Cary Knoop
Veteran
"8k? We don't need no stinkin' 8k!"
We need a better quality 4k sensor with 14 to 15 stops of dynamic range.
We need a better quality 4k sensor with 14 to 15 stops of dynamic range.
GH5 GH6 rumors
- Thread starter joe1946
- Start date
Thomas Smet
Veteran
Wasn't it Sony that said about a year ago that 8k on m43 was a bad thing? Funny how they not only changed their mind but clearly were working on such a sensor at the time they actually said that.
Thomas Smet
Veteran
And at the very least external raw support. I see little reason for the film community to move away from BMD at this point for a Panasonic camera if they don't. I love the look of Panasonic cameras but the market has shifted in much the same way it shifted away from highly compressed 8bit HD on Canon DSLRs with line skipping. If Panasonic doesn't play their cards right they will suffer. It is the indie film makers that sparked the DSLR revolution and without them camera sales will suffer.
It was Panasonic ... because they didn't want Sony to hurry too much with 8K until they moved a lot of 4K MFT's.
Of course, this sets up more dominoes to fall. If MFT has 8K and 41 MPX for stills, what does APS-C get? Eyeball a 1.5 difference. Which then will move full frame beyond 60 (Canon is supposed to be working on a 80 MPX, Nikon on a 70 MPX model).
Because 41 is a trifling number.
Cary Knoop
Veteran
Indeed, all BMD needs now is a pocket camera with IBIS and Panasonic can have their idea of a vast full-frame only pocket indie market realized.
joe1946
Veteran
Even though the GH5s had significantly larger image sensor pixels (4.68 µm) than the GH5 (3.33 µm), this didn't seem to translate into a significant difference in dynamic range, although the noise performance in low light conditions was improved over the GH5.
This new 8K sensor has pixels that are 2.3 µm in size which is very similar to the 2.4 µm pixels of Sony's 20-megapixel 1-inch sensors that are used in cameras like the RX10 and RX100. Which suggests that huge breakthroughs in imager sensitivity weren't really required for this new sensor. It might mostly be a matter of scaling up existing sensor technology to a slightly larger sensor area.
This new 8K sensor has pixels that are 2.3 µm in size which is very similar to the 2.4 µm pixels of Sony's 20-megapixel 1-inch sensors that are used in cameras like the RX10 and RX100. Which suggests that huge breakthroughs in imager sensitivity weren't really required for this new sensor. It might mostly be a matter of scaling up existing sensor technology to a slightly larger sensor area.
Realdvxuser
Well-known member
I shot with both and I found the roll-off in the highlights quite a bit nicer with the GH5s actually. It provides a really nice 'full' image.
BobKo
Well-known member
Just being pedantic for a moment, 8192 x 4320 is 32 megapixels, which is four times the resolution of 4096 x 2160 (8 megapixels), not double the resolution.
Switching pedantic mode off now...
Bob
You shot HLG footage ?
Because the V-LOG L kneecaps dynamic range to 12 stops, where HLG can record 1-1.5 more stops of dynamic range with the GH5s. ( this came up in the EBU review of the GH5s ) It's too bad that Panasonic hasn't upgraded the GH5s to shoot the same V-LOG as the S1 cameras. The GH5 can't produce more dynamic range than 12 stops of DR so it makes no difference if you shoot HLG or V-LOG L.
When measuring DR, the protocol is to start at the white clipping point and count stops going down, the final few stops are fairly deep into the noise making the final number somewhat subjective depending on which stop you find the noise too objectionable. DR is determined by the sensor, which has no highlight “rolloff.” The highlight rolloff characteristic is a function of the gamma curve, the knee in the case of 709 and no rolloff in the case of S-Log2/3. Highlight and shadow rolloff can be included in log raw such as BMD BRAW, as well as extracting a little more highlight recovery when only one of the RGB channels has clipping. Sony 16 bit Linear RAW has shadow rolloff at the toe in accordance with the application of S-Log2/3 transfer, selectable in post one or the other. Highlight rolloff can be manipulated in post with Resolve. Software like imatest attempts to apply a consistent floor to the DR measurement by specifying an RMS noise %. It can be fooled by NR.
combatentropy
Veteran
I see no value in this subjectivity and am surprised by how often it is said, even in articles.
There is a standard way to measure dynamic range, ISO 21550. You just divide well capacity by noise. For example, if a photodiode can hold 50,000 electrons, and noise is 5 electrons, then your dynamic range is 10,000:1. To convert it to f-stops, you have to use logarithms: 2 to the what equals 10,000? Answer: 13.29.
---
People complain that the last couple of stops are unusable, so filled they are with noise. True. Measured this way, the final stop ends when noise equals signal. Any detail at that level is positively drowned in noise. But this is not really a problem.
This would only be a problem if this is how we work: You show up at a set, light it, take out your light meter, measure every nook and cranny, measure 13 stops of range, go to the shop, find a camera whose dynamic range is 13 stops, rent it, shoot your scene, go into the editing suite and find out, oh no, you didn't fully capture the scene. No one does this.
This is how we really shop for cameras: Watch videos of a certain camera, say "Hey, I like how that one handles dynamic range," look up its dynamic range (let's say 13 stops), and say to ourselves, "13 seems like enough." Look at other videos, see one where we don't like the dynamic range, look up that camera's dynamic range (let's say 12 stops) and say to ourselves, I need a camera with at least 13 stops!
In the first scenario, deciding on "usable" stops would be important. In the second scenario, agreeing on a standard by which all cameras can be easily compared would be important.
I thought twice about contradicting you, since your post also contains this nugget of truth, seemingly impossible for many to understand.
Last edited:
Thomas Smet
Veteran
I'm not sure if cramming 8k onto a m43 sensor is a great idea but 8k does finally represent the best balance for a hybrid camera. No limitation on stills or video.
For example the GH5s does video extremely well due to the extra size for each pixel but the stills suffer. HD video from larger sensors typically had some level of compromises due to primitive down conversion.
8k is a solid number for stills and its a solid perfect even factor for 8k, 4k and 2k video. No funny math needed at all and should yield very good down conversions.
I also see no reason why a m43, APS-C and FF sensor can't all stay at 8k with each size increase increasing the sensitivity and potential dynamic range.
While 8k on a m43 sensor may end up with more noise a funny thing about the GH5s vs GH5 when it comes to raw stills is just how little those larger pixels actually matter. For example if printing a 8x10 photo the GH5s may be slightly cleaner but the GH5 with more pixels can scale down for that 8x10 print which eliminates a lot of the visible noise. So images from a 8k m43 sensor may have more noise compared to a APS-C or FF sensor but the noise might be so small and since a lot of people will still down convert to 2k or 4k a lot of that extra noise may get eliminated. I think it would really suffer for those that want to pan&scan on 8k video thinking they could delivery a virtual 4x zoom for 2k but if we use 8k to view as is or down converted it might be fine.
Still seems like a giant step backwards however for what really matters to people. Less noise, better signal to noise ratio and better dynamic range. All the things people tend to criticize m43 for. Just seems like all of that is being ignored to hit slightly better quality in perfect lighting conditions.
For example the GH5s does video extremely well due to the extra size for each pixel but the stills suffer. HD video from larger sensors typically had some level of compromises due to primitive down conversion.
8k is a solid number for stills and its a solid perfect even factor for 8k, 4k and 2k video. No funny math needed at all and should yield very good down conversions.
I also see no reason why a m43, APS-C and FF sensor can't all stay at 8k with each size increase increasing the sensitivity and potential dynamic range.
While 8k on a m43 sensor may end up with more noise a funny thing about the GH5s vs GH5 when it comes to raw stills is just how little those larger pixels actually matter. For example if printing a 8x10 photo the GH5s may be slightly cleaner but the GH5 with more pixels can scale down for that 8x10 print which eliminates a lot of the visible noise. So images from a 8k m43 sensor may have more noise compared to a APS-C or FF sensor but the noise might be so small and since a lot of people will still down convert to 2k or 4k a lot of that extra noise may get eliminated. I think it would really suffer for those that want to pan&scan on 8k video thinking they could delivery a virtual 4x zoom for 2k but if we use 8k to view as is or down converted it might be fine.
Still seems like a giant step backwards however for what really matters to people. Less noise, better signal to noise ratio and better dynamic range. All the things people tend to criticize m43 for. Just seems like all of that is being ignored to hit slightly better quality in perfect lighting conditions.
Bassman2003
Veteran
I think the camera companies are getting drunk on the software manipulation tech that makes them think the megapixel race can march on like Moore's law. We think it is nuts but they seem to have marketing on the brain. I would love to see a sub-20 mpx GH6 variant like the GH5s. Big pixels just can't be beat. Yes there is a divide between stills and video resolution needs but we are leaving "needs" with the FF cameras going over 50+ mpx anyway.
Cary Knoop
Veteran
Provided the AD converters and amplifiers are not the limiting factors larger pixels have a larger dynamic range because they catch more photons.
combatentropy
Veteran
They're just measuring the wrong numbers.
"That camera is 4K. This camera is 8K."
- or -
"Those pixels are 2 microns. These pixels are 4 microns."
"That camera has 13 stops. This camera has 15 stops."
"That camera is usable to ISO 6400. This camera is usable to ISO 25600."
"That camera records up to 60 fps. This camera records 120 fps."
"That camera compresses 20:1. This camera compresses 5:1."
"That camera offloads footage in 20 minutes. This camera offloads footage in 5 minutes."
"That camera requires a $2,000 editing system. This camera requires a $1,000 editing system."
"That camera's media is $100 an hour. This camera's media is $25 a hour."
"That camera runs 1 hour per charge. This camera runs 2 hours."
"That camera requires $5,000 lenses. This camera requires $500 lenses."
"That camera costs $3,000. This camera costs $2,000."
"That camera has rolling shutter. This camera has global shutter."
Resolution is the one spec that is the enemy of all other specs.