The sensors are totally different . Why would you expect the noise generated to be the same and then assume that because it isn't that noise reduction was the reason. Neither the sensor or the processing is the same. That is the difference between the two cameras. Other than those two things the cameras are almost identical ( Well there are some Red coloured buttons on one !!! ) The output is what matters.
GH5s? New low light monster?
- Thread starter Brandon Rice
- Start date
Joshua Cadmium
Well-known member
I never said anything like that. I do not think that the GH5s is just a marketing trick. For low light video, I think that the GH5s is leaps and bounds ahead of the GH5, and for general video usage, I would gladly take it over the GH5, except for limited circumstances.
My conclusion is logically based on the raw images that were posted on DPReview and an understanding of sensor engineering.
The facts of the matter can plainly be seen on DPReview: the raw image of the GH5s looks better than the GH5, but no more than about one stop better. Based on physics, that is what is to be expected.
The GH5s has about a single stop advantage in total light collected. The photosite size on the GH5s is about 4.70um, while the photosite size on the GH5 is about 3.34um. Total surface area for each photosite is therefore 4.70um x 4.70um = 22.1um^2 for the GH5s and 3.34um x 3.34um = 11.2um^2 for the GH5. That's 1.97x as much light or 0.986 stops advantage to the GH5s.
When looking at video comparisons, though, you can see something like a 3-5 stop difference in noise between the GH5 and GH5s. There is no way that is coming from just the sensor, otherwise we would see it in the raw stills, too. DPReview said that the stills are going through the same dual iso pipeline.
So, if the raw images are going through the same dual iso pipeline that the video is going through, what else can account for the massive difference between what is seen in raw stills versus what is seen in video? There is only one thing that will do that and that is temporal noise reduction, like Neat Video, which can only be used when you have multiple frames of information, like in video.
This isn't some crazy logical leap, it's the best fit for all the data.
It's a misunderstanding of sensor technology to think that any difference at the photosite level would only be seen in video and not at the less processed level that a raw image would be at. A raw image is much closer to the actual sensor readout than a video that has be demosaiced in camera on the fly, then have a color profile alter it further, then be compressed by a lossy codec.
For instance, let's say that the GH5s is using a Sony Exmor R sensor that is backside illuminated and the GH5 is using a standard Exmor sensor that is front illuminated. That difference in sensor technology would clearly be seen in raw images as well.
All that being said, it doesn't necessarily matter what is going on under the hood, it's the final image that ultimately matters and that final image looks great, especially in low light. However, it's not just the sensor which is contributing to that fantastic low light performance.
Last edited:
Samuel H
Major Contributor
I can agree with all of that ^
Except you need one more piece to fill your argument, and you're very unlikely to get it.
A few years ago I ran some DR tests with my a7S, in RAW stills and in video mode, and concluded that Sony could offer much more DR than slog2 was showing. What I failed to see there was that RAW stills were using the full ADC pipeline, whereas video was probably using the reduced-bit-depth pipiline that the camera switches to when it does continuous shooting.
I have no idea, but probably Panasonic is doing the same, and the extent of this can be different from one camera to another. So, if the GH5 is using a 10-bit ADC pipeline for video and the GH5s is using a 12-bit ADC pipeline, or even the full 14-bit ADC pipeline, then that comparison across RAW stills that use the full bit depth pipeline can mislead you.
The way to know how big the sensor-related improvement is, is to apply noise reduction to the GH5 footage, until it gets to the same level of NR-related detail loss and artifacts as the GH5s, then see how big the improvement is.
Except you need one more piece to fill your argument, and you're very unlikely to get it.
A few years ago I ran some DR tests with my a7S, in RAW stills and in video mode, and concluded that Sony could offer much more DR than slog2 was showing. What I failed to see there was that RAW stills were using the full ADC pipeline, whereas video was probably using the reduced-bit-depth pipiline that the camera switches to when it does continuous shooting.
I have no idea, but probably Panasonic is doing the same, and the extent of this can be different from one camera to another. So, if the GH5 is using a 10-bit ADC pipeline for video and the GH5s is using a 12-bit ADC pipeline, or even the full 14-bit ADC pipeline, then that comparison across RAW stills that use the full bit depth pipeline can mislead you.
The way to know how big the sensor-related improvement is, is to apply noise reduction to the GH5 footage, until it gets to the same level of NR-related detail loss and artifacts as the GH5s, then see how big the improvement is.
Thomas Smet
Veteran
The same is also true of most cameras. Compare raw stills from an APS-C camera to raw stills from the GH5 and you will also see at best there is maybe a one stop advantage. All cameras use a heavy amount of noise reduction including APS-C and FF. If the raw from the GH5s is one stop better than the GH5 that in fact puts it in the same ballpark as a larger APS-C camera which is incredibly impressive for a tiny sensor.
Combine that with the fact that m43 users get to use just about any lens out there on a focal reducer and that pushes the GH5s another stop ahead of what larger sensor cameras can do. There is really no longer an argument that APS-C is superior to m43. The GH5s has the same sensitivity and once you add a Speedbooster you have the same FOV, DOF of APS-C with the raw sensitivity of every FF camera out there minus the A7S. In fact if you shoot DCI and use FF lenses with a SpeedBooster XL you only have a 1.2x crop compared to FF so you have a FOV and DOF much closer to FF and the sensitive to match. That is before we even factor in the temporal noise reduction and the dual native sensor.
Arash Ayrom
Cataclysmo
I know it's possibly a baseless rumor, but the idea of near-14 stops of DR, to me, along with the low-light capabilities of the GH5s make it a must-buy to accompany my GH5. I'd feel less excited if it's just the additional low-light powers. But Matt Frazer is pretty emphatic and clear in the interview above, and there are other sources quite clearly saying they're getting nearly 14 stops DR.
Now, unfortunately, I'm pretty herpaderp when it comes to understanding some of the intricacies and innards of camera tech. For example, what DOES affect DR?
- Does photodiode size help with DR? This bit from Imaging Resource:
"Each photodiode has grown in size by around 1.96 times, and according to Panasonic that change translates to a 1.5-stop (9.7dB) improvement in the signal-to-noise ratio."
- Is it a processing issue? The GH5s seems to have all the 10-bit / 4:2:2 goodness of the GH5. This bit from Digital Trends:
"Panasonic states the GH5S can hit 14 stops of dynamic range for RAW still photos, but hasn’t actually listed dynamic range for video, which is likely a tad lower due to the lack of a RAW video format."
- Does the HLG format allow for more DR than V-log L?
I know I'm kinda trying to convince myself that the 14 stop DR rumor is true, so any thoughts / comments welcome.
Now, unfortunately, I'm pretty herpaderp when it comes to understanding some of the intricacies and innards of camera tech. For example, what DOES affect DR?
- Does photodiode size help with DR? This bit from Imaging Resource:
"Each photodiode has grown in size by around 1.96 times, and according to Panasonic that change translates to a 1.5-stop (9.7dB) improvement in the signal-to-noise ratio."
- Is it a processing issue? The GH5s seems to have all the 10-bit / 4:2:2 goodness of the GH5. This bit from Digital Trends:
"Panasonic states the GH5S can hit 14 stops of dynamic range for RAW still photos, but hasn’t actually listed dynamic range for video, which is likely a tad lower due to the lack of a RAW video format."
- Does the HLG format allow for more DR than V-log L?
I know I'm kinda trying to convince myself that the 14 stop DR rumor is true, so any thoughts / comments welcome.
Bassman2003
Veteran
I read a post on DPReview where a user in Japan received a v1.0 firmware a few days ago and posted a video with improved AF performance. So I think it is safe to say the camera is still being formulated right up until the release date. So this might also apply to DR as well. Looking to be a great camera.
Given the sources saying that 14-stops is true, I strongly suspect it's the real deal, but like anything you will have to get everything right when shooting and processing to see the full DR of this camera. It won't replace my current GH5, but I will add it as a second b-roll camera too my shoots. ( I might rent a G9 too, just to see if the still-photo performance is worth it over a GH5 )
fboonzaier
Well-known member
If it truly is a 14stops DR camera, that would be the biggest news and put it in the must buy category....BUT...as with Filmguy's IBIS love story re another thread....not at the expense of my GH5....I love shooting with IBIS and will simply build the GH5s out into a bigger, heavier camera using an Atomos ETC.....effectively a heavy camera on sticks or heavy shoulder mount....PS nice analogy filmguy!
fboonzaier
Well-known member
Well this is one is the rub...unless it comes with VLog...not VlogL...there was similar uncertainty with the EVA and some specs including DR was not called till the last moment, though Matt Frazer may have boldly gone there, because they already know the sensor is capable?....I find Panasonic is quite a conservative company in terms of their spec claims....except their AF���� (just a little Canon humor....never touch it)...and if one looks at the camera they packed into the EVA body which is still very small, PERHAPS this is not impossible?...it would open my wallet lol!
While I've heard that GH5 V-Log L would capture 12 stops of DR, I've never heard that the V-LOG L format was capped at 12 stops. It's possible that the 12 stop limitation was entirely the GH5 camera and not the actual recording format.
It's the other way around. V-Log-L is capped at 80 IRE/12-stops. Link to white paper by Barry Green.
http://pro-av.panasonic.net/en/dvx4k/pdf/ag-dvx200_tech_brief_vol6_en.pdf
Thomas Smet
Veteran
Correct. V-log L was designed for 8 stops under and clips at 4 stops over middle gray for 12 stops. It is a modification of the full V-log curve that has the same 8 stops under but the full 8 stops over. Pretty sure V-log L itself will clip anything over 12 stops. Perhaps there is a way to make a little bit more fit but the whole point of V-log and V-log L is that they could be a universal log format between cameras and luts. There is only so many stops over middle gray you can have due to that 80IRE clip.
Now HLG may be a totally different story since as far as I am aware HLG doesn't have the same hard capped limit. It may actually be possible to push past 12 stops with HLG.
Remember what Barry Green says in the white paper. With VLog-L all the scene files are turned off, it's more than a gamma, it's a mode that gets you closest to the sensor, debayering but adding the least amount of processing. HLG is a simple gamma curve layered on top of the same background processing happening as a scene file. It's likely VLog-L outperforms HLG (in camera) yielding a better signal to noise.
Also agree the true sensitivity difference at the sensor compared to the GH5 like with raw stills, is 1 stop. It's just that when you amplify by adding gain (higher iso) that number multiplies. So for example at base iso gain 0, there's 1 stop lower noise in the GH5s that you can hardly notice if at all. Double the iso 6db gain and 1 stop difference becomes 2. Double it again 12db and 2 becomes 4. At that point, 12800 iso in the GH5s is as clean as 3200 in the GH5, but at 400 iso the noise is so low you can't tell the difference.
Saying 14 stops DR is just trying to put a number to describe what you should already be seeing if it were so. From the posted videos, it looks identical to the GH5 when side by side. That said, even 12 stops is a super good number and even that I feel could be exaggerated, I say more like 10.5-11.0, still really really fine. Your benchmark is the Alexa. It claims only 14 stops and if it is, then next best is 12. Many would say the Alexa is more like 16 stops. The important thing is not the number, it's the image quality you get.
Also agree the true sensitivity difference at the sensor compared to the GH5 like with raw stills, is 1 stop. It's just that when you amplify by adding gain (higher iso) that number multiplies. So for example at base iso gain 0, there's 1 stop lower noise in the GH5s that you can hardly notice if at all. Double the iso 6db gain and 1 stop difference becomes 2. Double it again 12db and 2 becomes 4. At that point, 12800 iso in the GH5s is as clean as 3200 in the GH5, but at 400 iso the noise is so low you can't tell the difference.
Saying 14 stops DR is just trying to put a number to describe what you should already be seeing if it were so. From the posted videos, it looks identical to the GH5 when side by side. That said, even 12 stops is a super good number and even that I feel could be exaggerated, I say more like 10.5-11.0, still really really fine. Your benchmark is the Alexa. It claims only 14 stops and if it is, then next best is 12. Many would say the Alexa is more like 16 stops. The important thing is not the number, it's the image quality you get.
fboonzaier
Well-known member
It would make business sense that if the GH5s has 14 stops of DR, Panasonic will release it with full Vlog instead of VlogL...and it would make no sense whatsoever to cripple a 14 stop latitude camera with a 12 stop limit Log....this may well be why Vlog is included with the GH5s and not offered as an option...only VlogL has ever been offered as a paid upgrade....the Varicams and EVA come with Vlog already included in the price of the camera, and as to my needs anyway, this camera looks like it's much more suitable for narrative (wedding photographers may well disagree) but I would be looking at the wider DR....used on an external recorder shooting Prores HQ, not the low light, except under the rare circumstances where at the end of s shoot day, you grab an emergency shot with no time to re relight before going into Golden Time for the crew.
Bassman2003
Veteran
One question: Is the dual ISO always on? Meaning that it can be used when shooting in VLog? Maybe Matt can clarify the 14 stop claim over time.
Thomas Smet
Veteran
Yes it works when shooting V-log. Thats one of the main advantages of the GH5s over the GH5. Not so much just for shooting in the dark but you can shoot V-log at ISO 2500 with the same roughly clean look as V-log at ISO 400 which is just under a 3 stop improvement. This opens up a couple of possibilities:
1. You can shoot indoors with a f5.6 zoom for the same roughly noise quality and dynamic range for a prime lens shooting at f2.2. This opens up a whole new world of possibilities using zoom lenses in tough situations where with the Gh5 you almost always had to use primes or extremely expensive zoom lenses with a very limited zoom range. The f2.8 zooms can be as clean as a theoretical f1.1 zoom lens.
2. You can shoot 3x slow motion with roughly the same noise quality as regular motion on the GH5. Slow motion in poor indoor lighting is now somewhat possible.
3. You can use less output LED lights 3 stops lower on an indoor set which means cheaper, smaller and lighter light kits.
4. You can shoot a production outdoors at ISO 400 and indoors at ISO 2500 with the same lenses and aperture for a consistent look throughout the production. Every lens looks different at different aperture amounts. F1.4 will never look the same as f4.0 so just adjusting aperture will never deliver consistent results. One solution is to use a boatload of ND filtering outside so you can keep the lens as the same aperture as you would use indoors. That only works if you use the lenses wide open or if you use a ton of light indoors. A wide open lens is never as good as when stopped down a bit and every lens has a sweet spot. FF lenses tend to have a sweet spot around f4-f5.6. Dual native ISO means it is now much easier to target the same sweet spot aperture on a production without a massive light kit.
5. You can use a lot more diffusion on lights to help soften them. Smaller LED lights create more of a harsh light. The more the lights are diffused the weaker the output. ISO 2500 means you can diffuse the daylights out of those lights to get very soft lighting without an impact on the quality.
Dual native ISO is not just a means to suck in more light to shoot in the dark. It is a highly advanced tool that can and should be leveraged to rethink how one shoots in different environments. It really is a game changer. ISO 2500 doesn't mean no longer having to use lights for production value but it means smaller more agile lights can be used that take minutes to setup vs hours. There are a slew of affordable LED panel lights out there now with CRI 97+ accuracy, dimmable and color temp adjustable for under $100 or $200. They don't have the massive output of traditional lights or even a Dracast LED but with 2500 ISO these lights become a more realistic option.
Can you post a link ? ...I have not seen it verified anywhere.
When you've got a well known and pretty technical savvy Panasonic rep, Matt Frazer, stating that the GH5s has 2 more stops of DR than the GH5, and then he goes on to say that this is close to 14 stops of latitude. And when you've got two fairly well known Indie filmmakers ( one who shot several shorts for Panasonic over the past year ) also stating that this new camera has 14 stops of dynamic range, I'm going to go with these guys and believe that this new beastie might actually have 14 stops of DR.
We should know a lot more once the GH5s firmware is finished ( currently it's at version 0.2 ) and the camera is finally released for anyone to test, which sounds like end of February or early March 2018.