GH5 GH5 vs GH4 - did the GH4 look better in some instances?

So interesting!

So Thomas do you believe this accounts for the difference in looks between the oversampled EVA1 vs the (1:1 readout?) on the C200?

Do you think shooting 4K Sensor crop on EVA1 (1:1 readout) would “fix” this on EVA1 if trying to avoid that look?

And as Dingo points out - perhaps introduces different problems - now, you don’t have the extra resolution to make up for debayering...

So really we come full circle and wish for a 3-chip sensor with 1:1 pixel read out and an OLPF to achieve this sort of non-digital, organic look.

Of course, it seems now too if shooting in 4K but viewing in HD, we’ll get edge enhancement from the downscale - so ironically a more organic look from 4K viewed at 4K vs 4K viewed at HD? Or am I misunderstanding that portion?
 
filmguy123 said:
Looping back from Moire/Aliasing, still really curious about this. I don't understand it fully.

Googled online and can't seem to find much of anything regarding 1:1 Sensor Readout VS Oversampling.

Thomas, what is it that makes 1:1 sensor readout so "special" as it may relate to the GH4/GH5S vs the GH5 (or the EVA1 vs a lower resolution C200, for example).

And to be clear I haven't much tested the 4K raw VS 5.7K oversample, but I am curious to do so now, if it becomes a 1:1 sensor readout.
Click to expand...

It is also possible the EVA1 is doing a better quality down conversion due to superior processing power. This is why the prospect of scaling in camera kind of scares me. Sounds awesome in theory and can yield great results for a lot of people but it really puts the look of the down conversion at the mercy of the engineering team. Did the engineers prefer a more consumer friendly video look or a more film like look when they developed the down conversion?
 
TheDingo said:
The new Canon EOS C700 Full-Frame cinema camera claims the same thing by debayering a 5.9K sensor to achieve full 4K resolution in the final image.
Click to expand...

Totally agree with you and its really the best option we have which is why movies like Ender's Game shoot 5k and down convert to 2k for their master and post work. Yields better 2k than what a 2k sensor camera could ever achieve. It all comes down to exactly how each camera handles that down conversion.
 
First, I think that the plastic-ish look from the GH5 (and even more so on the GH5s) comes more from aggressive noise reduction than from over-sampling. The GH5s doesn't oversample, and you can still really see it.

Second, the color science is SO much better on the GH5, and even a little better than that on the GH5s. I know this is subjective, but I do think that our customers are actually more sensitive to color - especially skin color - than texture/detail. I spent my entire time with the GH4 (still have 2, actually) wrangling the color, fighting the codecs/profiles, etc. trying to get something that approached what it gave me in stills.
 
Oversampling is important, but there's probably more to it as we see with the beloved ALEXA.

Those cameras provide a beautiful 2K-ish image from a barely 3K-ish+ sensor.

While other cameras like the F35 does it from a 6K sensor and REDs may do it from an 8K sensor...yet - in this resolution race - the 3K-ish sensor has dominated.
 
I do think the NR on GH5 plays a big part, because what I notice seems to have to do more with that plasticky smooth look vs a more natural raw look on the GH4 footage. Which I suppose makes sense, there’s only so many ways to get a sensor on the GH5 with more photosites than the GH4... to look cleaner than a GH4 at higher ISO... so, yeah. Even with NR -5 it’s more.

To clarify, I’m also in the camp that the GH5, holistically, so far outshines the GH4 that I wouldn’t use my GH4 for anything but emergency backup now. Not because it’s bad, it’s just as stated, every other image parameter on the GH5 wins. But yes. The camera overall looks less natural, sort of a less filmic “fake” look (it’s subtle) on many of the shots. Even after film convert and playing with different film stocks and curves, adding in noise, etc. Even matching things via Leeming LUTS and comparing. It’s sort of just baked into the image.

Oh well.

If there’s one thing this thread points out it’s that cameras are immensely complex in how they arrive at their image, and for all the chat on sensor size and the way it affects the image, it may actually be the net result of a dozen other things in actuality.

The GH5 looks better than the GH4 as a whole, but the GH4 looks better in some parameters. I’m going to pay more attention to the GH5S. It’s hard because these things tend to get noticed a year later, not right away. Scientific tests, home tests, etc. don’t reveal it as well. It’s when you start looking back over tons of footage over a year and notice certain trends.
 
Last edited:
filmguy123 said:
I do think the NR on GH5 plays a big part, because what I notice seems to have to do more with that plasticky smooth look vs a more natural raw look on the GH4 footage. Which I suppose makes sense, there’s only so many ways to get a sensor on the GH5 with more photosites than the GH4... to look cleaner than a GH4 at higher ISO... so, yeah. Even with NR -5 it’s more.
Click to expand...


It is both sharpening and nr.
The temporal noise reduction on the GH5s is way worse than on the GH5 even. There's is literally no fine low contrast detail left, faces look like wax. Things are worse on the Sony A7 III though - the mother of plastic.

The forced sharpening at -5 even pushes the footage further into plastic (high contrast halos, harsh micro contrast areas, no low contrast detail). The further downscaling on the GH5 sharpens the footage more than it should. 6K anamorphic footage from the GH5 doesn't have this issue when scaled down in post.

Take EVA-1 V-Log or C200 RAW footage on the other hand - a completely different world! Very organic.
 
Last edited:
joe12south said:
First, I think that the plastic-ish look from the GH5 (and even more so on the GH5s) comes more from aggressive noise reduction than from over-sampling. The GH5s doesn't oversample, and you can still really see it.

Second, the color science is SO much better on the GH5, and even a little better than that on the GH5s. I know this is subjective, but I do think that our customers are actually more sensitive to color - especially skin color - than texture/detail. I spent my entire time with the GH4 (still have 2, actually) wrangling the color, fighting the codecs/profiles, etc. trying to get something that approached what it gave me in stills.
Click to expand...

Agreed. I despise the NR on the Panasonic cameras which is why I tend to shoot it at -5. It never totally gets rid of it however and this is one area where the Pocket may have a large advantage. I still stand by the 150mbps on the GH cameras holding up well to ProRes and really when factoring in just the format itself we are splitting hairs. It is that lack of over processing that is great about the Pocket and one significant reason why I am considering the Pocket.

I honestly had hoped the GH5S was better at it. To be fair most of us have seen material online where we don't know what settings were used. How exactly does it look at -5 NR? Regardless it is always there and that sucks. I would pay money to Panasonic to buy a firmware update to just totally tun off the NR on the camera. I can understand they want to appeal to a wide range of users including videographers but at least give us the option to turn it off.

I honestly don't mind a bit of noise. I think we have become obsessed with a zero noise look when some of my favorite movies shot on 35mm film look like a giant grain fest.

Joe do you know of any examples of the GH5S with the NR at -5 where it still looked like plastic? Its almost impossible to find any actual image quality tests of the camera out there.
 
Last edited:
No noise reduction at all from sensor data makes no sense to me, I think you are better off cleaning sensible things up first than to send everything to the encoder and have the encoder 'smear' it out more unfavorably anyway.

Compressing 4K@30p 4:2:2 10 bit to 150Mb/s is a compression ratio of about 33:1, that is very high, that translates into maintaining 3% of information per pixel.
Note that due to the fact that most sensor noise is high frequency noise with high entropy and has no inter-frame motion it is not going to be compressed, it is going to be mostly thrown away during encoding. That's why encoding with for instance an Atomos at the best settings often shows a noisier image, a little bit more of that noise is maintained.

The plastic you complain about is due to DCT based compression, retaining noise will not change that, and potentially it will make it even worse for the encoder.
 
Last edited:
Cary Knoop said:
Compressing 4K@30p 4:2:2 10 bit to 150Mb/s is a compression ratio of about 33:1, that is very high, that translates into maintaining 3% of information per pixel.
Click to expand...

Interesting.

Any idea what the figures are when going to the higher 400AI bit rate? Is it a linear scaling?
 
stoneinapond said:
Interesting.

Any idea what the figures are when going to the higher 400AI bit rate? Is it a linear scaling?
Click to expand...
Inter-frame and intra-frame compression togetter compresses far better than intra-frame only but it won't change the compression rate. Rate wise 400 is 2 2/3 as much as 150.

To calculate you take 3840 x 2160 UHD (4K is slightly different) x 3 colors x 10 bits x 2/3 4:2:2 sub-sampling x 30 frame rate for the total bit rate per second.
 
Thomas Smet said:
Joe do you know of any examples of the GH5S with the NR at -5 where it still looked like plastic? Its almost impossible to find any actual image quality tests of the camera out there.
Click to expand...
No, i can't point to an online example that I know for sure used -5 ...but... I did do my own tests. The GH5s gets really aggressive with NR as the ISO increases, noticeably more so than the GH5, even at -5.

Regardless it is always there and that sucks. I would pay money to Panasonic to buy a firmware update to just totally tun off the NR on the camera. I can understand they want to appeal to a wide range of users including videographers but at least give us the option to turn it off.
Click to expand...
Part of the challenge here is how hard it is for a codec to deal with noise. Without any baked in NR, that the GH-series codecs would be severely taxed. You'd really have to go to an external recorder.
 
Cary Knoop said:
The plastic you complain about is due to DCT based compression, retaining noise will not change that, and potentially it will make it even worse for the encoder.
Click to expand...
Absolutely understand how compression effectively acts like NR...but...test this yourself with a live HDMI feed: the GH5, and especially GH5s, bake-in a lot of NR even before compression, no matter the setting.
 
With the ability of BM to record RAW. With ProRes RAW launching and being able to record to a v90 SD card. I wonder how the RAW signal out of a GH5S would fair. I wonder if Panasonic would ever enable this to compete with a successful BM camera.

I wonder if the GH relies so much on internal processing trickery and optical lens correction in some of the glass etc that showing the world a RAW signal would be something they are unwilling to do until a GH6 more tuned for such. Even if it made financial and technical sense today.
 
joe12south said:
Absolutely understand how compression effectively acts like NR...but...test this yourself with a live HDMI feed: the GH5, and especially GH5s, bake-in a lot of NR even before compression, no matter the setting.
Click to expand...

Makes sense. Just see how much better EVA-1 footage looks. No signs of temporal noise reduction, a rich looking image, no oversharpening either...
 
DeeZiD said:
Makes sense. Just see how much better EVA-1 footage looks. No signs of temporal noise reduction, a rich looking image, no oversharpening either...
Click to expand...
The EVA-1 is $7,345, the GH5 is $1,997.99.

You need to do processing to make what comes out of the sensor look palatable. For an expensive sensor that is doing "just a little" for a cheaper sensor that means "do a lot more".

I am certain that if Panasonic gives a firmware option that sends unaltered debayered data to the encoder the result will be far more unsatisfactory than when it is processed to make it look as good as possible.
 
filmguy123 said:
With the ability of BM to record RAW. With ProRes RAW launching and being able to record to a v90 SD card. I wonder how the RAW signal out of a GH5S would fair. I wonder if Panasonic would ever enable this to compete with a successful BM camera.

I wonder if the GH relies so much on internal processing trickery and optical lens correction in some of the glass etc that showing the world a RAW signal would be something they are unwilling to do until a GH6 more tuned for such. Even if it made financial and technical sense today.
Click to expand...

If the sensor is the same as the Pocket I would expect the raw to more of less look the same. It is the DSP and processing that make the look. Raw is raw and if two sensors are the same then the raw should be the same.
 
You must log in or register to reply here.
Back
Top