This is a two-page article.
Page one covers:
Page two covers:
For some photographers a downside of the Micro Four Thirds system is that the sensors available can’t match the pixel count of some full frame and APS-C systems. The fact that the majority of full frame cameras in use have 24 million pixels or fewer, only marginally more than the 20 million pixels of the Lumix GX8, GH5 and G9, doesn’t seem to dampen the perception that full frame cameras are in the league of 33-50MP. Only half the current range of full-frame DSLRs on the market have a resolution greater than 30MP.
This ‘limited’ resolution of micro four thirds cameras turns off some landscape photographers and still life workers who demand as much detail as possible for their shots.
Fortunately, this doesn’t need to be so much of an issue now, as one of the many great things about the Lumix G9 is its ability to shoot 80-million-pixel images that can record four times the pixel-count of standard pictures. Using its finely tuned in-body image stabilisation system the camera is able to perform a neat trick that allows us to record much finer detail than usual and to produce images that out-resolve the best full frame cameras.
Shot at 1/13sec and ISO 200. Click to see full size
How it works
The G9 has a highly sensitive image stabilisation system that is able to shift its sensor by tiny distances and, crucially, to stop and start that movement very quickly and accurately. With an amazing degree of control, the IS system can move the sensor to the right by exactly half a pixel, and move it to the left, up and then down. By recording an image with each tiny movement the camera can collect images that record four times the amount of data in the area that one pixel usually occupies.
To do this the camera takes eight images as the sensor shifts in a loop pattern, and combines all the information collected into one 80-million-pixel file. The movement of the sensor creates what are effectively off-set pixels that together record four times that amount of data – just as flatbed scanners work with off-set rows of pixels in their scanning heads (but a little bit more complicated.)
The idea of shifting pixels isn’t new and has been used to create high resolution images for a long time, but usually in expensive medium format backs that often record different colours in separate exposures as well.
The effect of using a shutter speed of 1/2000sec. Click to enlarge
Implementation in the Lumix G9
In this camera Panasonic uses the shifting sensor technique to record eight images half a pixel apart. The eight images are then combined to produce a raw or a JPEG file, or both. Users can elect to save a normal version as well as the composite file.
The resultant image measures 10368×7776 pixels, and as such will comfortably print to 35x26in using the photo-quality 300ppi print sizing standard.
Users also have the choice of a 40MP mode that produces images of 7296×5472 pixels in what the camera describes as ‘a balance between resolution and file size’, though the drop in resolution doesn’t save as much space as you might think if you shoot raw.
Resolution Image dimensions JPEG file size Raw file size
20MP 5184×3888 10MB 23.11MB
40MP 7296×5472 17MB 125.4MB
80MP 10368×7776 28MB 125.4MB
These quotes file sizes are approximate as file size depends on the content of the image. However, these are all for the same subject matter, so they give you an idea of how file sizes compare.
This shows the difference shutter speed makes to motion. The frame on the left was shot at 1/160sec, and the one on the right was shot with 1sec exposures. Click to enlarge
Shutter speed and working silently
The High Resolution mode uses the camera’s electronic shutter to avoid the internal vibrations that might occur if eight images were recorded in quick succession using a mechanical shutter. This ensures the camera remains absolutely still so the maximum detail can be taken from the scene.
Working in electronic shutter mode means that our shutter speeds are restricted to between one second and 1/32,000sec – though for most such a range won’t sound like too much of a restriction.
The key thing to note though is that we are effectively working with a multiple exposures so we have to take care of how motion in the scene will be portrayed.
When photographing static subjects we don’t have to pay too much attention to our shutter speed, but if there is something moving in the frame how it that motion will be represented in the final image needs to be considered. We have a number of choices but they basically come down to whether we attempt to record the moving object as sharply as possible with a series of short shutter speeds, or allow the motion to blur by using long shutter speeds. If something really is moving and we use a series of short shutter speeds we should expect the edges of the moving object to be staggered, while a long shutter speed creates a smoother edge because the camera is blending slightly blurred edges rather than sharp ones.
I experimented with longer shutter speeds for some landscapes with water in them to see what effects could be achieved. Using a neutral density filter I was able to record images with 1 second exposures which made the motion very smooth and the water surface as you would expect from a long exposure of several seconds. With shorter exposures the water surface looks much more textured – as we would expect from a multiple exposure. The looks are just different and well suited to different occasions.
Click through to the next page to see comparisons with the Nikon D850
This is a two-page article.
See page one for:
The camera I am using doesn’t seem to allow apertures smaller than f/8. I suppose this is to ensure we use the best of the sharpness of the lens, as there is no point creating a high resolution image of soft detail. Also, ISO settings can’t be turned beyond 1600, which may have something to do with asking the camera to perform advanced noise reduction on each image before they are blended, and the amount of detail we lose at high ISO settings.
Comparing relative image sizes between the Lumix G9 (left) and the Nikon D850 (right). Click to enlarge
Does it really work
Panasonic says that the XL mode in High Resolution gives us 80 million pixels, and when you count the pixels there are indeed 80 million of them. But that is different to having 80MP-worth of resolution.
I compared 80MP files from the Lumix G9 with the 45MP files from the Nikon D850 and found that the G9 images really do contain a lot more detail though whether there is almost twice as much could be debated. The Nikon D850 images measure 5504×8256 pixels and the raw files (of the same scene as those shown above for the G9 files) weigh 55MB.
I suppose rather than concentrating too much on exactly how many pixels the results of this mode compare to it is more useful to consider how much of an improvement over the G9’s normal shooting mode this provides. The answer to that is ‘a great deal’.
This screen shot shows the difference in image size and detail between the Lumix G9 (left) and the Nikon D850 when the images are enlarged to 66%. Both pictures were taken with the Nikon 60mm f/2.8 macro lens. Click to enlarge.
Is this comparison with the Nikon D850 science?
No, this isn’t a scientific test for a number of reasons. I can’t use the same lens in the same way on both bodies to get the same optical resolution for a start. When I mount Nikon lenses on the G9 I’m only using the central area of the lens and I have to move the camera further from the subject. Also the Nikon files shown here are from raw files and the G9 files are JPEGs – I don’t have raw conversion software yet.
Using the same lens on both cameras would provide some element of consistency, but in real life most people will be using a Panasonic lens, so I have compared images using a Nikon 24-70mm f/2.8 with those shot on the G9 with the X-Vario 12-35mm f/2.8.Using the same lens on both cameras would provide some element of consistency, but in real life most people will be using a Panasonic lens, so I have compared images using a Nikon 24-70mm f/2.8 with those shot on the G9 with the X-Vario 12-35mm f/2.8.
Small areas of the above image enlarged for comparison. On the left is the G9 with the Leica 25mm f/1.4 and on the right is the Nikon D850 with the Nikkor 50mm f/1.4. Click to enlarge.
However, the idea of this exercise was to get some idea whether Panasonic’s claims for 80MP resolution stack up, and how the files look in a general way compared to one of the best high resolution DSLRs on the market. It isn’t supposed to be a scientific comparison.
I should also point out that I’m not trying to imply that the G9 is a better camera than the Nikon D850. They are very different cameras, and the D850 achieves its high resolution in a single shot, and for all subject types. The idea of this comparison is to see just how the detail gathered by the G9’s High Resolution mode compares to one of the best full frame DSLRs currently on the market.
Lenses and resolution
Clearly for any sensor to be able to resolve lots of detail it needs to have been fed lots of detail by a capable lens. The quality of your results using High Resolution mode will depend very much on the lens that you put on the front of your camera. If you use a low cost zoom you will not get the same benefits that you will get if you attach one of the best fixed lenses. I guess that is pretty obvious, but worth pointing out all the same.
A further word
The camera I used to illustrate this piece is running firmware v0.3, so it is some way off being finished. That’s why you can’t see the full resolution images. But, once I get a final version camera, I’ll make the tests again and you’ll be able to see more clearly what this mode is capable of.
If you have any comments, questions or requests please use the Lumix G9 thread in the forum.
Shot at f/8 and 0.4sec using the Leica 12mm f/1.4 lens. Click to enlarge
If you are interested in the Lumix G9 you might like these articles: