Focus Chart, 4k Tiff

[Shawn Nelson]

It's boring, but no one has posted one, so here we go. This was shot with my Red using the Red Zoom by the DP on my last short, Ryan Walters.

http://brainspasm.com/red/FocusChart4k.tiff

[LEON]

It's boring, but no one has posted one, so here we go. This was shot with my Red using the Red Zoom by the DP on my last short, Ryan Walters.

http://brainspasm.com/red/FocusChart4k.tiff

Thank you for posting this.
Would you know what were the lens settings ?

[Shawn Nelson]

I don't, perhaps Ryan will chime in.

[Simon Blackledge]

have you noticed the error all the way round the outside of the image? Some form of offset on the boarder?

S

[Fix]

It looks little soft. Or is it just me?

[NuclearNerd]

Way soft. To my eyes the resolution maxes out about 1000 lines / picture height (10 on the chart), which is only 36 lp/mm at the sensor. The nyquist resolution of the mysterium is 92 lp/mm. I would have expected at least 50% nyquist from a good lens in sharp focus.

[Michael Mann]

Yes, it's too soft, especially on the right side of the frame. Any chance to share another shot with settings, Shawn? Thanks in advance.

[Fix]

Yes I can also see its softer at the right side. At first i thought "my good, is this a result of the lens?!". But when you take a look at the left it seem pretty much alike the middle, so i guess it could be an angle problem.

Since its obviously very difficult to focus, wouldn't it be an idea to slowly focus while record. Than pick out the sharpest frame...

[Álex Montoya]

Shawn's footage, while nice, has been reliably soft. I don't know what's the problem though.

[Ken Corben]

Sorry,

Didn't download Shawn's chart and I am in no way attacking Shawn or his DP - Shawn's a great guy and it's cool he offered us a look at his test.

I can say that the RED 18-50 lens on a burst chart shot at 4K 2:1 24 fps 180deg 18mm and in the f4-8 stop range yielded a resolution chart result that is ABSOLUTELY STUNNING when viewed at 1:1 in photoshop from a 4K DPX or TIFF file generated from RED ALERT.

No, I'm not posting the umpteen MB file but SOFT is not a word to describe the image I saw. It's a guess, but perhaps the posted file may have software processing issues rather than what Shawn is really seeing?

[Kevin Halverson]

Thanks for posting this Shawn, greatly appreciated. If we could find out the aperture used, that would be most useful in evaluating the results. Looking at it, I don't think it was a focus issue, but more likely a wide open lens test. However, rather than speculate it would be good to know how it really was shot.

[Shawn Nelson]

Shawn's footage, while nice, has been reliably soft. I don't know what's the problem though.

Part of that is that most of the clips I have posted were at T3. Post anything at T3 and you're going to have softness.

However, I have heard that several of the Red's were delivered without having the PL mount properly collimated, thus I am somewhat afraid that I might have this issue, but I am not sure, thus one of the reasons I posted this, to see the reaction. Honestly, it felt soft to me.

[Stephen Williams]

Part of that is that most of the clips I have posted were at T3. Post anything at T3 and you're going to have softness.

However, I have heard that several of the Red's were delivered without having the PL mount properly collimated, thus I am somewhat afraid that I might have this issue, but I am not sure, thus one of the reasons I posted this, to see the reaction. Honestly, it felt soft to me.

Hi Shawn,

How did you focus the camera? If by eye at the focal length you were at it's probably OK. If you zoomed in to focus then the camera FFD & lens would have to be accurately set. Setting the focus by tape could also be an issue as there are no marks on the lens where the exact focus is. relative to the distance.

Stephen

[Kevin Halverson]

If improperly collimated, the image wouldn't necessarily be soft, unless the procedure used was to 'zoom' in to get focus and then 'zoom' out to fill the frame (if improperly collimated, this would lead to focus errors). What would be most useful is to see the various chart grabs with the lens stopped down over a 3 or 4 stop range. Just have the ability to compensate lighting (pull some scrims or turn on additional banks) to keep the exposure reasonably close.

[Shawn Nelson]

Hi Shawn,

How did you focus the camera? If by eye at the focal length you were at it's probably OK. If you zoomed in to focus then the camera FFD & lens would have to be accurately set. Setting the focus by tape could also be an issue as there are no marks on the lens where the exact focus is. relative to the distance.

Stephen

I'm usually focusing by eye, Graeme's focus assist, or by pulliing tape and then trying to guess where that is on the Red Zoom (since the Red Zoom will jump from 3 to 4 feet and then from 10 to 20 without any tick marks in between). I have never done a zoom in focus.

[Kevin Halverson]

I'm usually focusing by eye, Graeme's focus assist, or by pulliing tape and then trying to guess where that is on the Red Zoom (since the Red Zoom will jump from 3 to 4 feet and then from 10 to 20 without any tick marks in between). I have never done a zoom in focus.

Then collimation errors won't effect the ability to get good focus except for a tape approach. Would still like to see grabs at different apertures.

[Stephen Williams]

I'm usually focusing by eye, Graeme's focus assist, or by pulliing tape and then trying to guess where that is on the Red Zoom (since the Red Zoom will jump from 3 to 4 feet and then from 10 to 20 without any tick marks in between). I have never done a zoom in focus.

Hi Shawn,

As the image seems softer on the right I think of 2 possible causes.

1) The sensor is not parallel to the mount, this can easily be set with shims, however it's one of the reasons I don't like to change mounts at home if it cam be avoided.

2) The lens is not set up properly and is soft at one side. If you can go to your local rental house to see your lens on their projector, it will show up straight away. Also check for focus shift as you zoom. If they have a light transmission test rig checking out the T stop markings can also be of interest.

To get the best out of any camera & lens collimation is unfortunately a issue even with digital cameras.

Stephen

[Stephen Williams]

Then collimation errors won't effect the ability to get good focus except for a tape approach. Would still like to see grabs at different apertures.

Hi,

If the lens was way out, then its possible that optimal focus cannot be achieved with a zoom lens (or any lens that focuses with floating elements). Please note I am not saying the lens is way out, the only way to know is put it on a collimator.

Stephen

[Shawn Nelson]

To get the best out of any camera & lens collimation is unfortunately a issue even with digital cameras.

Stephen

I strongly feel that 100% of the Reds and Red Zooms need to come from the factory properly collimated and tack sharp. I'm sure you, and others, expect to have to tweak, but just look at how many of the Red purchasers don't even know what 'collimate' means. Jim told me himself that he wants to ship 5 to 10 thousand Reds by the end of 2008. If they aren't all tack sharp from the factory, I'm afraid they will have thousands of angry customers.

Now I am not saying mine is off, that remains to be seen. It's strongly possible that this chart was shot wide open at T3 (to test the worst possible case). Ryan's analysis of this shot was,



Using the RED 18mm - 50mm Lens, I was more then satisfied with the results- This is a sharp lens and imager that is able to resolve more then enough detail. In fact, This test chart that I had printed at the local Kinko’s does not go far enough in the horizontal and vertical scales to adequately show the detail capable in this camera. Both of those scales go up to 1,000 and you can clearly see the defined lines in both scales at the 1,000 mark. Using the flutes, it appears that the lines are blending together at around the 1,600 marks.

[Kevin Halverson]

Hi,

If the lens was way out, then its possible that optimal focus cannot be achieved with a zoom lens (or any lens that focuses with floating elements). Please note I am not saying the lens is way out, the only way to know is put it on a collimator.

Stephen

It would have to be "way out" for this to be probable. Parallelism to the imager is an issue that can be checked, but a wide open lens is going to partially swamp the ability to make this determination. Without knowing the aperture, this is all speculation.

[Graeme Nattress]

I use a combined 2k and 4k linear sinusoidal zone plate for measuring resolution - it's a custom job made for me. It's very good as you can see resolution at any angle, and by plotting the graph is intensity over distance (ImageJ is good for this) you can practically read off precise resolution by measuring at what distance from the centre of the chart the amplitude hit's zero, and you can also see the amplitude of any aliasing. It's a real torture test for any imaging system.

Graeme

[Shawn Nelson]

Graeme, can I get a digital copy of this zone plate? It sounds like I'll also need Red to mail me a pack of shims so I can get my Red mount properly dialed in.

[Graeme Nattress]

Sorry, it was made specially just for me and I can't give it out. If I find a source for you to buy them, I'll post it.

Graeme

[jbeale]

Just FYI, the Imatest program can generate a variety of test patterns including linear and circular sinusoidal zone plates. If you're printing these out yourself, you'd have to put them pretty far away so they're small in the frame in order to be sure it's the camera you're testing, and not the printer.
http://www.imatest.com/docs/tour_testcharts.html#charts

If you want to play with one right away, I put some PNGs here:

http://www.bealecorner.org/red/test-patterns/
Linear zones w/contrast ramp: Linear-ZonePlate.png (560 k)
Circular zone plate: Zone-600ppi.png (13 MB)

Good luck getting a clean printout! For a fine sinusoidal pattern, an inkjet printer will probably do better than a laserprinter since they can do better greyscales.

[Kevin Halverson]

A cheap and simple test that I have used for decades to evaluate optics is something that everyone can do at home without any special materials.

Just take wall and cover it with pages form any newspaper. Its an uncalibrated test, but if you make it large enough (I once did one that was about 10' x 40') it is really revealing of performance. At various distances, you can test just about every aspect of a camera system and its certainly good enough for relative comparisons of one lens to another (I used to do film tests this way too).

[Graeme Nattress]

Other thing is to print the chart as big as you can....

Graeme

[NuclearNerd]

Norman Koren makes a pretty useful sinusoidal plate which you can download for printing here: http://www.normankoren.com/Tutorials/MTF5.html. He also gives instructions on how to measure the resolution from it using ImageJ.

You can also analyse the resolution using any sharp black to white edge (the funny angled bars in your original photo are there for this purpose). You'll need a software mtf analyser such as imatest or quickmtf (http://www.quickmtf.com/) which will generate a curve relating resolution to contrast. Usually the "system resolution" is quoted at either 50% contrast (MTF50) or 10% contrast (MTF10)

I wish more Red owners would post test chart images. It isn't too hard to find free hosting sites for big files.

[Andrew Benz]

Norman Koren makes a pretty useful sinusoidal plate which you can download for printing here: http://www.normankoren.com/Tutorials/MTF5.html. He also gives instructions on how to measure the resolution from it using ImageJ.

You can also analyse the resolution using any sharp black to white edge (the funny angled bars in your original photo are there for this purpose). You'll need a software mtf analyser such as imatest or quickmtf (http://www.quickmtf.com/) which will generate a curve relating resolution to contrast. Usually the "system resolution" is quoted at either 50% contrast (MTF50) or 10% contrast (MTF10)

I wish more Red owners would post test chart images. It isn't too hard to find free hosting sites for big files.


This thread, these responses, your post and links are all greatly appreaciated.

[I Bloom]

Wouldn't the Graeme's Magic Focus Assist, be able to tell you if it was slightly out... Maybe this isn't public knowledge...

IBloom

[Ryan E. Walters]

Thank you for posting this.
Would you know what were the lens settings ?

I did this test to see what the lens looked like at it's worst. The chart I used while decent- it is a 30" chart printed out at Kinkos using one of their high quality B&W printers - it is not a full quality resolution chart. But it is good enough for what I wanted to test.

Using the Red 18-50mm lens, the lens was set at 50mm, at T3 (f2.8). All lenses when WFO are soft. Most lenses start performing their best when they are stopped down one or two stops. Shawn set up the camera, and I pulled focus via measuring and double checking through the LCD. (Although I'm not sure that I would trust such a small LCD to resolve a 4k image.)

Hope that helps. :)

[Kevin Halverson]

Thanks for the confirmation of the wide open lens test Ryan. That's exactly what I was assuming based upon the frame grab posted.

So, any interest in posting some grabs at other apertures?