Noise and DR roundup for Sony A9 and EM1x

I have posted several earlier articles about noise and dynamic range (DR) compared between the the Olympus EM1x and the Sony A9. They have created quite a lot of controversy in various camera enthusiast forums. It’s a very emotive subject it appears. It’s not just Sony enthusiasts who have a problem with the data. Many Olympus owners cannot accept that the noise and DR of an OMD camera can often be comparable to an A9. It seems to run counter to their own direct experience with their cameras. In this post I want to show how their experience and my data can both be correct. If you think low noise is not as important as a good image, I agree, and you can skip to the end of this article.

A balanced system to reduce Noise and DR

First of all let’s be absolutely clear. The 20 Mpx sensor used in the Olympus EM1ii, EM1iii and EM1x is around 1-1.5 stops behind a Sony A9 in Noise and DR performance. It’s a miracle that they got it that close but there is an undoubted and real gap in noise performance between the two cameras sensors.

For most people, and very specifically just about all internet pundits, the noise and DR story ends there. The Olympus is worse than the Sony, full stop. But actually that is not the end of the story. The noise and DR in the sensor is directly related to the ISO at which the photograph is taken. At low ISOs there is almost no noise in either camera, and what there is can be eliminated with technologies like DxO deep prime or (my favorite) Topaz denoise AI. As the ISO increases, the noise increases, and this is more noticeable for the Olympus cameras than for the Sonys. But there are advanced system capabilities that can be employed to reduce the ISO. These include faster lenses, better exposure management, image stabilization, and computational photography techniques like software ND filters.

The Olympus engineers are very conscious of the sensor shortcomings of even their high-end cameras and have built all of the above capabilities into their camera system. If they are used correctly, the Olympus system can absolutely compete on noise and DR with the Sony system. If they are not used, or used incorrectly then noise and DR will be visibly worse, just as many Olympus owners suspect.

Using the right lenses

Let me make a fairly obvious point. The way to reduce the ISO required is to let in more light. The most obvious way to let in more light is to use faster lenses. So the Olympus engineers provided very fast Pro f1.2 prime lenses at the key focal lengths a photographer might require (in 35mm equivalent, these are 35mm, 50mm and 90mm). In addition they built two very fast Pro telephoto lenses (the 40-150mm f2.8, and the 300mm f4).

To get the best out of the Olympus system in terms of noise, you have to use these lenses. When you do so, the Olympus system is absolutely comparable in noise terms to a Sony A9 system built with the Sony G lenses as I show at that link. If you combine the pro lenses with Topaz noise reduction software, the EM1x even manages a respectable performance compared with the A9 using Sony G Master glass (at least at ISOs up to around 3200 – see same link as above).

But the majority of Olympus owners and many of the the internet pundits who test the cameras do not use the pro glass. That’s because one of the great merits of the Olympus system is the small and light lenses, However these typically come with a penalty in lens speed. If you put an f1.8 prime lens, or a f3.5-5.6 zoom on your Olympus camera you will be getting a beautifully portable system, but losing the ability to compete with the big boys on noise and DR. The fabulous thing about the Olympus cameras is that you can do both of course. But when Olympus users express disbelief that you can get low noise out of an EM1x for example, it’s because their personal experience is frequently based on on using slower lenses.

Getting the exposure right

The worst possible thing to do with the Olympus system, particularly in low light, is to underexpose and then try and recover shadows in post processing. You will get far better results by very precisely exposing for the subject and keeping the the overall brightness of the image as far to the right of the histogram as you can get without blowing the highlights. This is a technique that has sometimes been called “expose to the right” or ETTR. The challenge is to be able to correctly expose for your subject and know that you have not blown critical areas of the image.

General matrix metering, or spot metering, does not provide an easy method of achieving this. However the Olympus system of highlight indication combined with exposure compensation provides a really simple and fast method of getting the subject exposed correctly. I describe it in some detail in this post, and I think it’s almost as important as fast glass to get the best out of your camera system.

Not properly exposing the subject in an image is another key reason why some Olympus users complain about the noise and DR of their cameras.

Image stabilization

The more light you let in, the lower the ISO, and one way of getting more light in is to have a slower shutter speed. Unless you are using a tripod with good skills, long lenses require high shutter speeds, and this can lead to high ISOs. Advanced sensor stabilisation can significantly reduce the shutter speed required, in the case of the Olympus by up to 8 stops with a corresponding reduction in ISO. In practice, amazingly long shutter speeds can be obtained with hand-held shots on the OMD cameras, and I have been able to hand hold at up to 6 seconds and still get pin sharp results.

This is typically of benefit for longer focal lengths in dim lighting situations, and and is only useful for slow moving or static subjects. But in those situations, the ability to use a longer shutter speed Is a real additional bonus in keeping ISOs low.

Of course all camera vendors now have some form of sensor stabilization. But most of them lie about the stabilization capabilities of their systems. In particular Sony’s sensor stabilisation is not a patch on the Olympus one and is several stops behind particularly at the longer focal lengths. So this can provide a genuine differentiating advantage and help to pull back the noise and DR disparity of the sensor. However it is only of use in limited circumstances, and is not much help for any kind of moving subject.

The biggest factor of all

By far the biggest reason people believe the Olympus system has significantly worse noise than a comparable Sony system, is that very few people use both systems together at the same time.

Internet pundits test one camera, hand it back, test another camera, hand that back, and so on. If they do any tests at all they will be in isolation on the camera they have in their hand. They always conclude that the Olympus cameras have comparatively poor noise performance. That’s true at the sensor level, but the Olympus system with fast Olympus lenses does not have comparatively poor noise, as we have shown above.

You will only ever find out the real facts by long term use of multiple camera systems in the same scenarios with lens attached, and no internet pundit ever does that. Very few Olympus enthusiasts ever do that. Even if an Olympus bird enthusiast also has a Sony full frame camera, he/she will use them for different roles, and rarely compare them in the same situation. For various reasons, I did take out multiple comparable ‘bird in flight’ camera systems to the same repeatable locations to do a degree of structured data collection. I think I am relatively unique in doing that, and for that reason my results are surprising and non-intuitive to many enthusiasts. But as I hope I have shown, these results are both well founded in the physics of the cameras, and in data, both my own and those of quality testing organisations like DXO and DP review. Despite that, I am sure that the majority of camera enthusiast won’t believe my conclusions. So be it……

To summarize: You will get noise comparable to the Sony systems in your Olympus Camera system if you:

• Use the fast Pro lenses
• Expose correctly for the subject
• Use Image stabilisation
• Generally have good technique.
• Shoot to the strengths of the camera
• If necessary, use Topaz denoise AI or DxO deep prime

You will get higher noise than Sony systems if you:

• Use slow Olympus glass
• Underexpose your subject
• Are shooting in low light at moderate to short focal lengths
• Are shooting at ISOs higher than 6400
• Are doing high resolution portrait or landscape work at base ISOs
• Are comparing to the $13,500 Sony telephoto prime lenses, or the GMaster primes or zooms.
  

Is noise that important?

Finally, it’s worth noting that a good image beats any level of low noise. My most popular image on Flickr, with nearly 24,000 views and almost 600 favorites was taken with the ancient Olympus EP5, with a 16Mp sensor and really poor noise, at an aperture of f11. Seems to have done OK.

And here is a little challenge for you. This link shows an album of long exposure shots of London bridges. There are shots from a Nikon D600, D750 and D800 in there and also a selection of shots from M43 cameras of great decrepitude by today’s standards. And I think one Fuji shot. They are all around 3-6 minute exposures, and no noise reduction has been done on any of these as that kind of software was not then available. See if you can tell which ones are the FF and which are the M43 shots. Then check from the EXIF data to see which image was taken with which camera (press the little i in the top right of the image screen). No prizes will be awarded, but you might be surprised.

Ah, but while those screen images might look OK, everybody knows that you cannot make a large 300dpi print from M43 cameras. Like for example the 90 cm x 60 cm (before framing) shot shown on the left. Taken hand held, wide open at 1/40 with the E-M1 mk ii. The actual images from this set are here. And I have made exhibition quality prints from many of them.

Note: I will cover software ND filters in a future post on long exposure photography with the Olympus system. It’s a limited use case but one that’s actually very pertinent to the kind of photography I do, so look out for a post exploring techniques and hardware for long exposure images with the EM1iii.