Edit: I just want to preface this blog, after rereading it, that I do know you can send an analog or digital signal over most cables since they are just electric pulses any way you slice it. One uses period of charge and one uses wavelength to determine data contents. I mention them in stark contrast a lot, and I mean in the practical utilization of cable standards (for example, TN phone cable is really crappy digital cable since it has terribly low refresh rate).
Today there are quite a few "standard" connections for a wide variety of technology. Most of them boil down to analog or digital, though. For example, 3.5mm audio is almost a hundred years old and is just differential voltage meaning height of a sound wave. Digital signals are barely different - they are still electrical signals over a wire - but they are analyzed with a discrete principle that the presence of absence of charge can indicate their state. It is why spdif requires a sound card on the recieving end to reencode the signal as an analog output and 3.5mm audio is just raw sound signals. For the same reasons, sound over distance becomes distorted if not modulated and digital signals can go longer distances (some ethernet standards can go kilometers). And you can tell the difference because all the digital standards are measured in bitrates, and analong cabling is always fixed to some interpretation standard (analog video off coax or rca cable, or off vga, for example).
One really silly reality is that modern computers almost always use analog modular sound and digital via hdmi or dvi video. For some dumb reason, dvi has some backwards compatability with vga and supports analog signals on some of its pins at some low video resolution modes. Hertz rates play a big part in cable distance in digital world, since dvi is usually 400hz, and runs on copper, so it is limited to a short distance compared to the often fiber based log 550mhz ethernet lines, but they carry less data rate.
The big deal is the end is the tradeoff of bandwidth to distance traveled in digital land. Same thing happens in wireless world, except with smaller numbers because of much higher environmental interference in the air. But I don't get why we can't just have standard digital interconnects that span the spectrum they really translate to. What we want is the greatest stable frequency distance with the lowest power draw transferring the maximum data rate with the lowest latency. Each of those 4 quantifies some use case, and each deserves a specialization - really, we have that already. distance is coax, power draw is ethernet cat (which blurs with distance, since they are synonymous in many ways), bandwidth is hdmi / dvi (6 gigabit, but ethernet is getting competitive, so like I said, lines are blurring) or latency (again, they are all mostly low latency - usb has a higher latency since it does byte conversion on all transfer nodes).
In honesty, usb sucks. It has a really short distance before distorting, it only has (at usb3 even) 600mbit bandwidth, and it has higher latency because it is a translation layer. Ethernet is already packet based, it just doesn't provide power. Is it really that expensive for a digital packet interconnect standard to carry power? If so, why does SAS work at 6gigabit (albeit, a SAS or SATA cable has a much shorter range limit than any of the above).
You really end up with two use cases then. Maximize bandwidth with minimum losses of distance, power, and latency; and maximum distance, with minimum loses of bandwidth, power, and latency. You want low latency / power draw, and preferrably the ability to prove 5v or some such power over the line (albeit that, by example of SAS and USB, really costs the distance, so a long distance interconnect is almost guaranteed not to also carry power due to noise problems, and shielding the two layers would be expensive).
So what do we end up with? I'd say something like SAS + ethernet. And cat ethernet cable is really starting to pique my interest, since we are now talking about 100 gigabit ethernet, with room for terabit. The bandwidth on these cables is huge, the maximum range is huge as well without signal degredation, and all it takes is intelligent enough tvs and monitors to decode packets. Really, it isn't hard for ethernet packet systems to just ask newly connected devices what they are for and provide accordingly. The only real issue would be in the realm of a network adapter thinking its on a router when its on a display, and you get noise on your screen, or vice versa. Just means you need good handshakes.
I guess that also validates the reasons for having divergent systems in place for each device. Even though usb blurs them all since it seems to not care what goes over the line as much. Ethernet controllers want packet data, hdmi and dvi want display packets, usb wants device packets, etc.
I really was thinking about this since usb is a train wreck. USB device drivers are erroneously complex, slow, and in general suck. They all draw from one device controllers bandwidth and power, they have terrible transfer rates and distance, and seem to only be a standard for the sake of it. They were first to market with plug and play powered ubiquitous devices and won where I am really surprised nothing is competing. Because USB is shit.
And the digital display (vga is too restrictive a medium, which is interesting in and of itself - with 4k resolution on the horizon, it is really interesting how analog can't support that) with analog audio (well, hdmi carries digital, but the idea of digital display data being decoded by screens while headphones and speakers get raw audio sine waves seems like an undue disparity, but they do represent fundamentally different things - after all, photons are particles and pixels are inherently discrete and digital, but sound is a wave and is inherently continuous and analog).
So it is an interesting thought experiment, at the least. Still can't believe we cant' do better than 3.5mm modulated voltage copper wires for sound transfer though, in a hundred years. Sound seems like it was always an easily solved problem, so why the hell can't ALSA get hardware mixing right.
No comments:
Post a Comment