Panasonic S1 4K 10bit video mode to be present at launch with Hybrid LOG Gamma

[Cliff Totten]

The ÷/- noise reduction parameters work perfectly fine on the GH5 and GH5-S. On my GH5, I run my NR settings all the pulled back. As we know, noise reduction plugins in post are WAY more accurate and surgical than any live in-camera algorithms are.

I really hope that Panasonic fixes the S1 noise reduction +/- settings in firmware 1.0. It's stupid to see it in the menu but have it disabled behind the scenes.

[mirekti]

I have quite a few M lenses, but prefer older mounts as one can put an adapter with built in ND variable filter in between (to be released).

Once we get organic sensor and don’t need to use NDs anymore M mount lenses will be great for these cameras.

[kye]

11 hours ago, Andrew Reid said:

The colour space is Rec.2100 which develops on several aspects of Rec.2020 introduced with the Sony A7R III’s HLG mode.

This is very interesting to me, especially the Like2100 mode shown on the LCD.

The reason I raise it is because I am not sure that the HLG on the GH5 is actually Rec2100.  I often want to adjust exposure or WB in post and I tried the method of turning the GH5 HLG from Rec2100 to Linear, making some kind of offset adjustment, then turning it back from Linear to Rec2100, but couldn't match the results from raising the exposure in-camera by one stop.  

I shot a test scene at two exposure levels and tried matching one with the other via this method.  Every adjustment I could think of resulted in complex non-linearities, and when I asked the question over at LiftGammaGain someone suggested that maybe the GH5 HLG mode isn't true Rec2100.  I tried Rec2020 and other colour spaces too, nothing worked.  I did a bunch of googling to try and find out what the GH5 HLG mode actually is and couldn't find any definitive answers, so I basically just concluded that it wasn't real Rec2100 and that I couldn't do an exposure or WB adjustment in post that completely matched having done it in camera.

Any idea what's going on here?

[thebrothersthre3]

3 hours ago, androidlad said:

It's unlikely to go beyond 12 stops in V-log, this is a stills-oriented camera that operate using 12bit ADC in video mode. Heavy NR also tricks the image analysis software for cleaner shadows (more absolute DR in shadow but not very usable).

Yeah it looks good on paper but I doubt its actually that close to the Ursa 4.6k in terms of usability.

[ssrdd]

After spending 4000plus dollars we get ''FREE'' 10bit out of the box for unreliable AF mode and proven history. WOw. Canon is not parameter here, we forgot it a long time ago. We are now fuji generation.  

[jack jin]

The problem I see with the s1, is that the color and the highlight roll off seems to be pretty bad, it just doesn't have the mojo that the gh5s has, it's better then the sony, but i wish it could be as good as the gh5s.

[ade towell]

This is all pre-production, let's see what it looks like when fully functioning and with proper v-log, AF has looked pretty good on latest firmware and hopefully NR can be turned completely off on production model -  there isn't enough footage to make any solid conclusions yet but the lowlight performance, and DR along with IBIS have got me intrigued

[levisdavis]

HLG is said to handle up to 16-stops DR.  That would make me think that the 12.1 or 12.2 stops is the max for the camera; even with VLOG.  Makes me wonder about the colors / tones / highlights roll-off and shadow detail though as it may be a more cinematic look in comparison to the HLG?

In that same line of thought, in terms of max DR...  Will the Z6/Z7 reach 14 stops DR w/ and HDMI out ATOMOS BRAW combination?  ... Thinking the sensor natively handles about 14-stops of latitude in stills...  The BRAW output could theoretically be the same.  However, it's possible that the full DR is being clipped in the output signal.

[Cliff Totten]

2 hours ago, levisdavis said:

HLG is said to handle up to 16-stops DR.  That would make me think that the 12.1 or 12.2 stops is the max for the camera; even with VLOG.  Makes me wonder about the colors / tones / highlights roll-off and shadow detail though as it may be a more cinematic look in comparison to the HLG?

In that same line of thought, in terms of max DR...  Will the Z6/Z7 reach 14 stops DR w/ and HDMI out ATOMOS BRAW combination?  ... Thinking the sensor natively handles about 14-stops of latitude in stills...  The BRAW output could theoretically be the same.  However, it's possible that the full DR is being clipped in the output signal.

About the S1's dynamic range....I have two questions:

1.) What in the Hell is this mystery sensor? Who's tech is in it? Some think it might be an IMX410 (A7-III & Z6) Meanwhile, some are saying its a  completely brand new sensor that nobody has ever seen before. There are rumors it's NOT even a back side illuminated chip. Samsung IsoCell??

2.) Panny might very well calibrate it's video A/D conversion and raw readout to NOT exceed 12stops. Supposedly, this could have been a demand that VariCam managers had placed on these S cameras. You can bet your life that the VariCam group did NOT want a high sensor read clock speed (intentional rolling shutter) and wanted it capped at 12stops.

Who knows? We will find out more in the months that follow. On the GH5/GH5-S,...HLG matches VLog-L in dynamic range.

[sanveer]

2 hours ago, Cliff Totten said:

About the S1's dynamic range....I have two questions:

I would start believing real dynamic range figures once DXOlabs get their hands on the S1 series. Many othe sites have infamous figures, both dubious as well as uninspiring. 

 

2 hours ago, Cliff Totten said:

1.) What in the Hell is this mystery sensor? Who's tech is in it? Some think it might be an IMX410 (A7-III & Z6) Meanwhile, some are saying its a  completely brand new sensor that nobody has ever seen before. There are rumors it's NOT even a back side illuminated chip. Samsung IsoCell??

It's not a BSI sensor. That was clearly listed at a few places including the Dpreview.com comparison with other 24MP mirrorless full frame cameras. Apparently the new micro lens arrangement would have an issues with a BSI sensor (light fall off due to angle of incidence, and introducing all kinds of colour and other noise). Also the improvement in low light between FSI and BSI could apparently, be easily negated by the new micro lens architecture (along with sensor design and some great processing).

Which also explains why Panasonic is completely averse to PDAF because they don't want to substitute missing portions of the image and improve parts of the image by heavy processing (I personally don't agree with their viewpoint). 

2 hours ago, Cliff Totten said:

Who knows? We will find out more in the months that follow. On the GH5/GH5-S,...HLG matches VLog-L in dynamic range.

Not according to the kind folks at Cinema5d.com. According to them the difference in more in the ballpark of 1 stop (or about 0.8 stops on the GH5s, being 9.9 stops for HLG and 10.7 stops for VLog-L). 

[Lars Steenhoff]

 

[thebrothersthre3]

9 hours ago, sanveer said:

I would start believing real dynamic range figures once DXOlabs get their hands on the S1 series. Many othe sites have infamous figures, both dubious as well as uninspiring. 

 

It's not a BSI sensor. That was clearly listed at a few places including the Dpreview.com comparison with other 24MP mirrorless full frame cameras. Apparently the new micro lens arrangement would have an issues with a BSI sensor (light fall off due to angle of incidence, and introducing all kinds of colour and other noise). Also the improvement in low light between FSI and BSI could apparently, be easily negated by the new micro lens architecture (along with sensor design and some great processing).

Which also explains why Panasonic is completely averse to PDAF because they don't want to substitute missing portions of the image and improve parts of the image by heavy processing (I personally don't agree with their viewpoint). 

Not according to the kind folks at Cinema5d.com. According to them the difference in more in the ballpark of 1 stop (or about 0.8 stops on the GH5s, being 9.9 stops for HLG and 10.7 stops for VLog-L). 

Thats odd, people I know have found more dynamic range in HLG. Which to me makes sense as Vlog dynamic range is capped on the GH5 and GH5S. HLG having a half stop more DR brings the GH5S closer to the BMPCC4K, which makes sense because they both share a similar sensor. 

IMHO the best test of dynamic range to me is over under tests on human subjects. You can see where skintones become unusable and its just more practical then those charts Cinema5d uses. Just in terms of noise for instance the XT3 shadows are cleaner than the original BM Pocket, however the pocket maintains colors much better where the XT3 starts looking washed out. 

[Cliff Totten]

So S1 HLG is capped at roughly 12 stops. We know full VLog is coming as a paid upgrade. However, if they keep it capped at 12 stop in VLog as well?....that makes the upgrade far less attractive. You might as well get the same 12stops in HLG and record to ProRes from the S1's 10bit HDMI.

I would love to be a fly on the wall in Panny's HQ meetings about the S1. I would love to listen to them debate each other on how they properly cripple the S1 to protect the EVA-1 and Varicam....as well as this new L-Mount camcorder in development.

[thebrothersthre3]

3 minutes ago, Cliff Totten said:

So S1 HLG is capped at roughly 12 stops. We know full VLog is coming as a paid upgrade. However, if they keep it capped at 12 stop in VLog as well?....that makes the upgrade far less attractive. You might as well get the same 12stops in HLG and record to ProRes from the S1's 10bit HDMI.

I would love to be a fly on the wall in Panny's HQ meetings about the S1. I would love to listen to them debate each other on how they properly cripple the S1 to protect the EVA-1 and Varicam....as well as this new L-Mount camcorder in development.

Why would they advertise full Vlog if it will be capped? I don't know if HLG is capped, it may be just the sensor in video mode doesn't get that great of dynamic range. Its only doing 10 bit HEVC. But yeah I agree, the upgrade wouldn't be worth it if Vlog isn't offering more dynamic range. I take that back, the good thing about Vlog is there are a ton of LUTS for it. The other good thing is H264 in vlog will be edit ready(for some systems). HEVC is kind of unusable without transcoding for most people. I wonder if 72mpbs HEVC is equal to H264 150mbps.

I kind of doubt the S1 will take away from Varicam sales much. I can't see anyone wanting to get a Varicam settling on an S1. 

[sanveer]

13 hours ago, thebrothersthre3 said:

Why would they advertise full Vlog if it will be capped? I don't know if HLG is capped, it may be just the sensor in video mode doesn't get that great of dynamic range. Its only doing 10 bit HEVC.

Most likely because the Full VLog (non L version) will be making it to the GH Series too. Also, the full VLog will add almost the full sensor dynamic range (which should be closer to 14 stops and over 13 stops in the least).

And with the hight bitrate external External VLog update, one may actually be able to use most of this dynamic range in post work, almost like highly compressed RAW.

I also suspect that the absence of PDAF may be glaringly obvious in video, but for photos it won't be as bad. If the new engine has improved, then the new autofocus may actually be very good (not necessarily high speed sports photography good, but even better than the GH5). 

[mirekti]

1 hour ago, sanveer said:

Most likely because the Full VLog (non L version) will be making it to the GH Series too

I was under impression VLog-L was made due to GH5 sensor’s dynamic range limitations. 

[deezid]

Here comes the 'still not getting it DXOMARK single raw frame performance is  irrelevant for video' crew ?

And also, since when is having a lower noise floor because of heavy noise reduction and detail filtering aka HLG on the GH5/GH5s equivalent to better dynamic range?
Tbh V-Log L (yes, even crippled V-Log L) shows way more detail in shadow areas than HLG on both the GH5 and GH5s.

If real V-log comes, things may change on the GH5s (same sensor as P4K, IMX294 based) at least if Panasonic doesn't mess it up with horrible temporal filtering like they do on the S1 pre-production model. Also real V-Log with V-Gamut would allow for a professional workflow using ACES and other color management systems.

[Cliff Totten]

6 hours ago, deezid said:

Here comes the 'still not getting it DXOMARK single raw frame performance is  irrelevant for video' crew ?

And also, since when is having a lower noise floor because of heavy noise reduction and detail filtering aka HLG on the GH5/GH5s equivalent to better dynamic range?
Tbh V-Log L (yes, even crippled V-Log L) shows way more detail in shadow areas than HLG on both the GH5 and GH5s.

If real V-log comes, things may change on the GH5s (same sensor as P4K, IMX294 based) at least if Panasonic doesn't mess it up with horrible temporal filtering like they do on the S1 pre-production model. Also real V-Log with V-Gamut would allow for a professional workflow using ACES and other color management systems.

I think you are right about the S1's noise reduction in these early firmware testing. Cinema5D said their noise reduction -/+ controls were not even working at all on their version. I REALLY hope the full 1.0 firmware has a functioning parameter set. To me, effective noise reduction -/+ settings are even more important than color fine tuning. Noise reduction is a destructive process,...color is not.

Actually, Panny.....PLEASE give us a noise reduction "off" setting! That would make many people really happy. (It would also free up some of your image processor CPU for other things too!)

[thebrothersthre3]

I don't think the H264 codec is capable of the same dynamic range the RAW codec is. Not a tech guy so obviously I could be wrong. 

[deezid]

29 minutes ago, thebrothersthre3 said:

I don't think the H264 codec is capable of the same dynamic range the RAW codec is. Not a tech guy so obviously I could be wrong. 

H264 can store 12 bit video. With a log gamma it can easily store 16 or even more stops.

1 hour ago, Cliff Totten said:

Actually, Panny.....PLEASE give us a noise reduction "off" setting! That would make many people really happy. (It would also free up some of your image processor CPU for other things too!)

Yep, that (and sharpening off as well of course) with real V-Log and V-Gamut would be a killer in its price class.
But wouldn't it rival a Varicam 35/LT/EVA-1 too much with all image "enhancing" bullshit turned off?