Results 1 to 12 of 12

Thread: Results from removing AA filter from E-510

  1. #1
    Join Date
    Dec 2007
    Location
    Canada
    Posts
    1,727
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Feedback Score
    20 (100%)

    Default Results from removing AA filter from E-510

    Had the AA filter removed from my E-510 and did some testing against the E-3 to see what the difference is. Thought everyone would like to see the results, since the strength of the AA filter in recent bodies has been a topic of discussion.

    Used the Leicasonic 25/1.4 lens, as it delivers the most resolution of any lens I have (including the 50/2) and outresolves the 10 MP Panasonic sensors, so as to remove the lens from the equation.

    Set up the camera on a Manfrotto X190PROB tripod, IS off, manual exp, 2 sec anti-shock, 2 sec timer, ISO 200, NF OFF. Used a slant-edge test target about 7.5 ft away and shot using magnified LV and manual focus to get an approximate sharpest focus setting. Then took dozens of RAW shots (over 200-some odd... used up entire battery for both the E-3 and E-510) between f/2.8-5.6 with micro adjustments to the focus each time and processed them through ACR 4.4 at the default settings except for contrast, which was set to 0. Discarded results that were not good (meaning they were not accurately focused) and ended up with about 77 files to analyze through Imatest. Saved the top 3 test results from each body and discarded the rest.

    Plotted a graph of the MTF curves from the two cameras using the corrected MTF values from Imatest up to the Nyquist frequency, which is about ~2700 lw/ph for a 10 MP sensor and corresponds to about 105 lp/mm for those of you used to that from the film era.

    Took all of Saturday and this morning to shoot and process the files through ACR and Imatest. Hope you enjoy the results...



    It is interesting to note that even the E-510 with no AA filter does not deliver MTF50 approaching Nyquist. This is in contrast to the L10, which delivers MTF50 up to ~2650. This suggests that there is more at work here than just the AA filter. I'm too tired to process and graph all the L10 files, so you'll just have to take my word for it

    Also, this is testing the E-3 vs the E-510 with its AA filter removed. It is not a direct test of the effects of the AA filter. There are too many variables that have changed, including the camera, sensor, and tuning of the RAW conversion. Even if it were two E-510s being tested with the only difference being the AA filter, it might not be considered definitive, but at least it would eliminate a lot of the variables.

    Nevertheless, it does give you an idea of the resolution limits of the two bodies, and one might say that the E-3 delivers lower ultimate resolution than an E-510 with its AA filter removed. You can say that a lot of this difference is likely due to the AA filter, but there is nothing concrete to back that up.
    Last edited by e_dawg; 08-03-2008 at 09:29 AM.
    4/3: Oly E-3, E-1, E-520, 9-18, 11-22, 14-54, 50-200, 70-300, 25/2.8, 35/3.5 | Leica 14-50, 14-150, 25/1.4
    m4/3: E-P3 | G1 | 14/2.5 | 20/1.7 | 14-42 IIR | 17/2.8 | 45/1.8 | Nik 20/3.5
    Nikon: D90, D700, 70-300/VR, 24/2.8, 35/1.8, 35/2, CV 40/2, 50/1.4G, 60/2.8, 85/1.8, 180/2.8 | f/2.8 zooms | Zeiss 25/2.8
    Sony: A55, A900, 24-85, 70-210/4, 20/2.8, 24/2.8, 30/2.8, 35/1.4G, 50/1.7, Zeiss 85/1.4
    P&S: Canon S90

  2. #2
    Join Date
    Sep 2005
    Location
    Sweden
    Posts
    6,393
    Thanks
    1
    Thanked 8 Times in 5 Posts
    Feedback Score
    5 (100%)

    Default Re: Results from removing AA filter from E-510

    Good work e_dawg!

    The results is pretty much as expected isn't it? Did you replace the AA filter with anything else to compensate for the changed optical path length?

    Cheers, Jens.
    Motto: Wildlife won't come to me unless I go to it.
    --------------------------------------------------------
    My Wildlife Photos: jensbirch.smugmug.com

    E-5, E-3, E-510, IR-E-1 ,E-P2
    ZD: 7-14, 14-54, 50, 50-200 SWD, 90-250/2.8, 300/2.8, EC-14, EC-20
    Peleng 8mm fisheye, shift Tamron SP 17/3.5, Tokina AT-X 300/2.8
    FL-50R, FL-40, FL-20, HLD-2, HLD-4, cleaved ZD EX-25 w. electric bypass, 250D, 500D, KatzEye Plus OptiBrite
    Feisol CT-3472LV and CM-1471

  3. #3
    Join Date
    Nov 2007
    Location
    Wisconsin, USA
    Posts
    1,066
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Feedback Score
    0

    Default Re: Results from removing AA filter from E-510

    Wow, now that's a thorough test!

    A couple questions:

    1) Did you go hunting for people in plaid shirts, to coax some aliasing or mazing out of the E-510?

    2) Any thoughts on the possible effect on the IS mechanism? You've lightened the mass it must shift, correct? I could imagine lightening the IS mechanism's load could either make it go completely haywire (slamming the now-lighter sensor around harder than it needs to), or slightly improve its capability (as there's less mass to move, the mechanism can resist more severe shocks than before).

    And of course this begs the question: what happens if you remove the L10's AA filter? (Sorry!)

    Thanks for posting your findings. Assuming the E-510 no better than matched the E-3 before the surgery, and you aren't suffering from heavy aliasing now, the improvement seems worth the effort. But it also makes me scratch my head as to what Panasonic did to achieve such higher numbers even with an AA filter in place!
    Chris


  4. #4
    Join Date
    Mar 2006
    Location
    BC, Canada
    Posts
    2,379
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Feedback Score
    2 (100%)

    Default Re: Results from removing AA filter from E-510

    Did you replace the AA filter with anything, or was the sensor left bare under the dust buster? Ideally one would try a filter swap with an L-10.

  5. #5
    Join Date
    Dec 2007
    Location
    Canada
    Posts
    1,727
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Feedback Score
    20 (100%)

    Default Re: Results from removing AA filter from E-510

    Quote Originally Posted by jebir View Post
    Good work e_dawg!

    The results is pretty much as expected isn't it? Did you replace the AA filter with anything else to compensate for the changed optical path length?

    Cheers, Jens.
    Personally, I would say the results are as expected, but given the level of doubt and aspersions surrounding dpreview's test results and their assertions that it is a relatively strong AA filter that seems to be responsible for reducing resolution of newer generation Olympus bodies, not everyone would say that.

    To clarify, I didn't do the procedure myself. Sent it to LDP/maxmax.com and had them work on my E-510 while I was on vacation. They replaced the AA filter with whatever optical glass they usually use for this type of thing. It does have a turquoise tint, however, which means you will have to either set up a custom profile in ACR or whatever photo editor you use, or do a custom WB when processing your shots.

    Due to the significant colour shifting going on, you lose about half a stop of sensitivity / noise performance and should shoot in RAW and process your images in 16-bit space.

    Due to the nature of the Bayer filter, per-pixel colour information, and higher potential for aliasing and multicolour moire, I suspect there would be a greater benefit when using this camera for B&W photography instead.
    4/3: Oly E-3, E-1, E-520, 9-18, 11-22, 14-54, 50-200, 70-300, 25/2.8, 35/3.5 | Leica 14-50, 14-150, 25/1.4
    m4/3: E-P3 | G1 | 14/2.5 | 20/1.7 | 14-42 IIR | 17/2.8 | 45/1.8 | Nik 20/3.5
    Nikon: D90, D700, 70-300/VR, 24/2.8, 35/1.8, 35/2, CV 40/2, 50/1.4G, 60/2.8, 85/1.8, 180/2.8 | f/2.8 zooms | Zeiss 25/2.8
    Sony: A55, A900, 24-85, 70-210/4, 20/2.8, 24/2.8, 30/2.8, 35/1.4G, 50/1.7, Zeiss 85/1.4
    P&S: Canon S90

  6. #6
    Join Date
    Mar 2006
    Location
    BC, Canada
    Posts
    2,379
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Feedback Score
    2 (100%)

    Default Re: Results from removing AA filter from E-510

    Quote Originally Posted by e_dawg View Post
    Due to the significant colour shifting going on, you lose about half a stop of sensitivity / noise performance and should shoot in RAW and process your images in 16-bit space.
    Sounds like the replacement filter doesn't have as high a throughput in the visual spectrum as the stock one does. Quite a high drawback to add on top of the inevitable aliasing, to gain a little resolution.

  7. #7
    Join Date
    Feb 2006
    Location
    Tequesta, Florida
    Posts
    4,115
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Feedback Score
    63 (100%)

    Default Re: Results from removing AA filter from E-510

    Shawn,
    I have absolutely no idea as to how to read MTF charts even though I've been doing this stuff for a few years.
    In layman's terms, what did you find? Better? Worse?
    Thanks.

  8. #8
    Join Date
    Mar 2007
    Posts
    1,652
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Feedback Score
    0

    Default Re: Results from removing AA filter from E-510

    Quote Originally Posted by ckrueger View Post
    And of course this begs the question: what happens if you remove the L10's AA filter? (Sorry!)
    i thought that was pretty funny, but also makes you wonder a bit, pretty clever observation

    Quote Originally Posted by ckrueger
    Thanks for posting your findings. Assuming the E-510 no better than matched the E-3 before the surgery, and you aren't suffering from heavy aliasing now, the improvement seems worth the effort. But it also makes me scratch my head as to what Panasonic did to achieve such higher numbers even with an AA filter in place!
    i wonder too about a property i call image acuity, a certain gritty sharpness you see it in 5D, M8 and SD14
    i think 5D might have a weakish AA filter to begin with, perhaps some of you that operate one agree (or not)

    over at dp, to recap, they posted horizontal resolutions of
    E3 2100 to 2400 (extinction)
    510 2050 to 2300
    L10 2200 to 2450
    L1 1750 to 2100
    M8 1900 to 2100
    5D 2300 to 2500
    400D 2200 to 2300
    SD14 was quite unfairly never tested

    just (just) my view but they seem to push *any* low statistic to the extreme if its Olympus
    I really want to see more of what 510 operates like given the modification
    Riley

    Olympus User, Pro Photographer since 2003

  9. #9
    Join Date
    Dec 2007
    Location
    Canada
    Posts
    1,727
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Feedback Score
    20 (100%)

    Default Re: Results from removing AA filter from E-510

    Quote Originally Posted by Bill_Turner View Post
    Shawn,
    I have absolutely no idea as to how to read MTF charts even though I've been doing this stuff for a few years.
    In layman's terms, what did you find? Better? Worse?
    Thanks.
    Here's an analogy that might help:

    If you've ever had the pleasure of comparing an optimized HDTV broadcast to regular cable, you know how much improvement to the viewing experience better resolution can provide. But HDTV is a huge jump in resolution from regular cable (330 lines to 720+).

    But ~20% higher resolution (if this test is actually representative) by removing the AA filter is more akin to comparing 1080p vs 720p feeds of the same content. You're probably not going to notice a difference on a 37" plasma or LCD. But on a 63", you might.

    With photos, it essentially means you're taking the equivalent of a 9 MP camera (if you subscribe to the notion that Oly uses a relatively strong filter and therefore does not deliver as much resolution as the average 10 MP camera) and "upgrading" it to an 11 MP camera. This means you're probably not going to notice a difference with anything on-screen (just like if you were upgrading from a 10 MP camera to a 12 MP one), but you will if you print large / crop. I'd say if you print 13x19 or larger, or do cropped 8x10's, you will notice a difference.

    We're talking about improvement in resolution at the limit here, so of course one comes back to the whole photographer vs equipment debate, using a tripod or not, filter or not, JPEG vs RAW, etc. And speaking of tripods, if you're not using a tripod (unless you're using shutter speeds MUCH faster than the 1/f rule), you won't be getting anywhere close to the resolution limit of your camera. (wanted to save Godfrey the trouble of having to chime in with that important tip )
    4/3: Oly E-3, E-1, E-520, 9-18, 11-22, 14-54, 50-200, 70-300, 25/2.8, 35/3.5 | Leica 14-50, 14-150, 25/1.4
    m4/3: E-P3 | G1 | 14/2.5 | 20/1.7 | 14-42 IIR | 17/2.8 | 45/1.8 | Nik 20/3.5
    Nikon: D90, D700, 70-300/VR, 24/2.8, 35/1.8, 35/2, CV 40/2, 50/1.4G, 60/2.8, 85/1.8, 180/2.8 | f/2.8 zooms | Zeiss 25/2.8
    Sony: A55, A900, 24-85, 70-210/4, 20/2.8, 24/2.8, 30/2.8, 35/1.4G, 50/1.7, Zeiss 85/1.4
    P&S: Canon S90

  10. #10
    Join Date
    Dec 2007
    Location
    I am based in Moorpark California
    Posts
    112
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Feedback Score
    0

    Default Re: Results from removing AA filter from E-510

    A fascinating read. Will you be working in the IR with your modified camera?

  11. #11
    Join Date
    Sep 2005
    Location
    Sweden
    Posts
    6,393
    Thanks
    1
    Thanked 8 Times in 5 Posts
    Feedback Score
    5 (100%)

    Default Re: Results from removing AA filter from E-510

    By the way e_dawg,

    I suspect that the IR light, which now is a significant part (maybe ~50%) of the recorded light, will noticeably affect your results. The IR light, due to its different refraction in the lenses, has optimum focus at a different focus setting than visible light and will therefore deteoriate the sharpness of this image.

    Also, I wonder if you ever will reach the Nyquist limit as long as you are using a Bayer interpolating RAW developer?

    Cheers, Jens.
    Motto: Wildlife won't come to me unless I go to it.
    --------------------------------------------------------
    My Wildlife Photos: jensbirch.smugmug.com

    E-5, E-3, E-510, IR-E-1 ,E-P2
    ZD: 7-14, 14-54, 50, 50-200 SWD, 90-250/2.8, 300/2.8, EC-14, EC-20
    Peleng 8mm fisheye, shift Tamron SP 17/3.5, Tokina AT-X 300/2.8
    FL-50R, FL-40, FL-20, HLD-2, HLD-4, cleaved ZD EX-25 w. electric bypass, 250D, 500D, KatzEye Plus OptiBrite
    Feisol CT-3472LV and CM-1471

  12. #12
    Join Date
    Dec 2007
    Location
    Canada
    Posts
    1,727
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Feedback Score
    20 (100%)

    Default Re: Results from removing AA filter from E-510

    Quote Originally Posted by ckrueger View Post
    1) Did you go hunting for people in plaid shirts, to coax some aliasing or mazing out of the E-510?
    No, but it can happen with a fair number of repeating patterns. The worst is probably denim. If I'm doing a fashion or architecture shoot, it won't be with the E-510

    2) Any thoughts on the possible effect on the IS mechanism?
    No change in the mass, as the AA filter was replaced by optical glass.

    And of course this begs the question: what happens if you remove the L10's AA filter? (Sorry!)
    Probably a minimal increase in resolution and a slight increase in aliasing and moire. We're already close to Nyquist here, so no need.

    Thanks for posting your findings. Assuming the E-510 no better than matched the E-3 before the surgery, and you aren't suffering from heavy aliasing now, the improvement seems worth the effort.
    As acme mentions above, there is no free lunch. The optical glass imparts a turquoise tint on your images unless you do a custom WB or profile. That resulting colour balance shift to reduce the green channel and raise red and blue gain reduces sensitivity and increases noise by up to half a stop. But that is not a big deal most of the time, as you wouldn't be after maximum resolution in poor lighting conditions in which you would have to crank up the ISO, nor is the higher potential for aliasing / moire a big concern unless I was a pro fashion photographer or if it was my only body.

    Ultimately, this is just my backup/fun camera now, since the resale value dropped a fair bit when the E-520 came out. I wasn't going to use it much after I got the E-520, so I thought I might as well get the AA filter removed to see what would happen. Someone had to do it And in fact the resale value probably went up if I wanted to sell it.

    But it also makes me scratch my head as to what Panasonic did to achieve such higher numbers even with an AA filter in place!
    The L10 could have an AA filter that is tuned to start just before Nyquist instead of significantly before it. Or, could it be possible that they could have used a higher order design with a steeper cut-off at a higher frequency? (have no idea, really)

    But I suspect part of the difference is just that the Oly RAW files still have a small amount of NR applied either on-sensor or in the imaging ASIC before saving to the card, or both. Most cameras these days can have NR applied on the sensor itself or on the RAW data before saving to the card, and manufacturers often design their cameras with a certain baseline level of NR applied to all files -- wouldn't be surprising if Olympus did that to mitigate noise complaints.

    Panasonic may have chosen to disable NR for RAW files, preferring to give the photographer full control over NR -- possibly influenced by Leica on that as they use Panasonic bodies.

    This seems the most plausible explanation to me, as I notice my L10 RAW files are a bit noisier than any of my Olympus files at the same settings.

    Another possibility is that the demosaicing / deBayering algorithms are better optimized for the Panasonic RAW files than they are for the Olympus files.
    4/3: Oly E-3, E-1, E-520, 9-18, 11-22, 14-54, 50-200, 70-300, 25/2.8, 35/3.5 | Leica 14-50, 14-150, 25/1.4
    m4/3: E-P3 | G1 | 14/2.5 | 20/1.7 | 14-42 IIR | 17/2.8 | 45/1.8 | Nik 20/3.5
    Nikon: D90, D700, 70-300/VR, 24/2.8, 35/1.8, 35/2, CV 40/2, 50/1.4G, 60/2.8, 85/1.8, 180/2.8 | f/2.8 zooms | Zeiss 25/2.8
    Sony: A55, A900, 24-85, 70-210/4, 20/2.8, 24/2.8, 30/2.8, 35/1.4G, 50/1.7, Zeiss 85/1.4
    P&S: Canon S90

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •