HDMI 2.1: The Need For Speed Continues

Ever since HDMI version 2.0 was announced in September 2013, I’ve been pretty vocal about criticizing its “not quite fast enough” speed upgrade from 10.2 to 18 Gb/s, which turned out to be barely adequate for transporting 4K (3840×2160) video at full color resolution (RGB, or 4:4:4 in the world) at a frame rate of 60 Hz – and only with 8-bit color.

Given how quickly the display industry is shifting to 4K and even higher resolutions, it was inconceivable that this new interface would in effect create a “speed bump” in the 4K chain, particularly since high dynamic range (HDR) and wide color gamut (WCG) enhancements were becoming part of the UHD ecosystem. And both enhancements require at least 10-bit color rendering, something that would be impossible to pass through the HDMI 2.0 interface if using a full-resolution color format.

It didn’t help that HDMI’s competitor – DisplayPort – had already broken the 20 Gb/s barrier way back in 2007 with version 1.2 and could easily interface a 2160p/60 signal with 10-bit RGB color @ 60 Hz, and earlier in 2013 had announced version 1.3, which saw a speed boost to 32.4 Gb/s.

For a time there, I thought the superMHL format, which had its debut at CES 2015, might be the successor to HDMI. It was faster (36 Gb/s), had a large, reversible connector, was compatible with USB Type-C Alternate Mode, and most importantly, supported Display Stream Compression.

Alas; it appears superMHL turned out to be mostly a science experiment. The MHL Forum was conspicuous by its absence at CES 2017, but the HDMI Forum more than made up for it by unveiling version 2.1. And now, we’ve got a real horse race.

High dynamic range support will be much easier with version 2.1, especially deeper color from RGB sources.

THE DETAILS

The public press release on HDMI 2.1 is sketchy on details, except to say that the maximum speed of the interface has now reached a mind-boggling 48 Gb/s (that’s faster than most network switches!). Quite the leap from 18 Gb/s, wouldn’t you say?

The release goes on to talk about a new generation of 48G cables, a greatly improved eARC audio return channel with auto-detect, and finishes with a discussion of high dynamic range and higher video resolutions, both of which are possible with faster data rates that enable higher frame rats and deeper color. And of all of this happened while retaining the familiar 19-pin Type A connector. (Wha-a-a-t?)

But what’s really going on here? How did HDMI accelerate to 48 Gb/s? Hold on, and I’ll provide the details missing from the press release.

First off, the current version of HDMI uses three connections – well call them lanes, like DisplayPort does – to transport red, green, and blue display pixels. There’s a fourth lane for the clock to synchronize frames, and the balance of the connectors are used for ‘hot plug detect’ connections, the Data Display Channel (EDID). That doesn’t leave much room for expansion.

But HDMI 2.1 adds another lane for TMDS data (although it’s not really TMDS anymore) by taking over the clock lane and embedding clock data within the existing signal, much the same way it’s done with packet-based signaling systems.

Next, the physical data rate over each lane has been raised from 6 Gb/s to 12 Gb/s. I don’t know how that 100% increase was achieved, but that’s an impressive achievement considering that we are still waiting for 12G SDI cables to come to market.

The 12G number may also be a function of jiggering the acceptable signal-to-noise (SNR) ratio, something proposed a year ago by Steve Lampen of Belden – but then again, we’re not likely to see 12 Gb/s of data traveling down any display pipes in the immediate future. (For comparison, DisplayPort’s HBR3 cap is 8.1 Gb/s per lane.)

That’s not all. The standard ANSI coding format for HDMI, DVI, and DP (not to mention numerous other interfaces) is known as 8b/10b, coding 8-bit words into 10-bit symbols, resulting in about 20% overhead. Example: A 4K/60 signal encoded as an 8-bit RGB signal requires 17.28 Gb/s, and 20% of that is overhead from 8b/10b coding.

HDMI 2.1 has adopted a more obscure form of coding known as 16b/18b. You can find a IEEE PDF from 1999 describing how it works here, and it’s formally known as “partitioned DC-balanced 16b/18b transmission code.” The net effect of moving from 8b/10b to 16b/18b is reducing the overhead to about 12% from 20%. What’s interesting though is that the HDMI 2.1 signal isn’t really TMDS we’ve come to know and love when in this mode – it’s something else, possibly more of a packet structure.

HDMI is now compatible with USB Type-C Alternate Mode – a”must have” feature for any new display interface.

Last but not least, HDMI announced last fall that it was compatible with the USB Type-C Alternate Mode format. And now, it appears that HDMI 2.1 is also compatible with DisplayStream 1.2 compression, which is a much more efficient way to transport signals like 7680×4320/60 (8K, for those not paying attention). Although at 48 Gb/s, version 2.1 could theoretically transport that signal uncompressed using 4:2:0 color.)

Compatibility with DSC wouldn’t be that much of a shocker – superMHL also offered it and it’s another TMDS format. In fact, at second glance, it appears that much of the engineering that went into superMHL has now migrated over to HDMI 2.1 (about time) and the most significant breakthrough is doubling the interface speed.

Given that 40 Gb/s is definitely optical fiber territory, the only remaining question is why we still haven’t seen a detailed HDMI specification for direct optical interfaces. 48G cables will be expensive and difficult to engineer, but multimode optical fiber can already do the job and is cheap. To come up with 50-foot and longer manufactured optical cables for HDMI would be a piece of cake – and it’s already been done in the past for HDMI 1.3/1.4.

So there you have it: HDMI 2.1; a faster, smarter, and more appropriate display interface as we head into the era of 4K and beyond. How soon will we see HDMI 2.1 interfaces and cables? Well, considering it took almost 3 years for version 2.0 to achieve any significant presence in commercial AV, I’d say maybe a year from now at the earliest…and perhaps not until 2019 in any quantity.

By then, a good deal of the industry may have already shifted to AV-over-IP for the bulk of its signal switching and distribution, using simple format conversion at the display end. And we still have to see who is going to adopt on DisplayPort 1.3/1.4, still a “no-royalty” interface that can hit 32 Gb/s and supports all the forward-looking necessities (Type-C Alternate Mode, DSC, HDR).

Gentlemen, start your engines…