FCPX Color Wheels Take 2

Prior to version 10.4, the color correction tools within Final Cut Pro X were very basic. You could get a lot of work done with the color board, but it just didn’t offer tools competitive with other NLEs – not to mention color plug-ins or a dedicated grading app like DaVinci Resolve. With the release of 10.4, Apple upped the game by adding color wheels and a very nice curves implementation. However, for those of us who have been doing color correction for some time, it quickly became apparent that something wasn’t quite right in the math or color science behind these new FCPX color wheels. I described those anomalies in this January post.

To summarize that post, the color wheels tool seems to have been designed according to the lift/gamma/gain (LGG) correction model. The standard behavior for LGG is evident with a black-to-white gradient image. On a waveform display, this appears as a diagonal line from 0 to 100. If you adjust the highlight control (gain), the line appears to be pinned at the bottom with the higher end pivoting up or down as you shift the slider. Likewise, the shadow control (lift) leaves the line pinned at the top with the bottom half pivoting. The midrange control (gamma) bends the middle section of the line inward or outward, with no affect on the two ends, which stay pinned at 0 and 100, respectively. In addition to luminance value, when you shift the hue offset to an extreme edge – like moving the midrange puck completely to yellow – you should still see some remaining black and white at the two ends of the gradient.

That’s how LGG is supposed to work. In FCPX version 10.4, each color wheel control also altered the levels of everything else. When you adjusted midrange, it also elevated the shadow and highlight ranges. In the hue offset example, shifting the midrange control to full-on yellow tinted the entire image to yellow, leaving no hint of black or white. As a result, the color wheels correction tool was unpredictable and difficult to use, unless you were doing only very minor adjustments. You ended up chasing your tail, because when one correction was made, you’d have to go back and re-adjust one of the other wheels to compensate for the unwanted changes made by the first adjustment.

With the release of FCPX 10.4.1 this April, Apple engineers have changed the way the color wheels tool behaves. Corrections now correspond to the behavior that everyone accepts as standard LGG functionality. In other words, the controls mostly only affect their part of the image without also adjusting all other levels. This means that the shadows (lift) control adjusts the bottom, highlights (gain) will adjust the top end, and midrange (gamma) will lighten or darken the middle portion of the image. Likewise, hue offsets don’t completely contaminate the entire image.

One important thing to note is that existing FCPX Libraries created or promoted under 10.4 will now be promoted again when opened in 10.4.1. In order that your color wheel corrections don’t change to something unexpected when promoted, Projects in these Libraries will behave according to the previous FCPX 10.4 color model. This means that the look of clips where color wheels were used – and their color wheel values – haven’t changed. More importantly, the behavior of the wheels when inside those Libraries will also be according to the “old” way, should you make any further corrections. The new color wheels behavior will only begin within new Libraries created under 10.4.1.

These images clarify how the 10.4.1 adjustments now work (click to see enlarged and expanded views).

©2018 Oliver Peters


What’s up with Final Cut’s Color Wheels?

NOTE: The information presented here has been superseded by the release of FCPX 10.4.1 in April 2018. With that release the color wheels model has been changed. Please read the linked blog post for updated information.

Apple Final Cut Pro X 10.4 introduced new, advanced color correction tools to this editing application, including color wheels, curves, and hue vs. saturation curves. These are tools that users of other NLEs have enjoyed for some time – and, which were part of Final Cut Studio (FCP 7, Color). Like others, my first reaction was, “Super! They’ve added some nice advanced tools, which will improve the use of FCPX for higher-end users.” But, as I started to primarily use the Color Wheels with real correction work, I quickly realized that something wasn’t quite right in how they operated. Or at least, they didn’t work in a way that we’ve come to understand.

In trying to figure it out, I reached out to other industry pros and developers for their thoughts. Naturally this led to some spirited discussions at forums like those at Creative COW. However, other editors have noticed the same problems, so you can also find threads in the Facebook FCPX group and at FCP.co. It is certainly easy to characterize this as just another internet kerfuffle, surrounding Apple’s “think different” approaches to FCPX. But those arguments fall flat when you actually try to use the tools as intended.

The FCPX Color Wheels panel includes four wheels – Master, Shadows, Midtones, and Highlights. The puck in the center of each wheel is a hue offset control to push hues in the direction that you move the puck. The slider to the right of the wheel controls the brightness of that range. The left slider controls the saturation. One of the main issues is that when you adjust luminance using one of these controls, the affected range is too broad. Specifically, in the case of the Midtones control, as you adjust the luminance slider up or down, you are affecting most of the image and not just the midrange levels. This is not the way this type of control normally works in other tools, and in fact, it’s not how FCPX’s Color Board controls work either.

“What’s the big deal?” you might ask. Fair enough. I see two operational issues. The first is that to properly grade the image using the Color Wheels, you end up having to go back-and-forth a lot between wheels, to counteract the changes made by one control with another. The second is that using the Midtones slider tends to drive highlights above 100 IRE, where they will be clipped if any broadcast limiting is used. This doesn’t happen with other color tools, notably Apple’s own Color Board.

A lot of the discussion focuses on luma levels and specifically the Midtones slider, since it’s easy to see the issue there. However, other controls are also affected, but that’s too much to dissect in a single post. Throughout this post, be sure to click on the images to see the full view. I have presented various samples against each other and you will only get the full understanding if you open the thumbnail (which is small but also cropped) to the full image. I have compared the effect using five different tools – the Color Board, the Color Wheels, a color corrector plug-in that I built as a Motion template using Motion effects, Rubber Monkey Software FilmConvert (the wheels portion only), and finally, the Adobe Lumetri controls in Premiere Pro.

I am using three different test images – a black-to-white ramp, a test pattern, and a demo video image. The ramp without correction will appear as a diagonal line (0-100 IRE) on the scope, which makes it easy to analyze what’s happening. The video image has definite shadow and highlight areas, which lets us see how these controls work in the real world. For example, if you want to brighten the area of the shot where the man is in the shadows, but don’t want to make the highlights any brighter, this would normally be done using a Midtones control. Be aware that these various tools certainly aren’t calibrated the same way and some have a greater range of control than others. The weakest of these is FilmConvert’s wheels, since this plug-in has additional level controls in other parts of its interface.

Color science models

In the various forum threads, the argument is made that Apple is simply using a different color science method or a different weighing of some existing models. That’s certainly possible, since not all color correctors are built the same way. The most common approaches are Lift/Gamma/Gain and Shadows/Mids/Highlights. Be careful with naming. Just because something uses the terminology of Shadows, Midtones, and Highlights, does not mean that it also uses the SMH color science model. Many tools use the Lift/Gamma/Gain model, but in fact, call the controls shadows (Lift), mids (Gamma), and highlights (Gain). Another term you may run across is Set-up in some correction tools. This is typically used for control of shadows (equal to Lift), but can also function is an offset control that raises the level of the entire image. Avid Symphony employs this solution. Finally, both Symphony and Adobe SpeedGrade use what has been dubbed a 12-way color corrector. Each range is further subdivided into its own subset of shadows, mids, and highlights controls.

An LGG model provides broad control of shadows and highlights, with the midtones control working like a curve that covers the whole range, but with the largest effect in the middle. An SMH model normally divides the levels into three distinct, precisely overlapping ranges. This is much like a three-band audio equalizing filter. A number of the color correctors add a luma range control, which gives the user the ability to change how much of the image a specific range will affect. In other words, how broad is the control of the shadows, mids, or highlights control? This is like a Q control in an audio equalizer, where you change the shape of the envelope at a certain frequency.

Red Giant’s Magic Bullet Looks offers both color correction models with two different tools – the 4-way color corrector (SMH) and the Colorista color corrector (LGG). When you adjust the midrange control of their 4-way, the result is a graceful S-shaped curve to the levels on the waveform.

To study the effect of an LGG-based corrector, test the ramp. The shadows control (Lift) will raise or lower the dark areas of the image without changing the absolute highlights. The diagonal line of the ramp on the waveform essentially pivots, hinged at the 100 IRE point. Conversely, change the highlights control (Gain) pivots the line pinned to 0 IRE (at black). When you adjust the midtones control (Gamma), you create a curve to the line, which stays pinned at 0 and 100 IRE at either end. In this way you are effectively “expanding” or “compressing” the levels in the middle portion of your image without changing the position of your black or white points.

How the various color correction tools react

Looking at the luma control for the Midtones, two things are clear. First, all of these tools are using the LGG color science model. It’s not clear what the Color Wheels are using, but it isn’t SMH, as there is no bulge or S-curve visible in the scope. Second, the Color Wheels quickly drive the image levels into clipping, while the other tools generally keep black and while levels in place. In essence, the Midtones control affects the image more like a master or offset control would, than a typical mids or Gamma control. Yet, clearly Apple’s Color Board controls adhere to the standard LGG model. The concern, of course, is clipping. In the test image of the man walking on the village street, the sunlit building walls on the opposite side of the street will become overexposed and risk being clipped when the Color Wheels are used.

What about color? As a simple test, I next shifted the Midtones puck to the yellow. Bear in mind that the range of each of these controls is different, so you will see varying degrees of yellow intensity. Nevertheless, the way the control should work is that some pure black and white should be preserved at the top and bottom of the video levels. All of these tools maintain that, except for the Color Wheels. There, the entire image is yellow, effectively making the hue offset puck function more like a tint control.

One other issue to note, is that the Color Wheels offer an extraordinarily control range. The hue offset control RGB intensity values go from 0 (center of the wheel) to 1023. However, the puck icon can only go to the rim of the wheel, which it hits at about 200. With a mouse (or numerical entry), you can keep going well past the stop of the wheel icon – five times farther, in fact. The image not only becomes very yellow in this case, but you can easily lose the location of your control, since the GUI position in no longer relevant.

The working theory

The big question is why don’t the Color Wheels conform to established principles, when in fact, the Color Board controls do? Until there is some further clarification from Apple, one possible explanation is with HDR. FCPX 10.4 introduced High Dynamic Range (HDR) features. One of the various HDR standards is Rec. 2020 PQ. In that color space, the 0-100 IRE limitations of Rec. 709 are expanded to 0-10,000 nits. 0-100 nits is roughly the same brightness as we are used to with Rec. 709.

Looking at this image of the man walking along the street – where I’ve attempted to get a pleasing look with all of the tools – you’ll see that the Color Wheels in Rec. 709 don’t react correctly and will drive the highlights into a range to be clipped. However, in the bottom pane, which is the same image in Rec. 2020 PQ color space, the grade looks pretty normal. And, in practice, the Color Wheels controls work more or less the way I would have expected them to work. Yes, the same controls work differently in the different color spaces – properly in 2020 PQ and not in 709.

But why is that the case? I have no answer, but I do have a wild guess. Maybe, just maybe, the Color Wheels were designed for – or intended to only be used for – HDR work. Or maybe there’s conversion or recalibration of the controls that hasn’t taken place yet in this version. If the tool is only calibrated for HDR, then its range and weighing will be completely wrong for Rec. 709 video. If you increase the Midtones luma of the ramp in both Rec. 709 and Rec. 2020 PQ, you’ll see a similar curve. In fact, if you overlay a screen shot of each waveform, placing the full Rec. 709 scope image over the bottom portion of the Rec. 2020 PQ scale, you’ll notice that these sort of align up to about 100 IRE and nits. It’s as if one is simply a slice out of the other.

Regardless of why, this is something where I would hope Apple will provide a white paper or other demonstration of what the best practices will be for using this tool effectively. If it isn’t intentional, and actually is a mistake, then I presume a fix will be forthcoming. In either case, put in your feedback comments to Apple.

A word about HDR

Over the course of testing this tool and this theory, I’ve done a bit of testing with the HDR color spaces in FCPX. If you want to know more about HDR, I would encourage you to check out these contrary blog posts by Stu Maschwitz and Alexis Van Hurkman. I tend to side with Stu’s point-of-view and am not a big fan of HDR.

The way Apple has implemented these features in Final Cut Pro X 10.4 is to allow the user to set and override color spaces. If you set up your project to be Rec. 2020 PQ (and set preferences to “show HDR as raw values”), then the viewer and a/v output (direct from the Mac, not through a hardware i/o device) are effectively dimmed through the Mac’s color profile system. When you grade the image based on the 0-10,000 nits scale, you’ll end up seeing an image that looks pleasing and essentially the same as if you were working in Rec. 709. However – and I cannot over-emphasize this – you are not going to be able to produce an image that’s truly compatible with Dolby Vision and actually look correct as HDR, unless you have the correct AJA i/o hardware and a proper display. And by display, I mean a top-end Dolby, Canon, or Sony unit, costing tens of thousands of dollars.

As I understand the PQ specs, the bulk of the higher range is for the highlights that are normally constrained or clipped in our current video systems. However, that 10,000 nits scale is weighed, so that about 50% of the image value is in the first 100 nits, making it of comparable brightness to the current 100 IRE. The rest of that range is for brighter information, like specular highlights. You don’t necessarily get more brightness in the shadow detail. Therefore, if you are grading a shot in FCPX in a 2020 PQ color space and you only have the computer display to go by, you’ll grade by eye as much as by scope. This means that to get a pleasing image, you will end up making the average appearance of the image brighter than it really should be. When this is viewed on a real HDR monitor, it will be painfully bright. Having a higher-nits computer display, like on the iMac Pro (up to 500 nits), won’t make much difference, unless maybe, you crank the display brightness to its maximum (ouch!).  “Mine goes the 11!”

Right now, HDR is the wild, wild west. If you are smart, you’ll realize that you don’t know what you don’t know. While it’s nice to have these new features in FCPX, they can be very dangerous in the wrong hands.

But that’s another matter. Right now, I just hope Apple (or one of the usual suspects, like Ripple Training, LumaForge, or Larry Jordan) will come out with more elaboration on the Color Wheels.

©2018 Oliver Peters

More 4K


I’ve talked about 4K before (here, here and here), but I’ve recently done some more 4K jobs that have me thinking again. 4K means different things to different people and in terms of dimensions, there’s the issue of cinema 4K (4096 pixels wide) versus the UltraHD/QuadHD/4K 16:9 (whatever you want to call it) version of 4K (3840 pixels wide). That really doesn’t make a lot of difference, because these are close enough to be the same. There’s so much hype around it, though, that you really have to wonder if it’s “the Emperor’s new clothes”. (Click on any of these images for expanded views.)

First of all, 4K used as a marketing term is not a resolution, it’s a frame dimension. As such, 4K is not four times the resolution of HD. That’s a measurement of area and not resolution. True resolution is usually measured in the vertical direction based on the ability to resolve fine detail (regardless of the number of pixels) and, therefore, 4K is only twice the resolution of HD at best. 4K is also not sharpness, which is a human perception affected by many things, such as lens quality, contrast, motion and grading. It’s worth watching Mark Schubin’s excellent webinar on the topic to get a clearer understanding of this. There’s also a very good discussion among top DoPs here about 4K, lighting, high dynamic range and more.

df_4kcompare_1A lot of arguments have been made that 4K cameras using a color-pattern filter method (Bayer-style), single CMOS sensor don’t even deliver the resolution they claim. The reason is that in many designs 50% of the pixels are green versus 25% each for red and blue. Green is used for luminance, which determines detail, so you do not have a 1:1 pixel relationship between green and the stated frame resolution of the sensor. That’s in part why RED developed 5K and 6K sensors and it’s why Sony uses an 8K sensor (F65) to deliver a 4K image.

The perceived image quality is also not all about total pixels. The pixels of the sensor, called photosites, are the light-receiving elements of the sensor. There’s a loose correlation between pixel size and light sensitivity. For any given sensor of a certain physical dimension, you can design it with a lot of small pixels or with fewer, but larger, pixels. This roughly correlates to a sensor that’s of high resolution, but a smaller dynamic range (many small pixels) or one with lower resolution, but a higher dynamic range (large, but fewer pixels). Although the equation isn’t nearly this simplistic, since a lot of color science and “secret sauce” goes into optimizing a sensor’s design, you can certainly see this play out in the marketing battles between the RED and ARRI camps. In the case of the ALEXA, ARRI adds some on-the-sensor filtering, which results in a softer image that gives it a characteristic filmic quality.df_4kcompare_2

Why do you use 4K?

With 4K there are two possible avenues. The first is to shoot 4K for the purpose of reframing and repositioning within HD and 2K timelines. Reframing isn’t a new production idea. When everyone shot on film, some telecine devices, like the Rank Cintel Mark III, sported zoom boards that permitted an optical blow-up of the 35mm negative. You could zoom in for a close-up in transfer that didn’t cost you resolution. Many videographers shoot 1080 for a 720 finish, as this allows a nice margin for reframing in post. The second is to deliver a final 4K product. Obviously, if your intent is the latter, then you can’t count on the techniques of the former in post.

df_4kcompare_3When you shoot 4K for HD post, then workflow is an issue. Do you shoot everything in 4K or just the items you know you’ll want to deal with? How will this cut with HD and 2K content? That’s where it gets dicey, because some NLEs have good 4K workflows and others don’t. But it’s here that I contend you are getting less than meets the eye, so to speak.  I have run into plenty of editors who have dropped a 4K clip into an HD timeline and then blown it up, thinking that they are really cropping into the native 4K frame and maintaining resolution. Depending on the NLE and the settings used, often they are simply blowing up an HD shot. The NLE scaled the 4K to HD first and then expanded the downscaled HD image. It didn’t crop into the actual 4K native resolution. So you have to be careful. And guess what, if the blow up isn’t that extreme, it may not look much different than the crop.

df_4kcompare_4One thing to remember is that a 4K image that is scaled to fit into an HD timeline gains the benefits of oversampling. The result in HD will be very sharp and, in fact, will generally look better perceptually than the exact same image natively shot in an HD size. When you now crop into the native image, you are losing some of that oversampling effect. A 1:1 pixel relationship is the same effective image size as a 200% blow-up. Of course, it’s not the same result. When you compare the oversampled “wide shot” (4K scaled to HD) to the “close-up” (native 4K crop), the close-up will often look softer. You’ll see defects of the image, like chromatic aberration in the lens, missed critical focus and sensor noise. Instead, if you shoot a wide and then an actual close-up, that result will usually look better.

On the other hand, if you blow up the 4K-to-HD or a native HD shot, you’ll typically see a result that looks pretty good. That’s because there’s often a lot more information there than monitors or the eye can detect. In my experience, you can commonly get away with a blow-up in the range of 120% of the original image size and in some cases, as much as 150%.

To scale or not to scale

df_4K_comparison_Instant4KLet me point out that I’m not saying a native 4K shot doesn’t look good. It does, but often the associated workflow hassles aren’t worth it. For example, let’s take a typical 1080p 50” Panasonic plasma that’s often used as a client monitor in edit suites. You or your client may be sitting 7 to 10 feet away from it, which is closer than most people sit in a living room with that size of a screen. If I show a client the native image (4K at 1:1 in an HD timeline) compared with an separate HD image at the same framing, it’s unlikely that they’ll see a difference. Another test is to take two exact images – one native HD and the other 4K. Scale up the HD and crop down the 4K to match. In theory, the 4K should look better and sharper. In fact, sitting back on the client sofa, most won’t see a difference. It’s only when they step to about 5 feet in front of the monitor that a difference is obvious and then only when looking at fine detail within the shot.

df_gh4_instant4k_smNot all scaling is equal. I’ve talked a lot about the comparison of HD scaling, but that really depends on the scaling that you use. For a quick shot, sure, use what your NLE has built in. For more critical operations, then you might want to scale images separately. DaVinci Resolve has excellent built-in scaling and lets you pick from smooth, sharp and bilinear algorithms. If you want a plug-in, then the best I’ve found is the new Red Giant Instant 4K filter. It’s a variation of their Instant HD plug-in and works in After Effects and Premiere Pro. There are a lot of quality tweaks and naturally, the better it does, the longer the render will be. Nevertheless, it offers outstanding results and in one test that I ran, it actually provided a better look within portions of the image than the native 4K shot.

df_4K_comparison-C500_smIn that case, it was a C500 shot of a woman on a park bench with a name badge. I had three identical versions of the shot (not counting the raw files) – the converted 4K ProRes4444 file, a converted 1080 ProRes4444 “proxy” file for editing and the in-camera 1080 Canon XF file. I blew up the two 1080 shots using Instant 4K and cropped the 4K shot so all were of equal framing. When I compared the native 4K shot to the expanded 1080 ProRes4444 shot, the woman’s hair was sharper in the 1080 blow-up, but the letters on the name badge were better on the original. The 1080 Canon XF blow-up was softer in both areas. I think this shows that some of the controls in the plug-in may give you superior results to the original (crisper hair); but, a blow-up suffers when you are using a worse codec, like Canon’s XF (50 Mbps 4:2:2). It’s fine for native HD, but the ProRes4444 codec has twice the chroma resolution and less compression, which makes a difference when scaling an image larger. Remember all of this pertains to viewing the image in HD.

4K deliverables

df_4K_comparison-to-1080_smSo what about working in native 4K for a 4K deliverable? That certainly has validity for high-resolution projects (films, concerts, large corporate presentations), but I’m less of a believer for television and web viewing. I’d rather have “better” pixels and not simply “more” pixels. Most of the content you watch at theaters using digital projection is 2K playback. Sometimes the master for that DCP was HD, 2K or 4K. If you are in a Sony 4K projector-equipped theater, most of the time, it’s simply the projector upscaling the content to 4K as part of the projection. Even though you may see a Sony 4K logo at the head of the trailers, you aren’t watching 4K content – definitely not, if it’s a stereo3D film. Yet, much of this looks pretty good, doesn’t it?

df_AMIRAEverything I talked about, regarding blowing up HD by up to 120% or more, still applies to 4K. Need to blow up a shot a bit in a 4K timeline? Go ahead, it will look fine. I think ARRI has proven this as well, taking films shot with the ALEXA all the way up to Imax. In fact, ARRI just announced that the AMIRA will get in-camera, on-the-fly upscaling of its image with the ability to record 4K (3840 x 2160 at up to 60fps) on the CFast 2.0 cards. They can do this, because the sensor starts with more pixels than HD or 2K. The AMIRA will expose all of the available photosites (about 3.4K sensor pixels) in what they call the “open gate” method. This image is lightly cropped to 3.2K and then scaled by a 1.2 factor, which results in UltraHD 4K recording on the same hardware. Pretty neat trick and judging by ARRI’s image quality, I’ll bet it will look very good. Doubling down on this technique, the ALEXA XT models will also be able to record ProRes media at this 3.2K size. In the case of the ALEXA, the designers have opted to leave the upscaling to post, rather than to do it in-camera.

To conclude, if you are working in 4K today, then by all means continue to do so. It’s a great medium with a lot of creative benefits. If you aren’t working in 4K, then don’t sweat it. You won’t be left behind for awhile and there are plenty of techniques to get you to the same end goal as much of the 4K production that’s going on.

Click these thumbnails for full resolution images.










©2014 Oliver Peters

Post Production Mastering Tips

The last step in commercial music production is mastering. Typically this involves making a recording sound as good as it possibly can through the application of equalization and multiband compression. In the case of LPs and CDs (remember those?), this also includes setting up the flow from one tune to the next and balancing out levels so the entire product has a consistent sound. Video post has a similar phase, which has historically been in the hands of the finishing or online editor.

That sounds so sweet

The most direct comparison between the last video finishing steps and commercial music mastering is how filters are applied in order to properly compress the audio track and to bring video levels within legal broadcast specs. When I edit projects in Apple Final Cut Pro 7 and do my own mixes, I frequently use Soundtrack Pro as the place to polish the audio. My STP mixing strategy employs tracks that route into one or more subgroup buses and then a master output bus. Four to eight tracks of content in FCP might become twenty tracks in STP. Voice-over, sync-sound, SFX and music elements get spread over more tracks and routed to appropriate subgroups. These subgroups then flow into the master bus. This gives me the flexibility to apply specific filters to a track and have fine control over the audio.

I’ll usually apply a compressor across the master bus to tame any peaks and beef up the mix. My settings involve a low compression ratio and a hard limit at -10dB. The objective is to keep the mix levels reasonable so as to preserve dynamic range. I don’t want to slam the meters and drive the signal hard into compression. Even when I do the complete mix in Final Cut, I will still use Soundtrack Pro simply to compress the composite mix, because I prefer its filters. When you set the reference tone to -20dB, then these levels will match the nominal levels for most digital VTRs. If you are laying off to an analog format, such as Betacam-SP, set your reference tone to -12dB and match the input on the deck to 0VU.

Getting ready for broadcast

The video equivalent is the broadcast safe limiting filter. Most NLEs have one, including Avid Media Composer and both old and new versions of Final Cut. This should normally be the last filter in the chain of effects. It’s often best to apply it to a self-contained file in FCP 7, a higher track in Media Composer or a compound clip in FCP X. Broadcast specs will vary with the network or station receiving your files or tapes, so check first. It’s worth noting that many popular effects, like glow dissolves, violate these parameters. You want the maximum luminance levels (white peaks) to be limited to 100 IRE and chrominance to not exceed 110, 115 or 120, depending on the specs of the broadcaster to whom you are delivering. In short, the chroma should stay within the outer ring of a vectorscope. I usually turn off any RGB limiting to avoid artifacts.

It’s often a good idea to reduce the overall video levels by about five percent prior to the application of a broadcast safe filter, simply so you don’t clip too harshly. That’s the same principle as I’ve applied to the audio mix. For example, I will often first apply a color correction filter to slightly lower the luminance level and reduce chroma. In addition, I’ll frequently use a desaturate highlights or lows filter. As you raise midrange or highlight levels and crush shadows during color correction, the chroma is also driven higher and/or lower accordingly. Red, blues and yellows are most susceptible, so it’s a good idea to tone down chroma saturation above 90 IRE and below 20 IRE. Most of these filters let you feather the transition range and the percentage of desaturation, so play with the settings to get the most subtle result. This keeps the overall image vibrant, but still legal.

Let me interject at this point that what you pay for when using a music mastering specialist are the “ears” (and brain) of the engineer and their premium monitoring environment. This should be equally true of a video finishing environment. Without proper audio and video monitoring, it’s impossible to tell whether the adjustments being made are correct. Accurate speakers, calibrated broadcast video monitors and video scopes are essential tools. Having said that though, software scopes and modern computer displays aren’t completely inaccurate. For example, the software scopes in FCP X and Apple’s ColorSync technology are quite good. Tools like Blackmagic Design Ultrascope, HP Dreamcolor or Apple Cinema Displays do provide accurate monitoring in lower-cost situations. I’ve compared the FCP X Viewer on an iMac to the output displayed on a broadcast monitor fed by an AJA IoXT. I find that both match surprisingly well. Ultimately it gets down to trusting an editor who knows how to get the best out of any given system.

Navigating the formats

Editors work in a multi-standard world. I frequently cut HD spots that run as downconverted SD content for broadcast, as well as at a higher HD resolution for the internet. The best production and post “lingua franca” format today is 1080p/23.976. This format fits a sweet spot for the internet, Blu-ray, DVD and modern LCD and plasma displays. It’s also readily available in just about every camera at any price range. Even if your product is only intended to be displayed as standard definition today, it’s a good idea to future-proof it by working in HD.

If you shoot, edit and master at 1080p/23.976, then you can easily convert to NTSC, 720p/59.94 or 1080i/29.97 for broadcast. The last step for many of my projects is to create deliverables from my master file. Usually this involves creating three separate broadcast files in SD and two HD formats using either ProRes or uncompressed codecs. I will also generate an internet version (without bars, tone, countdown or slate) that’s a high-quality H.264 file in the 720p/23.976 format. Either .mov or .mp4 is fine.

Adobe After Effects is my tool of choice for these broadcast conversions, because it does high-quality scaling and adds proper cadences. I follow these steps.

A) Export a self-contained 1080p/23.976 ProResHQ file from FCP 7 or X.

B) Place that into a 720×486, 29.97fps After Effects D1 composition and scale the source clip to size. Generally this will be letterboxed inside of the 4×3 frame.

C) Render an uncompressed QuickTime file, which is lower-field ordered with added 2:3 pulldown.

D) Re-import that into FCP 7 or X using a matching sequence setting, add the mixed track and format it with bars, tone, countdown and slate.

E) Export a final self-contained broadcast master file.

F) Repeat the process for each additional broadcast format.

Getting back there

Archiving is “The $64,000 Question” for today’s digital media shops. File-based mastering and archiving introduces dilemmas that didn’t exist with videotape. I recommend always exporting a final mixed master file along with a split-track, textless submaster. QuickTime files support multi-channel audio configurations, so building such a file with separate stereo stems for dialogue, sound effects and music is very easy in just about any NLE. Self-contained QuickTime movies with discrete audio channels can be exported from both FCP 7 and FCP X (using Roles).

Even if your NLE can’t export multi-channel master files, export the individual submixed elements as .wav or .aif audio files for future use. In addition to the audio track configuration, remove any titles and logos. By having these two files (master and submaster), it’s very simple to make most of the future revisions you might encounter without ever having to restore the original editorial project. Naturally, one question is which codec to use for access in the future. The preferred codec families these days are Avid DNxHD, Apple ProRes, uncompressed, OP1a MXF (XDCAM) or IMX. FCP editors will tend towards ProRes and Avid editors towards DNxHD, but uncompressed is very viable with the low cost of storage. For feature films, another option to consider would be image sequences, like a string of uncompressed TIFF or DPX files.

Whichever format you standardize on, make multiple copies. LTO data tape is considered the best storage medium, but for small files, like edited TV commercial masters, DVD-ROM, Blu-ray and XDCAM media are likely the most robust. This is especially true in the case of water damage.

The typical strategy for most small users who don’t want to invest in LTO drives is a three-pronged solution.

A) Store all camera footage, elements and masters on a RAID array for near-term editing access.

B) Back-up the same items onto at least two copies of raw SATA or SSD hard drives for longer storage.

C) Burn DVD-ROM or BD-ROM copies of edited master files, submasters, project files and elements (music, VO, graphics, etc.).

A properly polished production with audio and video levels that conform to standards is an essential aspect of delivering a professional product. Developing effective mastering and archiving procedures will protect the investment your clients have made in a production. Even better, a reliable archive routine will bring you repeat business, because it’s easy to return to the project in the future.

Originally written for DV magazine/Creative Planet/NewBay Media, LLC

©2012 Oliver Peters

Why 4K

Ever since the launch of RED Digital Cinema, 4K imagery has become an industry buzzword. The concept stems from 35mm film post, where the digital scan of a film frame at 4K is considered full resolution and a 2K scan to be half resolution. In the proper used of the term, 4K only refers to frame dimensions, although it is frequently and incorrectly used as an expression of visual resolution or perceived sharpness. There is no single 4K size, since it varies with how it is used and the related aspect ratio. For example, full aperture film 4K is 4096 x 3112 pixels, while academy aperture 4K is 3656 x 2664. The RED One and EPIC use several different frame sizes. Most displays use the Quad HD standard of 3840 x 2160 (a multiple of 1920 x 1080) while the Digital Cinema Projection standard is 4096 x 2160 for 4K and 2048 x 1080 for 2K. The DCP standard is a “container” specification, which means the 2.40:1 or 1.85:1 film aspects are fit within these dimensions and the difference padded with black pixels.

Thanks to the latest interest in stereo 3D films, 4K-capable projection systems have been installed in many theaters. The same system that can display two full bandwidth 2K signals can also be used to project a single 4K image. Even YouTube offers some 4K content, so larger-than-HD production, post and distribution has quickly gone from the lab to reality. For now though, most distribution is still predominantly 1920 x 1080 HD or a slightly larger 2K film size.

Large sensors

The 4K discussion starts at sensor size. Camera manufacturers have adopted larger sensors to emulate the look of film for characteristics such as resolution, optics and dynamic range. Although different sensors may be of a similar physical dimension, they don’t all use the same number of pixels. A RED EPIC and a Canon 7D use similarly sized sensors, but the resulting pixels are quite different. Three measurements come into play: the actual dimensions, the maximum area of light-receiving pixels (photosites) and the actual output size of recorded frames. One manufacturer might use fewer, but larger photosites, while another might use more pixels of a smaller size that are more densely packed. There is a very loose correlation between actual pixel size, resolution and sensitivity. Larger pixels yield more stops and smaller pixels give you more resolution, but that’s not an absolute. RED has shown with EPIC that it is possible to have both.

The biggest visual attraction to large-sensor cameras appears to be the optical characteristics they offer – namely a shallower depth of field (DoF).  Depth of field is a function of aperture and focal length. Larger sensors don’t inherently create shallow depth of field and out-of-focus backgrounds. Because larger sensors require a different selection of lenses for equivalent focal lengths compared with standard 2/3-inch video cameras, a shallower depth of field is easier to achieve and thus makes these cameras the preferred creative tool. Even if you work with a camera today that doesn’t provide a 4K output, you are still gaining the benefits of this engineering. If your target format is HD, you will get similar results – as it relates to these optical characteristics – regardless of whether you use a RED, an ARRI ALEXA or an HDSLR.

Camera choices

Quite a few large-sensor cameras have entered the market in the past few years. Typically these use a so-called Super 35MM-sized sensor. This means it’s of a dimension comparable to a frame of 3-perf 35MM motion picture film. Some examples are the RED One, RED EPIC, ARRI ALEXA, Sony F65, Sony F35, Sony F3 and Canon 7D among others. That list has just grown to include the brand new Canon EOS C300 and the RED SCARLET-X. Plus, there are other variations, such as the Canon EOS 5D Mark II and EOS 1D X (even bigger sensors) and the Panasonic AF100 (Micro Four Thirds format). Most of these deliver an output of 1920 x 1080, regardless of the sensor. RED, of course, sports up to 5K frame sizes and the ALEXA can also generate a 2880 x 1620 output, when ARRIRAW is used.

This year was the first time that the industry at large has started to take 4K seriously, with new 4K cameras and post solutions. Sony introduced the F65, which incorporates a 20-megapixel 8K sensor. Like other CMOS sensors, the F65 uses a Bayer light filtering pattern, but unlike the other cameras, Sony has deployed more green photosites – one for each pixel in the 4K image. Today, this 8K sensor can yield 4K, 2K and HD images. The F65 will be Sony’s successor to the F35 and become a sought-after tool for TV series and feature film work, challenging RED and ARRI.

November 3rd became a day for competing press events when Canon and RED Digital Cinema both launched their newest offerings. Canon introduced the Cinema EOS line of cameras designed for professional, cinematic work. The first products seem to be straight out of the lineage that stems from Canon’s original XL1 or maybe even the Scoopic 16MM film camera. The launch was complete with a short Bladerunner-esque demo film produced by Stargate Studios along with a new film shot by Vincent Laforet (the photographer who launch the 5D revolution with his short film Reverie)  called Möbius.

The Canon EOS C300 and EOS C300 PL use an 8.3MP CMOS Super 35MM-sized sensor (3840 x 2160 pixels). For now, these only record at 1920 x 1080 (or 1280 x 720 overcranked) using the Canon XF codec. So, while the sensor is a 4K sensor, the resulting images are standard HD. The difference between this and the way Canon’s HDSLRs record is a more advanced downsampling technology, which delivers the full pixel information from the sensor to the recorded frame without line-skipping and excessive aliasing.

RED launched SCARLET-X to a fan base that has been chomping at the bit for years waiting for some version of this product. It’s far from the original concept of SCARLET as a high-end “soccer mom” camera (fixed lens, 2/3” sensor, 3K resolution with a $3,000 price tag). In fact, SCARLET-X is, for all intents and purposes, an “EPIC Lite”. It has a higher price than the original SCARLET concept, but also vastly superior specs and capabilities. Unlike the Canon release, it delivers 4K recorded motion images (plus 5K stills) and features some of the developing EPIC features, like HDRx (high dynamic range imagery).

If you think that 4K is only a high-end game, take a look at JVC. This year JVC has toured a number of prototype 4K cameras based on a proprietary new LSI chip technology that can record a single 3840 x 2160 image or two 1920 x 1080 streams for the left and right eye views of a stereo 3D recording. The GY-HMZ1U is derivative of this technology and uses dual 3.32MP CMOS sensors for stereo 3D and 2D recordings.

Post at 4K

Naturally the “heavy iron” systems from Quantel and Autodesk have been capable of post at 4K sizes for some time; however, 4K is now within the grasp of most desktop editors. Grass Valley EDIUS, Adobe Premiere Pro and Apple Final Cut Pro X all support editing with 4K media and 4K timelines. Premiere Pro even includes native camera raw support for RED’s .r3d format at up to EPIC’s 5K frames. Avid just released its 6.0 version (Media Composer 6, Symphony 6 and NewsCutter 10), which includes native support for RED One and EPIC raw media. For now, edited sequences are still limited to 1920 x 1080 as a maximum size. For as little as $299 for FCP X and RED’s free REDCINE-X (or REDCINE-X PRO) media management and transcoding tool, you, too, can be editing with relative ease on DCP-compliant 4K timelines.

Software is easy, but what about hardware? Both AJA and Blackmagic Design have announced 4K solutions using the KONA 3G or Decklink 4K cards. Each uses four HD-SDI connections to feed four quadrants of a 4K display or projector at up to 4096 x 2160 sizes. At NAB, AJA previewed for the press its upcoming 5K technology, code-named “Riker”. This is a multi-format I/O system in development for SD up to 5K sizes, complete with a high-quality, built-in hardware scaler. According to AJA, it will be capable of handling high-frame-rate 2K stereo 3D images at up to 60Hz per eye and 4K stereo 3D at up to 24/30Hz per eye.

Even if you don’t own such a display, 27″ and 30″ computer monitors, such as an Apple Cinema Display, feature native display resolutions of up to 2560 x 1600 pixels. Sony and Christie both manufacture a number of 4K projection and display solutions. In keeping with its plans to round out a complete 4K ecosystem, RED continues in the development of REDRAY PRO, a 4K player designed specifically for RED media.

Written for DV magazine (NewBay Media, LLC)

©2011 Oliver Peters

Levels – Avid vs. FCP

One of the frequent misconceptions between Avid and Final Cut editors involves video levels. Many argue that FCP does not work within the proper video level standards, which is incorrect. This belief stems from the fact that FCP is based on QuickTime and permits a mixture of consumer and professional codecs. QuickTime Player often changes a file’s appearance as compared with FCP, when it is used to play the file directly. QuickTime Player is trying to optimize the file to look its best on your computer monitor; however, it isn’t actually changing the file itself. Furthermore, two identical clips will appear to be different within each NLE’s interface. Avid clips look flatter and more washed out inside Media Composer. FCP clips will be optimized for the computer display and appear to have more contrast and a different gamma value. This is explained well by Janusz Baranek in this Avid Community thread.

Contrary to popular opinion, both NLEs work within the digital video standards for levels and color space – aka Rec. 601 (SD) and Rec. 709 (HD). Digital video levels are generally expressed using an 8-bit/256-step scale. The nominal black point is mapped to 16 and the white point to 235, which permits level excursions without clipping: 0-16 for shadow detail and 235-255 for highlight recovery. This standard was derived from both camera design and legacy analog NTSC transmission. On most waveform monitors digital 0, analog 7.5 IRE and 16 on this scale are all the same level. Digital 100 (700 millivolts on some scopes), analog 100 IRE and 235 on the scale are also equal. Higher and lower levels will be displayed on a waveform as video above 100/100IRE/235 and below 0/7.5IRE/16.

I want to be clear that this post is not a right/wrong, good/bad approach. It’s simply an exploration in how each editing application treats video levels. This is in an effort to help you see where adjustments can be made if you are encountering problems.

Avid Media Composer/NewsCutter/Symphony

Video captured through Avid’s i/o hardware is mapped to this 16-235 range. Video imported from the computer, like stills and animation can have either a full range of 0-255 (so called “full swing”) or a digital video range of 16-235 (so called “studio swing”) values. Prior to AMA (Avid Media Access), Avid editors would determine these import values in the Media Composer settings, by selecting to import files with RGB values or 601/709 values. You can “cheat” the system by importing digital camera files with an expanded range (spreading the levels to “full swing” of 0-255). Doing so may appear to offer greater color grading latitude, but it introduces two issues. First, all clips have to be color corrected to adjust the levels for proper output values (legal for broadcast). Second, some filters, like the BCC effects, clip rendered files at 16 and 235, thus defeating the original purpose.

It has now become a lot more complex in the file-based world. The files you import are no longer just stills and animation, but also camera and master files from a variety of sources, including other NLEs, like FCP – or HDLSRs, like the Canon 5D. Thanks to AMA in Media Composer 5, this is now automatically taken care of. AMA will properly import files at the right levels based on the format. A digital graphic, like a 0-255 color bar test pattern, is imported at the full range without rescaling the color values from 0-255 to 16-235. A digital video movie from a Canon 5D will be imported with values fitting into the 16-235 range.

Because of the nature of how Avid handles media on the timeline, it is possible to have a full range (0-255) clip on the same timeline next to a studio range clip (16-235) and levels will be correctly scaled and preserved for each. Avid uses absolute values on its internal waveform (accessed in the color correction mode), so you are always able to see where level excursions occur above 235 (digital 100) and below 16 (digital 0).

I would offer one caveat about AMA importing.  Apparently some users have posted threads at the Avid Community Forums indicating some inconsistencies in behavior. In my case, everything is working as expected on multiple systems and from various Canon HDSLR cameras, but others haven’t been so lucky. As they say, “Your mileage may vary.”

Apple Final Cut Pro

If you capture video into FCP using one of the hardware options and a professional codec (uncompressed, ProRes, IMX, DV/DV50/DV100), then the media files will have levels mapped to 601/709 values (16-235). From here, the waters get muddy, because the way in which those levels are handled in the timeline is based on your processing settings. This affects all imported files, as well, including graphics, animation and media files from cameras and other NLEs.

Confusion is compounded by FCP’s internal waveform monitor, which always represents video with a relative 0-100 percent scale. These display numbers do not represent actual video levels in any absolute sense. When you process in YUV, then the full display area of the waveform from top to bottom equals a range of 0-255. The “legal” digital video standard of 16-235 is represented by the area within the 0-100% markings of the scope. However, when you process in RGB, then the portion within the 0-100% marks represents the full 0-255 range. Yes – in an effort to make it simple, Apple has made it very confusing!

When you set the sequence processing to YUV, with “white” as “white”, then all timeline video is mapped to a “studio swing” range of 16-235. On the scope 0% = 16 and 100% = 235. If you import a “full swing” color bar pattern (0-255), the values will be rescaled by the sequence processing setting to fall into the 16-235 range.

When you set the sequence processing to YUV, with “white” as “superwhite”, you’ve extended the upper end of the range, so that the 16-235 scale now becomes 16-255. The 0-255 color bar pattern is now effectively rescaled to 16-255; however, so is any video as well. Digital video that used to peak at 100% will now peak at 120%.

The YUV processing issues are also affected by the 8-bit, versus “high-precision” rendering options. When you elect to process all video as 8-bit, excursions above 100% and below 0% caused by color correction will be clipped. If you change to “high-precision YUV”, then these excursions are preserved, because they fall within the 0-16 and 235-255 regions. Unfortunately, certain effects and transition filters will still clip at 0% and 100% after rendering.

One way to fully protect “full swing” 0-255 levels is to work in RGB processing. A 0-255 color bar pattern will be correctly displayed, but unfortunately all video is spread to the full range, as well. This would mean that all clips would have to be color corrected to adjust for proper video levels. The only way that I’ve found for FCP to display both a 0-255 and a 16-235 clip on the same timeline and maintain correct levels is to apply a color correction filter to adjust the levels on one of these clips.

For general purposes, the best way to work with FCP is to use the ProRes family of codecs and set your sequence settings for YUV processing, white=white and high-precision rendering. This offers the most practical way of working. The only caveat to this is that any “full swing” file will be rescaled so that levels fall into the 0%-100% (16-235) “studio swing” range. If you need to preserve the full range, then FCP’s color correction filters will allow you to expand the range. The levels may appear to clip as you make the adjustment, but the rendered result will be full range.

Real world examples

I’ve done some quick examples to show how these level issues manifest themselves in actual practice. It’s important to understand that for the most part, the same clip would appear the same in either Media Composer of Final Cut as viewed on a broadcast monitor through output hardware. It will also look the same (more or less) when displayed to a computer screen using each app’s full screen preview function.

The following screen grabs from my tests include a 0-255 color bar chart (TIFF original) and a frame from an H.264 Canon 5D clip. The movie frame represents a good spread from shadow to highlights. I imported the files into both Avid Media Composer 5 (via AMA) and Apple Final Cut Pro 7. The FCP clips were rendered and exported and then brought into MC5 for comparison. The reason to do this last step was so that I could check these on a reasonably trustworthy internal scope, which displayed an 8-bit range in absolute values. It is not meant to be a direct comparison of how the video looks in the UI.

Click any image to enlarge.

Imported into Final Cut. ProResHQ with YUV processing. White as white. Note that the peak white of both images is 100%.

Imported into Final Cut. ProResHQ with YUV processing. White as SuperWhite. Note that peak white of both images exceeds 100%.

Imported into Final Cut. ProRes4444 with RGB processing. Note the boundary limits at 0% and 100%.

Imported into Media Composer 5 using AMA. Note that the color bars are a 0-255 range, while the Canon clip is 16-235.

FCP7 YUV export, imported into MC5 via AMA. Note that the color bar pattern has been rescaled to 16-235 and is no longer full range.

FCP7 YUV export with SuperWhite values, imported into MC5 via AMA. Note that the color bar pattern has been rescaled to 16-255 and is no longer full range. It has a higher top-end, but black values are incorrect. This also alters the scaling values of the levels for the Canon clip. Color correction filters would have to be applied in FCP for a “sort of correct” level match between the bars and the Canon clip.

FCP7 RGB export, imported into MC5 via AMA. Note that the color bar pattern exhibits the full 0-255 range. The Canon clip has also been rescaled to 0-255. Color correction filters would have to be applied in FCP to the Canon clip to bring it back into the correct relative range.

I have revisited the YUV settings in FCP7. This is a ProResHQ sequence rendered with high-precision processing. I have applied a color corrector to the color bars, expanded the range and rendered. Note the regions above 100% and below 0%.

FCP7 YUV export (color correction filter applied to the color bars), imported into MC5 via AMA. Note that the color bar pattern spreads from 0-255, while the Canon clip is still within the standard 16-235 range.

©2010 Oliver Peters

Codec Smackdown

Modern digital acquisition, post and distribution wouldn’t be possible without data rate reduction, AKA compression. People like to disparage compression, but I dare say that few folks – including most post production professionals – have actually seen much uncompressed content. In fact, by the time you see a television program or a digitally-projected movie it has passed through at least three, four or more different compression algorithms – i.e. codecs.

Avid Media Composer and Apple Final Cut Pro dominate the editing landscape, so the most popular high-end HD codecs are the Avid DNxHD and Apple ProRes 422 codec families. Each offers several codecs at differing levels of compression, which are often used for broadcast mastering and delivery. Apple and Avid, along with most other NLE manufacturers, also natively support other camera codecs, such as those from Sony (XDCAM-HD, HD422, EX) and Panasonic (DVCPRO HD, AVC-Intra). Even these camera codecs are being used for intermediate post. I frequently use DVCPRO HD for FCP jobs and I recently received an edited segment as a QuickTime movie encoded with the Sony EX codec. It’s not a question of whether compression is good or bad, but rather, which codec gives you the best results.

Click on the above images to see an enlarged view. (Images from Olympus camera, prior to NLE roundtrip. Resized from original.)

I decided to test some of these codecs to see the results. I started with two stills taken with my Olympus C4000Z – a 4MP point-and-shoot digital camera. These images were originally captured in-camera as 2288-pixel-wide JPEGs in the best setting and then – for this test – converted to 1920×1080 TIFFs in Photoshop. My reason for doing this instead of using captured video, was to get the best starting point. Digital video cameras often exhibit sensor noise and the footage may not have been captured under optimum lighting conditions, which can tend to skew the results. The two images I chose are of the Donnington Grove Country Club and Hotel near Newbury, England – taken on a nice, sunny day. They had good dynamic range and the size reduction in Photoshop added the advantages of oversampling – thus, very clean video images.

I tested various codecs in both Avid Media Composer 4.0.5 and Apple Final Cut Pro 7. Step one was to import the images into each NLE. In Avid, the conversion occurs during the import stage, so I set my import levels to RGB (for computer files) and imported the stills numerous times in these codecs: 1:1 MXF (uncompressed), DNxHD145, DNxHD220, DNxHD220x, XDCAM-EX 35Mbps and XDCAM-HD422 50Mbps. In Final Cut Pro, the conversion occurs when files are placed on the timeline and rendered to the codec setting of that timeline. I imported the two stills and placed and rendered them onto timelines using these codecs: Apple 8-bit (uncompressed), ProRes LT, ProRes, ProRes HQ, DVCPRO HD and XDCAM-EX 35Mbps. These files were then exported again as uncompressed TIFFs for comparison in Photoshop. For Avid, this means exporting the files with RGB levels (for computer files) and for FCP, using the QuickTime Conversion – Still Image option (set to TIFF).

Note that in Final Cut Pro you have the option of controlling the import gamma settings of stills and animation files. Depending on the selection (source, 1.8, 2.20, 2.22) you choose, your video in and back out of Final Cut may or may not be identical to the original. In this case, choosing “source” gamma matched the Avid roundtrip, whereas using a gamma setting of 2.2 resulted in a darker image exported from FCP.

Click on the above images to see an enlarged view.

You’ll notice that in addition to various compressed codecs, I also used an uncompressed setting. The reason is that even “uncompressed” is a media codec. Furthermore, to be accurate, compression comparisons need to be done against the uncompressed video image, not the original computer still or graphic. There are always going to be some changes when a computer file is brought into the video domain, so you can’t fairly judge a compressed video file against the original photo. Had I been comparing video captured through a hardware card, then obviously I would only have uncompressed video files as my cleanest reference images.

I lined up the exported TIFFs as Photoshop layers and generated comparisons by setting the layer mode to “difference”. This generates a composite image based on any pixel value that is different between the two layers. These difference images were generated by matching a compressed layer against the corresponding Avid or FCP uncompressed video layer. In other words, I’m trying to show how much data is lost when you use a given compressed codec versus the uncompressed video image. Most compression methods disproportionately affect the image in the shadow areas. When you look at a histogram displaying these difference results, you only see levels in the darkest portion of an 8-bit scale. On a 0-255 range of levels, the histogram will be flat down to about 20 or 30 and then slope up quickly to a spike at close to 0.

This tells you that the largest difference is in the darkest areas. The maximum compression artifacts are visible in this range. The higher quality codecs (least compressed), exhibit a smaller histogram range that is closer to 0. The more highly-compressed codecs have a fatter range. This fact largely explains why – when you color grade highly compressed camera images – compression artifacts become quite visible if you raise black or gamma levels.

The resulting difference images were then adjusted to show artifacts clearly in these posted images. By adjusted, I mean changing the levels range by dropping the input white point from 255 to 40 and the output black point from 0 to 20. This is mainly for illustration and I want to reiterate that the normal composite images DO NOT look as bad as my adjusted images would imply. In fact, if you looked at the uncorrected images on a computer screen without benefit of a histogram display, you might think there was nothing there. I merely stretched the available dynamic range for demonstration purposes.

Of these various codecs, the Apple DVCPRO HD codec shows some extreme difference results. That’s because it’s the only one of these codecs that uses horizontal raster scaling. Not only is the data compressed, but the image is horizontally squeezed. In this roundtrip, the image has gone from 1920-pixels-wide (TIFF) to 1280 (DVCPRO HD) back to 1920 (exported TIFF). The effects of this clearly show in the difference image.

Click on the above images to see an enlarged view.

There are a couple of other things you may notice, such as level differences between the Avid and Apple images and between each of these and the originals. As I said before, there will always be some differences in this sort of conversion. Plus, Apple and Avid do not handle color space, level and gamma mapping in the same way, so a round trip through each application will yield slightly different results. Generally, if 2.2 gamma is selected for imported stills, the Apple FCP image will have a bit more contrast and somewhat darker shadow areas when compared to Avid on a computer screen – even when proper RGB versus Rec. 709 settings are maintained for Avid. This is mainly a result of the various QuickTime and other conversions going on.

If I were to capture video with Avid DX hardware on the Media Composer and AJA, Matrox or Blackmagic hardware on FCP – and compared these images on a video monitor and with scopes – there would likely be no such visible difference. When I used “source” gamma in FCP, then the two matched each other. Likewise, when you review the difference images below, 2.2 gamma in this case resulted in a fault difference composite between the FCP uncompressed and the original photo. The “source” gamma version more closely resembles the Avid result and is the right setting for these images.

The take-away from these tests should be that the most important comparisons are those that are relative, i.e. “within species”. In other words, how does ProRes LT compare to ProRes HQ or how does DNxHD 145 compare to DNxHD 220x? Not, how an Avid export compares with a Final Cut export. A valid inter-NLE comparison, however, is whether Avid’s DNxHD220x shows more or less compression artifacts than Apple’s ProRes HQ.

I think these results are pretty obvious: Higher-data-rate codecs (less compression) like Apple ProRes HQ or Avid DNxHD 220x yield superb results. Lower-date-rate codecs (more compression) like XDCAM-EX yield results that aren’t as good. I hope that arming you with some visible evidence of these comparisons, will help you better decide what post trade-off to use in the future.

(In case you’re wondering, I do highly recommend the Donnington Grove for a relaxing vacation in the English countryside. Cheers!)

Click on these images to see an enlarged view.

©2010 Oliver Peters