exposure at all times, sometimes at ETTR you will have a pretty dark image because there's a small very bright area in your image, this means the dynamic range of the scene is just too high. In this situation, ETTR would be upping exposure until the point just before that bright small area is clipped, and leaving the rest of the image dark. In this scenario this IS correct ETTR. You're just doing your best by upping exposure (ETTR) to help the shadows bring up, and the best you could do (upping exposure/ettr) is not enough to get clean shadows, then your camera needs more dynamic range. In this case if the shadow areas are more important than the small bright area, you can give up and blow out the unimportant highlight, then up the shadows as high as possible, getting the cleanest image in the important area. It's a compromise you'll have to do at many times when you have a camera with a lower dynamic range. 

Thanks, makes perfect sense now. Sometimes I have been shooting some shots which had a lot of range and I felt like I had to choose if I should allow something to clip in the background because my subject was still too dark. Obviously, in the perfect scenario I would use a light to supplement, but sometimes when I'm the middle of the woods, on a sunny day, I don't remember to throw an LED in my bag.