Upcoming hands-on comparison of F3, FS100 and AF100 from Philip Bloom

[Mark David Williams]

Alister

Detail IS turned off completley. This is a film about Vampires hence the pale skin and choice of colour.
The girls hair is wet and so does look darker. The black line is her hair going over shoulder.

However the point is re resolution.

Mark

[Brian Drysdale]

Given that it's HD video, looking at the frames, I'd have assumed a light diffusion filter had been used. Nothing wrong with that, camera people have been using light black promist & pro mist filters to give a similar feel for many years. Black nets behind the lens are rather nice as well.

[Alister Chapman]

Can't see any decent resolution of anything in those shots. Resolution allows you to see those subtle textures in hair and dark areas that are lost in your images altogether. Resolution is all about contrast, resolution allows you to see subtle contrast differences and textures that are otherwise lost. It is not all about image "sharpness". In a high resolution image you would be able to see the stubble on the mans face as opposed to the nondescript fuzz in your shots.

At NAB at the Sony F65 demo the most striking thing about the film that was shown was not how sharp it was, in fact it didn't look particularly sharp at all. But then you could see incredible subtle shadow details and textures where you wouldn't normally see them, the actors iris's or the texture of leaves on plants for example, subtle, low contrast areas. This made the images look incredibly real.

When you buy a high end, high resolution lens what you are paying for is how that lens handles high frequency contrast. Sadly the one thing that the Letus does is completely kill high frequency contrast, thats why it's MTF50 results are always poor as it lacks both resolution and contrast and while you can have one without the other, unless you have both together the image will not contain those subtle details that film is so good at capturing.

If you have a high resolution, high contrast image you can compress the blacks and you will still see the textures.

Now this may be controversial, but it's my opinion that one of the reasons why so many low/no budget films are horror flicks of some sort is because it's easy to hide short comings in the budget with the typical crush blacks, ultra dark grade. It's harder in my view to make a great looking film that involves more natural lighting, natural vista's or conventional daytime scenes. These must be shot well as you can't hide the flaws with a dark. crushed grade.

[Mark David Williams]

We're talking about very shallow dof here You have to look at areas that are in focus and judge that
Alister

This is a test for resolution Not a test for contrast or how bad you think my colour correction or crushed blacks and overblown highlights which are if that is the case "choices" and not because I'm trying to hide the Letus shortcomings but because I liked the mood atmosphere and tone.

Any shadows in any format can be crushed out of existance. To say the Letus completely kills high frequency contrast is not true.

[Liam Hall]

Mark, I have used lens adapters a lot. They are an effective tool to get a particular look. Your shots look nice and moody. I like the look, but they do show a lot of technical flaws too. Chromatic aberration, lack of sharpness and loss of resolution being the most obvious ones.

For my money, adapters have had their day and there are better methods, both technically and artistically, to get the cinema look if you are on a budget.

[Piotr Wozniacki]

Mark,

I realize this kind of look might have been your intentional choice, but nevertheless - in accordance with what Alister said - I remember using the same trick to mask imperfections of my old good V1E, HDV camera years ago...Sending everything noisy and / or lacking full resolution to oblivion by just crushing blacks severely is all so easy to do.

With the EX1 picture (without the Letus), I never need to do it any more. With the Letus on though, I'm often forced to use the old trick again! Now compare this to the artistic freedom the S35 cameras are giving you, thanks to their phenomenal low-light capabilities AND the total DOF control...

I understand your situation very well; we've both invested into our adapters and glass, and we'd love to preserve the investment. But why rationalize it by saying this is a better overall solution to the new S35 machines? C'mon, I'm sure you know better than this :)

[Mark David Williams]

Hi Piotr

I've attached another picture for you to check for noise. This is straight from the camera as my colour correction is not being appreciated. I seem to be fairly lucky in that I don't seem to have had the same issues you have with noise. I would be worried about noise with the FS100 though as the camera is likely to have considerable noise reduction applied and could be the reason for its low resolution etc.

May I say the last thing I'm trying to do is protect my investment because as I've mentioned I really would like to get into the big sensor game.

Unfortunatley so far the choice is between the AF101 which is four thirds 8bit out and 680 lines of resolution but it does have ND filters and HDSDI out.


The FS100 is 8bit internal processing 8bit hdmi out with no filters an LCD on top making it very awkward for some shots.


Then again I have my EX1 1000 lines 10 bit out with ND filters that is ahead in the game in all areas except low light. I simply can't find justification for moving sideways and down. I wish I could and looking to be persuaded but so far I can't see a reason apart from low light capability but how many times do you need to film in extreme darkness if you want a chap lighting a match unless you want overblown flares you will have to still light the dark.

EDIT
One thing I have noticed is fringing around the girls shoulders. Same thing Alister noted as detail settings. Maybe in this shot the EX1 was zoomed in to far for some reason anyway not worth going into it on this thread.

[Alister Chapman]

The fact that the letus kills high frequency contrast is completely true. It's built in to the design. It uses a vibrating ground glass which randomly refracts the light passing through it which will reduce the resolution of any image projected onto it through dispersion and reduce the contrast through scattering. In addition the vibration adds additional scatter which will also reduce contrast and resolution at the frame rates typically used in video. A ground glass is the tool of choice in optics for reducing contrast. The only way the letus can not have an effect on contrast is by using a clear glass, but then of course it won't work as designed.

Consider this. If the GG stops vibrating what do you see? You see the grain structure of the GG. This appears as a texture. What is a this texture? It's small variations in brightness across the image caused by the varying dispersion and scattering across the GG. To eliminate this undesirable texture the GG is vibrated at a high frequency to blur the texture. This means therefore that subtle variations in brightness or contrast are getting blurred together to even out the image. So it's clear to see that the vibrating process alone must reduce the contrast in the image as its blending small changes in light and dark. To try to claim otherwise is nonsense in my opinion.

Are you really trying to say that the images don't look sharp just because your using an ultra shallow DoF? So what is supposed to be in focus. Ultra shallow DoF can make one part of the image perhaps appear sharper than it really is because of the dramatic difference between in and out of focus. This is one of the key areas people don't really consider. A higher resolution image needs less shallow DoF to achieve a similar effect as the ratio between in and out of focus increases with any true resolution increase. The circle of confusion decreases as resolution increases. This also applies to screen size and screen resolution, the larger the screen the greater the effect. The Circle of Confusion is governed by resolution and screen size as much as aperture and focal length.

Again going back to the Sony F65 demo footage, the scenes had shallow DoF, yet you could still see detail in the background due to the foreground starting off at a much higher true resolution.

You cannot ignore contrast when you talk about resolution. I wish people would understand that the two go hand in hand. It is ONLY through contrast that you see differences in resolution as it is only though contrast that you can see edges etc. If you don't have the contrast, you can't see the resolution.

Sure it's your choice if you want to crush your blacks etc. What I am saying that if you've got good high frequency contrast then you can crush your blacks but still retain subtle details that prevent blacks from just becoming solid blocks of nothing.

[Steve Mullen]

Originally Posted by Alister Chapman

I though Steve that you were suggesting that the "magic" was happening at the sensor level with pixels groups being read together off the chip as opposed to singly to overcome all the normal readout speed and heat issues associated with trying to read every pixel with DSLR type pixel counts?.

Roberts' finding of the Green signal being equal to Red and Blue, unlike that normally associated with a Bayer filter, suggested to me the Bayer chip was sending all photosite data to the DSP. (Yes, I'm now assuming the parts can all run at 60Hz, which given the Sony $400 camera can record 1080p60 seems very reasonable.)

I believe it is the DSP which groups 8x8 blocks into one of 3.7 MPixels that are then de-bayed down to 2MP. By combining the 8x8 in the DSP, the signal is boosted by +12dB (electronic binning) and all of the pixels output (2MP) have an equal strength of each color.

This new DSP process would yield a better than a Bayer balance amoung all three colors and a +12dB gain increase and, perhaps, a -6dB decrease in noise.

My assumption is that the DSP chips are the same and run versions the same firmware. The F3 DSP firmware does all processing at 12bits (from your reports) and does 4:4:4 processing which is output HD-SDI. The FS100 firmware does all processing at 8bits (David's reports) and does 4:2:2 processing for output via HDMI. When recording 4:2:0, both 4:4:4 and 4;2;2 are converted to 4:2:0.

I can imagine the conversion from 4:4:4 to 4:2:0 will look better than a conversion from 4:2:2 further increasing the quality of the F3. And, of course, the HD-SDI output will be really be much better.

One could calculate the data flow difference at 60Hz between 4:4:4 at 12bits and 4:2:2 at 8bits to see how much faster the F3 DSP would need to run. But, I'm not sure one could say much more than the FS100 should consumer less power. I don't think this will prove anything. Just be of interest.

PS: Both would look much better than the current VG10 which I assume simply pulls about 12 million photosite 16:9 window from a approx 14 million photosites (these are Sony specs) into the DSP where they are processed in the traditional way. No increase in gain at all. Green should be stronger.

[Alister Chapman]

I prefer the straight from the camera look, but you've still lost most of the texture from the darker areas of the girls hair, eyebrows, eylashes etc and you can't see any detail in her iris's. It all looks slightly soft. These are the low contrast areas that a high resolution image should reveal.

It does still look like you have some detail correction or sharpening going on. I wonder if turning detail on but setting the level to -20 would look any different? Or setting frequency to +99 to get the thinest possible edges.

[Brian Drysdale]

Originally Posted by Mark David Williams

Then again I have my EX1 1000 lines 10 bit out with ND filters that is ahead in the game in all areas except low light. I simply can't find justification for moving sideways and down. I wish I could and looking to be persuaded but so far I can't see a reason apart from low light capability but how many times do you need to film in extreme darkness if you want a chap lighting a match unless you want overblown flares you will have to still light the dark.

That's the EX1 alone, once you put something into the optical path, unless it's of the highest quality, without particles to scatter the light, there's going to an impact on quality of the image. In certain circumstances you can be putting this in front of your built in zoom lens when the aperture is wide open, so you have things like drop off in the edge resolution to factor in, plus an overall drop in the image quality from the lens.

As I mentioned earlier, you should test the camera and compare the results with your EX1. 8 bit can produce great results and many high end productions have been shot and continue to be shot using 8 bit, "Monsters" was shot 8 bit using the internal codec.

It's understandable that you don't want to buy another camera, but the EX1 Letus combo does have it's own limitations. You might like the images created, but that's another matter to the resolution figures.

[Alister Chapman]

Sorry Steve but I just don't buy it that Sony have suddenly found a way to read 14 million pixels worth of data 60 times a second and then group and process that data in an 8 watt camcorder when every other camera reading more than 4 million pixels suffers from heat and power issues requiring fans, heatsinks and a lot more power. The Canon DSLR's overheat even with pixel skipping. Red Epic has some of the noisiest fans I've ever heard (although they don't run when shooting, unless you have long takes). Arri's Alexa (which has pretty much the same pixel count as the F3/FS100) contains a massive heat sink. The F65's CMOS 20.4 MP sensor has a massive fan cooled heatsink to keep it cool, why doesn't the FS100 or F3 need fan cooling?

Considerably more data with considerably less power and heat are two things that just don't normally go together.

[Mark David Williams]

Obviously you will lose some resolution from the Letus combo but the point is how much? I've posted frame grabs that although unscientific show the differences between the Letus/EX1 and just the Letus. Unscientific because I downloaded a chart printed it off on bargain bucket paper and shot it Now I posted it with a warnig that it was only a comparison between the two However Alister measured its resolution using the MTF 50 test which somehow seems to be lower for all cameras than other tests seem to show.

Alister you want to mix resoltion up with contrast as part and parcel so I am sure that resolution and contrast is diminished by using the EX1/ Letus combo which the EX1 has a staritng resolution of 1000 lines or whatever test you use. So do we lose as much resolution and obviously contrast if your going to say resolution is part and parcel as the 730 lines you resolved on the FS100 well your test proves maybe a little even though it was a test on my printed card downloaded from the net.

Is your explanation of high frequency light being killed meaning all of it some of it as this is not a percentage figure more just a suggestive use of words that all high frequency light is killed. Have you done tests can you link to something that explains this more scientifically?

QUOTE
Are you really trying to say that the images don't look sharp just because your using an ultra shallow DoF? So what is supposed to be in focus.
---------------------------------
Your criticism of the lads facial hair not being in focus wheras his nose was In other words you have to look for the areas on a picture that is in focus. Therefore I supplied you a picture where the facial hair was in focus.
--------------------------------------
QUOTE
Ultra shallow DoF can make one part of the image perhaps appear sharper than it really is because of the dramatic difference between in and out of focus. This is one of the key areas people don't really consider. A higher resolution image needs less shallow DoF to achieve a similar effect as the ratio between in and out of focus increases with any true resolution increase. The circle of confusion decreases as resolution increases. This also applies to screen size and screen resolution, the larger the screen the greater the effect. The Circle of Confusion is governed by resolution and screen size as much as aperture and focal length.
---------------------------------------
Alister can we please stick to the subject these educational diversions kind of deflect from the topic?
--------------------------------------------------
QUOTE
Again going back to the Sony F65 demo footage, the scenes had shallow DoF, yet you could still see detail in the background due to the foreground starting off at a much higher true resolution.
----------------------------------------------------
Which F65 demo?
-------------------------------------------------
QUOTE
You cannot ignore contrast when you talk about resolution.
----------------------------------------------------
Of course you can. Maybe the MTF50 test uses contrast and maybe why it gives different results to other tests. It is unfair to add other variables as an answer to a specific line of investigation In this case resolution.
---------------------------------------------------
QUOTE
I wish people would understand that the two go hand in hand. It is ONLY through contrast that you see differences in resolution as it is only though contrast that you can see edges etc. If you don't have the contrast, you can't see the resolution.
-----------------------------------------------------
This doesn't seem to make sense of course you need contrast and the best glass on your lens but contrast is used as a term to how much something is seperated from something else whereas resolution is used as a term to describe how much is resolved. Of course the two go together but you can measure one without the other on a black and white chart. Are you saying the resolution tests are wrong?
------------------------------------------------------------
QUOTE
Sure it's your choice if you want to crush your blacks etc. What I am saying that if you've got good high frequency contrast then you can crush your blacks but still retain subtle details that prevent blacks from just becoming solid blocks of nothing.
------------------------------------------------------------------
If we're talking about blacks, the big sensors win because they see in the dark better.
----------------------------------------------------------------
QUOTE
I prefer the straight from the camera look, but you've still lost most of the texture from the darker areas of the girls hair, eyebrows, eylashes etc and you can't see any detail in her iris's. It all looks slightly soft. These are the low contrast areas that a high resolution image should reveal.
----------------------------------------------------------------------
This is subjective and the contrast comes back when you colour correct.
---------------------------------------------------------------
QUOTE
It does still look like you have some detail correction or sharpening going on. I wonder if turning detail on but setting the level to -20 would look any different? Or setting frequency to +99 to get the thinest possible edges.
-----------------------------------------------------
Detail is turned OFF.

[Mark David Williams]

Brian
It's understandable that you don't want to buy another camera, but the EX1 Letus combo does have it's own limitations. You might like the images created, but that's another matter to the resolution figures.

I got very enthusiastic about these new cameras and wanted in until I looked a little closer and realised they don't match up the the EX1 letus combo.

"Monsters" was shot 8 bit using the internal codec.

Is this justification for buying an 8bit camera over a 10 bit? 10 bit really only comes into its own for colour correction. Monsters was also transferred to film which would have helped make it look more filmic. It also had a very good post house work on it.

[Brian Drysdale]

It depends where you do your colour correcting, if you do it mostly in camera, as DITs often do on the HDCAM range, you can get a wide range of looks in camera. The advantage being you can see how the costumes and art direction respond and you can make creative decisions there and then. I used an Amelie look on a short film and the director was delighted because the camera was responding to colours in unexpected ways and we could push some visual aspects further than we would've otherwise have dared.

I'm not sure how you can compare cameras looking at heavily compressed on line videos