Topic: Are Bits Just Bits? (Read 2898 times)

DaveBSC · « **Reply #10 on:** April 15, 2013, 03:58:49 AM »

Quote from: AstralStorm on April 14, 2013, 02:43:59 PM

SPDIF coax has the clock signal provided with it instead, but the DAC and/or interface chip needs to keep very well in sync with it or nasty sound happens.
Optical is about the same, but you also need fast precise photodiodes - and the cable is more of a pain. On the upside, it enables full galvanic isolation.

S/Pdif was always a quick and dirty series of compromises. At least with BNC connectors and proper 75Ohm terminations it works reasonably well. Breaking ground loops and immunity to RF/EMI is great.. but in my experience Toslink causes more problems than it solves, and support above 24/96 is limited. In general I think Toslink should be used only as a last resort and is otherwise best avoided.

Interestingly, I've read that ST (Toslink's technically superior yet nearly extinct cousin) is only effective at extremely long lengths where it beats everything else short of Ethernet. Using ST at say 1M apparently causes all kinds of nasty reflection problems which may explain its near extinction. Most people don't need 10M digital cable runs.

DaveBSC · « **Reply #11 on:** April 15, 2013, 04:03:10 AM »

Quote from: gelocks on April 15, 2013, 03:28:51 AM

Bits are just Bits....

Is not that different from saying "feedback is feedback" or "harmonic distortion is harmonic distortion". All imply a nearly total level of ignorance on the subject.

Tari · « **Reply #12 on:** April 15, 2013, 04:10:13 AM »

And East is East, and West is West, and if you take cranberries and stew them like applesauce they taste much more like prunes than rhubarb does.

Now, uh... tell me what you know.

ultrabike · « **Reply #13 on:** April 15, 2013, 04:41:43 AM »

I'm much more familiar with digi-comm than audio applications, but here are some of my thoughts: AFAIK it is desired to have the analog part of a system isolated as much as possible from the digital part. Space constraints result in trade-offs (fertile ground for innovative solutions.)

In the digital side of some commercial JAT (Jitter ATtenuator) and CLAD (CLock rate ADapter) that I've worked with, there are more than just digital buffers in the digital path (such as filters, comparators, multipliers, and stuffs) than may inject noise to the analog path.

The bits may not necessarily get corrupted inside or at the digital interface of the DAC (digital buffers might be practically error free.) However, the analog part of a DAC (some clock elements, S/H, output lowpass filters, analog buffers and driver i.e. amps, etc...) responsible for the reconstruction of the digital signal into the analog domain might get corrupted by noise generated by the digital circuits themselves. The noise may result in relatively low ENOB (effective number of bits) numbers, and possibly high jitter values.

Furthermore, from the little I have learned in audio, the drivers (amps) in an audio system might need to support relatively large dynamic ranges. In PCM (BPSK), the dynamic range is fairly low, which makes the life of the driver amps a little easier (specially in high data rate cases.) However, in WiFi systems using OFDM, the dynamic range needed out of the amps is much higher than in PCM, and the drivers may "compress." Audio signals may have even higher dynamic range than OFDM signals... and the noise requirements might be more stringent due to the dynamic range requirements.

Solderdude · « **Reply #14 on:** April 15, 2013, 04:54:19 AM »

Quote from: DaveBSC on April 15, 2013, 03:58:49 AM

Interestingly, I've read that ST (Toslink's technically superior yet nearly extinct cousin) is only effective at extremely long lengths where it beats everything else short of Ethernet. Using ST at say 1M apparently causes all kinds of nasty reflection problems which may explain its near extinction. Most people don't need 10M digital cable runs.

ST (Straight Tip / Bayonet Fiber Optic Connector) is one of many different, and quite common, connector types and still used a LOT in fiber optics today. ST connectors are intended for glass fibers with a core diameter of 9 micron, 50 micron or 62.5 micron. the fibers themselves are all 125 micron in diameter.
The only thing glass fibers and plastic fibers (like TOSLINK) have in common is... they use light.
ST is not anywhere near extinct, but better (push/pull) connectors are out there and nowadays are more often used as the 'old' bajonet mount of the ST. Before the ST connector was there screw mount connectors were mostly used.

Using real optical fiber (glass not plastic) even at 0.1m is no problem unless one uses lasers or high levels of light intended for long distances which can 'overexpose' the detectors. glass fiber can easily cross distances of hundreds of km's (smaller 9 micron so called Single Mode (SM) fibers that is). Gigabit speeds is no problem as isn't multiplexing with different 'colors'.
In most cases MM (Multi Mode) connections are several 100 meters to a few km's long. speeds in general are lower upto 255Mbit in general.
It is less jittery than 1mm plastic fiber but for 'normal' data transfer over long distances jitter is not a problem, smearing (limited bandwidth) can be with MM fibers over longer distances, also dependent on the 'color' of the infrared wavelengths used and the core diameter and type of multimode fibers.

mkubota1 · « **Reply #15 on:** April 15, 2013, 08:01:43 AM »

Quote from: Analixus on April 14, 2013, 09:52:33 AM

So that's what happens when you take the red pill in Japan.

Haha... funny. I didn't realize it until now- that is (mostly) Japanese in a mirror.

kkl10 · « **Reply #16 on:** April 15, 2013, 12:06:09 PM »

Precious source of information so I linked this thread on the How to get clean audio from the PC to a USB powered Dac? first post.

This makes me even more tempted to test my ODAC with an USB over Ethernet server...

kkl10 · « **Reply #17 on:** April 15, 2013, 12:23:32 PM »

Quote from: roBernd on April 14, 2013, 11:30:21 AM

skipped trough it, more like random syllables, some mirrored, some Latin.

So, TL;DR
-COAX is the way to go (with the right drivers installed)?
-USB drivers and ports aren't by default designed for audio transmission? (similar to TCP vs UDP in network standards?)

USB standard was not originally designed with real time audio streaming in mind in the first place.
And that shows by the way how the digital and analog parts on most typical USB hosts are not well separated and by it's numerous sources of Jitter (even when Asynchrounous mode transmission is employed which the OP quoted text seems to suggest).
Ethernet is more reliable for any digital data transmission which requires "bit perfect" quality.

BUT:
what if the DAC has a buffer of his own, where he stores the (corrected) data given from usb and plays everything with a short delay.
If this is the case the whole signal would be bit-perfect without any checking or corruption in the end, because the checking happened before it was buffered in the DAC?

In Bulk mode there's a buffer and error checking and correction function, I think, but this still won't solve all issues that can affect sound quality.

AstralStorm · « **Reply #18 on:** April 15, 2013, 03:55:45 PM »

The piece is somewhat filled with junk, but otherwise good.

Clock phase modulation is visible on IMD chart as a spreading of the central peak and (relatively large) increase in IMD in general, so can be seen in good specs.

The fastest isochronous update frequency is every 2 ms, giving highest clock modulation frequency of 500 Hz. It can even be done every 512 ms if the buffer allows - 2 Hz. (that's some 256 KiB at 24 bit packed, not really astonishing) Even if data fill is done low latency in small chunks (e.g. those 2 ms frames), that is still not too bad as the signal amplitude is low - it's differential for exactly this reason and that's why USB cables cannot be too long.
Even if you do send those 2 ms chunks, NRZI encoding with bit stuffing nicely improves noise capability by shifting the frequency at least to 1.5 MHz, maybe 60 MHz even at high speed. It has a vanishingly small low frequency component. USB 3 is even better, with linear feedback shift register plus 8b/10b - making the data transfer a violet noise - no DC, no carrier frequency peak.

USB is noisy for an entirely unrelated reason - it is a nice channel for all kinds of ugly EMF straight from the computer. The only way to fix that is to install a regenerator in the middle, preferably far from the target device.

SPDIF unfortunately requires you to use that dirtied clock or run clock injection back. Actually optical is better in this regard, as there are no transformers in the path to mangle the clock waveform - as long as the photodiodes and laser diodes are fast - this should mean lower jitter, but unfortunately most everyone uses cheap chinese crummy laser.

--
Bulk transfers (usually) aren't used for audio because:
1) Such mode is not defined in USB Audio Class, thus anything like that is proprietary.
2) There is no bandwidth reservation support.
3) You still need a protocol for synchronization.

--
I forgot to mention: Windows default USB drivers aren't stellar. Linux (when it works) is by far superior. OS X, I can't tell, I don't own any stuff made by an orchard company.

DaveBSC · « **Reply #19 on:** April 15, 2013, 06:18:04 PM »

Quote from: Solderdude on April 15, 2013, 04:54:19 AM

ST (Straight Tip / Bayonet Fiber Optic Connector) is one of many different, and quite common, connector types and still used a LOT in fiber optics today. ST connectors are intended for glass fibers with a core diameter of 9 micron, 50 micron or 62.5 micron. the fibers themselves are all 125 micron in diameter.
The only thing glass fibers and plastic fibers (like TOSLINK) have in common is... they use light.
ST is not anywhere near extinct, but better (push/pull) connectors are out there and nowadays are more often used as the 'old' bajonet mount of the ST. Before the ST connector was there screw mount connectors were mostly used.

Using real optical fiber (glass not plastic) even at 0.1m is no problem unless one uses lasers or high levels of light intended for long distances which can 'overexpose' the detectors. glass fiber can easily cross distances of hundreds of km's (smaller 9 micron so called Single Mode (SM) fibers that is). Gigabit speeds is no problem as isn't multiplexing with different 'colors'.
In most cases MM (Multi Mode) connections are several 100 meters to a few km's long. speeds in general are lower upto 255Mbit in general.
It is less jittery than 1mm plastic fiber but for 'normal' data transfer over long distances jitter is not a problem, smearing (limited bandwidth) can be with MM fibers over longer distances, also dependent on the 'color' of the infrared wavelengths used and the core diameter and type of multimode fibers.

Whether or not the piece I read was accurate or not in terms of the limitations of ST as a digital audio transmission format, it's true that it's effectively dead in the audio space. In the early to mid '90s pretty much all high-end PCM63 or UA DACs had ST inputs. That's all over. Bel Canto is still using it, pretty much no one else is. BC only recently started using it as well, there was no ST input on their DACs from a few years back. It allows you to separate a converter (and attached computer) and DAC at distances far longer than USB can support without an active cable. If you want your computer 50 feet from your DAC, your options are pretty much Ethernet, or a Bel Canto ST converter and DAC.

Author Topic: Are Bits Just Bits? (Read 2898 times)

DaveBSC

Re: Are Bits Just Bits?

DaveBSC

Re: Are Bits Just Bits?

Tari

Re: Are Bits Just Bits?

ultrabike

Re: Are Bits Just Bits?

Solderdude

Re: Are Bits Just Bits?

mkubota1

Re: Are Bits Just Bits?

kkl10

Re: Are Bits Just Bits?

kkl10

Re: Are Bits Just Bits?

AstralStorm

Re: Are Bits Just Bits?

DaveBSC

Re: Are Bits Just Bits?

(adsbygoogle = window.adsbygoogle || []).push({});

Author Topic: Are Bits Just Bits? (Read 2898 times)