I stumbled across something interesting on micro-4/3 camera forums, and thought it worth pursuing. It’s been said (here, and here, and here) that the Olympus E-M1 has a problem with long exposures. I’d noticed what looked like grain in my E-M1 images, so thought maybe it had a similar cause.
Since I’ve got an E-M1 and an E-M5 camera, and a bit of history with debugging, I thought I was well-placed to do some characterization testing. To get a fair comparison, I first ran pixel mapping on my E-M5 and E-M1 bodies (to take care of any obviously “hot” pixels on the sensors), then took dark images (with the lens cap on) under the same conditions with both bodies:
- ISO 200
- 60 second exposure
- Noise Filter set to Standard
I did this twice for each camera body — once with Noise Reduction off, and once with it on. I took all 4 RAW images, converted them to JPG using Olympus Viewer 3, then grabbed a 400×400 pixel crop out of the center of each.
Here’s how they compare:
You can see that Noise Reduction does a pretty good job on the E-M1, but still — the un-filtered image is dramatically noisier than was the case for the model’s predecessor. Not a good thing for a model being billed as Olympus’ “flagship” camera.
Meanwhile, Noise Reduction works by taking a second image of equal exposure (but with the shutter closed) and subtracting it from the first — so a 60 second exposure will be followed by 60 seconds in which you can’t do anything with the camera. Not good for astrophotography or fireworks or low-light HDR.
Given that Olympus’ cameras turn off Noise Reduction when sequential shooting is being done, I suspect this issue is also at fault for the “grain” I’ve seen in a lot of my low-light images. I take low-light photos in 3-shot bursts using low-speed sequential shooting, in order to get at least one frame unblurred by my button pushing.
Bottom line — this is a big problem for long exposures that can’t (for whatever reason) use Noise Reduction, even at low ISOs.
Update 18 November: the data, sliced in more directions…
I did a few more tests over the weekend, just on the E-M1. First, I wondered if the Noise Filter settings would make any difference in the results — so at ISO 200, taking 60 second exposures (approximately, timed by the display on the camera’s screen), here’s what things look like (all slices are 100% crops):
So, setting the Noise Filter to High takes some of the edge off — but still, not a huge improvement.
Finally, I was wondering if changing the sensor ISO had any effect on the sensor noise — so I took additional images assuming that sensor reciprocity (doubling ISO and halving exposure time) still held:
So at first blush, ISO seems to have little impact on the noise (I was expecting this noise would increase at higher ISOs), while the exposure time appears to be the dominant factor. I’ll have to run more tests to double-check this. Stay tuned…
Thanks for your just and fair post.
Unfortunately there seem to be quite a couple of people out there at the moment who defend their beloved E-M1 they just spend a lot of money for…
… using arguments like “Well, most of the time this is not relevant. No real photographer would shoot without the automatic dark frame [called NR by olympus] anyway, this is only relevant for astrophotographers”, while it is a REAL problem for quite a few photographers who love long exposures, bracketing, low light photography.
Again, thx for your objective un-emotional post 😉
I know what you mean — all the fanboys and haters get the most attention on camera forums. I’ve used Oly gear for years, and to be honest am more than a little disappointed with them thanks to the E-M1 noise issue. Granted, Nikon and Canon have had their own occasional misfires (cameras that put oil on their sensors, or have splotchy sensor response). Still, for a camera that’s supposed to be top-of-the-line, this seems like a big thing to miss in pre-sale testing.
very nicely done. I love my E-M5 and know the dark frame subtraction well. I actually thought the E-M5 was slightly noisy without the dark frame and it kind of irked me, but the camera is so great in many other ways I got over it. But the E-M1 would have to be a pretty mind blowing camera in other respects for me to be able to live with that amount of noise. I hope the E-M6 follows the path of the E-M5 and not the E-M1. Otherwise I am going to have to switch camps.
Thanks! Again, this seems to primarily impact either long exposures, or sequential shooting in low light — so most folks won’t notice or care. Still, it’s annoying that you have to keep such obvious limitations in mind when shooting with a $1300 body.
I have a bit of a work around for this, if Olympus is listening. The Dark frame subtraction works well. The main obstacle is that if you are using sequential shooting long exposures for something like star trails you will get bit gaps as you wait for each dark frame to be captured. When I do a dark frame subtraction manually I notice that dark frames taken under the same conditions (same day and temperature) are very very similar. So similar they you don’t need to take a dark frame after each shot. So Olympus could simply put the dark frame at the end of the sequential shooting group and apply that to all images. This may be done in camera or in Olympus Viewer software.
Good idea. The only problem I see with it is that it only works so long as you only shoot sequentially for a few frames (until you fill the camera’s buffer). If you shoot sequentially for longer, the frames will start being written to the SD card, and it’d be hard to make the camera read them back in / subtract dark frame / write back out.
I’m still hoping there’s a firmware fix / patch for this. Even just allowing NR in sequential shooting below some frame rate (4 fps? 5 fps?) would do most of what I need.
@ Eric
a) The method of just one dark frame to clean more than one image is problematic. Especially when you are bracketing and have different exposure times (and of course different ISOs), you also need different dark frames, for every shutter time (and ISO of course) a different dark frame.
One dark frame for several images is especially problematic if it’s done automatically (in cam). You would have to adjust the dark frame for optimal cleaning depending on settings (see above) and current sensor temperature.
b) I don’t see the urgent need for NR when shooting 4fps or 5fps. It’s most imortant with long exposures. But maybe I didn’t understand your argument?
Star trails require long expsoure sequential shooting. With an E-M5 you don’t need dark frame subtraction, with an E-M1 it looks like you really do, or perhaps it creates many more stars 😉
I believe those are hot pixels. It’s only natural you found they’re related to exposure time because they’re a consequence of sensor temperature increasing as the exposure gets longer. Incidentally, I felt the same problem with an E-P1. I had it repaired and now looks OK.
Como que el iso no afecta al ruido. ¿Puede subir una foto a ISO 1600 y 60s para demostrarlo, o el resultado es demasiado lamentable para enseñarse?
The problem is you are zoomed in to far, when viewed from a distance or a slight squint it has a filmic look.
very interesting! can you post up some raw files from both cameras? I don’t have either of these cameras (I use an E-PL3), but I’d like to take a look at the noise statistics from these sensors.
Great post. With the imminent release of firmware 1.1 for E-M1, which includes some changes for long exposures, please try it and see if it actually helps.
Thank you for taking time to make E-M1 vs E-M5 comparisons and for your candid review. I started the “Long Exposure issue with OM-D E-M1?” thread on Flickr (thanks for mentioning it), and guess that any contribution added to the knowledge base of “long exposures issue” is highly important for Olympus fellows.
Thank you one more time!