SETTING SOURCE DEVICE OUTPUT IN DISTRIBUTED VIDEO AND AV OVER IP SYSTEMS
- Published , by Tom Devine
Let’s draw a distinction at the outset:
- Dolby Vision 60Hz (output by any device at 60Hz) is a hardware relationship with a display that has a native refresh rate of 60Hz (or a derivative thereof, such as 120Hz).
- Dolby Vision at 60fps is High Frame Rate (HFR) source content. Currently, no mainstream Hollywood creative content is native 60fps (special technical projects not included).
As of this writing, the only applicable practical use for Dolby Vision 4K/60fps is for the Apple iPhone 14 camera in video mode (and perhaps other brands), which have acquired content natively at that frame rate. But what might that setting in the new Apple TV impose upon distributed video systems?
With the noblest of intentions, integrators always seek to do what they interpret is best for their clients and this invariably includes attempts at delivering the best system image fidelity from the latest generation of assorted source devices, in which Apple TV figures prominently. In systems (or signal pathways which use supplemental products, like HDMI extenders) where bandwidth is design-limited such as AV over IP, care must be taken to avoid attempting to send signals that exceed bandwidth limitations.
Without video calibration training, it is hard for integrators and end-users to resist temptations to select settings, which numerically, depict what are perceived to be the best (read: biggest) possible. When perusing a device menu, initial inclinations may be to choose RGB output instead of Y’CbCr, thinking perhaps since all display types are RGB, selecting RGB represents a synergistic match. At the content creation level, or at least for signal transfer between studio-level professional devices, this presumption would be correct, as the signal would remain bandwidth uncompressed. For most devices classified as consumer level, RGB output, where found, may prove to be less fortuitous than envisioned. An explanation is forthcoming, however let’s proceed to one more parameter first: Chroma Subsampling.
Here again is where Big Number-itis may creep in to afflict integrators. In the menu tree of a source device, it is natural to presume that 4:4:4 must be indicative of best performance. If working in Hollywood post-production, that presumption is valid. But after Hollywood post-production, in content distribution, 4:4:4 is no more…it ceases to exist, except in highly extraneous isolated circumstances (Kaleidescape may be the most notable example, where content is capable of native delivery).
All media delivered at the consumer level, whether live broadcast via television, CATV, and satellite, or from streaming, are all transported in Y’CbCr, 4:2:0. Professionally, 4:2:0 is generally considered to be a transport and storage format, containing color information plus the black & white signal, yet is highly efficient in terms of bandwidth consumption. Not all devices can or will comfortably handle 4:2:0, instead performing upconversion to 4:2:2. Blu-ray and UHD Blu-ray are prime examples. Content on the disc is stored at 4:2:0 however, the disc player performs upconversion at playback. Broadcast is no different. Cable boxes and satellite receivers input signals at 4:2:0, then upconvert to 4:2:2 for output (though some EDID readings may show boxes issued by some providers upconvert to 4:4:4).
All display technologies (CRT, plasma, LCD, DLP, OLED, Micro LED, and LCOS) are RGB devices. Prior to displaying content, upconversion into RGB is necessary at some point in the signal chain.
Let’s refer back to UHD Blu-Ray. Content is still sub-sampled at 4:2:0, but unlike Blu-Ray, the enormous storage capacity allows for 10- and 12-bit color depth, plus a wider color gamut (P3 or BT. 2020) with High Dynamic Range (SMPTE ST 2084). Until HDMI Version 2.0, introduced in 2013, HDMI did not support 4:2:0 chroma subsampling. Even with HDMI 2.0, most players continued to upconvert to 4:2:2, while some offered 4:4:4 as an option.
It is at this juncture where we start to confront the central dilemma: Where is the best place to perform upconversion? In distributed systems, especially those with AV over IP limited to 1Gbps bandwidth, two interplaying factors arise. First, what does conversion better, a $129 streaming box or the $3,000 top-end display? Secondly, what is necessary to ensure the signal gets to the display?
While Apple TV, Roku and similar sundry devices all look good, is their true function designed for best image fidelity or to provide the most content? Satellite is perhaps the quintessential example of quantity over quality.
For Blu-ray and UHD Blu-ray players the past two decades, discs with test patterns were available to determine whether the player did a better job of upconversion, or if it was the display. Ultimately, the signal requires conversion into RGB for an image to appear.
Given the refinements incorporated into today’s displays, there is no compelling reason in this era to presume a display from a major manufacturer will not provide the best upconversion into RGB, so let us address getting the signal to the display.
A premium source product directly connected to a display from a major manufacturer is not the subject of this writing. The venerable Oppo 205 set to output RGB, or the aforementioned Kaleidescape, may give any well-designed display’s electronics a run for its money. But a configuration such as that (direct connection to the display) is obviously impossible with distributed video/AV over IP systems. The focus here is optimizing source devices for distributed scenarios.
As outlined previously, settings with the highest numbers may not likely deliver the best image fidelity with distributed video systems. For example, by selecting RGB output, the source product will perform all upconversion, forcing the largest amount of data through the pipeline, leaving the display with literally nothing left to do except create the image – from the ingredients delivered to it. In this mode, nearly all image processing within the display is bypassed, recognizing the RGB signal as having no need for upconversion, sending the signal into the display controller (different display types require controlling measures, such as LCD flat panels with multiple screen imaging tiles, or LCOS and high-end DLP projectors with three imaging panels). Presuming the pathway was capable of faithfully transporting such a heavily bandwidth-laden signal, minus infrastructure degradation, upconversion quality is entirely dependent on the capabilities of the source device. Intermediate, gradient permutations will yield varying results. Instead of RGB, sending Y’CbCr at 4:4:4, the display merely needs to convert Y’CbCr into RGB, avoiding color scaling. If Y’CbCr 4:2:2 is selected in the source device, the display will need to perform upconversion scaling of Cb and Cr into 4:4:4 and then convert this to RGB.
In almost every instance, the optimum scenario is to select 4K/30Hz Y’CbCr, 4:2:2 for a signal containing all the original information, yet consuming the lowest possible bandwidth, especially for content that includes dynamic metadata (which is layered on top of transport and not embedded inside the signal).
What about 4:2:0? While it represents the original signal by definition, and the latest versions of HDMI support it for transport, odd combinations of settings for most of the equipment in a signal chain is required for pass-through without some degree of difficulty.
For mixed-generation systems containing legacy displays, pre-HDMI 2.0 displays will not be able to process 4:2:0 signals. 4:2:2 is better preferred as default chroma subsampling.
NextGen systems with 8K signals (using 8K-capable matrix switchers and/or 8K extenders) the scenario is different, as these source devices will eventually contain uncompressed content.
To date, only gaming consoles are NextGen sources, on which games present different challenges (for example, Variable Refresh Rates) and in typical residential settings are direct connected.
In summary, the best practice for distributed video systems is to configure source device output at the lowest denominators possible, for ease in signal transport. Upconversion should be done by displays which in nearly every practical sense, perform this process with better precision than when configured outboard and upstream. You will eliminate infrastructure headaches from attempting to force signals that are on the bleeding edge of eclipsing system thresholds, with ultimately zero loss of fidelity to the original signal. Resist the urge to configure sources using their largest bandwidth settings…you may find your troubleshooting time plunge to near zero.
