Panasonic S1 V-LOG -- New image quality king of the hill

[KnightsFan]

 

21 hours ago, sanveer said:

We even had a discussion here somewhere and someone curiously compiled a long list of cameras which he claimed were consistent, which it probably wasn't, since I just did a random Google search with cinema5d dynamic range chats and there were all sorts of contradictory results. Plus, I realized later, that while some of their charts were on other sites (like NFS), they didn't seem to have those charts or writeups. It was very curious.

That "someone" was me. I compiled a list of every DR rating I could find on C5D, and found the only inconsistency was with one Sony camera. I contacted C5D about it, and they said that the measurement differences were due to one being UHD and the other UHD downscaled to 1080p. The discrepancy was consistent with the difference you'd expect from downscaling.

In my list, I included links to the articles I got the DR ratings from. I encourage anyone to look over the list and double check--that's why I posted it in the first place. I'm happy to post it again. Until you check my work, saying "he claimed were consistent, which it probably wasn't" isn't justified. I asked people to help me expand the list, and am still happy to do so if anyone is interested in actually figuring out the degree to which we can trust those DR results, rather than say "it probably wasn't" without looking.

A few years ago, C5D changed their test methodology, and edited their past articles to remove the numbers that reflected their old methodology. However, other sites (such as NFS) still quote the old methodology and have old chart images, but if you follow NFS's links you will see that the C5D articles have been edited and those charts have been removed.

[sanveer]

26 minutes ago, KnightsFan said:

 

That "someone" was me. I compiled a list of every DR rating I could find on C5D, and found the only inconsistency was with one Sony camera. I contacted C5D about it, and they said that the measurement differences were due to one being UHD and the other UHD downscaled to 1080p. The discrepancy was consistent with the difference you'd expect from downscaling.

In my list, I included links to the articles I got the DR ratings from. I encourage anyone to look over the list and double check--that's why I posted it in the first place. I'm happy to post it again. Until you check my work, saying "he claimed were consistent, which it probably wasn't" isn't justified. I asked people to help me expand the list, and am still happy to do so if anyone is interested in actually figuring out the degree to which we can trust those DR results, rather than say "it probably wasn't" without looking.

A few years ago, C5D changed their test methodology, and edited their past articles to remove the numbers that reflected their old methodology. However, other sites (such as NFS) still quote the old methodology and have old chart images, but if you follow NFS's links you will see that the C5D articles have been edited and those charts have been removed.

I am sure they will come up with another methodology in another few months or years. The problem is not even a new methodology. It is that the error rate is so high or so vast. To be fair to them, some greater expert than them, who was once with the BBC, has rated the GH5s for 14.6 or 14.7 stops of dynamic range.

My reference to those figures was not to suggest that you are a certified idiot, contradicting yourself. I did glance through your list, and I thought you did put quite a bit of effort into compiling it. I merely mentioned it, to say that the very edifice of the list is untenable, which is proven by the fact that they took down those figures (and the theory of discrepancy by downressing 4k to 1080p on the A7s, creating a vast anomaly in results, among 9ther theories of theirs), and the fact that the discrepancies were enormous. I do apologise if it sounded otherwise. You never conducted those tests. You just documented it or compiled it into a list. 

Also, don't bother researching or compiling anything from their figures, because I believe it's an exercise in futility. I am sure someone is extremely accurate or acceptably inaccurate, with dynamic range figures. It's just that that person has not been found yet or hasn't become famous enough yet. 

[Andrew Reid]

Where's the link to your list?

[KnightsFan]

3 minutes ago, Andrew Reid said:

Where's the link to your list?

 

5 minutes ago, sanveer said:

I am sure they will come up with another methodology in another few months or years. The problem is not even a new methodology. It is that the error rate is so high or so vast. To be fair to them, some greater expert than them, who was once with the BBC, has rated the GH5s for 14.6 or 14.7 stops of dynamic range.

As long as they say how they are measuring, changing the method is more confusing than inaccurate. I don't know of any other site that has nearly as extensive a list of "objective" DR measurements for video mode (and I say "objective" with quotes since we are trying to verify to what extent they are accurate). It would be great if we had a couple other sources to compare against.

7 minutes ago, sanveer said:

My reference to those figures was not to suggest that you are a certified idiot, contradicting yourself.

I appreciate it

7 minutes ago, sanveer said:

I merely mentioned it, to say that the very edifice of the list is untenable, which is proven by the fact that they took down those figures (and the theory of discrepancy by downressing 4k to 1080p on the A7s, creating a vast anomaly in results, among 9ther theories of theirs)

Taking down the old figures is a good thing. Pushing forward with inconsistent, uncontrolled tests would have been the bigger mistake.

Downsampling increases SNR. That's what they are measuring: the contrast ration between signal clipping in whites, and SNR = 2 in the shadows. So while you can argue that SNR isn't a "true measure" of dynamic range, it's what C5D decided to document. Downsampling and noise reduction could both give falsely high ratings using C5D's method (depending on how you look at it), which is something you have to keep in mind for sure. But both would come at the expense of resolution and detail, which is something we can judge from other tests.

Again, it would be nice if more people did lab tests, maybe even measuring in different ways, so we could compare results.

[Lars Steenhoff]

Here's some S1 Vlog test footage I just shot on the ninja v, pro res hq, 

Shot with Zeiss 50mm zf.1

Sharpening at 0 

noise reduction at 0

Stills graded with FilmConvert fuji eterna

https://www.dropbox.com/s/can0ot1k98ogd7a/NINJAV_S001_S001_T066.MOV?dl=0

[currensheldon]

10 minutes ago, Lars Steenhoff said:

Here's some S1 Vlog test footage I just shot on the ninja v, pro res hq, 

Shot with Zeiss 50mm zf.1

Sharpening at 0 

noise reduction at 0

Stills graded with FilmConvert fuji eterna

https://www.dropbox.com/s/can0ot1k98ogd7a/NINJAV_S001_S001_T066.MOV?dl=0

Awesome! How did you get a hold of the update already? 

[Geoff_L]

6 minutes ago, Lars Steenhoff said:

Here's some S1 Vlog test footage I just shot on the ninja v, pro res hq,

Thank your Lars ! If it is not too much to ask, can you deliver you first impressions about the S1 ?

[Lars Steenhoff]

11 minutes ago, Geoff_L said:

Thank your Lars ! If it is not too much to ask, can you deliver you first impressions about the S1 ?

Sure, Its a great camera.

Plus points, footage looks very organic with vlog.

10 Bit is so nice for grading.

Dynamic range is awesome
Ibis is awesome

Autofocus Is good enough for stills for me

For filming I use mostly manual lenses.
And I love the 96 megapixel pixel shift mode, those shots are so clean, no noise.

 

There are a only few things I miss compared to my Canon 5d3 with raw.

1- On the canon I could crank up the sharpness to the max for in camera for monitoring only, and because it was raw the sharpening is not baked in the footage. I wish Panasonic will add this to firmware. ( Edge sharping for preview only )

2. I cannot zoom into the footage while recording, only before pressing record can I zoom in. Again firmware request

3. There is no record trigger for the Ninja. also firmware request

Besides those points I love the camera.

 

16 minutes ago, currensheldon said:

Awesome! How did you get a hold of the update already? 

I'm in the Netherlands, they send it to me after I send in my invoice last week. very fast turnaround. ( and yes it was free, thank you Panasonic)

[thebrothersthre3]

I wonder how well that matches to Eterna on a Fuji. 

[Jak234]

@Lars: did you see a difference when getting nr to -5?

Also would be -5 for sharpening better in terms of in camera sharpening?

Most people on the gh5(s) hat nc -5 an sharpening -5

[Lars Steenhoff]

12 minutes ago, Jak234 said:

@Lars: did you see a difference when getting nr to -5?

Also would be -5 for sharpening better in terms of in camera sharpening?

Most people on the gh5(s) hat nc -5 an sharpening -5

When using vlog there is no minus, it starts from 0 and goes to 10

So you can compare my settings with -5

[Andrew Reid]

3 hours ago, Lars Steenhoff said:

1- On the canon I could crank up the sharpness to the max for in camera for monitoring only, and because it was raw the sharpening is not baked in the footage. I wish Panasonic will add this to firmware. ( Edge sharping for preview only )

2. I cannot zoom into the footage while recording, only before pressing record can I zoom in. Again firmware request

3. There is no record trigger for the Ninja. also firmware request

Good suggestions!

4 hours ago, KnightsFan said:

 

As long as they say how they are measuring, changing the method is more confusing than inaccurate. I don't know of any other site that has nearly as extensive a list of "objective" DR measurements for video mode (and I say "objective" with quotes since we are trying to verify to what extent they are accurate). It would be great if we had a couple other sources to compare against.

I agree there should be better measurements, and all in one place under set conditions.

According to the C5D results, the Sony A5100 in standard Creative Style has 1 stop more dynamic range than a GH5 shooting V-LOG

The X-T2 is 1 stop higher than X-T3. Eh!

Also what matters are not even the numbers, when it comes to DR, but the usability of the image when pushed to the max, the colours, the lack of artefacts in the pushed dark regions and the look of highlight roll off.

[KnightsFan]

1 hour ago, Andrew Reid said:

According to the C5D results, the Sony A5100 in standard Creative Style has 1 stop more dynamic range than a GH5 shooting V-LOG

The X-T2 is 1 stop higher than X-T3. Eh!

Those results are certainly against conventional wisdom. However, the A5100 is HD, whereas the GH5 is UHD. The GH5 should gain another ~stop of DR if it was downscaled to HD, putting it above the A5100 in this test. Noise reduction and/or sharpening could play a role as well. The article says "[the a5100] performs slightly worse than the GH4." Not idea what picture profile, nor could I find any measurements of the GH4 on its own that weren't taken down after the method change.

Perhaps noise reduction gives the XT2 a lower value also? Interestingly, the XT2 and XT3 match when downscaled to HD, maybe equalizing the NR done on the 4k image. I'm sure tweaking camera settings would make different values, on all the tests done. I wish they were more rigorous in documenting what settings were used, though in recent articles it's been better.

1 hour ago, Andrew Reid said:

Also what matters are not even the numbers, when it comes to DR, but the usability of the image when pushed to the max, the colours, the lack of artefacts in the pushed dark regions and the look of highlight roll off.

Certainly true! That's why I'm in this topic in the first place, to get a crack at the non-test chart footage you have so kindly uploaded

Anyway, apologies for going off topic.

[Deadcode]

8 hours ago, thebrothersthre3 said:

I wonder how well that matches to Eterna on a Fuji. 

Poorly. F-Log curve is like SLOG2 on Sony's +1EV (check lutcalc). 

V-LOG curve on the other hand is like SLOG3/LogC/Cineon curve.

F-Gamut naturally shifts to green which should be corrected with gamut mapping in post.

V-Gamut do not shifts to magenta or green.

So technically this "use FLOG to Eterna" Filmconvert simulation on VLOG is a creative choice of grading nothing more.

[jagnje]

On 7/14/2019 at 6:00 AM, Mako Sports said:

I saw a video that was posted by Wex Photo Video and was seriously impressed. I think we have finally made it to true 14 stop mirrorless cameras (in video). The "Varicam Look" is starting to grow on me as well. ?

 Read an article on Newshooter "Panasonic told us at Cinegear that their strategy is to unify the Varicam, EVA1, and Lumix divisions and try and bring them all closer together."

I`m sorry, did not read the whole thread, but this is the video you were impressed with? The pink/green clouds, like crap looking skintones? I mean, this is just about as nasty as a low level ENG camera. I`m sure that its a user error, poorly shot, poorly grade, but seriusly, this clip???

[Lars Steenhoff]

1 hour ago, Deadcode said:

Poorly. F-Log curve is like SLOG2 on Sony's +1EV (check lutcalc). 

V-LOG curve on the other hand is like SLOG3/LogC/Cineon curve.

F-Gamut naturally shifts to green which should be corrected with gamut mapping in post.

V-Gamut do not shifts to magenta or green.

So technically this "use FLOG to Eterna" Filmconvert simulation on VLOG is a creative choice of grading nothing more.

I did not use Flog to eterna, I used the varicam preset with vlog in filmconvert.

 

 

 

[Jimmy G]

3 hours ago, jagnje said:

I`m sorry, did not read the whole thread, but this is the video you were impressed with? The pink/green clouds, like crap looking skintones? I mean, this is just about as nasty as a low level ENG camera. I`m sure that its a user error, poorly shot, poorly grade, but seriusly, this clip???

My initial impression and suspicion in watching this, um, "test video" was that this video was perhaps not shot entirely on the S1? (Ignoring the decidedly "different look" of the interspersed stills images) I was noticing a variety of poor IQ clips throughout. However in re-reviewing the VLog grading swipe at the 2:40 Mark...

Panasonic LUMIX S1R and S1 V-Log firmware update | Southern Italy real-world test | Pt. 2 - YouTube: (t=160)

 

...(partial screenshot showing severe sky banding from that 2:40 wipe...click to enlarge)...

...I'm now thinking that there are (again, perhaps) A. some post grading problems (choices) going on with this presentation which are B. being exacerbated by YT/google's insistence on serving up 1080P-max footage to Apple users (above screenshot from a non-Retina 27" iMac) and worse, still, only 360P-max available for this clip on my iPhone 8 Plus! Haw!

So bad, IMHO, were the grading/YT platform choices for hosting this video that this screenshot (click to enlarge) from the 7:09 Mark...

...shows a combination of lens-vignetting banding cross-hatching with the sky-gradation banding along with a sorry mix of sky color errors tossed in for good measure! Oofah!

I would be most curious to see what this review looked like on Vimeo or some other hosting service that would serve up this S1-sampling video in 4K without the VP9 reinterpretation and added compression that YT/G seem insistent on shoving down everyone's throat. As it is, IMHO, the kind folks at WEX are doing both their customers and Panasonic a disservice by hosting this test in its current location...put simply, I would not make a purchase order for this camera based on these results.

On the plus side, these test results have prioritized what my first tests will be when the DMW-SFU2 firmware upgrade eventually arrives at my Post Office.

Edit: grammar

[deezid]

This blows the Pocket 4K out of the water since its anti DNG update.
So much detail and texture despite the "worse" codec. Wow.
No sharpening artifacts either.

[Andrew Reid]

Yep. Just don't judge it off YouTube adverts.

[Geoff_L]

On 7/16/2019 at 10:27 PM, Lars Steenhoff said:

Sure, Its a great camera.

Plus points, footage looks very organic with vlog.

10 Bit is so nice for grading.

Dynamic range is awesome
Ibis is awesome

Autofocus Is good enough for stills for me

For filming I use mostly manual lenses.
And I love the 96 megapixel pixel shift mode, those shots are so clean, no noise.

 

There are a only few things I miss compared to my Canon 5d3 with raw.

1- On the canon I could crank up the sharpness to the max for in camera for monitoring only, and because it was raw the sharpening is not baked in the footage. I wish Panasonic will add this to firmware. ( Edge sharping for preview only )

2. I cannot zoom into the footage while recording, only before pressing record can I zoom in. Again firmware request

3. There is no record trigger for the Ninja. also firmware request

Besides those points I love the camera.

 

I'm in the Netherlands, they send it to me after I send in my invoice last week. very fast turnaround. ( and yes it was free, thank you Panasonic)

Thank you for your impressions

Reading that from someone who used ML raw is encouraging ! One last question : do you record on SD or XQD ? If sd, would you mind sharing which model ? I wonder too if pany will release an update for allowing compatibility with CFexpress (as Nikon promised), even if I believe there is not a lot of news on this front.

Ah, I just checked my drawers and found a complete kit of LEE gnd filters + filter holder to sell, plus a lens or two... that way I will have no more excuse to continue to resist buying a S1 ?