Those terrible Samsung sensors (or are they?)

Forums: 
Here is the Samsung NX20 20 MP sensor compared to the Fuji XP1 both @ ISO 3200, f/4, 1/60. I used my typical back light weak front light scene with the baby doll (Pizza delivery guy commented on it when he dropped off the food). These scene is typical of a low light restaurant and street shooting. A situation where you would be using ISO 3200. Both manual exposure. I bracketed ISO settings just to make sure both cameras were exposing similarly (they are). I used the 30 mm f/2 on the Samsung and 35 mm f/1.4 on the fuji, which is why the Fuji has a bit more magnification (I left camera in same position for test). Since the Fuji is zoomed in more it has an advantage of appearing to have more detail (this is not a detail test so I wouldn't go by that, I didn't attempt to make the focus plane identical, but did focus on the same spot on subject). Outputs were cropped to the same sensor area, not the same number of pixels. This prevents the higher MP on the Samsung from penalizing it (you don't view images at the pixel level). So they were cropped 870x580 for the Samsung and 778x519 for the Fuji, Lightroom does this automatically when you sync the crop. My thoughts going into this test were that I would see a large advantage to the Fuji, and while this is partially true looking at the RAW data with default settings, the Samsung appears to respond better to noise reduction in Lightroom than the Fuji. I set the values where I liked, if I went much higher with the Fuji it got plasticky. The bottom line is the Fuji is not significantly better in RAW at high ISO when some noise reduction is applied to both cameras. Out of camera JPEG is definitely going to favor the Fuji at ISO 3200. I am not trying to prove anything here (if anything was trying to justify the purchase of the XP1 I got for a good deal used), but what I found was, and as I have found before, the attacks on the "Terrible samsung sensors" are not founded. Granted, Lightroom is not the ideal RAW converter for Fuji yet, but it might not be for Samsung either, and noise levels should be impacted much either way. I picked the Fuji as there are several fervent Fuji fans that think it is the best APS-C sensor for high ISO (and in JPEG it might very well be the best, but I don't shoot enough JPEG to know or care). Overall Shot IMAGE(http://dpzen.com/dpzattaches/dpzattachesi//201409231846003154.jpg) Samsung with some noise reduction original sized image: http://farm9.staticflickr.com/8047/8149760077_46ac0b0c08_o.jpg IMAGE(http://dpzen.com/dpzattaches/dpzattachesi//201409231846023155.jpg) Fuji with some noise reduction original sized image: http://farm9.staticflickr.com/8054/8149760629_25823d4d1e_o.jpg Now some crops: IMAGE(http://dpzen.com/dpzattaches/dpzattachesi//201409231846053156.jpg) Samsung no noise reduction. IMAGE(http://dpzen.com/dpzattaches/dpzattachesi//201409231846073159.jpg) Fuji no noise reduction IMAGE(http://dpzen.com/dpzattaches/dpzattachesi//201409231846103160.jpg) Samsung Noise reduction 29, detail 47, color 31, detail 50 (I went with what I liked best) IMAGE(http://dpzen.com/dpzattaches/dpzattachesi//201409231846133161.jpg) Fuji with Luminance 8, detail 62 (I just went with what I liked best) IMAGE(http://dpzen.com/dpzattaches/dpzattachesi//201409231846153164.jpg) Samsung no noise reduction (noticeably grainier) IMAGE(http://dpzen.com/dpzattaches/dpzattachesi//201409231846173167.jpg) Fuji no noise reduction IMAGE(http://dpzen.com/dpzattaches/dpzattachesi//201409231846193169.jpg) Samsung with Luminance 35, detail 44 (again, just what I happened to like) IMAGE(http://dpzen.com/dpzattaches/dpzattachesi//201409231846203171.jpg) Fuji with Luminance 18, detail 50 (again, what I happened to like)

I studied the noise under the doll's chin on each image before NR, and the fuji one looks a little better to me, but after you have done your PP tidy up, I prefer the samsung one. I think it has a smoother gradiant from light to dark, this can be seen on the tip of the nose where dark meets light. So I don't think there is much in it, and I would not place much weight on the sensor if I was choosing between the two. Brian

The samsung shot looks underxposed by -1/3 ev the curtains in the background (left side of the frame) and the doll's eyes are the most noticeable areas. IMHO People shouldn't sweat the noise, the more processing to get rid of noise the more the image looks "plastic fantastic"

SunnyFlorida wrote: The samsung shot looks underxposed by -1/3 ev the curtains in the background (left side of the frame) and the doll's eyes are the most noticeable areas. IMHO People shouldn't sweat the noise, the more processing to get rid of noise the more the image looks "plastic fantastic"
Just make sure you are comparing in the same spot, the samsung has a wider field of view so the corners are darker because of light falloff from the fluorescent. I do think it is slightly darker, but when I compared 1/3 stop EV apart it was a bit too much. Agreed, most of the current sensors are very good. Eric

brianj wrote: I studied the noise under the doll's chin on each image before NR, and the fuji one looks a little better to me, but after you have done your PP tidy up, I prefer the samsung one. I think it has a smoother gradiant from light to dark, this can be seen on the tip of the nose where dark meets light. So I don't think there is much in it, and I would not place much weight on the sensor if I was choosing between the two. Brian
Exactly my thoughts, at least if you shoot RAW. I noticed the gradient as well. For some reason the noise reduction the Samsung does isn't great. APS-C sensors are all fairly close. Technological improvements make small differences, and their are certainly sensors that might not perform as well, but this one seems good enough. Same with Canon, etc. Eric

Very interesting finding. Thank you !

I don't think anyone is questioning Samsung's ability to make a good product. The new sensor is good and their lenses are quite good as well. In terms of product, I really like Samsung. I still use my NX100 and like the images I get from it, either with Samsung lenses or adapted lenses...so for me I've never really questioned Samsung from an image quality perspective. My difficulties with Samsung stem from how they seem to be developing their NX line. I have very little confidence in Samsung from that perspective, not from the actual products themselves.

crsantin wrote: I don't think anyone is questioning Samsung's ability to make a good product. The new sensor is good and their lenses are quite good as well. In terms of product, I really like Samsung. I still use my NX100 and like the images I get from it, either with Samsung lenses or adapted lenses...so for me I've never really questioned Samsung from an image quality perspective. My difficulties with Samsung stem from how they seem to be developing their NX line. I have very little confidence in Samsung from that perspective, not from the actual products themselves.
Yes, agree, totally different issues. I only brought this up is I will see posts on rumor sites, etc, saying how bad Samsung sensors are, etc. They really are good enough, just need a little better processing from the camera (which has improved a lot) and some refinements, etc. Listen to enthusiasts more, etc. You are right, I would like to know what direction they are taking the system, especially from an enthusiast point of view. Fuji is fairly clear for the next year or so at least, and they are definitely listening to their target audience. Eric

Here are the original RAW files if you want to play with your own RAW developer or noise reduction settings. These are huge, only download if you really want to use them (32 MB for the Samsung and 26 MB for the Fuji): http://erphotoreview.com/RawFiles/SAM_8009.SRW http://erphotoreview.com/RawFiles/DSCF7341.RAF Eric

I appreciate you posting some samples without NR. I'm against all forms of NR and to see how much detail the Samusung sensor retains at high ISO is astonishing.

Fuji has RAW based NR. They are doing it since X100 as far as I remember. Thats why even if you switch NR off and process RAW file, you still have wierd blotches. Cause it does NR on both chroma and luminace noise. Samsung obviously has RAW NR too, but Samsung was clever, no luminace noise reduction = much more resolution at cost of bit of noise. Ofc colors will suffer, but on other hand at ISO 3200 colors will suffer either way, no matter if you clean RAW or just do NR later in PS. Hope you didnt expect those RAW files to be actually "RAW". Whats important is, that Samsung is still "the clever" one. Btw. they sorta inherited this from Pentax (they do same since K20D or so I think). Oh and thanks for RAWs.

viking79 wrote: Here is the Samsung NX20 20 MP sensor compared to the Fuji XP1 both @ ISO 3200, f/4, 1/60.
Super, thanks. Matches my observations. I prefer a more natural "grainy" look.

thanks Eric. I'm just wondering if you used the 4.2 version of LR? I have to say that using LR for the XP1 might not the best way for evaluating raw noise and noise reduction. for some reason adobe seems to have some problem on the matter of displaying noise signal in the DNG conversion. the better converters for the XP1 raw files are the SilkyPix, RPP and DCRAW. although it is surprising to know that you can pull a lot of detail from the XP1 using ACR which is of course unrelated. I have to look at the files myself and see the degree of difference. so far, I would agree that the noise difference is minimal as far as using LR version ? is concerned. the Samsung sensor is definitely not bad but just needs some fine tuning or some upgrading with the image processor if it wants to go side by side with the Fuji. 20MP + NX lens gives a heck of a detail (again unrelated), but could be better with better NR engine or converter. it is also true however that the XP1 Jpegs are unrivaled as of the moment, so that is one area the XP1 makes it's bearing and one that Samsung needs to improve on. besides, it helps a lot that you can do continuous shooting without fearing delays and having very good results immediately. as far as people that are saying bad things about Samsung sensor, I'm not sure what they are talking about nor they know what they are talking about. the only possible thing that I know of is that the other sensors have a bit more DR headroom at base and lower ISOs. but that doesn't make the NX sensor a terrible sensor. but having that certainly would make the NX a truly truly truly fantastic one.

ok, I just finished downloading the 2 files. I upgraded my LR to 4.2 just recently and I'm not sure if there is any impact with regards to noise amount during DNG conversion but so far here is what I did and saw. upon conversion, I removed both NR application on both images and focused on the dark background area which I think is a door. what I noticed is that there is definitely a significant amount of RAW noise difference. I have to say this but the Fuji is even way better than any of the best APS-C camera out there, including the current dslrs. I think if there was a camera that has been way overrated than what it actually does, it's not the Fuji but the NEX-5n.

Ariston wrote: ok, I just finished downloading the 2 files. I upgraded my LR to 4.2 just recently and I'm not sure if there is any impact with regards to noise amount during DNG conversion but so far here is what I did and saw. upon conversion, I removed both NR application on both images and focused on the dark background area which I think is a door. what I noticed is that there is definitely a significant amount of RAW noise difference. I have to say this but the Fuji is even way better than any of the best APS-C camera out there, including the current dslrs. I think if there was a camera that has been way overrated than what it actually does, it's not the Fuji but the NEX-5n.
Well, try both with noise reduction and see how that goes, it seems to me that at high ISO the NX20 is definitely noisier, but has more detail, and maybe responds to noise reduction better (with a corresponding drop in detail). Also, the NX20 loses a bit to having more pixels if you are comparing at the pixel level, so I would expect maybe 1/3 stop from that. I think the other issue is that without noise reduction, the Fuji has the advantage of having more green pixels in the RAW (but fewer red and blue?) so it won't show luminence noise as much. Also, the pseudo-random pattern will probably help with making the noise look more random. Here are a couple from ISO 6400 from both the Fuji and 20 MP Samsung, different times, different areas, but both have visible noise at this level. I probably had to process the Samsung shot a bit more to get it how I liked. The Fuji shot has the advantage of much better lighting, and you might be right, where the Samsung has a lot more shadows. IMAGE(http://dpzen.com/dpzattaches/dpzattachesi//201409231846223173.jpg) Fuji XP1 at ISO 6400 http://www.flickr.com/photos/erictastad/8154408146 IMAGE(http://dpzen.com/dpzattaches/dpzattachesi//201409231846253175.jpg) NX200 at ISO 6400 http://www.flickr.com/photos/erictastad/6963225069/

hmmmm.....let's just say that the XP1 RAW result is an NX20 RAW with about almost half luminance NR (maybe a 1/3) and about full color NR. so it's more like all you have to do with the XP1 raw is enhance the detail further if someone wants to without having to worry about NR or detail lost from it. what I noticed is that the XP1 has always been consistent on smoothing out the grain while the NX has harsher ones. I can't exactly comment fully on which camera has better detail retention as they are both totally different implementation. the XP1 has no AA filter while the NX has larger MP, the image processor and lens output difference also plays some part as to what is perceived better. although personally from what I have seen, I would say both cameras are not far-off and maybe I'll give the XP1 a bit of an edge on this department. the NX is really good til 3200 and only it's poor NR RAW implementation at over 3200 and poor jpegs makes it a bit of work but certainly has nothing to do with the notion of terrible sensor. I think the key point here is the image processor of the XP1 and DR flexibility at the lower ISOs. for some reason, Fuji has somehow able to exploit the sensor's capability and might be able to do the same with the NX20 sensor. anyway, I saw a difference with regards to how the 2 images behave under such lighting. is there any chance that the doll's real color is blue or green? I guess it's blue if I'm not mistaken?

Ariston wrote: anyway, I saw a difference with regards to how the 2 images behave under such lighting. is there any chance that the doll's real color is blue or green? I guess it's blue if I'm not mistaken?
Yes, the Fuji has the WB on the doll better, but the reds are wrong at the back of the room where the Samsung did a better job. I don't know how much of this is WB, I set using a dropper on the chart in the background. The front light is tungsten balanced fluorescent (poor quality, low CRI, typical consumer CFL bulb) and the back light is more of a daylight balanced fluorescent. I could probably tweak the colors to get them to be more similar. I am in the predicament that I really love both cameras. I love the articulating screen and the overall function of the NX20 (it is a bit smoother operating and faster responding overall, but at least the fuji lets you change focus point when it is writing ), but the Fuji is just classic. Nice mechanical film like body, solid feeling (solid but light, definitely not Leica feeling) I am just blown away with how sharp the Fuji is, but than distant focus or fine textures are funky with the water color texture, even in JPEG there is some of that. Also, the computer is slow to deal with the files, takes several times as long to process them. I really need to get an SSD drive to work from with the RAW files, my internal HDD is so slow. The main things that get me on the fuji are the delay from focused to shot taken, and the camera makes a lot of racket as it does stuff. Steps down all the time and the viewfinder makes noises as it changes lenses/brings up EVF, etc. With my Samsung in AF-S if I press and hold the shutter button all the way down I am pretty much guaranteed to get a shot, and with the 50ms shutter lag with electronic first curtain shutter I am pretty much guaranteed to get a shot in focus if the camera can focus. The Fuji when I do this half the times I get a totally blurry shot as it takes the photo even though it couldn't focus. However, I was still impressed I was able to get some dance shot photos in focus with the Fuji Once I get the photos for the event posted tonight or so I will send you a PM with the link and you can check them out. I liked the Fuji as it is more unassuming than the D7000 with 50-150 f/2.8 OS. I had the Nikon on my black rapid strap with that lens, and the Fuji around my neck (could probably attach it to the black rapid strap some way) Eric

I think the water color effect is something that should be fixed, but otherwise, the XP1 is great. I believe there was a firmware update that has fixed this issue? although you convinced me further into getting an XE1, lol. my decision on it is just partially based on jpeg engine and particularly lowlight shooting at High-ISO, but ultimately has to do with using Leica lenses. so a different particular need. and the NX is for NX lens used. at lower ISO, both cameras are both neck and neck in IQ.

Ariston wrote: I think the water color effect is something that should be fixed, but otherwise, the XP1 is great. I believe there was a firmware update that has fixed this issue? although you convinced me further into getting an XE1, lol. my decision on it is just partially based on jpeg engine and particularly lowlight shooting at High-ISO, but ultimately has to do with using Leica lenses. so a different particular need. and the NX is for NX lens used. at lower ISO, both cameras are both neck and neck in IQ.
Hehe, that wasn't my intent :), but the XE1 looks like a great camera. I got this XP1 used for a good price, but it has a couple small scrapes on the body, etc. Still good operationally though. I am seriously thinking of selling my D7000, but I still need that 50-150 mm f/2.8, and wedding guests feel better about the photographer if they see them with a large camera/lens. Should just buy a D2h to put at my waist and never use it Eric

how bout the RX1? lol. too bad it's selling for 3 grand. I think of sticking with a D800E with flash setup for such, atleast for now.

I have to make a correction here since I can no longer edit the previous statement. this is regarding the color noise reduction adjustment from both cameras. both cameras needed a some amount of application to reduce or eliminate the few residues left from the luminance noise. it is true that the XP1 has way lesser reds and blue which are certainly the worst type of noise displayed and are more obvious to see compared to the green although the green was still well managed considering they can be eliminated without adjusting the luminance. combining the degree or amount of NR applied, is only about half the amount without really sacrificing much of detail. although there is still some variance with regards to the quality of the grain to some degree and did a bit more NR application to atleast replicate the XP1 result. so far, I applied these combination on the images. for the XP1 I applied +13 amount of color NR. on the NX20, I applied +40 color NR and +23 luminance NR. I pushed the color NR further inorder for the NX20 to avoid increasing the luminance NR which would reduce the detail. I added a bit just enough luminance nr to smoothen the image to acceptable or identical level with the XP1. the XP1 is certainly cleaner looking. as far as detail, I don't see any problem with the NX20 at all, especially with B&W shots. one could choose to be grainy or smooth. just don't go over 3200.

Add new comment

Attachment
More information
  • Files must be less than 2 MB.
  • Allowed file types: txt.
By submitting this form, you accept the Mollom privacy policy.