ISO still the weakess of MFT Format

<p>Hello all<br>

I was pleased, if a little surprised, to read about the Panny GH4</p>

<p>http://www.dxomark.com/Reviews/Panasonic-Lumix-DMC-GH4-sensor-review-Heavyweight-contender/Panasonic-Lumix-DMC-GH4-Versus-Olympus-OM-D-E-M1-Versus-Sony-NEX-7-Competitive-performance</p>

<p>This format is getting very close in performance to that of the best of APS-C. I think I remember dpreview stating in their NEX7 review that a FF sensor would be needed to get significantly better performance. Add some top glass and you've made up much of the difference between MFT and APS-C? I do note however, that MFT's ISO performance seems static. Although we can feel re-assured that we have a wonderful selection of lenses to choose from for this format, including a couple of Leica-designed speed monsters, it is still seems like we've hit a bit of a barrier over the last few years.<br>

I would like to see Olympus and Panasonic concentrating instead on the<em> look</em> and<em> texture</em> of the noise their cameras produce. I suspect few of us would have a problem with an 'authentically' creamy grain. It could in fact become a USP. How many hobbyists would be pleased with the film look from their digital files? I think most would. </p>

<p>Good point about the aesthetics of the noise. I don't worry much about noise with the Nikon V1, which can be rather noisy at high ISOs - it's good, but limited by the CX/one-inch sensor. The luminance noise is palatable, very comparable to film grain. I usually add a bit of chroma noise reduction and little or no luminance noise reduction, unless the noise interferes with some fine details.</p>

<p>That's a huge improvement over my older teensy sensor digicam and even over the Nikon D2H, which is barely usable above ISO 800 due to banding.</p>

<p>With the Olympus dSLR-style mirrorless models I've noticed the high ISO raw files are remarkably good, very comparable to APS sensor cameras. And the in-camera JPEGs are clean and crisp almost to ISO 800, often looking better than larger sensor rivals on dpreview's sample photos.</p>

<p>But above ISO 800 Olympus seems to impose weird noise reduction that on the one hand over-smooths some textures, while creating clumpy pepper grain noise.</p>

<p>I know from messing around with some highly flexible and customizable noise reduction software that it's tricky trying to find a setting that's suitable for all photos at a given ISO. But Olympus seems to favor superficially "clean" looking JPEGs that begin to look odd around ISO 800.</p>

<p>Not a big deal if for folks who prefer to edit from raw. But I enjoy the convenience of good in-camera JPEGs and keep hoping for that perfect compact camera that will let me take a break from editing.</p>

<p>Ideally, I suppose, Olympus would offer more in-camera control over noise reduction to allow users to choose the NR for the conditions. Even at the same high ISO, daylight and nighttime or indoor photos may demand different NR techniques; same with photos of people rather than scenery. The clumpy pepper grain noise that might be intrusive in faces might go unnoticed in photos of foliage or textured subjects.</p>

Nope, no barrier :)

 

From the Olypmus E-P2 to the E-M5 I saw a big leap in high ISO performance, about equivalent to what I saw between Nikon D200 and D300. The E-M1 has no low-pass filter and therefore slightly different noise characteristics that could be seen as "more noise" when pixel peeping, but otherwise it is more or less the same as the E-M5.

 

Look at it this way: APS-C and FF will always have either bigger pixels at the same resolution, or more pixels of the same size. As nobody has access to exclusive technology, you can always expect bigger sensors to perform slightly better at high ISO. There is no reason to expect MFT to ever "catch up".

 

On the other hand, there is something that YOU use your camera for, and that is taking YOUR images. My images are mostly made of static subjects, I am no action shooter. Sometimes I like to play with shallow DOF (and I can with MFT), but more often I like the whole scene in focus. On MFT, f5.6 is fine for deep focus, at least if I don't focus too close. Think of architecture, landscape or street photography. On FF you'd have to stop down to f11 for the same DOF. These are two stops. Now compare this with a FF camera like, for instance, the Nikon D610.

 

On the E-M1, thanks to five-axis sensor stabilization, I can hold my 25/1.8 at 1/10s all the time, and that without a proper steady stance. In fact I have just tried with the 12-40/2.8 at 25mm: I can repeatedly take images of critical sharpness at 1/10s, and that while holding the camera in one hand and away from my body. Taking a proper stance I can always hold 1/6s.

 

You can't do that with an unstabilized 50mm prime lens on the unstabilized D610. The rule would be one over focal length, but let's assume you have steady hands and can reliably hold 1/30s. These are at least two more stops, and in sum that translates to the difference between ISO 200 and ISO 3200.

 

Is the D610 at ISO 3200 as clean as the E-M1 at ISO 200? It's not. It has slightly more resolution and with a little bit of noise reduction and scaling down to the same size you may come near. Dynamic range will suffer though.

 

Thus, for my applications the Nikon D610 has zero advantage in image quality. In all other respects it is worse. It's heavier, bigger, and the more lenses you add, the worse is the difference in size and weight.

<p>Only today I sold my D5100. I am now Nikon-free! This did have the nicest high ISO files with very creamy, small, unjagged noise. Over the cameras limit, I almost always convert to monochrome and the D5100 sensor was actually quite lovely at around 2,000 in low light. Of course, it all depends on <em>when</em> you use a high setting. ISO 1,600 on a sunny or overcast day with a small aperture for D.o.F. and +1/3 exp. comp. dialled in looks<em> very</em> different from the same ISO in the corner of a museum at max Av and slow Tv. I find with my 25/1.4 that very little post-sharpening is required, which also helps a lot. But sometimes, you've just got to take the shot at whatever ISO is needed or lose it. However, at it's worst, the noise of the GX1 looks really bad. As I've said before, it takes a lot to get the best out of the GX1, but it is worth the effort. I was particularly unhappy at the Canon noise. Dreadful, blotchy chroma noise destroying faces and only full stop adjustments allowed. Not right.</p><div></div>

<p>Andreas, I think both your maths and reasoning skills are at fault. I'm a big fan of MFT, but this is going too far.</p>

Stephen, would you care to explain where the fault lies?

<p>Love the title: ISO is still the weakess...</p>

<p>Look, the subject of noise has been beaten to death, and there's little difference between m4/3 and APS-C; nearly nothing in terms of IQ. Like the film days, the bigger the "sensor" the better the IQ, generally speaking, so that hasn't really changed. Also, like the film days, we didn't all shoot 8x10 view cameras, did we? Nor do we all need anything more than m4/3.</p>

<p>Honestly, these comparisons are truly meaningless. If an m4/3 cam meets your needs, get it. If not, get what does. Who gives a rat's ass about sensor size, whether or not the camera has a mirror, etc? It's all about the tool being suitable to the need. Same as always.</p>

<p>At the pixel level. the EM1 is on par (or better) than the D800. And with DXO Prime NR, ISO 6400 from the EM1 is noise free and retains reasonable detail and more than good enough for most typically sized prints, and perhaps even small poster size. Needless to say, I no longer own a D800 (and don't miss it) and never worry about shooting high ISO with the M43 bodies I have.</p>

<blockquote>

<p>I do note however, that MFT's ISO performance seems static.</p>

</blockquote>

<p>Not sure why you say it as if it's something bad. APS-C ISO performance has also peaked 3 years ago and there has been no significant improvement to it since the Pentax K-5 was released. If anything, MFT peaked later than the other two larger formats with the E-M5.<br /> <br /> The difference that you're seeing at dxomark is pretty much the difference between the sensor sizes. MFT low light score is about ISO 800 and APS-C is about ISO 1000 (1200 if you would compare to A6000). Big deal? Not really, but some get a kick out of making it so. Why they think APS-C is just good enough but MFT falls apart is the part they never manage to explain.</p>

<blockquote>

<p>Andreas, I think both your maths and reasoning skills are at fault. I'm a big fan of MFT, but this is going too far.</p>

</blockquote>

<p>The technical details of what Andreas is talking about are not easy to grasp, but if you have an interest, read about <a href="http://josephjamesphotography.com/equivalence/index.htm">equivalence</a>. The point is this: a smaller format can obtain the same results as a larger one when used <strong>at lower ISO</strong> with <strong>faster lenses</strong>. In practice, you can get faster lenses, but the lower ISO is the limiting factor. And what this means is that you can match the high ISO results of larger formats but you cannot match their <strong>low ISO</strong> results. Of course, good understanding being such a luxury, most people think that large formats have an advantage at high ISO, not at low ISO. It's both funny and sad that such misconceptions are still being held. If you disagree, read about <a href="http://josephjamesphotography.com/equivalence/index.htm">equivalence</a>.</p>

<p>The question really is <em>"do you need that high performance at low ISO?" </em>Having gone from 100 ISO to 400 ISO as my basic usage on changing from bridge to MFT I am very happy with my choice<br />My current camera is so far advanced beyond my needs I am very happy. The GH4 looks fascinating but really I don't need it being with the GH2 ... which apparently was a vaste improvement on the GH1 from what I read.</p>

<p>note .. funny to be talking about a Panny on the Oly forum :-)</p>

<blockquote>

<p>"The question really is <em>"do you need that high performance at low ISO?" "</em></p>

</blockquote>

<p><em><br /></em>Yup, I do, Stephen does, just about anyone who engages in candid photography does - whether it's street photography, documentary or happy snaps of family and friends.</p>

<p>The alternative is to limit ourselves to mostly daylight photos, which worked wonderfully well for Vivian Maier. But I'm a night owl and my taste runs to Brassai and Weegee.</p>

<p>Data and sample photos indicate Olympus has achieved some remarkable high ISO performance with MFT. But they're up against increasingly tough competition from some very affordable APS sensor mirrorless models.</p>

<p>Some of the reactions here to Stephen's comments seem to focus on measurebator minutiae rather than the gist of his message. My takeaway is that he's referring to the overall performance and aesthetics, notably the subjective look of high ISO noise. That's not something DxO can measure. And from what I've seen of the Olympus high ISO in-camera JPEGs, he's right. But I'm betting all it would take is a different in-camera noise reduction method, with less emphasis on oversmoothing and clumpy pepper noise, in favor of a more randomized texture reminiscent of film grain.</p>

<p>These responses are slightly surprising to me, especially the one about using the right tool for the job and not worrying about the capabilities of other formats. How are you supposed to <em>get</em> the right tool for the job if you don't look at what else is available? Also I note that so far, Lex and I seem to be the only members interested in Street or, as he puts it more accurately Candid Photography. I really can't see the point of not buying something like a K-5 IIS if you're into static subjects. The shot I posted as an example of an exposure I knew was going to be technically inadequate but wanted to get anyway, was taken in an extremely dimly-lit aquarium at the base of the SH Pearl Tower. I was on a school trip on Wednesday and getting nice shots of the kids has sort of become my function on these trips. Thank goodness I bought that 25/1.4. It has never been off my camera since buying it as a Christmas present to myself and I would frankly have been unable to take <em>any</em> acceptable pictures without it, but even this lens could not focus at this light level so I had to MF at 100% mag on a fuzzy, indistinguishable blob. I guessed the FD! ISO 1,600 is the limit for this camera and even at f/1.4 it was tough going. Given the fact that the A7 and XT-1 exist, the mirrorless world is getting very competitive indeed and so with the release of the GH4 I had hoped to see an improvement over my GX1 when it came to noise levels. As to the whole notion of equivalence, it is not difficult, but I simply cannot understand why anyone would limit themselves to ISO 200 and <em>still</em> refuse to bring along a tripod if they are into architecture. This seems to be missing the point entirely. Most SPers I am aware of on Flickr and other fora are using small, mirrorless cameras and the low light performance and AF speed of Fuji in particular seems very impressive. I have a lot of affection for MFT, which is why I would like to see them progress. If it is simply not possible at present for them to be designed with less noise from signal amplification, I feel another solution is in order. I would love to be able to do more clean Night Street, but at the moment this is rarely possible with MFT. I am seriously considering an XT-1 or XE-2 for myself but am also waiting for the GX7 successor, primarily to see a bump in high ISO performance. It is an issue. </p>

 

<p><em>Tool for the job</em> doesn't mean anything if the <em>job</em> doesn't include a context for how an image will be used. The joy of pixel peeping at 200% isn't making a print of some specific size, or posting on the web.</p>

<p>I personally have been reasonably content with the noise level of 4/3s format since my E30 but there ARE times that it needs help. I think the newest NR software does just fine with getting rid of noise and in making the remaining noise aesthetically pleasing. I am particularly impressed with the combination of ACDsee Pro 7 and Topaz DeNoise. YMMV</p>

<p>@JC:</p>

<blockquote>

<p>The question really is <em>"do you need that high performance at low ISO?" </em></p>

</blockquote>

<p>I don't and I think most people don't either because they often state that performance at base ISO is good enough. But as Lex and Stephen point out, some people do care.<br /> <br /> @Lex:</p>

<blockquote>

<p>the subjective look of high ISO noise. That's not something DxO can measure.</p>

</blockquote>

<p>Yes, DxO does not measure JPEG noise, they only measure RAW performance without any noise reduction. Even if the camera "bakes" noise reduction in RAW files, like Pentax does, they'll point that out in their graphs.</p>

<p>But still, I never saw a problem with the Olympus noise at the default settings I use (I shoot RAW but I process files with the Olympus software using the camera settings).</p>

<p>And generally, I avoid going above ISO 800 by using relatively fast apertures (f/2-f/2.8) and relying on the camera image stabilization. I often end up with apertures in 1/2-1/15 range, but I get no shake blur.</p>

<p>For example, this is shot at 1/6, ISO 800:<br /> <a href="https://www.flickr.com/photos/laurentiucristofor/13653326885"><img src="https://farm8.staticflickr.com/7009/13653326885_a9d12fa487.jpg" alt="" width="500" height="375" /></a></p>

<p>I'll attach a couple of 100% unprocessed crops saved at 97% compression. This quality is good enough for me.</p><div></div>

<p>I am also attaching a crop from the darker area at the top of the image.</p>

<p>This was shot at 26mm, which is about 50mm FF equivalent. So if we would go with the conventional wisdom that says that 1/50 would be the recommended aperture for a handheld shot at this focal length, image stabilization gave me 3 stops over that: 1/50->1/25->1/12->1/6.</p>

<p>Oh, and this was shot with the E-M1. I don't know how the noise on Panasonic cameras looks, but I never had a complaint with Olympus, even going back to the E-PL2.</p><div></div>

<p>You're using NR tools at ISO 800?<br>

You're making my case for me Laurentiu.</p>

 

<blockquote>

<p>You're using NR tools at ISO 800?</p>

</blockquote>

<p>No. What gave you that idea in what I said? I only use the Olympus software to get a JPEG from the RAW file and then gimp for processing the JPEG. The crops are a straight conversion just cropped in gimp, so they're as close to in-camera JPEGs as I can get, because I only shoot RAW.</p>

<blockquote>

<p>You're making my case for me Laurentiu.</p>

</blockquote>

<p>I wish I knew what your case was. You're not helping me understand it.</p>

<p>Lex Jenkins wrote<em>:</em></p>

<blockquote>But I'm a night owl and my taste runs to Brassai and Weegee.</blockquote>

<p>Both of these photographers used large format film. Brassai used a tripod and long exposures. Weegee used flash. Their techniques are still available to you. Expecting to do the same without flash or a tripod is unrealistic.</p>

<p>Or rather it <em>would be </em>unrealistic, except that incredible advances in digital cameras allows us to perform this magic. Images from Olympus MFT cameras at ISO up to 1600 look perfectly good to me, so long as exposure and other niceties are handled correctly by the photographer. We save on the insane weight of their systems. But perhaps cannot make up for their talent, dedication, and access privileges.</p>

<p>I am also pretty sure that neither candid photographer used "high ISO in-camera JPEGs", so this is an apples and oranges argument. If you want the best from high ISO you need to take the time to develop.</p>

<p>I am finding that the micro 4/3 give enough sharpness and depth of field to use the lenses wide open, lowering the average ISO that I use relative to APS-C sensors. I can use the 25 1.8 wide open with no worries about the sharpness or the image getting "dreamy" on me. </p>

<p>Stephen, when I said to just choose the right tool for the job, what I meant was that you shouldn't necessarily choose the camera with the lowest noise, highest pixel count, etc. Yes, compare, but don't expect smaller sensor cams to ever match larger sensor cams. But that doesn't mean you should always choose the largest sensor cam on the market, does it? I'm just saying that if m4/3, or any other format does the job for you, use it and don't worry if camera C or N delivers slightly lower noise. I mean, really, we don't all drive the largest SUVs on the market, do we? Don't we buy a vehicle suited to our needs, if we need one?</p>

<p>This is the point David. I really like my GX1 and hope to have a reason to ugrade to the GX8 or whatever it is called. Given that the MFT makers are fighting a losing battle over noise, they should play to their strengths and make the noise look good, like an integral part of the image. I know that when to comes to tone, valves are king. Despite valves being very old tech, most guitarists playing electric instruments won't even consider a transistor amplifier since the innate tone of a valve amp is so much nicer. Similarly in the HI FI world, many people are prepared to spend a lot of money for the warmth of a valve amplifier. Look at the watch world. Mechanical is King despite its very real shortcomings. I think that Panasonic and Olympus should try to do something similar.<br>

Laurentiu, I was mistaken about you using NR at low ISOs. Please forgive me. But 800 is considered a low ISO these days, so it doesn't say much for the capabilities of the format. As has been said, a good fast lens brings the ISO down and also reduces the occurence of PP artefacts, but camera buyers, like so many others, are fickle. Given the high price of both lenses and bodies (the last time I visited the Shanghai Photography Mall they were asking 6,000 for a GX7 and <em>6,200</em> for a K-3), the threat from other APS-C /mirrorless models cannot be ignored. I would have to pay 6,000 for a GX7 and 3,000 for the 25/.4. Or I can get an XT-1 plus kit lens for a few hundred yuan more.<br>

Would any of us really blame someone for going for the Fuji? What would be the <strong>advantage</strong> of the GX7?</p>

<p>I get it now:</p>

<blockquote>

<p>I really like my GX1</p>

</blockquote>

<p>and</p>

<blockquote>

<p> read about the Panny GH4</p>

</blockquote>

<p>So you use the GX1 and are concerned about the performance of the GH4. I thought you actually used the GH4 and found something fishy about it.</p>

<p>If you'll compare the GH4 with the GX1 at dxomark, you'll see that there is a significant difference between them, but that is mainly in dynamic range and color sensitivity at low ISO. If you also check the graphs to compare MFT with APS-C you'll see that the difference is very small and doesn't get worse at higher ISO. <a href="http://www.dxomark.com/Cameras/Compare/Side-by-side/Panasonic-Lumix-DMC-GH4-versus-Panasonic-Lumix-DMC-GX1-versus-Sony-NEX-7___943_754_736">See for yourself</a> - click on the graphs in each category and compare them.</p>

<blockquote>

<p>But 800 is considered a low ISO these days, so it doesn't say much for the capabilities of the format.</p>

</blockquote>

<p>It is considered low by people that don't understand <a href="http://josephjamesphotography.com/equivalence/index.htm" rel="nofollow" target="_blank">equivalence</a>. You don't buy a small sensor camera to use it at high ISO. You don't buy a large sensor camera for that either, but you're forced to use the high ISO when you're forced to stop down the aperture to gain DOF.</p>

<blockquote>

<p>I would have to pay 6,000 for a GX7 and 3,000 for the 25/.4. Or I can get an XT-1 plus kit lens for a few hundred yuan more.<br />Would any of us really blame someone for going for the Fuji? What would be the <strong>advantage</strong> of the GX7?</p>

</blockquote>

<p>I got into MFT because I liked a MFT lens. I still like the MFT lens lineup better than that of any other system - that is the advantage that drew me in.<br>

<br>

I don't buy into a system because of a camera model looking good because I know I'll spend a lot more on lenses, so I look at what lenses that system offers. When I got into Olympus, I started with the E-PL2 which is probably similar in performance to your GX1. I also picked Olympus rather than Panasonic because of the in-body image stabilization - I know the GX7 has this feature too, but it's first generation and I don't expect it to match the Olympus system, especially the 5 axis one. These are my reasons and if I had to start from scratch today, I'd still pick MFT over anything else - it just offers me more right now.<br>

<br>

Why don't you just rent a GH4 for a weekend and see how much better you find it over your GX1? Rent an XT-1 too and then compare them. Then buy whichever you liked better and share your experience here.</p>

<blockquote>

<p>"Would any of us really blame someone for going for the Fuji?"</p>

</blockquote>

<p>Depends on personal preferences and photography styles. I'm trying an X-A1 but I'm not sure it's right for me. In decades of photography I've never before been tempted to return or exchange a camera for another model, but I may return/exchange the X-A1. And I'm really conflicted about it because the X-A1 image quality is outstanding. But the laggy rear screen and other factors make it less than ideal for the main type of photography I intended for the X-A1: low light and nighttime.<br>

</p>

<p>At this point, with the new micro 4/3, the new Fujis, and the new DSLRs, it is really challenging to complain about image quality. We are now down to splitting hairs about differences in senor characteristics that usually don't matter much. Personal preferences for ergonomics and control layouts are the more important issue now, IMHO. I like the micro 4/3 because of the size and some of the lenses. Other people will prefer the optical viewfinder on the X-Pro 1. </p>

<p>I really like the photos form the OMD EM 1, and I am sure I would like the photos from a Fuji XT-1. </p>

<p> </p>