Monday, May 11, 2015

Samsung NX1 moiré comparison

Last month I did a sort of "first impression" review on the Samsung NX1, concentrating on its 1080p video and things I liked and did not like about the camera. I touched on places I was seeing the dreaded DSLR moiré tendency, which is one of the many reasons I'm not a DSLR fan in general. But as stated then, I can not deny how many producers have an appetite for cheap, reasonably descent video.

Again, my decision to buy the NX1 over other DSLRs stems from a specific project I have starting in June, shooting finished architecture for a construction company wanting to show samples of their work at presentations and on their web site. My original pitch was to document their buildings in UHD & 1080p video as well as high resolution stills for print or any other graphic application. This way they would somewhat future proof their accomplishments in formats of the future. How dynamic would it be for them to display their work at a trade show in UHD, for instance, setting themselves apart?

My original plan was to shoot Arri Amira at both 1080p and UHD, (as that capability is now available through licensing and clip re-calibration). But that approach proved to be out of the client's budget range so another solution had to surface. I needed a DSLR camera to shoot the stills at high resolution but also shoot UHD and 1080p video. The Sony A7s only shoots 12MP stills, so despite it's superior video capabilities, I had to rule it out (this time). The Panasonic GH4 could shoot very sharp 4K video internally at impressive 100 Mbps, but again the stills were only 16MP, not quite enough for architecture. And aside from the needs of this project, I wanted the camera to be useful for other projects as well.

The Samsung NX1 caught my eye because it has an impressively sharp sensor, shoots UHD & 4K internally as well as 1080p as H.265 files, which are impressively compact files that can be easily sent to clients over the internet and stored using far less space compared to drive hungry uncompressed codecs. Of course, the caveat is that H.265 currently cannot be used as an editing codec and thus must be transcoded into a codec that is. So there is a time and machine dedication requirement, that to some, might be objectionable in their workflow. In my case, I was willing to accept that caveat.

For the most part, relative to DSLRs, I like the video quality coming out of the NX1. It's very sharp and colors are pretty good for 4:2:0 color space internally. It's dynamic range is average for DSLR cameras, which seems to be acceptable to producers and clients wanting a low cost solution. It has the fastest 1080p rolling shutter performance for DSLRs. The biggest issue I have with the camera comes down to moiré performance. I am to shoot architecture of brick and mortar that create very thin horizontal and vertical lines that asks a lot of any video camera's sensor, especially DSLRs. Moiré has been a very difficult tendency to overcome. Some DSLRs place anti-aliasing filters over the sensor to help minimize this problem but don't seem to eliminate it in all modes of shooting. The Samsung NX1 has no such filter, thus it produces ultra-sharp images but suffers from moiré under certain visual conditions. Changing the sharpness setting in the camera might tighten moiré but does not eliminate it in challenging situations.

So to understand what I was in for on the architecture project, I had to test the NX1 and see what it was capable of. At the very least, I expected I could use the NX1 for the stills (in RAW mode only) but could I shoot the video with it? 

Thus today's review is centered around a moiré comparison between the Samsung NX1, in both 1080p & UHD modes, and the Sony EX1 in 1080p mode and Arri Amira in 2K mode. I realize this is kind of an odd comparison. Ideally, I'd want to compare the NX1 against the A7s and GH4, in UHD and 4K modes, DSLR cameras with 4K capabilities. But that's not what I needed to know. I own the Samsung NX1, Sony EX1 and the Arri Amira (currently up to 2K) only, and one of these cameras would provide the video I needed to deliver an acceptable 1080 final product now. For 4K delivery, being secondary to my client, I could simply upgrade my Amira for 4K acquisition for less money rather than buying another DSLR that may or may not handle moiré better.

Again, for what I'm seeing in reviews, the GH4 and A7s still suffer from moiré to some degree in some modes and neither provided the still resolution I needed.

My first architecture subject is a high school with a red brick facade, fine tight lines, that would prove to be the biggest challenge for the camera sensors. I shot from a slider on the ground with mostly side-to-side motion to see how moiré would appear in movement. I shot at different distances from the building knowing the further away I got the harder it would be for the sensor to interpret the detail. I was using my Canon Cinema Zoom 30-105mm on both the Samsung NX1 and Arri Amira at 30mm (both having similar crop factors). I also used Canon FD 24mm photo prime on the NX1 to see if lens choice affected the moiré significantly.

The Sony EX1 fixed lens had a wider angle of view, and it's ability to resolve detail would be much less, given lens quality and smaller 1/2" sensor. I shot two different days. Samsung NX1 in 1080p and UHD modes, both lenses, day 1, (which was sunny), and the Sony EX1 in 1080p mode and Amira in 2K mode day 2 (which was overcast). All video was shot 29.97 fps.

1080 frame grab

1080 frame grab from down converted UHD
1080 frame grab from fit-to-scale 2K

1080 frame grab

Obviously these frame grabs do not tell you much about moiré, but can show a little bit how well the lenses resolve detail. You have to see the video in motion and the link is at the bottom of this article. It is worth noting that moiré can appear differently on different monitors. For instance, what I saw on my 24" Dream Color monitor was not the same as what I saw on my 27" View Sonic, playing the same H.264 file at full size. What you see on your monitor could be very different. My observations are based on what I saw on the Dream Color, this being the more accurate display. Note the UHD images were placed on a 4K timeline (thus the black border) and then exported to a 1080 file, which was placed on a 1080p timeline along with the other camera files. I did not wish to scale it up first to 4K for this test and I don't have a UHD preset option within Premiere Pro CC 2014. The Amira, shot in 2K 4444 mode using C-log and corrected, was scaled-to-fit within 1080.

I did not worry too much about leveling the slider, and I will be using a wider angle PL lens on the actual shoot day in June.

In 1080p mode, the Samsung certainly suffered from moiré with either Canon Cine Zoom or Canon 24mm FD photo prime, but not as badly as I would have expected from a camera with no anti-aliasing filter, but not acceptable either. However, in UHD mode (which was down-converted to 1080) moiré was far less, almost gone. This was very encouraging to see! This being, what I expect is the worse case scenario, the sensor handled it pretty well; I would say acceptable. And for the speed of motion I am planning, rolling shutter is not an issue in UHD mode.

In 1080p mode, the Sony suffered moiré less than the Samsung in 1080p mode. Though, as expected, the fixed lens could not resolve the detail as well as the Samsung using the Canon Cine Zoom or even the FD 24mm prime. So even though there was slight moiré evident, I could see using this camera for all the 1080p video once I send the 10bit 4:2:2 SDI output to my AJA KiPro Mini. It will gain color fidelity and some sharpness while still matching the project budget level.

As expected, the Amira outperforms them all with virtually no sign of moiré at any distance. The image is as good as it can get, but the weight of the camera on this slider limits tilt movement; and the class level of the camera doesn't match the client's budget. But it served as a great comparison and benchmark. I'm anxious to see how this camera will handle the same test in UHD when I have that upgrade. (blog for that to come)

For the budget of this project, I am content with the Samsung NX1 shooting UHD for the future-proofing file, plus all the high resolution stills in RAW; and the Sony EX1/AJA Ki Pro combo for all the 1080p video needed now. I even have the option to down-convert all the Samsung UHD files to 1080p and use them instead of the Sony files, should I find a much better level of resolved detail. So I will be switching out cameras from the slider on every setup, one to record UHD and the other 1080p. And should the client decide to up the image quality down the road, I will have my Arri Amira 4K ready.

My last task will be to play the Samsung UHD architecture footage out on a 4K monitor to see how that really looks. The only reason I'm not considering using the true 4K out of the NX1 is because of the reported poor rolling shutter performance in that mode.

Again, moiré evidence seems to vary from monitor to monitor. I can't say there won't be moiré on whatever monitor the client views. It's not just how well the camera sensor can handle faults detail, but also how well a monitor can handle it. All I can do is confirm that the camera sensor was able to produce (in some recording mode) a low or non-incidence of moiré in a quality monitor, much the way we access color. We use color accurate monitors to grade by but can't say what lesser displays will do to that image.

I feel I will be able to provide my client reasonable image quality in both 4K and 1080p and stills at the price level they were looking for.

Here's the link to the video.

As an added bonus, I have uploaded the lens comparison between the Canon Cine Zoom 30-105mm, set at 30mm, and the Canon FD 24mm photo prime mounted on the Samsung NX1. This shows how different the high-dollar Cinema Zoom can resolve detail over the photo prime FD. This is why I'll be using a PL lens for the real shoot day. Also worth noting, in this instance I took the original UHD files and place them directly on the 1080 timeline and scaled down to fit. Although this is a direct way of dealing with UHD footage, it is more difficult for the HP Z800 to work with compared to down-converting first. Here's the link.

No comments:

Post a Comment