|
 |
NOTE: this article is now outdated, but has been retained for archival purposes.
HDMI 2.1 and 48G Cables --
What We Know and Don't Know
Well, the HDMI organization has surprised us all by announcing a 2.1 specification update, and with it, support for such things as 8K/60 and 4K/120. While most specification updates have not involved a new cable (this has happened only once - when HDMI 1.4 introduced the "Ethernet" option on HDMI cables), this one makes reference to a new cable capable of handling 48 Gbps, considerably more than the 18 Gbps data rate which is the top end of HDMI 2.0.
We've been receiving a lot of questions from customers about this, and we are, as always, eager to clear up whatever is unclear. Unfortunately, while HDMI Licensing announced that the 2.1 spec is coming, it's not yet available, and our efforts to get any inside hints on this from HDMI Licensing have been unavailing. We are an HDMI Adopter and we will have the specification as soon as it is published, but all we know for certain about that is that it is promised for "early in Q2 2017." When we have it, we'll explain precisely what it involves. For now, however, we can mostly only tell you what little we know, what we don't know, and what our own educated guesses are as to what the spec will involve.
What We Know:
The HDMI press release tells us:
48G cables enable up to 48Gbps bandwidth for uncompressed HDMI 2.1 feature support including 8K video with HDR. The cable is backwards compatible with earlier versions of the HDMI Specification and can be used with existing HDMI devices.
A few things can be gleaned here. First, video under 2.1 will remain, as in previous versions of the HDMI spec, uncompressed. One easy way to get more throughput would have been to go to some sort of compression, but we can rule that out. Second, the new HDMI spec will NOT use the "dual-link" 29-pin HDMI connector which has been a spectre haunting the HDMI spec since 1.0. It will use the same 19-pin connector, and we know that its pinout will remain the same (HDMI has no crossover pins - pin 1 maps to pin 1 on the other end, and so on, through pin 19) because if it didn't, the cable would not be backward compatible.
What We Don't Know:
Extent of Forward-Compatibility of existing cable:
The first big question, of course, is whether the cables used under previous versions of the HDMI spec will be, in any respect, forward compatible. It's a bit hard to work out the precise answer from what HDMI Licensing has said. On the one hand, we have this in the Q&A:
Q: Will 8K@60 or 4K@120 require a new cable or cable connector?
A: Yes a new cable is required
But here, it gets a bit wishy-washy:
Q: Is this cable required for delivering HDMI 2.1 Specification features?
A: The cable is the best way to ensure the high-bandwidth dependent features are delivered including the enhanced video and audio performance, and accounting for the new EMI characteristics
Q: Will existing HDMI High Speed cables deliver the HDMI 2.1 features also?
A: While existing HDMI High Speed Cables with Ethernet can deliver some of the new features, the new cable is the best way to connect HDMI 2.1 enabled devices to ensure delivery of all the features.
The best way to make all of this consistent is to assume that HDMI is making a clear distinction here between resolutions and features, and to suppose that something about the cable design precludes existing cables from supporting the new higher resolutions, but that other new features will indeed be supported by existing cable. We assume that indeed something will be done to the cable design which will render existing cables incapable of supporting the new resolutions, but that leaves a big question: what will that change consist of?
Manner of Accommodating 48G:
The permissible data rate of HDMI signals has been increased twice before: from 1.0 to 1.2, it was 1.65 Gbps/channel (4.95 Gbps total); in 1.3 and 1.4 it was 3.4 Gbps/channel (10.2 Gbps total); and in 2.0 it went to 6.0 Gbps/channel (18.0 Gbps). In both of these increases, the basic layout and structure of the cable remained the same. Four shielded data pairs in the cable carry the main signal -- one of these carries nothing but clock pulses, however, while the other three share the load of video and audio data. The encoding involves only two voltage levels -- in other words, it's all "ones and zeros," as they say.
How will this new cable accommodate 48 Gbps? We are pretty sure of one thing: the answer will NOT be to simply increase the permissible data rate, under the existing binary encoding system, through the existing data pairs. 6 Gbps/pair is already very difficult; it's posing considerable difficulty for some users as it's becoming harder and harder to ensure performance, especially over distance, of HDMI signal links. To run that to 16 Gbps/pair would be a disaster. At 16 Gbps, one would want the pairs to perform well out to 24 GHz, as this gives good results at the fundamental frequency (half the data rate, 8 GHz) and at the third harmonic, which is where much of the sharpness of the transitions resides (24 GHz). But no pair performs well at 24 GHz, and the return loss difficulties become immense. The wavelength of a 24 GHz signal is only 1.25 cm (in air -- somewhat longer in insulated wire). Impedance discontinuities and periodicities generally are significant it they are at least a quarter-wave long, and nobody can expect a paired cable to perform well where lengths under half a centimeter become critical. At these short wavelengths, even the diameter of the wire starts to become a factor as signals have multiple paths of differing length around the surface of the wire - meaningless at low frequencies, and quite a headache at 24 GHz.
All right, then, how to do it? Well, there are some possibilities that come to mind, and we think the answer is probably some combination of the following. Bear in mind that these are our educated guesses; what the spec actually says is something we won't know for sure till we see it.
Possibility 1: More Data Pairs, And More Use of Existing Pairs
As we've noted, there are still 19 pins, and 19 wires, in the new cable. Right now, twelve of these wires correspond to the plus, minus, and shield connections of four twisted pairs, leaving seven miscellaneous conductors for other purposes. A fifth twisted pair exists in any cable "with Ethernet" -- in HDMI 1.4, the specification added Ethernet and Audio Return Channel functions to some of those seven miscellaneous conductors, which now must be organized into another twisted shielded pair in order to support Ethernet over HDMI. That leaves four conductors still available in the cable for "multi-tasking," and it's entirely possible that three of these have been given new, additional function as the members of a sixth data pair.
And while the electronics to do it get a tad complicated, it's quite possible that more functions are being added to the clock and Ethernet pairs on the cable as well. The clock pair is the most underutilized of the existing pairs, carrying nothing but clock pulses, and it's entirely possible that some share of the actual data burden will now be distributed to it. The Ethernet pair isn't very heavily burdened either, though it does have Ethernet potentially running on it in full duplex mode -- so it's quite possible that some portion of the data, though likely not a "full" share, may be pushed onto it as well.
Our best guess is that there will indeed be a pair added, using three of the four currently-unpaired conductors. We base this guess upon the fact that the HDMI press release does seem clear that one WILL need a new cable for the highest resolutions -- if the internal cable architecture were not changing, one would not know that all existing cables would be insufficient.
Possibility 2: Other Internal Structural Changes
We don't think this one's very likely, but will throw it out there. Pairs aren't the only way to make a data cable, and if one is looking for high impedance stability at extremely high frequencies, coaxial cable is considerably superior to pairs. There are some radical changes that could be made -- a 100 ohm balanced data pair is equivalent to two 50 ohm coaxes sharing a common ground connection, and so cables could be made using an internal mini-coax structure to improve high-frequency performance while still maintaining backward compatibility with the 100 ohm paired version of the cable. In such a configuration, the coaxes could be de-paired -- turning what is now a single data channel into two separate, unbalanced data channels.
Possibility 3: Multilevel Encoding
A Cat 6A cable can carry 10 Gigabit Ethernet; but it's tested only to 500 MHz, well below the 3.75 GHz level of the third harmonic of a 2.5 Gbps (10 Gbps divided by the four pairs, which are run in full duplex mode) binary signal. How come? The answer is multilevel encoding. Instead of dealing in "ones and zeros" a 10GBaseT circuit sends and receives in a sixteen-level encoding system, PAM-16, which allows one to cram a good deal more data in to the signal without needing the same extreme high-frequency performance.
It's unlikely that we'll be seeing anything quite as extreme as PAM-16, but some type of multilevel encoding, using intermediate voltages apart from simple ones and zeros, would reduce the bandwidth demands on the data pairs. We think this is very likely to be part of the solution here -- even if there were six data pairs running full-bore, 48G is still 8 Gbps/pair, which is a lot to ask of a data pair running binary-encoded signals.
Is multi-level encoding the entire answer? It could be, in principle -- one could use pairs that perform acceptably under existing HDMI standards to convey the whole signal on the same three data pairs currently in use. We think this cannot, however, be the entire answer because if it were, HDMI Licensing would not be assuring us, in the press release, that a new cable will be needed to handle the highest resolutions. If the only change were to introduce multilevel encoding on the existing data pairs, one could not be sure that an existing cable would be insufficient for the job and HDMI Licensing would probably have said that existing cables might work but might not.
Don't Fear The Future!
Whatever the answer to all of this is, we'll know it, and publish a summary of it, as soon as we have got our hands on the specification -- and if that requires cable redesign, we'll be pushing ahead as quickly as we can to get new cable stocks built for it. We'll keep you posted as that process moves along.