Review Olympus OMD E-M1II Sensor

Machi

Mu-43 Top Veteran
Joined
May 23, 2015
Messages
766
Few days ago I finally get my hands on Olympus camera with newer 20 MPix sensor (OMD E-M1II) and
as a nerd I cannot resist to try measure parameters of this sensor and compare it to my older measurements of 16 MPix sensors.
It's true that those parameters were already measured by others (Photonstophotos, DXOMark) but as I found, they've ignored some aspects.
For example DxoMark published only results for ISO64, 200, 400, 800 and so on and only for mechanical shutter.
Photonstophotos ignored extended ISO100 and silent shutter mode.

I've measured basic properties of sensors at all ISOs with analog amplification +
few beyond that (digitally amplified) for both mechanical shutter and electronic (silent) shutter.

Basic parameters are System Gain, Full Well Capacity and Read Noise and
from those Dynamic Range (DR) and Signal-to-Noise Ratio (SNR) can be derived.
For those interested how it can be measured, good sources are
Signal to Noise: Part 3 - Measuring your Camera by Craig Stark and
shortened version of basic procedure is also here.

First table shows basic parameters for E-M1II camera for both shutter modes.
Read noise is in electrons, gain is in electrons per ADU (analog-to-digital unit) and
full well capacity is in kiloelectrons (thousands of electrons).
I've looked also on behaviour of ADC (Analog-to-Digital Converter) and
interestingly there is important difference between mechanical and silent shutter.
ISO64 up to ISO6400 are analog amplified in case of mechanical shutter but
in case of silent shutter it's only between ISO64 and ISO3200.
Digitally amplified values are in this table with red font.

Parameters.png
Subscribe to see EXIF info for this image (if available)


From those basic parameters dynamic range can be derived as ratio (full well capacity)/(read noise) and it's given in EV (stops).
I've added for comparison 16MPix Olympus OMD E-M10II and for fun also old 7MPix point-and-shoot camera Canon A570IS (1/2.5" sensor).
It's already visible that E-M1II has highest DR (~12.4EV) at ISO250, some 1/3EV higher than at ISO200.
That btw also means that landscape score (scaled for 8MPix normalized camera) of E-M1II at DxoMark should be ~13.1EV and not 12.8EV.
Another interesting result is that using mechanical shutter mode gives small boost (1/3EV) of DR between ISO2000 and ISO6400.
Silent mode doesn't have this advantage.

DR.png
Subscribe to see EXIF info for this image (if available)


Another graph shows SNR which is arguably even more important aspect of IQ than DR because
it determines how pronounced is noise in the image.
SNR can be given for any level of brightness in the resulting image (from shadows to highlights).
Here is SNR (in decibels) given for 18% of entire brightness scale (=grey) as it's favorite value,
it's in midtones and allows direct comparison with DxoMark results.
E-M10II and Canon A570IS are added for comparison.
Here one surprising result is that extended ISO100 has relatively bad SNR, it's comparable to the ISO320.

SNR.png
Subscribe to see EXIF info for this image (if available)


Last graph shows again total DR for both mechanical and silent shutter and more importantly
it shows tradeoff between improvement of shadows and loss of highlights.
Loss of highlights is the same for both silent and mechanical shutter modes but
there is 1/3EV advantage of mechanical shutter from ISO2000 to ISO6400 in shadows improvement.

DRHS.png
Subscribe to see EXIF info for this image (if available)


Some practical tips based on the results (mainly for RAW shooters):

1. There isn't much difference between IQ of silent and mechanical shutter mode for ISO64 to ISO1600 and ISO8000+.

2. For daylight photography, best IQ (SNR) is achievable at ISO64 and 200.
(ISO250 gives best DR for some special scenes with bright parts together with extreme dark shadows).

3. For night outdoor photography (especially astrophotography), ISO2000 with mechanical shutter is good compromise as it gives practically full improvements in shadows (~2EV) with smallest loss of highlights.

4. For night indoor photography, there will be no significant difference between mechanical and silent shutter in IQ as SNR is practically the same and better DR of mechanical shutter is practically observable only in very dark shadows (common outdoor at night, not so much indoor).

5. I don't see much value of extended ISO100 as it looks like ETTR at ISO320.

6. Anything beyond ISO6400 in case of mechanical shutter and ISO3200 in case of silent shutter is digitally amplified and it's better to avoid it when one is shooting in RAWs as it can be amplified in the computer later without unnecessary loss of highlights.

7. Despite all differences mentioned I doubt that one can see most of them practically without very detailed and specific comparisons as 1/3EV difference is on the edge what is visible under controlled conditions but if someone wants to squeeze all the IQ from their camera, there is a way! :)
 
Last edited:

caonidayeah

Mu-43 Rookie
Joined
Oct 7, 2020
Messages
16
thank you very much for your evidence-based conclusion that EM1 MK2 is such a capable camera and I can safely take photos up to ISO 6400. I'm more concerned with shaky image and loss of decisive moments than noise any day.
 

scb

Mu-43 Regular
Joined
Oct 21, 2017
Messages
66
Location
North Ridgeville, OH
My last camera body purchase was the EM5 Mark II. I'm getting this GAS pain that may foretell the purchase of the EM1 Mark II. But, I'm going to suppress that urge for quite some time! The question I have in my mind is whether I'm willing to part with my EM 10 Mark II to help subsidize the purchase. I still love that camera!
 

Machi

Mu-43 Top Veteran
Joined
May 23, 2015
Messages
766
...The question I have in my mind is whether I'm willing to part with my EM 10 Mark II to help subsidize the purchase. I still love that camera!
I definitely don't plan to get rid of my E-M10II. I can actually imagine that I would buy rather another one if my current one would be broken!
 

retiredfromlife

Mu-43 Hall of Famer
Joined
May 15, 2016
Messages
4,539
Location
Sydney, Australia
Thanks for the review, finally one I can nearly understand. :thumbup:

Most reviews I have read in the past on related subjects have had results/conclusions as convoluted as the theory of the subject matter [for me anyway, no technical knowledge of anything electonics]

Your graphs and conclusions I can actually use, and I like the comparisons to the other two cameras for reference
 

RichardC

Mu-43 Hall of Famer
Joined
Mar 25, 2018
Messages
3,851
Location
The Royal Town of Sutton Coldfield, UK.
Real Name
Richard
Thanks for the review, finally one I can nearly understand. :thumbup:

Most reviews I have read in the past on related subjects have had results/conclusions as convoluted as the theory of the subject matter [for me anyway, no technical knowledge of anything electonics]

Your graphs and conclusions I can actually use, and I like the comparisons to the other two cameras for reference
Do you have any particular author in mind?
 

Drdul

Mu-43 Veteran
Joined
May 16, 2015
Messages
382
Location
Vancouver, BC
Real Name
Richard
Great information! I have the E-M5iii which I gather is the same sensor as the E-M1ii. I’ve set an ISO limit of 6400, but my experience with DxO’s DeepPrime has been so good that I was considering going higher. Based on your tests and your conclusion that anything beyond 6400 is just digitally amplified, I’ll leave it at 6400. Thanks!
 

retiredfromlife

Mu-43 Hall of Famer
Joined
May 15, 2016
Messages
4,539
Location
Sydney, Australia
I always shoot at ISO 250 since that is the true base ISO, where DR is highest. You can see it on the graphs, ISO 250 has higher DR than ISO 200. E-M1 mk2 shifts everything 1/3 stop. More graphs here: https://www.dpreview.com/forums/thread/4108345
According to the official Olympus Australia training videos during the COVID 19 lockdown they indicated many times the base ISO is 200.
They mentioned this to try and clear up the issue. Unfortunately these were live events and only available for a few days after if you had the youtube link so I cant point to them.

Since there are tests by other suggesting other base ISO's there maybe shooting conditions that influence the best ISO for DR under certain conditions.

I have no idea and just use whatever I need to get the shot, and not over 3200.
 

Machi

Mu-43 Top Veteran
Joined
May 23, 2015
Messages
766
I always shoot at ISO 250 since that is the true base ISO, where DR is highest. You can see it on the graphs, ISO 250 has higher DR than ISO 200. E-M1 mk2 shifts everything 1/3 stop. More graphs here: https://www.dpreview.com/forums/thread/4108345
Base ISO is where camera has the largest achievable full well capacity.
That means at ISO200 (~30,000 electrons).
Also that DR is most important aspect of IQ is myth. SNR is more important.
SNR tells one how pronounced noise is in the image. Higher DR of ISO250 is because lower read noise and that can be observable only when there is practically no light in the scene ("photon starved regime") and that happens practically only at night or if one shoots black cats in the entrance of tunnel! :)
In all other cases shot noise is dominant source of noise and shot noise is dependent on quantity of detected photons (which will be higher at ISO200).

But overall differences between ISO200 and ISO250 are 1 db in SNR and 1/3EV in DR.
That's negligible under most scenarios but on the other hand, from my experience, even 1/3EV loss of highlights can be annoying and highlights are best preserved at ISO200.
 

ibd

Mu-43 Top Veteran
Joined
Jun 5, 2016
Messages
534
Great information! I have the E-M5iii which I gather is the same sensor as the E-M1ii. I’ve set an ISO limit of 6400, but my experience with DxO’s DeepPrime has been so good that I was considering going higher. Based on your tests and your conclusion that anything beyond 6400 is just digitally amplified, I’ll leave it at 6400. Thanks!
I'm going purely from memory here, but IIRC, the e-m5 III and the e-m1 III have shifted ISOs by 1/2 a stop or so to get better shadow recovery at the cost of worse highlight recovery, compared to the e-m1 ii. Can anyone confirm and would that directly translate to a (left/down) shift of the graphs presented here as well?
 

Machi

Mu-43 Top Veteran
Joined
May 23, 2015
Messages
766
I'm going purely from memory here, but IIRC, the e-m5 III and the e-m1 III have shifted ISOs by 1/2 a stop or so to get better shadow recovery at the cost of worse highlight recovery, compared to the e-m1 ii. Can anyone confirm and would that directly translate to a (left/down) shift of the graphs presented here as well?
Yes, they've did it for E-M1X, M1III and M5III. That is well supported by multiple sources (photonstophotos, optyczne.pl, mirrorlesscomparisons).
 

Machi

Mu-43 Top Veteran
Joined
May 23, 2015
Messages
766
I have now different E-M1II as I've returned the first one because it was in combo with bad 12-40 f/2.8 (new one is good). That allowed me to do basic measurements for the second copy.
I found that differences are minimal. Differences are at max 0.05 EV in DR and 0.2 dB in SNR.
But I did now measurements for the high resolution mode (HR mode).

Here is table which shows results for read noise, gain and full well capacity for the second E-M1II (I didn't measured ISO500, 640, 1000 and 1250 for normal mode) and comparison with high res mode.
High res mode works here as "synthetic sensor" with given parameters.

TB2.jpg
Subscribe to see EXIF info for this image (if available)


Next is the graph for dynamic range (DR). It's the graph from the first post but with added results for HR mode.

DR2.jpg
Subscribe to see EXIF info for this image (if available)


The last graph is for signal-to-noise ratio (SNR) and again it's the graph from the first post but with added HR mode.

SNR2.jpg
Subscribe to see EXIF info for this image (if available)


Measurements for the HR mode show modest improvement (~0.7EV) of DR where best results are now at ISO100 and ISO320 (~13EV) and not ISO250 as for the normal shots.

More interesting is almost 5 dB improvement of SNR. That's really significant improvement and it means that daylight images should be at least comparable to those from Sony A7III or Canon 1DXIII.

Another interesting finding (for RAW shooters) is that optimal ISO for night hi-res photography is ISO1000 as it combines lower read noise with better highlight protection than maximum ISO1600.
 
Last edited:

L0n3Gr3yW0lf

Mu-43 Top Veteran
Joined
Jul 31, 2013
Messages
928
Location
UK
Real Name
Ovidiu
Thank you so much for the in-depth research on this. I gives me a lot of confidence on what already was a very good experience with Olympus OM-D E-M1 Mark III and Olympus OM-D E-M5 Mark II. I would love to see an opportunity to test the HHHR of wither E-M1 X or E-M1 Mark III.
Given the huge improvement from the 80 MP HR I wonder if we can get 50 to 75% of the 80 MP HR which would still be a significant improvement in terms of noise and IQ.
I have been using HHHR as much as I could even at very high ISO because it gives a lot of improvements that can be seen.
 

Machi

Mu-43 Top Veteran
Joined
May 23, 2015
Messages
766
Thank you so much for the in-depth research on this. I gives me a lot of confidence on what already was a very good experience with Olympus OM-D E-M1 Mark III and Olympus OM-D E-M5 Mark II.
You're welcome!
I would love to see an opportunity to test the HHHR of wither E-M1 X or E-M1 Mark III.
Given the huge improvement from the 80 MP HR I wonder if we can get 50 to 75% of the 80 MP HR which would still be a significant improvement in terms of noise and IQ.
I have been using HHHR as much as I could even at very high ISO because it gives a lot of improvements that can be seen.
HHHR should be even better as it is combining more images together (up to 16).

If you want to I can do the measurements. Only things which I need for it are two light frames and one bias frame (all RAWs) at the same ISO setting.

Two light frames must be taken at the same exposure (same exp. time, same aperture, same ISO) with camera metering around 0EV (0 to +0.7EV).

I'm using as light frames flat fields with short to medium tele lens (>85mm).
Here is described method how to obtain such frames using LCD monitor.
Personally I'm using tripod, laptop, camera + lens covered with white clear paper tissue.
On monitor I have open in full screen mode, empty, homogenous image with white or grey backround color.
Camera must be so close to the monitor that entire field of view is covered by that image on the monitor.
Focus must be set to infinity.
Aperture of the lens, brightness of the monitor and/or white/grey background color of the image is set so exposure time for the light frame is 1/10s - 1s. That mitigates problems with the monitor's refresh rate frequency. As camera should photograph only light from the monitor, it's good to do this under such conditions that monitor is the brightest light source in the room.

Bias is a dark frame taken at very low exposure time (ideally 1/8000s, at night, with lens covered and with lowest f) because bias should "image" only the noise of the camera (not light).

That said, I'm not entirely sure if this works for HHHR as it can be using some details in the image as guides for combining them together and that is not possible with practically empty frame.
So such experiment can also show if HHHR mode uses for final combination:

a) only data from IBIS -> then flat field and bias should be without problems

b) data from IBIS and details in the individual images -> then it can work or not

c) only details in the individual images -> then it probably should not work
 
Last edited:
Links on this page may be to our affiliates. Sales through affiliate links may benefit this site.
Mu-43 is a fan site and not associated with Olympus, Panasonic, or other manufacturers mentioned on this site.
Forum post reactions by Twemoji: https://github.com/twitter/twemoji
Copyright © 2009-2019 Amin Forums, LLC
Top Bottom