Compression Ratio Advice for a first time RED user

[Adam Griffith]

Hello Everyone,

I'm Adam, I've been stalking this forum for a long while now and have actually found myself a used Epic-X Dragon 6K here on the forum that I'll be picking up soon. It'll be my first RED and I'm rather excited.

I have a question about compression as I'm yet to shoot with this thing and test it out first hand. I will be shooting at 50fps (Australia) as I'd like to slow it down a bit in post but I want to know what will give me the best results in regards to compression and resolution options.

I plan to deliver in 4K WS as a final product.
So do I A) shoot in 6KWS at 50fps which will give me a compression ratio of 9:1 (128GB REDMAGS)
or B) shoot in 4K WS with a 4:1 ratio.
Keeping in mind my end product will be 4KWS, which will give me the cleanest image? 6K with higher compression downscaled to 4K or shoot 4K at lower compression and leave it 4K throughout the whole process?

My theory is shoot 6K and any artifacts or noise etc from the higher compression will not be obvious once I export it out at 4K. Plus 6K will give me some post stabilization options over 4K source. Is that what you'd do?
Everything I read here says "Shoot as high res as you can even if you only need 1080 in the end".
Will I notice a significant difference between the compression ratios mentioned above once both @ 4K?

Thanks in advance for any input.


Adam

 

[Akin A]

I would get a Mini-Mag module, 512GB/1TB Mini-Mags and shoot 6K, lowest compression.

 

[Adam Griffith]

I would get Mini-Mag module, 512GB Mini-Mags and shoot 6K, lowest compression.

Yea I would too...however a MiniMag side module plus a single 512GB mini-mag would put me out of pocket a further $5800 AUD with todays exchange rates...

At this point in time I'm trying to work out what will net me the best results with what I've got right now.

 

[Michael Tiemann]

I shoot 8:1 all the time, and compression is not my #1 problem. I think you should be fine at 9:1. That said, if you shoot 5K WS, you can split the difference compression-wise, and still have some over-scan, which is often very nice to have.

 

[Maik Müller]

Some ideas to save space and get more runtime.

Wide open shots with blury background and not an extreme amount of details can easily be shot with higher compression, no 4:1 needed here. 9:1 6kWS will be fine, you can even compress a little more. 10-12:1 often works just good.

You have also access to other resolutions. 5.5k, 5k, 4.5k. It's not just 4kWS vs 6kWS. In regards of noise and sharpness you are fine with all of those options. The more the better, depending on your shot more or less visible. You'll often have a hard time to see a difference between 6kWS and 5kWS in terms of sharpness and noise (given it's not a special shot, like lowlight at night, where every bit for noise-reduction counts).

More resolution gives you other advantages (stabilizing, re-framing, a little for compensating focus-errors) and a wider FOV of course.

Better compression can handle more details better, so 5k with lower compression might be worth to consider over 6k with higher compression.

It's actually not really necessary to always shoot 50fps, hell you don't even have to shoot 50fps (or 25). I know, PAL and all, but we live in an international world. You could shoot for cinema and dial-in flickerfree shutterspeed if needed. speed-up to 25/50 for tv in post, like it's done with cinema, for everything else you can deliver 23.98/24. So shooting 24/48 would save some bandwidth too.

You say you want to slow-down a little in post, but everything??? Have a look into the ramping feature and shoot HFR when needed, otherwise standard fps. Test around with other framerates (you said "a little"). There is some room between 24 and 48. You can make presets, which you can load up with just one button on the side-handle. So jumping from 6kWS24 to 5kWS48 and 4.5kWS with even more fps is pretty easy.

EPIC is pretty flexible. 128GB Mags might be a bit limiting for what it can do and especially if you try to shoot everything in HFR/50fps.

 

[Adam Griffith]

Thank you for your suggestions, they've definitely provoked some extra thought.
Run time and saving space isn't a huge deal as I have the ability to offload as I shoot on most of my jobs. Hard drives are cheap and I've always had a "shoot more, never delete" mentality in my photography and cinematography.
My attraction to RED was the frame rates at exceptional resolutions in RAW. I like to shoot at 50 for the flexibility in post as the majority of my shoots aren't super planned out and structured. The option is always there and like I said - Hard drives are cheap.

I'm mainly concerned with the quality of the end product. Do I sacrifice resolution over compression or the other way around?
I've just watched Mark Toia's video (THE REAL DEAL) which shows plenty of footage at 17:1 at 100fps. It's given me the confidence that the Dragon can be pushed pretty far and still give great results (assume i nail exposure at the time).

 

[Tom Gleeson]

Adam,

Generally its just safer to use lower compression rates but I am always amazed how good really high compression rates can look. I am forced to high compression rates when I use high frame rates but its never been an issue just don't underexpose at high compression rates as it seems to exagerate the noise.

The nature of R3D Wavelet compression is the more you compress the more detail you throw away so if you plan to zoom in and crop your images the high compression will see the image go softer sooner.

But go with the higher resolution settings and live with the higher compression it is the best way to go. Remember one reason your 6K Epic looks so good is that large sensor "look" so I avoid cropping down unless I want to increase DOF

 

[Patrick Tresch]

Don't shoot 4k for 4k when you shoot RGB CMO's. Allways shoot higher res to get a good downscale. RGB CMO's loose 30% of res due to the way they handle colors and create an image.

I would shoot up to 12:1.
8:1 is a good way to go.
And lower than 8:1 if you can afford.

My 2 cent.

Pat

 

[Adam Griffith]

Don't shoot 4k for 4k when you shoot RGB CMO's. Allways shoot higher res to get a good downscale. RGB CMO's loose 30% of res due to the way they handle colors and create an image.

I would shoot up to 12:1.
8:1 is a good way to go.
And lower than 8:1 if you can afford.

My 2 cent.

Pat

Might have to help me with your terms "RGB CMO's". What does that mean? I'd perfer not to shoot 4k for 4k due to the crop factor. However there will be some shots I want at higher frame rates that I'd shoot in 4K WS to achieve the rate for super slow-motion. But 100fps with 6K WS (17:1) should give a better result?

As far as down-scaling 6K in post to 4K....I've only ever shot 4K before (DJI phantoms, a7s etc) and achieved the downscale to 1080p (for Vimeo) in Adobe Premiere by just setting the scale % and/or position (for cropping 16:9 to different ratio project) in premiere. Is this the best way to do it to ensure the best possible quality from the downscale or should I downscale using another method PRIOR to the edit and VFX etc in Premiere? My other thoughts were to make a 6K project and edit everything in 6K and just export to 4K at the very end, I've assumed (with no factual references) this to be the better way.

 

[jake southard]

Here's an interesting question i've always wondered, say you're shooting in my case, with a Scarlet Dragon, with 1.8" SSD's, 5K slow motion at 16:1 (minimum compression) ...
Does switching to Mini Mags allow for lower compression rates? and if so, what would it move my current 16:1 compression to in 5K? Has anyone done tests for this comparison?

 

[Akin A]

Here's an interesting question i've always wondered, say you're shooting in my case, with a Scarlet Dragon, with 1.8" SSD's, 5K slow motion at 16:1 (minimum compression) ...
Does switching to Mini Mags allow for lower compression rates? and if so, what would it move my current 16:1 compression to in 5K? Has anyone done tests for this comparison?

No, Scarlet Dragon compression rates are the same for 1.8" Mags and Mini-Mags.

http://s3.amazonaws.com/red_3/downl...ps_Guide/DSMC_Media_Operation_Guide_Rev-N.pdf
Page 85

 

[Marcus Friedlander]

RGB CMOS sensors use a Bayer pattern filter (Red, Green, Blue, Green) in order to achieve color. The cavet to this method is that approximately 30% of the captured resolution is lost in order to obtain the color information from the light hitting the sensor. This 30% resolution loss is one of the main reasons to shoot over scanned images. If you shoot a bigger then 30% overscan, by the time you deliver you should be measuring the full resolution of the deliverable.

Also, this is why the DRAGON MONOCHROME can produced a sharper image then the normal DRAGON sensor.

 

[Adam Griffith]

Here's an interesting question i've always wondered, say you're shooting in my case, with a Scarlet Dragon, with 1.8" SSD's, 5K slow motion at 16:1 (minimum compression) ...
Does switching to Mini Mags allow for lower compression rates? and if so, what would it move my current 16:1 compression to in 5K? Has anyone done tests for this comparison?

As per Akin - Scarlet the same, however the DRAGON will improve by 1 at most resolutions and frame rates. So 1:9 ratio on the 1.8"Mag will be 1:8 on the MiniMag

 

[Patrick Tresch]

+1 from Markus's words.

...Is this the best way to do it to ensure the best possible quality from the downscale or should I downscale using another method PRIOR to the edit and VFX etc...

This is a good question but you should probably talk to your finishing house. I you use Resolve for ex. do it at the end of your post production and stay as long as possible with the R3D. But if you have a more complexe route only testing is the way to go. All depends on the tools you choose and the algorithm they use.

4k low compression slow mo VS 6k high compression slowmo for 4k delivery is also a test you should do by yourself. You'll learn a lot and much more than what I could give you (or the forum).

Pat

 

[Mohammed Almedfa]

I would definitely tend towards at least 5 ->5.5K, shooting a bigger-effective sensor is worth a lot in and of itself. But a side-note I think you'll really appreciate when shooting on Dragon, especially at 6K is the dramatic reframing capability, which is still quite solid when delivering at 4K.

For example in a recent shoot I did on Dragon, we shot 6K WS & had frame guides set for 5.5K WS & 5K WS on our monitors. It was an intense 12-hour overnight shoot on a tower top and we had to be in & out within that timeframe - plus the camera was constantly moving - it was on the Ronin or on the Ronin-head mounted to a crane 90% of the time (couple of handheld shots, and 1 tripod shot otherwise) so speed was integral and being able to have the flexibility to know you're shooting such high resolution that you can seamlessly 'choose' to do the next shot at the 5K or 5.5K frame when its already balanced & mounted on a crane and in position and every minute counts - those luxuries of 6K shooting are really amazing

 

[Adam Griffith]

+1 from Markus's words.



This is a good question but you should probably talk to your finishing house. I you use Resolve for ex. do it at the end of your post production and stay as long as possible with the R3D. But if you have a more complexe route only testing is the way to go. All depends on the tools you choose and the algorithm they use.

4k low compression slow mo VS 6k high compression slowmo for 4k delivery is also a test you should do by yourself. You'll learn a lot and much more than what I could give you (or the forum).

Pat

I am the finishing house. It's just me. I don't use resolve currently, I may in the near future but for now its just the Adobe CC package for me, Premiere with Lumetri and speedgrade. I think I'll have to try the various methods like you say and see which is best.
Thanks

I would definitely tend towards at least 5 ->5.5K, shooting a bigger-effective sensor is worth a lot in and of itself. But a side-note I think you'll really appreciate when shooting on Dragon, especially at 6K is the dramatic reframing capability, which is still quite solid when delivering at 4K.

For example in a recent shoot I did on Dragon, we shot 6K WS & had frame guides set for 5.5K WS & 5K WS on our monitors. It was an intense 12-hour overnight shoot on a tower top and we had to be in & out within that timeframe - plus the camera was constantly moving - it was on the Ronin or on the Ronin-head mounted to a crane 90% of the time (couple of handheld shots, and 1 tripod shot otherwise) so speed was integral and being able to have the flexibility to know you're shooting such high resolution that you can seamlessly 'choose' to do the next shot at the 5K or 5.5K frame when its already balanced & mounted on a crane and in position and every minute counts - those luxuries of 6K shooting are really amazing

You've given me another option to think about, thats a clever way of speeding things up and with the shoot I've got in mind that method could save me a lot of time. Thanks for that, I've just got to see if 100fps is enough for a couple scenes we have planned. I might shoot two takes - one at 6k WS 100fps and then at 4K ws at the higher frame rate. I can decide in post which works best.

Thanks

 

[sam karr]

This is my personal opinion. I would shoot 5k for the reframing capability and extra FOV. 6k is nice for stills but overkill for motion. Too much data, and downscaled almost nobody will notice the difference. It seems Hollywood still shoot big projects in 5k on Dragon. Also people should be more relaxed with compression, it's getting a bit extreme sometimes. The noise on my red is looking nice. Whenever I watch a movie shot on the alexa in HD/bluray I see more noise in the background that I never had on my scarlet D which can "only" shoot @7:1 or 8:1 /24p, even in slowmo. It's hard for me to understand why people shoot 6k @3:1 when even international blockbusters don't go beyond 5k @5:1. And many times to just stay on vimeo and youtube.

 

[Marcus Friedlander]

This is my personal opinion. I would shoot 5k for the reframing capability and extra FOV. 6k is nice for stills but overkill for motion. Too much data, and downscaled almost nobody will notice the difference. It seems Hollywood still shoot big projects in 5k on Dragon. Also people should be more relaxed with compression, it's getting a bit extreme sometimes. The noise on my red is looking nice. Whenever I watch a movie shot on the alexa in HD/bluray I see more noise in the background that I never had on my scarlet D which can "only" shoot @7:1 or 8:1 /24p, even in slowmo. It's hard for me to understand why people shoot 6k @3:1 when even international blockbusters don't go beyond 5k @5:1. And many times to just stay on vimeo and youtube.

There are a lot of incorrect statements or assumptions here that I would like to fix.

Since 5k falls in the super35 image circle, it would be better considered as "normal" with 6k being there for the extra FOV. Also, its about a 30 percent increase in data to go from 5k to 6k. For major shows like the hollywood blockbusters you are describing, a 30% increase in data means a roughly ~30% increase in cost PER camera. Since most major shows shoot on multiple cameras, that extra 30% is a major cost. However, on smaller shows ironically, 6k makes more sense. I just got done shooting a project aimed for theatrical delivery that was shot with one camera, mostly at night, in a forest. For those reasons, I decided to shoot at 6k with a 5:1 Redcode. The reframing was essential to us making our days, and that 5:1 Redcode will make sure the detail in the trees, barks, and leaves will hold up throughout the post process.

Also, it seems like there are two other concepts you are not entirely clear on. The first, is the degraded differences between the image captured and the image projected, and the second is the method in which Redcode actually compresses.

The first is simple, even in the age of file based workflows, almost every post process is lossy in some way shape or form. So, if we start with the cleanest possible image, we are more likely to end with an acceptably clean image. You even said it yourself, the image you see out of your Scarlet Dragon looks better then Bluray DVD's and that is absolutely to be expected.

The second is a bit more complicated but I highly recommend you read Gavin Greenwalt's terrific thread on compression. http://www.reduser.net/forum/showthread.php?124071-Compression-Wedges
Redcode is a compression scheme so by definition it has to throw away data somehow. The difference between Redcode and the average compression scheme however is that Redcode is not a blanket compression scheme, it tries to throw away the least important data first, and then maintain the rest. What's important to remember is that the higher your compression, the less color detail you have. When you are heavily color correcting, and/or projecting onto a theatrical screen, that extra color data is the difference between the footage looking eye popping, and the footage looking meh. However, because the difference between 3:1 and 5:1 usually isnt worth the extra money, most productions land at around 5:1 for theatrical delivery.

 

[Anton Shavlik]

What you'll notice right away with higher redcode compression is:

1. The noise is blockier.
2. Shots of leaves and other things with lots of high frequency random detail throughout the frame will suddenly look much worse than other shots, even shots with the same compression settings. Suddenly there is a maximum amount of detail you can fit into a frame.
3. In post you will pull less clean keys for green screen and color correction. Technically, extremely fine details will soften and your color resolution will worsen but I don't think those are noticeable without A/B pixel peeping at 800%.

8:1 is pretty good for most shots, but I have had shots looking up into tree tops that ate up all 8:1 had to offer. IMO shooting 3k and below can't afford to lose detail or have blocky noise so keep the compression lower there as well.

I recall a thread a while back where the European Broadcasting Union (?) tested and found that the bayer sensor of the dragon cleanly resolved about 3.6k resolution at 6k. I think they required black and white line pairs be resolved with perfect attenuation or something similar, in other words an extreme definition of 'resolution'.