Scarlet sharpness workflow question

[Matt Ryan]

Heres a few more R3D's:
This shot was as the sun was going down on an overcast day in SF:
http://matthayslett.com/wp-content/uploads/2012/02/Alcatraz-Shipovercast.zip

Shot wide open on a CP.2. Truckyard at night with available light only:
http://matthayslett.com/wp-content/uploads/2012/02/truckyardavailable.zip

Dark street shot from a bridge overhead. Available light only:
http://matthayslett.com/wp-content/uploads/2012/02/Darkstreet.zip

I know this is off from the main topic but I wanted people to have access to some other R3D's.

 

[Matt Ryan]

The only lenses I saw that resolve 4K are the Master Primes. Even RPPs are not "poping" sharp at 4K. I think you can find some examples on-line.

A lot of people would disagree with you:
http://www.reduser.net/forum/showthread.php?72863-Ready-Aim-Fire&p=944831&viewfull=1#post944831
http://www.reduser.net/forum/showthread.php?72863-Ready-Aim-Fire&p=944978&viewfull=1#post944978

And from Ketch who owned Master Primes as well as RPP's:
http://www.reduser.net/forum/showthread.php?72863-Ready-Aim-Fire&p=945121&viewfull=1#post945121

Ultra Primes were chosen by Andrew Lesnie for The Hobbit and by Peter Lyons Collister ASC for The Amazing Spiderman. There's plenty of lenses that "resolve 4K" and are sharp other than Master Primes.

 

[Matt Ryan]

This thread is about sharpness at 100% crop. Your examples are of downscaled images that will look sharp no matter what lens.
The Ultra Primes are even more soft, but for a feature nobody is looking for an ULTRA-POPING-ELECTRONIC-SHARP look. That is why many people prefer to shoot with something that softens things out.
I own RPPs and a Scarlet, and I'ved shot with RPPs in R1s and they are too sharp most of the time. That is because I'm going for a cinematic look most of times. But if you shoot 2K crops, you start to have problems.

RPPs are great, but MPs are sharper. That is all. And this proves that the problem is not with the camera, it's a limitation of the lens.
If you shoot for stills and need a 100% crop, use MPs and you'll have a better result in sharpness and detail wise.

If you can't test it yourself, there are examples here in this forum. For example take a look here:

Quote from the thread: "Master Primes are unbelievably sharp! Damn. RPPs are second (from what I could judge) and Cookes are a tad soft."
http://www.reduser.net/forum/showth...omparison-frames-Cooke-Master-Prime-Red-Prime

Or download the 3 dpx and look in photoshop at 100%.
http://www.cinematography.net/lens-comparisons.html

Where are you seeing downscaled images? I posted links to the full res and the R3D's. Of course down scaled would appear sharper, that's not what we're discussing.

The thing we are discussing here is about the original posters question. Even downscaled the image lacks focus and even though he is using Canon lenses, there are plenty examples of sharper images at 100% crop such as Phil Hollands image a few posts back.

I'm not trying to argue with you, I'm just stating that there are other lenses that match up to the Master Primes close enough to get a sharp looking image, Canon glass included as found by Mark Toia in his tests (he owns a set of MP's). On his tests with Canon lenses even a 400% crop looks sharper than what Miles is getting. Master Primes are superior, of course, but there are others that can get you a sharp image.

http://www.reduser.net/forum/showthread.php?71717-FINALLY!!!-Got-my-Canon-mounts-and-just-tested-the-hell-out-of-it!&highlight=

All I'm trying to say is Miles should be getting a sharper, not completely soft image like he's been getting.

 

[Jacek Zakowicz]

5.4 micron pixel calculates to about 80-90 line pairs per mm. Every modern half decent lens resolves that easily. I project them all day long and see 200.
Now from the posted graphs I've seen OLFP starts falling off at about 50 until cutting off completely before the 90 or so pixel pitch. the total system MTF is a function lens * OLFP. So we are seeing a combination of the two.
Not surprising really all things considered.
That's why RED leaves it up to the end user to apply proper sharpness...

 

[Matt Ryan]

I agree with you that there are many lenses that are good enough to even cut with MPs in the same scene and nobody notice. I also agree that you can and should work sharpness in post. And you can solve many problems with soft images coming from lenses like Canon, Nikkor, Tamnron, etc. But, to get an ULTRA sharp and DETAILED image coming out of a 100% still image from Epic or Scarlet, at least as seen in my 30" monitor, I have only seen Master Primes doing it.

I believe you are a nice guy trying to help Miles, and I'm sure you did not meant anything bad when you said that there are many people that would disagree with my remark about only the MPs resolving 4K.

Just to get it clear I'll point out what I was referring to.

The images on both links that you used in your first reply to my statement about MPs resolving 4K. Both are downscaled jpgs from Jim's posts. There are no R3Ds in your original reply to my statement. You posted R3Ds earlier. I have no reason to be referring to you earlier posts.



Also I believe that the original post was the one bellow, and is very clear to me that he was talking about it being "a bit soft when viewed at 100%."



I believe that he could get away with that footage without a problem if he was finishing anything in movement. Specially if he would deliver it in 2K or 1080p. But like him, I was shocked to look at the 100%, full debayer images that came out of my Scarlet PL Mount with Red Pro Primes and find it a bit soft. Also I was shocked to see the 2K 60 fps footage that looked very soft and grainy, but in that case without the possibility to downscale. That is why I went out of my way to find out that the camera itself is capable of producing real 4K images but the lenses are not. In 2K shoots MPs make a lot of difference.

Since that, I realized that I should have gotten an Epic-X that was only 17K more expensive but I would be able to shoot 5K 120 fps and use my RPPs without a problem. But with a Scarlet I need a set of MPs to shoot 2K 60 FPS. So it's Cheaper to have an Epic if you shoot more than 24 fps often.

What makes Canon lenses so good for Toya is that he has an Epic and can Shoot 5K anytime. For me and Miles 2K is a reality and the flaws of the lenses are much more pronounced.

Hey Paulo,
Firs I'd like to apologize if anything I said offended you, it wasn't my intention. Im merely trying to state that his image is softer than normal. Of course in movement or downscaled it will be fine, but if you look at other examples here with users who shoot with Canon lenses they are getting sharper results than Miles at 100%.

Also, I've been shooting with the Red One MX for years now and consistently get sharp images at 100%. The Scarlet contains the same sensor, upgraded electronics/processing, better compression, which to me would make it seem like you should at least get the quality that we were all getting on the Red One MX if not better.

That is where my argument is coming from. I understand that glass plays a huge role in getting the most resolution possible, but in this case he should be getting a sharper image than what he is getting. My apologies for any offense.

This is what I am referring to, Phil Hollands example:

I've been following this thread from my dirigible fortress in the sky.

Not having been there, inspecting the lens, or shooting these tests I personally have no real idea what's going on. However, I thought I'd show a 100% crop from a 4k motion frame for some comparison. This is a processed image.

Even this 2K grab from the Scarlet is sharper:

I seem to get mixed results. Some shots I am "Wow". Some I am "Meh".

This is a grab from 2k. 8:1 on a Nikon 105 2.5. This 2k is perfect for me. My wife would be so mad if she knew I posted these. Just a little s curve in REDCine-x, a little alchemy and clarity. I think it looks great. Highlight rolloff is wonderful. So nice compared to the absolute death of the 7d rolloff.

2k!
Full size jpg: http://mattjohnstonsfp.com/pics/A002_C020_020807.jpg

R3D grab if you'd like:
http://mattjohnstonsfp.com/pics/A002_C020_020807.0000200F.R3D

But I can't always seem to replicate this. Especially the Scarlet's noise.

Once again, this is the type of sharpness I'd expect for a processed image, and even when Miles image was processed it was nowhere near this.

Your right that lens flaws are an issue and I've already stated that earlier on in this thread as well as several other possibilities for his outcome. I don't want to carry out an argument, we can agree to disagree but personally I think Miles should be getting better results even with the Canon glass.

 

[Paulo Kobayashi]

Cool. Thank you for all the info. Agreed. :drool:
All I want is to get the good images. I'll keep shooting and learning.
I hope Scarlets do not differ from one to another because of difference in quality grade of the sensors.

Thank you anyway. It wasn't offensive, we all just want to learn.

 

[Peter Majtan]

Just to point out that Canon still glass may lack many cine features - such as manual and de-clicked smooth aperture, solid focus and zoom movement, etc - but sharpness is not one of them, especially with the L glass. After all these lenses resolve easily higher megapixel sensors for stills in DSLRs (and I'm not talking just the FF35 either), 4K is walk in a park for them. Some softness is the result of the OLPF, but as I have quoted before from Greame's post - unlike other cameras, RED leaves the sharpening to the user in post. No offense to anyone, but 99% of soft images I have seen are focus and/or shutter related issues. After all most stills are taken at shutter speeds exceeding 1/500 unlike cine standard of 1/48. Even the slightest movement of either the camera or the subject (let alone both) will result in softer image...

Just my two yen...

 

[Lliam Worthington]

It's an excellent two yen.

 

[Phil Holland]

I agree with what Peter had to say. Besides shutter speed and focus; aperture, motion blur, filtration, etc. can all effect final image sharpness.

Another not, remember 100% of a 4k frame and 100% of a 2k frame are identical in terms of what Scarlet can output at 1:1 pixel resolution.

If you are interested in testing sharpness and want a tack sharp image off of a Red camera: find a stationary detailed subject, lock off the camera on a stable surface or tripod, stop the aperture down to f/8-f/16, shoot with a shutter of 1/250th or higher, lock critical focus with the 1:1 pixel zoom, and let it roll for 12 seconds without touching camera during that time. Try to keep the subject within 10-20 feet of the camera (if you shoot something a mile away atmospheric distortion can effect the image).

I have a photographer friend who once wondered why all his product shots he captured on a very precise motion control rotating surface were always moving a bit after he started rolling. He most have checked out the smoothness and level on that motion control base for days. Upon entering his shooting are he was in a room with a bit of a bendy floorboard effect on thick carpet. Any time he moved his feet near the tripod he was effecting the shot, which is especially noticeable at the extreme macro levels he was shooting with a telephoto lens. He just never picked up on that slight movement and that he in fact was creating it.

 

[Miles Heckendorn]

Thanks everyone for the comments and suggestions so far!

I'm sorry I haven't been able to respond the last few days. I've been out on shoots and tomorrow I'm off for a week long shoot outside the country.

But as soon as I get back I'm supposed to get together and test with another Scarlet user.

Hopefully the PL mount will be here by then, but even if it isn't I'll place the same lens on both scarlets and post the results. I'll also make sure to have a test chart in frame as well.

 

[Matt Ryan]

Thanks everyone for the comments and suggestions so far!

I'm sorry I haven't been able to respond the last few days. I've been out on shoots and tomorrow I'm off for a week long shoot outside the country.

But as soon as I get back I'm supposed to get together and test with another Scarlet user.

Hopefully the PL mount will be here by then, but even if it isn't I'll place the same lens on both scarlets and post the results. I'll also make sure to have a test chart in frame as well.

Your not alone in getting soft images from the Canon mount:
http://www.reduser.net/forum/showthread.php?73225-What-I-learned-today-getting-my-Scarlet-and-lenses-checked-by-a-lens-tech

It should be fixed by adjusting flange depth on the adjustment screw on top of the camera. Same procedure as adjusting the pl mount because it's adjusting the position of the sensor, not the mount.

 

[Craig Wright]

Thanks everyone for your contribution. It is certainly beneficial to read through such discussion.

Miles it would be very helpful to give the results of your adjustments when done - it would certainly give a final perspective to much of the thought shared.


I have been considering the apparent issue of 'back focus' and a few thoughts have come to mind. I may be way off the mark. I would be interested in getting some opinion on the validity of this.

The broadcast camera world is no stranger to back focus. Indeed each time a new lens is attached the back focus is carefully checked. My understanding of its function is to calibrate the position of the 'rear focus element' on the lens to the position of the camera sensor - ensuring correct hardware setup to enable sharp focus. RED (and possibly others?) introduces a paradigm shift. Instead of adjusting the 'position' of the rear lens element, it is the sensor position that is adjusted. Both achieve the same result but from different perspectives. If this is right, the implications are interesting.

Simplistically it would suggest that each time a different lens is placed on the camera a back focus check should be performed. This in its self indicates there is a possibility of a different setting for each lens. Looking a bit deeper this is not necessarily as bad as it seems. RED have the back focus adjustment tool which suggests that all RED lenses are calibrated identically - and once set there is only a need to fine tune periodically. Also being PL glass and of higher quality, it is what you would expect. Further I am assuming most cine glass, also being of high quality will be OK with the RED tool calibration. However, it is the still glass that raises the interesting question. Given the understood, sometimes significant variation in QC, as well as the variety of manufacturers, and given the ability of the RED to expose the weakness of 'average' quality lenses, does it not stand to reason that to ensure the sharpest picture possible from each lens one would need to check the back focus of each lens - potentially resulting in a different setting for each? (It is possible that any differences may be so small so as to be essentially un-noticeable or unnecessary to adjust - however this could only be determined by the user AFTER each lens had been individually checked) I guess what I'm saying is that the back focus check is not so much a necessity for the camera as such but rather for the lens - and will be determined by the number/variety of lenses used . . . .

 

[Jacek Zakowicz]

Craig the broadcast tech and motion picture tech for BF are completely different worlds. They have very little in common- and for a good reason.
In the MP world typically both lenses and camera are set in prep and left alone for the duration of the job. It applies to all lenses and cameras. That may change but as of now it stands. We make tools to check lenses and cameras but skill is still needed to do it correctly. And that's the biggest hurdle right now IMO- educating users how to perform tasks traditionally left for techs in labs.
Good luck with that!!!!

 

[Craig Wright]

In the MP world typically both lenses and camera are set in prep and left alone for the duration of the job.

I imagine you're suggesting A lens being matched with A camera and left for the duration of the job. Therefore there are real implications for changing lenses mid shoot, or for that matter between shoots on the basis that 'I had back focus checked recently so all should be well'. To ensure the 'sharpest possible image', back focus should be checked at least every time a lens is changed? And depending on the lens/quality the adjustment may be non existent to significant?

Thanks in advance for your input. This is a new concept for me and one that appears crucial to the ability to capture sharp images.

 

[Jacek Zakowicz]

I imagine you're suggesting A lens being matched with A camera and left for the duration of the job. Therefore there are real implications for changing lenses mid shoot, or for that matter between shoots on the basis that 'I had back focus checked recently so all should be well'. To ensure the 'sharpest possible image', back focus should be checked at least every time a lens is changed? And depending on the lens/quality the adjustment may be non existent to significant?

Thanks in advance for your input. This is a new concept for me and one that appears crucial to the ability to capture sharp images.

No they are not matched. They all are set to baseline standard so they are truly interchangeable. They typically last the whole job and when they return they are inspected and corrected so they are ready for the next job. This discipline is not maintained in the owner/operator realm for various reasons and hence the issues..

 

[Craig Wright]

Thanks Jacek. Appreciate your help.

 

[Jesper Sichlau]

Thanks everyone for the comments and suggestions so far!

I'm sorry I haven't been able to respond the last few days. I've been out on shoots and tomorrow I'm off for a week long shoot outside the country.

But as soon as I get back I'm supposed to get together and test with another Scarlet user.

Hopefully the PL mount will be here by then, but even if it isn't I'll place the same lens on both scarlets and post the results. I'll also make sure to have a test chart in frame as well.

Have you cheched your camera yet? I'm beginning to wonder if my camera has the same issue.