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

[filmguy123]

Now that it's been a while, has there been much analysis or thought on where the GH5 actually gets outshined by the GH4?

As I look back on footage, some of the GH5 footage can have a way of looking a bit more plasticky than the GH4 footage, which can tend to look a bit rawer (in a good way).

This might just be settings... but appears to be when both cameras are set to -5 / -5 Sharp/NR and on same (non native, non electronic) glass.

Anyone else notice this?

 

[Thomas Smet]

I think the GH4 was nice with its 1:1 pixel crop of the sensor. The GH5 has insane detail but it also doesn't have a OLPF and in rare cases can display some minor moire or aliasing. The GH4 has a OLPF and used a 1:1 portion of the sensor which is about as clean as it gets for a moving image.

Thats one reason why I really want to get a GH5S vs the GH5. None of the reviews out there cover this since they are all obsessed at testing the low light over, and over, and over, and over, and over again. How many Youtube tests do we really need to show the different ISO levels?

I think the GH5S could eliminate some of the overly video look criticisms the GH5 has because it once again has an OLPF and a 1:1 pixel readout. All while no longer having an even larger crop factor. We have been so focused on the low light and kind of ignoring the other areas we would like to see improved on the GH5. The GH5 gets around the lack of OLPF by using a very good method of scaling the full down to 4k video size. This acts as a filter in a way but its not perfect.

So from that perspective its possible the GH4 or GH5S could be a better choice. Other than that I can't really think of anything I would consider better on the GH4.

 

[filmguy123]

Ahh interesting, good insights Thomas.

True, the GH5 is so much better in every regard I couldn't bring myself to pull out the GH4 for anything now except a backup camera. But I think you may be onto something - I noticed right away that I prefered the GH5S image (vs the GH5) in a way that went beyond lowlight. It just has a better look that's hard to put a finger on. (And I agree, outside of lowlight, there's virtually nothing of value online to compare GH5 vs GH5S).

This is all subtle, for sure, but I've been reviewing footage from last year and I stacked up my favorite GH4 vs GH5 shots and the GH5 just looks overly smooth in comparison. Actually, out of my top selects, if I rank them, *all* of my favorite shots are GH4 first, followed by GH5. It's just the nice raw rendering of the image actually looks a bit more filmic. When you get the GH4 right, some of the images with manual glass really have a great, organic look VS the GH5. I also wonder if the GH5 is doing a bit more to clean up noise that can't be turned off vs the GH4?

I haven't shot much with the GH5S yet, just some tests, but I'm going to pay attention to this.

What is the pixel read out of the EVA1 with it's 5.7k oversampling? And, when it moves to 4K crop mode (such as in 4K RAW), does this mean it becomes a 1:1 readout?

I know OLPF helps with Moire as well as with IR pollution, but what exactly is it, and how else does it effect the image (such as in the instance we're talking about here)?

Thanks for the thoughts and insights! I realize this will remain largely a mystery, but it's fun at least trying to understand what may account for the differing image characteristics.

 

[Thomas Smet]

Ahh interesting, good insights Thomas.

True, the GH5 is so much better in every regard I couldn't bring myself to pull out the GH4 for anything now except a backup camera. But I think you may be onto something - I noticed right away that I prefered the GH5S image (vs the GH5) in a way that went beyond lowlight. It just has a better look that's hard to put a finger on. (And I agree, outside of lowlight, there's virtually nothing of value online to compare GH5 vs GH5S).

This is all subtle, for sure, but I've been reviewing footage from last year and I stacked up my favorite GH4 vs GH5 shots and the GH5 just looks overly smooth in comparison. Actually, out of my top selects, if I rank them, *all* of my favorite shots are GH4 first, followed by GH5. It's just the nice raw rendering of the image actually looks a bit more filmic. When you get the GH4 right, some of the images with manual glass really have a great, organic look VS the GH5. I also wonder if the GH5 is doing a bit more to clean up noise that can't be turned off vs the GH4?

I haven't shot much with the GH5S yet, just some tests, but I'm going to pay attention to this.

What is the pixel read out of the EVA1 with it's 5.7k oversampling? And, when it moves to 4K crop mode (such as in 4K RAW), does this mean it becomes a 1:1 readout?

I know OLPF helps with Moire as well as with IR pollution, but what exactly is it, and how else does it effect the image (such as in the instance we're talking about here)?

Thanks for the thoughts and insights! I realize this will remain largely a mystery, but it's fun at least trying to understand what may account for the differing image characteristics.

This is one of the reasons why I held off upgrading my GH4. That and a massive case of indecision made even worse now by the Pocket 4k.

The GH4 has its fair share of faults but it really doesn't get any better than a 1 to 1 sensor to pixel image. Over sampling is one of those things that sounds sexy in theory but in practice yields too digital of a looking image. The GH4 can create some stunning material. I'm just getting tired of always having to use an external recorder in order to shoot V-log.

The GH5 does have ETC which will do a 1:1 but that sensor doesn't have an OLPF so you can still end up with some moire and aliasing. If one can live without IBIS the GH5S really is the better all around video production choice of camera. Thats why I'm not entirely convinced the Pocket and GH5S go head to head directly. For film makers they do but the GH5S is not meant for just fil making and is more of a videography camera with cinema features thrown in.

I didn't know you bought a GH5S. Man you have a lot of cool toys.

Moire and aliasing is caused by high frequency detail. Thats why it shows up in fine patterns like bricks or roof tiles in the distance. The finer the detail the more potential for aliasing to show up. An optical low pass filter OLPF basically filters out some/all of that high frequency detail and passes through the lower frequency detail. Its really just a way to soften the image so there is no such thing as fine detail to make the image freak out.

http://www.xdcam-user.com/old-stuff/aliasing-and-moire-what-is-it-and-what-can-be-done-about-it/

 

[filmguy123]

This is interesting and likely explains most of my complaints on certain cameras. I really like the EVA1 and it's miles better and smoother than the GHx line, really fantastic. But on the zacuto shootout, I have to say, I tended to prefer the rendering of the C200. Not talking about color. Just the edges. They're smoother. I haven't read up much on this but I'm going to go out on a limb and guess the C200 is a 1:1 readout, vs the over sampling of the EVA1? Same thing, eh? Though, the EVA1 has the OLPF as I understand, so it seems that helps VS the GH5 oversampling issue...? (As I understand, both EVA1 and GH5 are around a ballpark 5.7k sensor?)

I need to test this more, but if so, I may find that I end up preferring using the EVA1 in 4K Crop mode for RAW output. It takes a 4K crop of the 5.7k sensor, thus ignoring the oversampling, and sends that out to either RAW or Video. You have to move to working in m43 crop mode but that doesn't really matter to me as I'm so used to it, and don't tend to love shooting much wider than 24mm (in terms of FF FOV) 90% of the time anyway.

The GH4 can create some stunning material. I'm just getting tired of always having to use an external recorder in order to shoot V-log.

Tangent, check out the thread here on run'gun solutions with Atomos. I've been quite pleased with what i've setup for that - Thinktank App House 10, throw that on a belt. Then attach a smallrig cage to the atomos with a shoulder strap, so it stays secure and can be pulled out of the bag one handed with no fear of a drop. Run your HDMI cable to the GH4, use a right angle HDMI adapter as needed to help. This will free up the GH4 to be mean and lean. Course, you have to use the LCD screen, but I don't think that's as bad as some people make it out to be. I shot dozens of hours of footage on it last year, run and gun, and didn't miss focus once using the focus peaking highlights. It's like flying with instruments. Just ETTR and trust the peaking, roll back and forth a couple times, all good. Well, YMMV, worked for me But I hear you. Or maybe the Ninja V would just solve that whole issue, more likely!

I didn't know you bought a GH5S. Man you have a lot of cool toys.

It's a disease, lol. No not really, just a lot of GAS the last year. Lucky to land some larger jobs and decided to invest back into the business. I say "business" but I really mean life long dream since a kid, so I just bought myself toys. Let's be honest. But at the end of the day, I do film for me, because I love it. So yeah, I could have bought a boat instead. But I'm still playing filmmaker. Hopefully they'll serve to elevate my work to the next level... it's the obsession that got me here in the first place, so, feet don't fail me now...

Ok so, looping back around. Why is it that over sampling, even WITH an OLPF, creates too digital of an image? Is it just because there's so much detail, or is it something in the actual process itself? And, does this translate if for example you were to shoot 5.7K RAW off of the EVA1 - no in camera oversample - and then just simple down-res in post?

I really need to look into this, because the 1:1 sensor/pixel read out may have been one of the key things all along that I've never been able to put a finger on that's always driven me a little crazy trying to figure out what I do or don't like on an image.

 

[TheDingo]

Moire and aliasing is caused by high frequency detail.

In almost 3 years of shooting with my GH4, I've had moire show up only once on person's shirt collar in a wide shot. It was pretty minor, and you really had to look for it to notice it. Over the past year of shooting with my GH5, I've only seen moire once, and again it was in a wide shot. It was also pretty minor, and something you would not spot unless you hunted for it.

Otherwise, I find the GH5 video image significantly better than my GH4 camera. Less noise, noticeably better color, better contrast in the overall image, especially the shadows. Being able to shoot entirely 10-bit footage opens up way more room in post to work the image, make corrections, etc...

I stopped using my GH4 cameras as soon as I saw GH5 footage from my first shoot with it. The difference was huge for me.

 

[Mark Williams]

I own both the GH4 and GH5 and like the image quality of both. What is great is someone with a low budget can buy the GH4 new at a ridiculous low price now and have a really decent camera.

 

[wwjd]

CAN'T STAND Moire!! It's 2018.... HD... 4k..... 8k... HOW is MOIRE even still a thing??? (obviously, filming certain PATTERNS will create some natural digital moire, but I mean from the sensor itself pixel binning and crap) JUST SAY NO!!

 

[Thomas Smet]

CAN'T STAND Moire!! It's 2018.... HD... 4k..... 8k... HOW is MOIRE even still a thing??? (obviously, filming certain PATTERNS will create some natural digital moire, but I mean from the sensor itself pixel binning and crap) JUST SAY NO!!

Ironically a lot of the high end cinema cameras really suffer from moire which is why companies have stepped in and created after market OLPF. I think in our quest for super high detailed images we lost touch of what is best for a moving image.

Thing with moire is it doesn't always show up and 90% of the time these cameras deliver amazing results. Its that 10% that bothers me. I agree cameras should no longer exist without an OLPF. I don't want to fart around and add one after market. These cameras that strive for lossless perfect quality are missing the mark by allowing such optical artifacts to exist. Robbing Peter to pay Paul. I'm sure a lot of people don't mind adding an OLPF and thats perfectly fine but I would feel much better if the manufacturer did the R&D and built their own specifically optimized for that particular sensor.

With that said if I do decide to get a BMD camera I will have to suck it up and get a OLPF as well. In my mind the camera just doesn't exist without one.

 

[Thomas Smet]

In almost 3 years of shooting with my GH4, I've had moire show up only once on person's shirt collar in a wide shot. It was pretty minor, and you really had to look for it to notice it. Over the past year of shooting with my GH5, I've only seen moire once, and again it was in a wide shot. It was also pretty minor, and something you would not spot unless you hunted for it.

Otherwise, I find the GH5 video image significantly better than my GH4 camera. Less noise, noticeably better color, better contrast in the overall image, especially the shadows. Being able to shoot entirely 10-bit footage opens up way more room in post to work the image, make corrections, etc...

I stopped using my GH4 cameras as soon as I saw GH5 footage from my first shoot with it. The difference was huge for me.

Totally agree. The GH5 shouldn't have a lot of moire either because of how they down convert the sensor with a bi-cubic type of scaling. This acts as a sort of filter and for the most part takes the place of an OLPF.

Only time I have seen the GH4 have potential moire is when using an overly sharpened electronic lens. I personally have never seen a single shot have aliasing and moire but I also shoot at -5 sharpness and that might help a bit. Edge enhancement can exaggerate aliasing and moire.

My initial post on this isn't that the GH5 can have more moire but that it is an overly sharp image due to the oversampling and thats something the film community is trying to get away from. The GH5 went a little too detailed for some people and there is no way to dial that back down without a brute force blur in post. A 1:1 sensor can yield a little bit more natural detail and transition between each pixel vs heavily image processed detail. Yeah its extreme hair splitting but its a thing and one of the reasons why the GH5 gets labeled as a "video" look.

I think the GH5S could be a little bit more natural looking for some of the film folks on here. Maybe not down to the BMD look but somewhere in between. If we are striving for more of a raw sensor like image then having over sampling is not the way to do that because no raw image would ever look like that. A raw image is a 1:1 image and therefore any GH camera that allows one to get that same type of 1:1 readout will get a bit closer to what the sensor can see. Thats one area where the GH4 and GH5S could have a slight advantage over the GH5.

 

[TheDingo]

CAN'T STAND Moire!! It's 2018.... HD... 4k..... 8k... HOW is MOIRE even still a thing???

Pretty much every camera can show moire under specific conditions, but the GH4 and GH5 have been the cameras with the least amount of moire for me. With the GH4 you MUST shoot in 4K because 1080p on the GH4 is a moire honey-trap. The over-sampling on the GH5 takes care of moire for both 4K and 1080p formats.

I am hoping that the BMPCC 4K camera has low moire, because the original BMPCC was prone to moire as it did not have an OLP filter. ( the GH5s doesn't have one either, but it seems to be ok )

 

[Thomas Smet]

Pretty much every camera can show moire under specific conditions, but the GH4 and GH5 have been the cameras with the least amount of moire for me. With the GH4 you MUST shoot in 4K because 1080p on the GH4 is a moire honey-trap. The over-sampling on the GH5 takes care of moire for both 4K and 1080p formats.

I am hoping that the BMPCC 4K camera has low moire, because the original BMPCC was prone to moire as it did not have an OLP filter. ( the GH5s doesn't have one either, but it seems to be ok )

Pretty sure the Pocket will not have a OLPF. Thats one way BMD keeps the cost down and it goes outside of their norm to suddenly add one. I hope they do but if they did I'm sure they would have marketed the daylights out of that spec. I may still get a Pocket but not until a reputable person makes an OLPF specifically for that camera. That could be good year away so it looks like I may invest in a GH5S instead. Any Pocket fans want to sell their GH5S?

 

[Tom Roper]

None of the reviews out there cover this since they are all obsessed at testing the low light over, and over, and over, and over, and over again. How many Youtube tests do we really need to show the different ISO levels?

I was just thinking this, no beauty shots from the GH5s, or not many. Mostly lots of flat lighting, low light demonstrations with all the interest of an unboxing video.

 

[filmguy123]

I really like the EVA1 and it's miles better and smoother than the GHx line, really fantastic. But on the zacuto shootout, I have to say, I tended to prefer the rendering of the C200. Not talking about color. Just the edges. They're smoother. I haven't read up much on this but I'm going to go out on a limb and guess the C200 is a 1:1 readout, vs the over sampling of the EVA1? Same thing, eh? Though, the EVA1 has the OLPF as I understand, so it seems that helps VS the GH5 oversampling issue...? (As I understand, both EVA1 and GH5 are around a ballpark 5.7k sensor?)

I need to test this more, but if so, I may find that I end up preferring using the EVA1 in 4K Crop mode for RAW output. It takes a 4K crop of the 5.7k sensor, thus ignoring the oversampling, and sends that out to either RAW or Video. You have to move to working in m43 crop mode but that doesn't really matter to me as I'm so used to it, and don't tend to love shooting much wider than 24mm (in terms of FF FOV) 90% of the time anyway.

Why is it that over sampling, even WITH an OLPF, creates too digital of an image? Is it just because there's so much detail, or is it something in the actual process itself? And, does this translate if for example you were to shoot 5.7K RAW off of the EVA1 - no in camera oversample - and then just simple down-res in post?

I really need to look into this, because the 1:1 sensor/pixel read out may have been one of the key things all along that I've never been able to put a finger on that's always driven me a little crazy trying to figure out what I do or don't like on an image.

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.

 

[Thomas Smet]

I was just thinking this, no beauty shots from the GH5s, or not many. Mostly lots of flat lighting, low light demonstrations with all the interest of an unboxing video.

I think the community has really done the camera a disservice by ignoring testing it as a all around camera. Everybody is so focused on the low light they have ignored tradability, sensor artifacts, detail and the overall electronic vs organic look of the camera. I understand a lot of people are obsessed with shooting in the dark but I personally don't know a lot of serious corporate or narrative work that doesn't use some kind of lights. I love the dual native ISO to make zooms more useful but I still never plan on going above ISO 6400 with the camera and even then its really when I have no other choice at all.

Every review is only focus on low light, bad auto focus, lack of IBIS and the low quality stills. Every single thing else has been ignored even though the camera has a whole new sensor we haven't seen before.

Honestly if the BMD is using the same sensor and call pull off the look then so can the GH5S.

 

[Thomas Smet]

Re: Moire/Aliasing:

https://www.nikonusa.com/en/learn-and-explore/a/products-and-innovation/moiré-false-color.html

This article offers insight and includes a few tips that can mitigate the issue while shooting. I personally found it helpful so I'm putting that here for others.

As others mentioned, it's tough to spot while shooting, but once you become aware of how your camera reacts to certain common patterns while using a certain lens at a specific distant from your subject/object, then using some of the techniques outlined in the article can help. For instance, there are times when I'm shooting with a hella sharp lens at the wide end and the artifacts are even visible on the monitor, but simply physically nudging the sticks forward by a mere inch or two will eliminate aliasing. It's important to have an intimate relationship with your gear, which is why I prefer to be an owner/operator in lieu of renting.

Or actually use a professional camera with a OLPF and not worry about hacks and tricks.

 

[Thomas Smet]

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.

Not really much to do with the moire and aliasing. My point about the oversampling is making a more digital share image. Thats kind of the entire point to over sampling. You get finer and higher detail than you would with a straight sensor readout. If some film makers don't like the insane detail at even -5 sharpness and adapter lenses then right there is the reason why. No electronic enhancement of the lens, no added sharpness. Just a sensor creating way too much natural detail.

We had the same problem when HD first came out and we needed to deliver DVD. Down converting the HD to SD created some insane level of details that were impossible to achieve on a SD camera. It was a massive amount of over sampling.

The moire and aliasing was a secondary discussion of things to look out for when there is no OLPF. How much the GH5 down scaling fixes the OLPF is hard for me to say but I have seen some users have some pretty ugly results. I do think it is extremely rare however. What might actually be worse on the GH5 is using ETC since its a 1:1 readout of a sensor with no OLPF. Not any worse than a BMD or other cinema camera with no OLPF but more potential to have issues. How often those artifacts pop up largely depends on how/what you shoot. You could shoot for years and not get any artifacts and then suddenly have a fashion shoot where half the shots are ruined. I personally like to use a camera where I don't have to worry about it but thats me.

It is true the GH5 doesn't have a OLPF but the scaling really does act like one so for all intents and purposes when shooting video it kind of does have an OLPF. Its when shooting full resolution stills on the GH5 that it has no OLPF.

 

[Thomas Smet]

In terms of the sharpness on the GH5 I believe when one adjusts the sharpness to -5 that is done before the scaling and image processing take place. Thats why even at -5 the sharpness may still seem very high to some users. -5 lowers the enhancement on the sensor image but when you down convert that to 4k you bump up that detail once again from the scaling process. Scaling can also introduce edge enhancement which is why programs like Fusion provide different scaling methods. Some scaling will enhance the edges more and it is possible the GH5 does create some slight edge enhancement on the scaled image. It might be better on the GH5 to shoot 6k and scale in post with a method that introduces less detail and less chance of edge enhancement.

Scaling uses similar math to how sharpness works and is all about taking a grid of pixels and figuring out a way to make a smaller grid of pixels by averaging pixels together. Thats just a ton of processing taking place which means the GH5 can never really yield a raw looking image even with the compression not factored in. It is that raw image before compression that is already heavily processed and enhanced.

For users that love detail land the video look the GH5 will be stunning. For users who prefer more of an organic look the Gh5S may yield a better look. Wish somebody would compare this between the two cameras since I'm basing this all off of theory and my understanding of image processing software development.

 

[TheDingo]

That's kind of the entire point to over sampling. You get finer and higher detail than you would with a straight sensor readout.

And it compensates for the resolution loss from the debayering process. The EVA1 5.7K sensor debayered down to 4K is supposed to achieve full 4K resolution from the 5.7K sensor.

 

[Thomas Smet]

And it compensates for the resolution loss from the debayering process. The EVA1 5.7K sensor debayered down to 4K is supposed to achieve full 4K resolution from the 5.7K sensor.

Thats the theory at least.

Not sure if it equals what a 4k 3 chip design could achieve but we will likely never know since 3 chip design is pretty much dead. Kind of ironic that we used to shoot with chip designs that were actually capable of true pixel accurate RGB and yet didn't have the compression methods to make that happen. 4:4:4 is now more of a reality but we no longer have a sensor design capable of a true pixel accurate 4:4:4.