Sony FX9 camera test
|
Dear colleagues, we have published our review of the Sony FX9 camera on the IMAGO website.
We hope it is of interest to you. https://www.imago.org/index.php/news/item/1127-sony-pxw-fx9-camera-test.html Regards Alfonso Parra ADFC www.alfonsoparra.com |
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
Jessica Gallant
Thank you for posting this, I found it informative and helpful.
toggle quoted message
Show quoted text
Jessica Gallant West Coast Administrator, CML Director of Photography | Los Angeles | CA http://jessicajgallant.com http://wb.imdb.com/name/nm0002680/ cell: 818-645-2787 email: jessicajgallant@... On Apr 16, 2021, at 7:49 AM, alfonso parra <info@...> wrote: |
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
Very good, Alfonso. Thank you for taking time to test and share. As usual, very informative.
Kind Regards,
Pawel Achtel ACS B.Eng.(Hons) M.Sc. “Sharp to the Edge”
ACHTEL PTY LIMITED, ABN 52 134 895 417 Website: www.achtel.com Mobile: 040 747 2747 (overseas: +61 4 0747 2747) Mail: PO BOX 557, Rockdale, NSW 2216, Australia Address: RA 913 Coles Bay Rd., Coles Bay, TAS 7215, Australia Location: S 42° 0'14.40"S, E 148°14'47.13" Email: Pawel.Achtel@... Facebook: facebook.com/PawelAchtel Twitter: twitter.com/PawelAchtel Skype: Pawel.Achtel _,_ |
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
alister@...
I’m curios Alfonso as to why the S-Cinetone tests were all done with +7dB of gain added? My own testing indicates that adding gain to the S-Cinetone curve reduces the highlight range, results in more abrupt highlight clipping and increases noise compared to using it at it 0dB/320 ISO.
toggle quoted message
Show quoted text
Alister Chapman Cinematographer - DIT - Consultant UK Mobile/Whatsapp +44 7711 152226 Facebook: Alister Chapman Twitter: @stormguy
|
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
We decided to use the dual ISO base settings
for all the curves and thus be able to compare them to the same ISO value, in
this case with the cinetone curve we use 800 as for the Slog3 curve so that we
can clearly see the differences. If it is interesting to observe how you have
made, the curve with different ISO values and compare the results. What
surprises me is that by raising the value of the S-cinetone to 800, detail is
lost in the highs when normally the opposite happens, by increasing the gain,
or the ISO value, the ability to collect detail in the highs increases, losing in the shadows.
I don't remember now, but does Sony make any recommendation on the ISO value to
use with the S-cinetone or does it indicate how much the middle gray value
should be? May I ask you how did you get to 320 ISO in relation to 0db? In the
theoretical sensibility test that we did, it gave us that the 800 ISO would be
0db considering the STD5 curve with gamma 2.4, which is actually quite similar
to S-cinetone, actually the most similar is the STD3 only that S-Cinetone
compresses something else in the highlights.
Regards Alfonso Parra ADFC
www.alfonsoparra.com Tel Colombia 57 311 5798776 Tel Spain 34 639109309 Instagram alfonso_parra_adfc |
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
alister@...
The base ISO’s for S-Cinetone on the FX9 are 320 and 1600. This is of course also 0dB. 0dB is virtually always the native ISO, so to discover the cameras native ISO if it has a dB mode, switch to dB and compare to the ISO value. This information is also contained in the clip metadata and the recommendation to use the base ISO’s of 320/1600ISO for the cameras 460% dynamic range curves (including S-Cinetone and HG1 to 4) is included in the manual.
toggle quoted message
Show quoted text
Raising the gain when you are using a power law gamma reduces the dynamic range because gain is a multiplier. If you multiply your 0 to 100% input range from the sensor by 2 it becomes 0 to 200%, but you can’t record 200%, you only have room for 100%. So you have a small change in the shadows but a much larger change in the brighter parts of the image and the dynamic range that can be recorded is reduced by 1 stop for every 6dB you add. Adding gain means you will clip earlier. If you stop down to compensate, bringing the highlights down to where they would be without the added gain you reduce the shadow range as the SNR will become worse, and the extra noise from the amplification will limit the shadow range. You gain no additional highlight range, you are simply returning it to where it would be without the extra gain, but the shadows suffer. With added gain the highlights are less pleasing because the gamma curve is designed to work with the sensors output range, when you add gain because you can now more easily exceed the recording range the highlights clip sooner and don’t look as nice as a result. Alister Chapman Cinematographer - DIT - Consultant UK Mobile/Whatsapp +44 7711 152226 Facebook: Alister Chapman Twitter: @stormguy
|
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
Thank you very much Alister for the clarification, I will try to compare the curves with that ISO value and see what it turns out.
toggle quoted message
Show quoted text
Alfonso Parra ADFC
www.alfonsoparra.com Tel Colombia 57 311 5798776 Tel Spain 34 639109309 Instagram alfonso_parra_adfc
|
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
alister@...
Hi again Alfonso.
toggle quoted message
Show quoted text
Sony’s logic when designing their cameras and determining base ISO appears to be based on finding the sensors clip point and working down from there. So where you have a 460% curve such as S-Cinetone you design the curve so that 100% sensor output = a recording level of 109IRE and then go down 11 stops from there to black. This will give a good SNR while making use of the sensors full highlight range, the shadow range is limited by the gamma curve, but the shadows remain clean because you are putting a good amount of light onto the sensor. Then for S-Log3 you put 100% sensor output at 94IRE (the clip point of the S-Log3 curve) and go down 15(ish) stops from there. This results in a reduced SNR, the impression of a greater highlight range (because you putting less light on the sensor) while the gamma curve extends deeper into the shadows and noise, so overall greater DR but at the expense of more noise than S-Cinetone. Because S-Log3 has 6 stops from clip to middle grey and then S-Cinetone has 4.5(ish) stops from clip to middle grey, you need to expose S-Cinetone 1.5 stops darker than S-Log3 to put middle grey in the right place and achieve the full dynamic range without clipping the highlights. So S-Cinetone ends up with a base ISO rating that is almost 1.5 stops lower than S-Log3. The base ISO for all the STD gammas as well as HG1-4 is 320/1600. HG7 & HG8 are 500/2500ISO (they have a greater highlight range than S-Cinetone so as above they are rated higher so you expose the sensor lower to gain the extra highlight range). Sony don’t give a middle grey value for S-Cinetone because the idea is that you adjust your exposure to make use of the roll-off that starts around 65IRE to alter the contrast in your upper mid range and brighter skin tones to provide more or less contrast depending on the look you want. There is also a very small change in the gain in the toe of S-Cinetone that has a similar effect to shadow contrast - brighter = less, darker = more. My own testing suggests that to match the 320ISO base rating that you would expose middle grey at 44IRE and a 90% reflectivity white card at 78IRE. There is a white paper on S-Cinetone: https://pro.sony/s3/2020/03/24095333/S-Cinetone-whitepaper_v2.pdf Alister Chapman Cinematographer - DIT - Consultant UK Mobile/Whatsapp +44 7711 152226 Facebook: Alister Chapman Twitter: @stormguy
|
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
alister@...
Alfonso. I look forward to reading your findings.
toggle quoted message
Show quoted text
Alister Chapman Cinematographer - DIT - Consultant UK Mobile/Whatsapp +44 7711 152226 Facebook: Alister Chapman Twitter: @stormguy
|
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
Cynthia Brett Webster
I too am very interested in this thread. I've been shooting with FS7 Cameras for a few years and just finished shooting a feature using FX9 Cameras for the first time. The budget didn't allow for Sony Venice Cameras, so I suggested using a pair of FX9s and the producers we're extremely happy with the results. I used S-Cinetone straight out of the box with no LUTs or adjustments or scopes. I simply relied on the camera built-in monitors for color and exposure and used larger monitors connected wirelessly only for framing and Camera Assistants pulling focus. Very pleased with the results. (Also, the low light capability of the FX9 cameras is phenomenal). I'm really surprised here by the results of the tests, as under normal set lighting condituons, I wouldn't have expected the differences between the FS7s and FX9s to be so small, (that is, if I understand the data correctly?) Cynthia Brett Webster DP, Los Angeles . On Sat, Apr 17, 2021, 10:47 AM <alister@... wrote:
|
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
alister@...
Cynthia.
toggle quoted message
Show quoted text
I think we need to be realistic about differences and improvements between cameras. Most of the current cameras are capable of producing great images. Sensor technology is not significantly changing, so any differences or improvements will only ever be small. You are not suddenly going to see a camera with dramatically greater image quality unless you do something radically different. But these small differences do count. Less noise is always nice and make footage easier to work with in post production. Improved color response can make faces and skin tones subtly more pleasing, especially for less experienced colourists. Alister Chapman Cinematographer - DIT - Consultant UK Mobile/Whatsapp +44 7711 152226 Facebook: Alister Chapman Twitter: @stormguy
|
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
Cynthia Brett Webster
toggle quoted message
Show quoted text
On Sun, Apr 18, 2021, 2:30 AM <alister@... wrote:
|
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
To me the two most noticeable improvements in the FX9 are noise and color. It improves the sensitivity with less noise in the shadows and also changes the colorimetry. The changes in the RD do not seem so significant to me and the gamma curves do not show a great difference compared to previous cameras.
Alfonso Parra ADFC
www.alfonsoparra.com Tel Colombia 57 311 5798776 Tel Spain 34 639109309 Instagram alfonso_parra_adfc |
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
>>>You are not suddenly going to see a camera with dramatically greater image quality unless you do something radically different.
I’m going to disagree with that.
Cameras still do not perform equally well under all conditions, or all look the same. And, generally, there are no financially-expedient shortcuts around certain aspects of physics.
|
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
alister@...
Well Art, we will have to agree to disagree.
toggle quoted message
Show quoted text
The differences that exist between the very best and the least good of the large sensor cameras are small. Small enough that give them to a competent operator, shoot some random typical scenes with the care you would normally take on a feature, grade all with equal skill and a blind test a general audience would have one hell of a hard time categorically identifying any individual camera. They are certainly unlikely to be saying “wow, look how dramatically better that camera is”. Different cameras have always behaved differently in extremes, perhaps for very low light you might get a better result from a $4K FX3 that can produce beautiful, low noise, high dynamic range images at ridiculously low light levels while other considerably more expensive cameras would really struggle to deliver anything useable. Then in extreme highlights perhaps a Venice/Arri LF will do much better. But really, for normal day to day shooting the differences are not vast as they are all using similar sensor technology and the majority allow us to record raw so we can move the image processing into post production. Most of what’s produced these days is graded heavily and often little remains of the cameras original colour palette. There’s a whole industry that revolves around the production of LUT’s and transforms that can very convincingly make almost any camera look like any other. Yes, in side by side, direct comparison tests or through careful evaluation with charts you will find differences, because there are differences. But as Alonso’s test shows the difference between Venice and the FX9 is not huge and nor is the difference between the FX9 and FS7. Often these differences only become clear when you dig very deeply into the image. As I said, these small differences can matter and they are real, so when you have the budget why would you choose to use anything but what you feel is the very best. That won’t ever change. But there is not a vast difference between what you can get from a $60K camera and a $6K camera. Alister Chapman Cinematographer - DIT - Consultant UK Mobile/Whatsapp +44 7711 152226 Facebook: Alister Chapman Twitter: @stormguy
|
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
The differences that exist between the very best and the least good of the large sensor cameras are small. Small enough that give them to a competent operator, shoot some random typical scenes with the care you would normally take on a feature, grade all with equal skill and a blind test a general audience would have one hell of a hard time categorically identifying any individual camera. They are certainly unlikely to be saying “wow, look how dramatically better that camera is”.
I think that depends. Is this the kind of “typical” scene you’d shoot on a feature when on a stage under controlled lighting, or shooting against the setting sun at the end of the day with the sun in the shot? (I just saw a stunning shot exactly like this at the ASC Awards that ppears in Nomadland—a film I have yet to see but looks to be beautifully shot). And other kinds of productions have different constraints. Is a “typical” feature scene the same as a “typical” episodic television scene, commercial scene, corporate scene or documentary scene? And isn’t it equally important to capture the atypical scenes? Often you don’t get to plan when those happen. They just happen.
There are a lot of cameras that will choke when shooting into bright clipped highlights. They most definitely do not respond the same. A general audience -might- notice consciously, but what really matters is that a cinematographer will notice. That’s who we have to keep happy. The general audience doesn’t choose the cameras or lenses or lighting. They are affected by those choices, but they don’t know why. They don’t notice consciously but they feel the impact of those choices.
Different cameras have always behaved differently in extremes, perhaps for very low light you might get a better result from a $4K FX3 that can produce beautiful, low noise, high dynamic range images at ridiculously low light levels while other considerably more expensive cameras would really struggle to deliver anything useable.
Once again, it depends. There are many cameras out there that have remarkable low light performance… if you don’t need to see much detail in the shadows because there’s a lot of noise reduction going on, and that kills shadow detail. There are some cameras whose dynamic range distribution changes a bit at those higher ISOs such that the mid-tones work nicely but highlights clip sooner, which can be problematic at night around streetlights and office windows. There are some cameras whose noise characteristics will produce a hue shift if you decide to pull the image up a bit later, because the noise tends to have color to it if you dig in too deeply so you really need to get the exposure right.
Then in extreme highlights perhaps a Venice/Arri LF will do much better. But really, for normal day to day shooting the differences are not vast as they are all using similar sensor technology and the majority allow us to record raw so we can move the image processing into post production.
Raw doesn’t save a clipped image. It may help in making it look less bad, but raw isn’t a magic solution for a lot of problems. And quite a few projects do not shoot raw for budgetary reasons: docs, commercials, even TV series and low budget features.
Most of what’s produced these days is graded heavily and often little remains of the cameras original colour palette. There’s a whole industry that revolves around the production of LUT’s and transforms that can very convincingly make almost any camera look like any other.
It's important to talk to colorists about this as they will tell you they absolutely have more control in pushing colors around with some cameras than others. That’s not just down to the sensor hardware alone. I’ve known colorists who will use ARRI’s LogC to grade material from other cameras because there’s something about the underlying math that makes it easier to manipulate color information.
They will also tell you that it is possible to make one camera look like another as long as you make the camera that looks better match the camera that looks worse. It doesn’t work the other way, and all cameras do not look the same. That is a myth that is easily dispelled through testing or speaking to a colorist.
And, lastly, not every project is graded by a master colorist.
If all cameras were equal we’d be watching Marvel movies shot on $5,000 cameras, and we aren’t.
Yes, in side by side, direct comparison tests or through careful evaluation with charts you will find differences, because there are differences. But as Alonso’s test shows the difference between Venice and the FX9 is not huge and nor is the difference between the FX9 and FS7.
Maybe it’s just me, but I saw a lot of differences.
Often these differences only become clear when you dig very deeply into the image. As I said, these small differences can matter and they are real, so when you have the budget why would you choose to use anything but what you feel is the very best. That won’t ever change. But there is not a vast difference between what you can get from a $60K camera and a $6K camera.
Once again, maybe it’s just me, but I see some significant differences. More expensive cameras do have obvious advantages if you know what to look for, or if you need to drag a shot back from the brink because, for whatever reason, the situation became “atypical.”
|
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
Noel Sterrett
On 4/19/21 3:49 PM, Art Adams wrote:
If all cameras were equal we’d be watching Marvel movies shot on $5,000 cameras, and we aren’t.The camera choice is all too often driven by ego and politics, rather than camera capabilities. Noel Sterrett | Admit One Pictures | Atlanta |
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
Noel Sterrett
On 4/19/21 3:49 PM, Art Adams wrote:
And quite a few projects do not shoot raw for budgetary reasonsI can shoot uncompressed 12 bit RAW on a Sigma fp for $120/hour. My first low budget in-the-can $128k movie spent 10 times that on film and processing alone. Red claims a patent on compressed RAW, and so we are left spending inordinate amounts of time and energy debating various encoding schemes invented by camera manufacturers to avoid the patent claim rather than just recording uncompressed RAW. A "Raw Film Camera" has no features which degrade the data which the sensor produces. Like a film camera, it has some type of viewfinder, a start/stop button, and optinally a speed control. "As simple as it can be and no simpler" - Albert E. Noel Sterrett | Admit One Pictures | Atlanta |
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
The camera choice is all too often driven by ego and politics, rather than camera capabilities.
This can be true. I once had to shoot a marketing product roll-out where the client mandated we shoot in 8K for 1080p release… because. There was no reason for it other than it was cool, and 8K was 4.2x greater than 1.9K.
But this is not always the case, or even mostly the case. I don’t know of any situations where someone chose to shoot with a cheaper camera because they liked the quality more than the more expensive camera.
There have been a lot of cameras sold with the promise of being as good as more expensive cameras for lower cost. I have yet to find one that I personally found to be as good as the more expensive option in one way or another. It still takes money and knowhow to tame physics.
|
||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||
|
Art Adams wrote: “There have been a lot of cameras sold with the promise of being as good as more expensive cameras for lower cost. I have yet to find one that I personally found to be as good as the more expensive option in one way or another. It still takes money and knowhow to tame physics.”
I tend to be with Art on this one. More often than not you get what you pay for. And, it is when you push the camera “out of the comfort zone”, the limitations become much more apparent.
What doesn’t help make informed camera choices is wide-spread misinformation and claims made by some manufacturers. This is why independent tests, like the one performed by Alfonso, are so valuable. (and, to be clear, I’m not saying that Sony is making misleading claims, they are in good books most of the time).
Through my own testing I found that most digital cinema cameras have very similar dynamic range (within a stop) despite vastly different claimed performance on paper. But, even though the actual dynamic range may be similar, there are other, more critical and more clearly discernible differences: like low light performance, noise, colour accuracy and range, contrast (MTF) and artefacts (compression and aliasing). In fact, some really striking differences and relatively easy to test side-by-side or, like Alfonso did, by meticulously measuring those properties in a consistent and repeatable manner.
Kind Regards,
Pawel Achtel ACS B.Eng.(Hons) M.Sc. “Sharp to the Edge”
ACHTEL PTY LIMITED, ABN 52 134 895 417 Website: www.achtel.com Mobile: 040 747 2747 (overseas: +61 4 0747 2747) Mail: PO BOX 557, Rockdale, NSW 2216, Australia Address: RA 913 Coles Bay Rd., Coles Bay, TAS 7215, Australia Location: S 42° 0'14.40"S, E 148°14'47.13" Email: Pawel.Achtel@... Facebook: facebook.com/PawelAchtel Twitter: twitter.com/PawelAchtel Skype: Pawel.Achtel
From: cml-raw-log-hdr@... [mailto:cml-raw-log-hdr@...] On Behalf Of Art Adams
Sent: Tuesday, 20 April 2021 9:22 AM To: cml-raw-log-hdr@... Subject: [cml-raw-log-hdr] Sony FX9 camera test
The camera choice is all too often driven by ego and politics, rather than camera capabilities.
This can be true. I once had to shoot a marketing product roll-out where the client mandated we shoot in 8K for 1080p release… because. There was no reason for it other than it was cool, and 8K was 4.2x greater than 1.9K.
But this is not always the case, or even mostly the case. I don’t know of any situations where someone chose to shoot with a cheaper camera because they liked the quality more than the more expensive camera.
There have been a lot of cameras sold with the promise of being as good as more expensive cameras for lower cost. I have yet to find one that I personally found to be as good as the more expensive option in one way or another. It still takes money and knowhow to tame physics.
|
||||||||||||||||||||||||||||||||||||||||||||||
|
|
