SLR Magic 8mm f/4.0 rectilinear lens for Micro 4:3 mount

[VidThreeNorth]

SLR Magic 8mm F4 M43 lens
Tested on Yi-M1

Primary Use:

I have been recording fireworks for a few years with varying success. Last year I had a problem so bad that I did not make a post-able finished version. I know the problem and I could probably correct it this coming year, but now that I have the Yi-M1, I thought it might be nice if I could get it done with this camera. To do this, I needed a "very wide" lens. My zooms cover 12mm, which is equal to 35mm camera 24mm ("eq 24mm") which is not really wide enough. A few months ago, I decided that I would probably want a Rokinon 10mm F2.8, manual lens. These cost around $400 US, which I could not afford, so I left it at that. There is also a new Venus Optics Laowa 7.5mm F2.0 lens selling for around $500.00 US, which I also cannot afford.
[2018-07-27 Correction: Laowa lens is 7.5mm, not 7.2mm]

B&H recently had a sale on the SLR Magic 8mm F4 for the Micro 4:3 mount which is intended for drone use. This lens normally sells for around $350 US. I bought one for around $180 US. I do not have a drone, but I felt that the specifications could be acceptable for my needs.

I have done some testing to verify its capabilities using the Yi-M1 which has a 20MP sensor, and I have mixed feelings about this. I am certainly glad I did not pay the full price.

I have tested it in UHD (3840 wide) video mode and Yi's unique 4:3 frame format 2K video mode, as well as in still images at full 20MP resolution.

First The Bad News:

The lens I received was assembled/adjusted improperly and the nominal "infinity" is not at infinity. I cannot tell precisely, but I think it might be focusing at around 1.0 M. Infinity is not even covered by depth of field as far as F11. I did not test F16.

Also, colour fringing is evident at all F-stops I tested (down to F11) and is even visible in the middle 1/3 of the image. There is some barrel distortion, and as expected, the sharpness decreases somewhat in the corners, depending on the F-stop.

Any Good News?:

I am not a technician, but I do think that I can probably correct the focus, if I can get the right tool. I think I need a metric 1.0 Allen key (hex key), or similar screw driver. I used to have that size, but I cannot find one now, and I do not know where I can get one.
[2018-02-18 Update]
I found my metric 1.0 hex key and it is too big. I compared the set screws to a ruler and the screw diameter seems to be a bit over 1mm (maybe 1.2mm?). The size of the hole for the hex key looks a bit bigger than 0.5mm (0.6mm?). I am still trying to find a vendor for any hex keys smaller than metric 1.0. Nothing is certain at this point.

Testing:

I have a lot of test images. I needed a lot of them because the depth of field is very large and the lens has issues, so it was very hard to tell what was going on. I am only going to post images from a couple of files. They indicate (somewhat weakly) the improper focus. I do not see any point in posting more at this time. If/when I get this lens adjusted to bring "infinity to infinity" then I expect to post more.

If you look at "industrial" (C-mount) lenses, they are generally rated as "xMP (megapixels)" for a sensor size. For example, a manufacturer might rate a lens at a "2MP lens for a 1/2.3" sensor".This implies that the manufacturer has rated the lens as having 2 megapixel image quality on a 1 over 2.3-inch sensor. That 1/2.3" sensor size will be a real limit, with the image often cut off beyond that size by vignetting. But the sharpness rating is up to the manufacturer and generally is the capability at full aperture. So a "1MP" lens will usually be good for 1280 x 720P ("HD") video work. A "2MP" lens would be recommended for 1920 x 1080P ("Full HD") video work. But a very good "1MP" lens might be good enough for "Full HD" when stopped down.
[2018-02-18]
A slight correction: If the sensor is 3:2, then a Full HD ("1980 x 1080") resolution will be about 2MP, which is what I stated above, but if the sensor is 4:3, then really you need 3MP performance to cover Full HD because the 16:9 frame of Full HD will not use the upper and lower rows of sensors.

My current estimate is that if/when I get it properly adjusted, this SLR Magic 8mm F4.0 lens will probably turn out to be something like a "1MP" lens. It might be usable for "Full HD", stopped down, but it is not going to be acceptable for "UHD".

I have not seen a review of the Rokinon 10mm F2.8 lens yet, but the company has a reasonably good reputation. So I think it is probably the better product. An early report indicates that the Laowa 7.5mm F2.0 lens is a better lens than this SLR Magic lens. It seems to me that it would be better to look for one of these other products instead. And it would be really nice if one of the major lens testers covered these products.


Original image format 5200 x 3902 Pixels
(20MP)

"P2060002.DNG"
Nominal settings:
F4.0, Focus Infinity
Partial EXIF data:
Date: Feb 6, 2018, 14:19:03
Pixel height 3888
Pixel width 5184
Color Space (irrelevant -- RAW)
Exposure program Normal
Scene capture Standard
Exposure mode Auto exposure
Exposure bias 0.00 ev
Exposure time 1/4000 sec.
ISO speed 200
Metering mode Center weighted average
Digital zoom ratio None
Gain control "Low gain up"

The following is not recorded in the EXIF data:
F number [nominal F4.0 -- fully open]
Focal length [nominal 8mm]
Focal length in 35mm [16mm]


All files are un-adjusted except for the conversion from RAW and then either cropping or resizing.

"P2060002-1a-rsz1640-C1.jpg"
- unadjusted conversion from DNG (raw) file, resized to 1640 x 1231

"P2060002-1b-Crop01-C1.jpg"
Crop from (2570,1400), size 1200 x 900
- look at the street light tops left from the billboard
- these are big enough to show the edges blurring from the left most to the right most. The blurring increases with distance.

"P2060002-1c-Crop02-C1.jpg"
Crop from (2100,1400), size -1200 x 900
- The "traffic island" sign and its post are not wonderfully sharp, but it does show the color fringing, and that the sharpness is not being lost in the nearer objects. They are probably a bit sharper, but the difference is so little that I will not claim to be able to tell.


"P2060020.DNG"
Nominal settings:
F8.0, Focus Infinity
Partial EXIF data:
Date: Feb 6, 2018, 14:25:31
Pixel height 3888
Pixel width 5184
Color Space (irrelevant -- RAW)
Exposure program Normal
Scene capture Standard
Exposure mode Auto exposure
Exposure bias 0.00 ev
Exposure time 1/1250 sec.
ISO speed 200
Metering mode Center weighted average
Digital zoom ratio None
Gain control "Low gain up"

The following is not recorded in the EXIF data:
F number [nominal F8.0 -- 2 stops down]
Focal length [nominal 8mm]
Focal length in 35mm [16mm]

"P2060020-1a-rsz1640-C1.jpg"
- unadjusted conversion from DNG (raw) file, resized to 1640 x 1231

"P2060020-1b-Crop01-C1.jpg"
Crop from (2570,1400), size 1200 x 900
- again, look at the street light tops left from the billboard
- The blurring still increases with distance.
This is also true of the hand rail on the far side of the bridge. I think it is more visible in this F8.0 picture than in the F4.0 picture

"P2060020-1c-Crop02-C1.jpg"
Crop from (2100,1400), size -1200 x 900
The "traffic island" sign and its post are sharper, but it still shows the color fringing, and that the sharpness is not being lost in the nearer objects. They are probably a bit sharper, but the difference is so little that I will not claim to be able to tell.

 

[VidThreeNorth]

Comparison F4 v F8 at 1 meter

I still do not have the tool I need to adjust this lens, so I am packing it away for now. Before I do so, I decided to make one last set of pictures to give an idea of how this lens is performing, which I think also indicates how it might perform after being adjusted.

The question on my mind for this set was where is the lens actually focused when set to nominally "infinity"? I previously stated that I think the lens is focused at about 1M. So what happens if I set up the camera 1M from something?

A secondary question is whether this lens can be considered good enough for, at least 720p and hopefully, for 1080p, and Yi's "2K" video mode. I will simply state that I have not seen any image yet that makes me feel that I would want to use it for "4K/UHD" on the Yi-M1, so I am not bothering with that possibility here.

As has been typical of my luck this past year, the weather was not bright and sunny, but was overcast with variable brightness and evenness. The camera was located roughly 1M from the biggest silver ball for this set of images and was steadied by a monopod. I took multiple images and selected the largest files that were available ("best of four" or more). The exposure times were 1/320 and 1/60, and though there was a breeze, motion blurring should not have been a problem.


Processing:
Corel PaintShop Pro 2018,
Asus M32BF, AMD A8-6500,
Windows 8.1 - 64-bit

NOTE: The ISO was forced to ISO-400 and exposure increased +1 in camera. Since the main subject matter was still dark, I increased the brightness by +0.3 during the conversion for a final total of EV = +1.3 backlight compensation. No alterations were made in these images after the conversion beyond resizing or cropping. Specifically, there were no alterations to reduce noise or to increase sharpness.


"P2210007" [F4.0]
Temperature 4858
Tint 19
Brightness: +0.3

Partial EXIF:
Date and time February 21, 2018, 14:38:44
Image width 5200
Image height 3902
Pixel height 3888
Pixel width 5184
Exposure program Normal program
Scene capture type Standard
Exposure bias 1.00 ev
Exposure time 1/320 sec.
ISO speed 400 (forced)
Metering mode Center weighted average
Gain control High gain up
Following not provided in EXIF:
F number [F4.0]
Max aperture [F4.0]
Focal length [8mm]
Focal length in 35mm [16mm]

Despite a 1/320 sec. exposure time which should have prevented it, the twig on the far left of the display shows a ghost image that resembles motion blurring, but examination of other files at full aperture show the same ghost image, so it must be caused by the lens. I think that this might be an effect of spherical abberation. The instruction sheet included with the lens does say that the lens has noticeable flare which goes away around F5.6, however, the sky conditions were somewhat variable, so the sky light flattened by cloud cover might be adding to the milky cast of flare. I think that most of the flare level is probably a result of the lens, but I think that more testing should be done to be certain. I think that spherical abberation is causing some of this softness. This should show up as a focus shift when changing the aperture, but it is hard to tell due to the depth of field.

"P2210007-1a-rsz900-C1.jpg"
This image has been resized to 900 x 1200 which gives a reasonable estimate of how this lens would perform in a 1280 x 720p video frame.

The general detail level is pretty good, however, I would not really want to use it wide open due to the "milky" appearance which, for now, I am attributing to flare. Using F5.6 reduces the flare to a more acceptable level (and sharpens the image), so I would probably rate this as an F5.6, 1MP lens.

"P2210007-1b-Crop01-C1.jpg"
Crop start @ (1770,2290)
See comments below.


"P2210016" [F8.0]
Temperature 4919
Tint 19
Brightness +0.3

Partial EXIF:
Date and time February 21, 2018, 14:39:49
Image width 5200
Image height 3902
Pixel height 3888
Pixel width 5184
Exposure program Normal program
Scene capture type Standard
Exposure bias 1.00 ev
Exposure time 1/60 sec.
ISO speed 400 (forced)
Metering mode Center weighted average
Gain control High gain up
Following not provided in EXIF:
F number [F8.0]
Max aperture [F4.0]
Focal length [8mm]
Focal length in 35mm [16mm]

"P2210016-1b-rsz-1231-C2.jpg"
This image has been resized to 1230 x 1640 which gives a reasonable estimate of how this lens would perform in a 1920 x 1080p (Full HD) video frame. This file has been compressed one level more than the other files (C2) in order to make it uploadable. There is some detail lost but it is still higher quality than most 1080p MP4 files, so it is still a good indicator of potential 1080p performance.

There is not much to say about this image beyond the obvious. It is sharper and more detailed than the F4.0 image, and does not have the "milky" appearance. If you want some idea of whether the lens could be used at 1080p you can tilt your monitor on its side and view this image. Of course, you really do not have to do that. If you zoom the image and look at the detail level you should have some idea of the performance potential. Personally, I think that the sharpness and detail, though not "wonderful" will probably be adequate -- at F8. But that makes this a very dark lens for such usage. I will be testing F5.6 for video usage as soon as I can because realistically, I do not record all my videos on sunny afternoons.

"P2210016-1c-Crop01-C1.jpg"
Crop start @ (1710,2280)

Comparing the detail cropped files. These are full resolution and at minimum compression. I think the only thing to take away from these are that higher contrast in in the subject translates in to more apparent sharpness in the final images, which is a generality of digital imaging. These detail crops actually make some of the distant objects look sharper than nearer objects, which is ironic. However, as a general observation I will say that I think my original estimate of the lens being focused at 1M seems to be just a bit off, though only "just a bit". Looking at these files and some others, it does seem to be a bit further. But only out to around 1.1M or 1.2M. It is still in this general zone.

The only thing I can add is that if I can adjust this lens, I do think it will be good enough for me to make a good enough fireworks video using Yi's 4:3 format "2K" video mode.

 

[VidThreeNorth]

I finally got a set of screwdrivers that included one that would fit the SLR Magic 8mm F4 lens. I was not really happy with the quality of the tools. Also looking at the screws in the lens, they were also badly made.

I think that B&H should put together a "Manual Lens Maintenance Kit" which includes the proper screwdriver and a couple of spare set screws for this lens, and maybe another lens. I have run into complaints about the Samyang 10mm lenses which sometimes also come mis-adjusted. If the parts are the same, then the kit could be sold for both.

A lens cleaning kit could be included to bring the price up to a level worth stocking.

The Tests:
Using Yi-M1 and tripod

I prefer a clear sky sunny day for lens testing. Photographers prefer somewhat overcast days for general photography. But when I am testing a lens, I prefer a maximum contrast scene to provide consistent performance baseline images. This lens did get tested on a fairly cloud-free day, but unfortunately, I cannot release those images. The day I took these images were completely overcast, though not with heavy rain clouds. The images are not ideal, but "pretty good".

I did not adjust the lens as I wanted to, but it is now focusing beyond infinity. That means I can focus it correctly, but it does take some time. Also, I might have accidentally moved the aperture control and I will have to check that later. I think the F-stops are close enough to where they should be, but I might have to post a correction note someday later.
[2018-07-26 02:31 it is now focusing ...]

As far as I am concerned, the performance of the lens is so generally poor, there was only one F-stop setting that really works, and that is roughly F8.0. It is stretchable from around 6.3 to around 9.5. When I say it "works", I mean for "Full HD" on a camera that can use close to the full sensor area at that resolution. If sensor cropping is substantial, I would not really want to use this lens even for that much.
[2018-07-26 02:31 clarification.]

As noted before, the lens should be quite good for use recording 720P, but I would not want to use it for UHD. When you see the uploaded images, you might feel that the lens might be good enough for UHD, but I feel that its performance is too close to borderline, and the results would not have enough image quality for re-processing.

Location:

York University, near west entrance and the pond.

The following images are intended to SIMULATE video conditions. Image "P7230011.jpg" is actually a still 20MP image which I have cropped to the sensor coverage of the Yi-M1 at UHD.


P7230011.JPG
- Not uploaded
I did not record what I had focussed on but I think it was somewhere on the tall grey cement building near the middle of the picture.

partial EXIF data from DNG:
Software ASDK-00141
FlashPix version 01.00
Date and time Jul 23, 2018 10:58:25
Pixel height 3888
Pixe width 5184
Component configuration YCbCr
Color space Uncalibrated (AdobeRGB)
Exposure program Normal program
Exposure bias 0.00 ev
Exposure time 1/640 sec
F number: [~F8.0]
Focal length 0.0mm [~8.0mm]
ISO speed 200
Metering mode Spot [my error, should have been Center Weighted]
Gain control Low gain up

"P7230011aJPG-rsz1900-C1.jpg"
This crop shows the theoretical sensor usage of the YiM1 when recording its UHD videos. It is encouraging that a true UHD frame size resulting in a file that was much too big to upload. I had to reduce it to 1900 pixels wide to get an small enough image file at C1 compression. The sensor crop compensation for an equivalent 35mm view was 2.4x. This makes it roughly equivalent to a 19.2mm lens view when recording UHD.


"P7230011bJPG-1640cropC1.jpg"
This detail crop shows the POTENTIAL sharpness and detail of this LENS when used on the Yi-M1. You will NOT actually get this level of sharpness and detail during an actual video recording because the compression method is different between a still image and a video.


"P7230019.MP4"
- Not uploaded
- Focus is on the farthest buildings near the middle of the frame.
- Exposure is +2 (spot metered)

"P7230019-Big-2048-C1.jpg"
I recorded videos using the Yi-M1 "2K" format. The videos were 2048 x 1536 frame size and used the full sensor area. I then captured frames. Both these image are full sized and unaltered except re-compressed using JPEG at full detail (C1) compression. This first is the full "2K" frame.
[2018-07-26 02:38 reworded for clarity.]

"P7230019-Big-FullHDcropC1.jpg"
This image file is cropped to represent the image coverage of the Yi-M1's "Full HD" video. It is probably a very accurate representation of the image quality potential of the lens and this camera for Full HD MP4 video.


Would the Lens Perform Better On a Different Camera?

The sensor crop for the Yi-M1 when recording UHD is greater than many other M4:3 cameras, and the image quality of the Yi-M1 is already pretty good. Would this lens be good enough for UHD on some other M4:3 camera? That is an interesting possibility. I would need to do such a test to say for sure. However, if you do not have this lens, then I think the real question is whether it is worth buying, and I think the obvious answer is still that if you can afford one of the better lenses, you might as well do so.

see: "Micro 4:3 "Very" Wide Angle Rectilinear Lenses"

[2018-07-30]
- adding 2 more image files:

"P7230011_JPG-rsz1640-C1.JPG"
This is the whole still image frame resized to 1640 x 1230. There are no adjustments beyond the resize.

"P7230011cJPG-Crop2Crnr-C1.JPG"
This is the upper right corner of the whole 20MP still image frame. This is at full resolution with no further adjustments.