Justify an F35 vs an FS7
- Thread starter steve phillipps
- Start date
mlysbakken
Well-known member
The color accuracy of the FS7 seems to be all over the place. That's the main thing, in my opinion. And it's only logical: a bayer sensor vs advanced pixel binning.
Liam Hall
Veteran
And you know this based on what exactly?
steve phillipps
Veteran
I must say that in this instance I tend to side with Ebrahim! If you stray away from technical considerations you're into the realms of religion and magic!!!
Steve
JMtheDP
Veteran
I don't think the bayer pattern is responsible for dodgy colour, a cheaper CFA or OLPF is more likely to be the culprit. I also don't think the pixel binning in the F35 is actually very advanced, I think its simplicity is one of the reasons it works so well.
Let's remember Sony have only made one camera with a higher sensor resolution than the F35, which was the F65. The F5 and F55 are good, but I don't think either is really suitable for top-end feature production. However, I do seem to be in a minority as the F55 in particular seems to be doing very good business in Hollywood, where I fail to understand why.
looks like an awesome piece of kit.
Matt Gottshalk
Resident Preditor
Viddovation
Veteran
Considering the same sensor is already in cameras on the market... more fair than it otherwise sounds.
mlysbakken
Well-known member
Didn't mean to seem harsh, but from the footage I've seen, the skin tones don't look very natural. It's got that sony look, IMO, in a bad way - smooth and milky colors, very digital.
FrameFarmMedia
Veteran
I may be in the minority but I think the F5 puts out a damn fine image. I have a huge sweet spot for the F3 image also. I may be looking for different qualities of the picture. I haven't used a sony camera since EX1. The FS7 is the large sensor EX1 I've wanted. If I was doing narrative or high end commercial I'd hire an F35 in a heart beat. For my needs the FS7 just covers more variety of shooting work.
I think a lot of the variance in Sony F footage looks is due to slog and exposing properly along with personal grade choices.
Last edited:
Grug
Veteran
I'm not sure I agree. Yes LOG exposure seems to throw a lot of people (particularly those who insist on "Exposing to the Right" without understanding how log gamma curves actually work), but having now shot and graded a lot of material, both raw and compressed, from a pretty wide range of digital cinema cameras (and Sony Cinealtas in particular), I now believe the primary difference between yielding either cinematic or video-y images has far less to do with exposure, and far more to do with the initial colour transform performed on your LOG footage in the grade. The Filmconverts, and Film-Emulation LUTs out there make it a cinch to yield attractive images from most of these modern cameras, people just need to learn how to use them.
Colour and mid-tone contrast - that's where the real magic lies.
FrameFarmMedia
Veteran
That's a good point.
I'm going to be using Slog3 for the first time when my FS7 arrives so I have been trying to read everything I can.
starcentral
Veteran
Exposure is definitely key as is color transform in post and also possibly use of some type of light strength halation filter(s)... none of which you really need to do or go through the efforts with the F35 to get a "cinematic" film like image.
That there I believe is the crux of the issue and hence difference.
Many of us since the F3 days have been spending a lot of time learning how to best work with slog and get more "filmic" results. Now we find ourselves getting into very advanced color correction and grading even trying to incorporate ACES and IDTs in our workflow... and now even the possible use of film emulation LUT's for a better starting point, etc..
So even though technology get's cheaper, etc.. there's a good reason the F35 cameras was +$200k originally (and the image proves it) and from what I've concluded over 4 years of working with the CineAlta's is man they do require a ton of work to look anything close to the F35 and although it's taken several years (for me at least) to get closer and I still feel like I am very far away.
When I first splurged for the F55 hoping that it would have some of these F35 mojo characteristics I was quite wrong. Out of the box it is going to be like any other Sony camera non-F35 and require a ****-ton of work all around to look good. If you now expect that the FS7 is going to be close to the F35, you are wrong again, but its not to say that without effort you can't get close - which I think is Mark's point.
That there I believe is the crux of the issue and hence difference.
Many of us since the F3 days have been spending a lot of time learning how to best work with slog and get more "filmic" results. Now we find ourselves getting into very advanced color correction and grading even trying to incorporate ACES and IDTs in our workflow... and now even the possible use of film emulation LUT's for a better starting point, etc..
So even though technology get's cheaper, etc.. there's a good reason the F35 cameras was +$200k originally (and the image proves it) and from what I've concluded over 4 years of working with the CineAlta's is man they do require a ton of work to look anything close to the F35 and although it's taken several years (for me at least) to get closer and I still feel like I am very far away.
When I first splurged for the F55 hoping that it would have some of these F35 mojo characteristics I was quite wrong. Out of the box it is going to be like any other Sony camera non-F35 and require a ****-ton of work all around to look good. If you now expect that the FS7 is going to be close to the F35, you are wrong again, but its not to say that without effort you can't get close - which I think is Mark's point.
Film Scientist
Active member
Here’s my take on the comparative list between an F35 and an FS7.
I agree there should be objective means to measure image quality, otherwise we can quickly fall into the audiophile trap of
attributing emotional feelings to ephemeral qualities. But it’s helpful to realize that CMOS and CCD chips are still inherently analog devices.
Chip technologies have their pros and cons, some are more sensitive others more subtle.
Please take this apple-to-pear-comparison with a grain of salt.
1-Resolution
The F35 yields over sampled 1080p signal, from a striped CCD sensor.
12441600 active pixels (x6 times oversampling of HD resolution~2,073,600 pixels)
5760 Horizontal resolution: 1920 red sub pixels , 1920 green sub pixels, 1920 blue sub pixels.
2160 Vertical resolution: 2160 red sub pixels, 2160 green sub pixels, 2160 blue sub pixels.
Each pixel generated, is summed from 6 sub pixels of full RGB data at 14 bits.
(3 times horizontal resolution oversampling, 2 times vertical resolution oversampling to yield 1920 x 1080p)
FS7 is a Bayer CMOS sensor, that outputs 4K raw and 1080p
11.6 pixels quoted, but seems more like 8.8k are active.
The press release doesn’t specify number of actual photo sites (maybe a purposeful marketing omission)
~4096 horizontal resolution: 2048 Green sub pixels, 2048 Blue sub pixels (assumes Bayer Pattern)
~2160 Vertical resolution: 2160 Green sub pixels, 2160 Red sub pixels (assumes Bayer Pattern)
1 pixel of RGB data is generated from 2 Green sub pixels, 1 blue sub pixel and, 1 red sub pixel.
Utilizes traditional 4:2:2 sampling sampling scheme to capture interpolated color for ‘4K’ or 1920x1080p
Each chip is designed with a purpose built low pass filter that when combined with a lens determines the ultimate real world resolution.
In terms of luminance resolution both cameras are comparable in resolution, but color resolution is often forgotten in the world of 4K, the F35 beats it hands down (It’s sampling more real color data)
2. Sharpness
Is a combination of many factors…
Density of CMOS or CCD chip, quality of debayer or processing + low pass filter + Lens + recording encoding method log 4:4:4 vs. raw 4:2:2
FS7 claims 11.7 million (not sure how many are active)
F35 has 12.4 million
I would expect the F35, to be slightly sharper if we had access to the raw data.
But with cameras outputs we downscaled to 1080p, sharpness will be pretty close to equal
3. Dynamic range
F35 - at native ISO 450, 12 stops (11.3 per Zacuto test), good highlight sensitivity (with seemingly more stops devoted to highlights like film)
FS7 - at native ISO 2000, (14 stops according to manufacturer), stops above and below middle grey still TBD
I wonder if the high ISO of the FS7 will affect highlight sensitivity above middle grey in favor of a lower noise floor.
4 - Color
F35 - True RGB 4:4:4 @ 1920x1080. 14 bit AD to 12bit conversion. Tungsten balanced chip (3200k)
FS7- 4:2:2 @ ‘4K’. RAW. Daylight balanced chip.
Most Bayer chips still rely on the brute force method of adding more sub pixels to interpolate color.
So as cameras increase in resolution, they will be able to extract more color from larger arrays of 4:2:2.
What the F35 misses in resolution and sharpness, it more than makes up for by capturing subtle color hues and natural grain
by utilizing a very efficient RGB stripe method processed in-camera.
5. Noise
Both cameras would need to be tested at similar ISO to properly evaluate.
F35 - Will naturally be noisier due to base ISO 450, but sampled information/noise represents real, not interpolated RGB data (so noise looks better.)
FS7 - The base ISO of 2000 equals more sensitivity, and less noise.
I tend to light everything to ISO 450 anyway, high ISOs tend to lose the ‘movie look’ for some reason.
Note, most movies shot on film the past 25 years never exceeded ISO 500 film stock (I like lighting to that look)
6. Noise Pattern
F35 - Will exhibit a ‘better’ more random noise (based on full RGB 4:4:4 data for each pixel generated)
FS7 - Noise will have more of a repetitive/patterned feel (based on 4:2:2 data for each pixel generated)
The lack of patterned noise in the F35 is probably the filmic quality that folks talk about.
7. Artifacts
F35 - CCD smear (vertical streaking), minimal color aliasing on any color channels, minimal moire.
FS7 - Rolling Shutter, color aliasing due to interpolated data, more moire due to 4:2:2 sampling method.
F35 does not exhibit motion artifacts, but prone to streak if sensor is oversaturated with light. FS7 will still have some motion artifacts and potentially more moire.
8 - Post Image Flexibility
F35 - Good highlight latitude, similar to Alexa and Film, and great color overall. It gets skin tone right, because the chip is natively balanced for tungsten, this is still a rarity in the chip world, and the reason why the camera was probably so expensive to produce.
FS7 - CMOS chips are typically natively 5000k, which is why things may feel cold or less organic from the outset without extensive grading.
I assume this camera will have good low light sensitivity, reasonable sharpness, but weaker at capturing subtle colors and hues.
9 - Sensor Size
F35 - 23.62 x 13.28mm (diagonal 27.1mm)
Fs7 - S35mm CMOS - size ?
10 - High Speed
F35 - maxes out at 50 fps, built for high end traditional narrative workflow - well engineered 2007 Panavision/Sony tech, spared no expense in terms of technology implemented (250k original price)
FS7 - maxes out at 180fps/(240 fps w/add on), built for lower end Videographer market - 2014 technology to hit a specific marketing price point.
R&D/chip technology is now optimized and refined to hit specific price point/marketing segment.
This comparison is like a hammer and a mallet analogy.
Both device’s are good for hitting things. Each tool is better tailored for a specific use. Choose wisely.
I agree there should be objective means to measure image quality, otherwise we can quickly fall into the audiophile trap of
attributing emotional feelings to ephemeral qualities. But it’s helpful to realize that CMOS and CCD chips are still inherently analog devices.
Chip technologies have their pros and cons, some are more sensitive others more subtle.
Please take this apple-to-pear-comparison with a grain of salt.
1-Resolution
The F35 yields over sampled 1080p signal, from a striped CCD sensor.
12441600 active pixels (x6 times oversampling of HD resolution~2,073,600 pixels)
5760 Horizontal resolution: 1920 red sub pixels , 1920 green sub pixels, 1920 blue sub pixels.
2160 Vertical resolution: 2160 red sub pixels, 2160 green sub pixels, 2160 blue sub pixels.
Each pixel generated, is summed from 6 sub pixels of full RGB data at 14 bits.
(3 times horizontal resolution oversampling, 2 times vertical resolution oversampling to yield 1920 x 1080p)
FS7 is a Bayer CMOS sensor, that outputs 4K raw and 1080p
11.6 pixels quoted, but seems more like 8.8k are active.
The press release doesn’t specify number of actual photo sites (maybe a purposeful marketing omission)
~4096 horizontal resolution: 2048 Green sub pixels, 2048 Blue sub pixels (assumes Bayer Pattern)
~2160 Vertical resolution: 2160 Green sub pixels, 2160 Red sub pixels (assumes Bayer Pattern)
1 pixel of RGB data is generated from 2 Green sub pixels, 1 blue sub pixel and, 1 red sub pixel.
Utilizes traditional 4:2:2 sampling sampling scheme to capture interpolated color for ‘4K’ or 1920x1080p
Each chip is designed with a purpose built low pass filter that when combined with a lens determines the ultimate real world resolution.
In terms of luminance resolution both cameras are comparable in resolution, but color resolution is often forgotten in the world of 4K, the F35 beats it hands down (It’s sampling more real color data)
2. Sharpness
Is a combination of many factors…
Density of CMOS or CCD chip, quality of debayer or processing + low pass filter + Lens + recording encoding method log 4:4:4 vs. raw 4:2:2
FS7 claims 11.7 million (not sure how many are active)
F35 has 12.4 million
I would expect the F35, to be slightly sharper if we had access to the raw data.
But with cameras outputs we downscaled to 1080p, sharpness will be pretty close to equal
3. Dynamic range
F35 - at native ISO 450, 12 stops (11.3 per Zacuto test), good highlight sensitivity (with seemingly more stops devoted to highlights like film)
FS7 - at native ISO 2000, (14 stops according to manufacturer), stops above and below middle grey still TBD
I wonder if the high ISO of the FS7 will affect highlight sensitivity above middle grey in favor of a lower noise floor.
4 - Color
F35 - True RGB 4:4:4 @ 1920x1080. 14 bit AD to 12bit conversion. Tungsten balanced chip (3200k)
FS7- 4:2:2 @ ‘4K’. RAW. Daylight balanced chip.
Most Bayer chips still rely on the brute force method of adding more sub pixels to interpolate color.
So as cameras increase in resolution, they will be able to extract more color from larger arrays of 4:2:2.
What the F35 misses in resolution and sharpness, it more than makes up for by capturing subtle color hues and natural grain
by utilizing a very efficient RGB stripe method processed in-camera.
5. Noise
Both cameras would need to be tested at similar ISO to properly evaluate.
F35 - Will naturally be noisier due to base ISO 450, but sampled information/noise represents real, not interpolated RGB data (so noise looks better.)
FS7 - The base ISO of 2000 equals more sensitivity, and less noise.
I tend to light everything to ISO 450 anyway, high ISOs tend to lose the ‘movie look’ for some reason.
Note, most movies shot on film the past 25 years never exceeded ISO 500 film stock (I like lighting to that look)
6. Noise Pattern
F35 - Will exhibit a ‘better’ more random noise (based on full RGB 4:4:4 data for each pixel generated)
FS7 - Noise will have more of a repetitive/patterned feel (based on 4:2:2 data for each pixel generated)
The lack of patterned noise in the F35 is probably the filmic quality that folks talk about.
7. Artifacts
F35 - CCD smear (vertical streaking), minimal color aliasing on any color channels, minimal moire.
FS7 - Rolling Shutter, color aliasing due to interpolated data, more moire due to 4:2:2 sampling method.
F35 does not exhibit motion artifacts, but prone to streak if sensor is oversaturated with light. FS7 will still have some motion artifacts and potentially more moire.
8 - Post Image Flexibility
F35 - Good highlight latitude, similar to Alexa and Film, and great color overall. It gets skin tone right, because the chip is natively balanced for tungsten, this is still a rarity in the chip world, and the reason why the camera was probably so expensive to produce.
FS7 - CMOS chips are typically natively 5000k, which is why things may feel cold or less organic from the outset without extensive grading.
I assume this camera will have good low light sensitivity, reasonable sharpness, but weaker at capturing subtle colors and hues.
9 - Sensor Size
F35 - 23.62 x 13.28mm (diagonal 27.1mm)
Fs7 - S35mm CMOS - size ?
10 - High Speed
F35 - maxes out at 50 fps, built for high end traditional narrative workflow - well engineered 2007 Panavision/Sony tech, spared no expense in terms of technology implemented (250k original price)
FS7 - maxes out at 180fps/(240 fps w/add on), built for lower end Videographer market - 2014 technology to hit a specific marketing price point.
R&D/chip technology is now optimized and refined to hit specific price point/marketing segment.
This comparison is like a hammer and a mallet analogy.
Both device’s are good for hitting things. Each tool is better tailored for a specific use. Choose wisely.
Last edited:
Viddovation
Veteran
Very nice Film Scientist, but please correct your high-speed findings: the F35 tops out at 50 fps. It has a 60fps video output, but the max frames from the sensor is 50 fps.
Ebrahim Saadawi
Carbonite Member
I think this is the answer we needed Filmscientist!
Ebrahim Saadawi
Carbonite Member
By the way what does a sensor balanced for tungsten mean? Is the sensor designed with a specific white balance? Can't you change it? Or is this something far from white balance at all?
Film Scientist
Active member
Yes, per Art Adams' article http://provideocoalition.com/aadams/story/my_first_shoot_with_the_sony_f35/ , the sensor is designed for the primary native white balance of the camera of 3200k (Which is why skin tones look so good). The other setting for white balance setting is 5600k (basically an in-camera digital transform). Sony designed the camera like traditional film stock, with one indoor and one outdoor setting. From what I can tell the camera really does seem to capture color gamut better than film stock if you treat the log transform (mid tones) with care.
It terms of car analogies, the F35 is somewhat like a the original F1 Maclaren, never to be repeated because of cost, one of the first true super cars of its kind, but through economies of sale we now have Nissan GTRs (or insert favorite 'super car' here) which are almost every bit as fast at a quarter or tenth of the cost. Only a small minority is usually crazy enough to appreciate/extract that last 5-10% of that performance envelope.
Last edited:
Viddovation
Veteran
To further clarify - just like film stock has an inherent color balance, so do digital image sensors. When you white balance a camera you are telling it what temperature to treat as neutral, or white. However, this setting doesn't change how the sensor captures light, merely how the camera interprets and recolors the image. All image sensors have a native color temperature. If the camera white balance setting doesn't match the native setting, then the camera has to push the image to the user desired color temperature. CCD imagers typically are 3200k or tungsten. CMOS imagers tend to have a native setting around 5000k, and therefore tend to perform best under daylight conditions.