I originally wrote:
there is probably an explanation in physics
Subtext: when probability fails, then try behavioral psychology. Heard behaviour?
Do I need an expensive digital cable?
I have been using a fairly inexpensive optical cable to connect my CD transport to my Moon 280D streamer. I was told that an SPDIFcoax cable would sound better. For an experiment I purchased an inexpensive Pangea coax cable. It didn't sound at all because its terminator ends did not fit snugly in my equipment. I consulted chatgbt who often gives me audio advice. It advised that for the short run of 1 meter, an RCA interconnect would work. It did. And sounded much better than the optical. Chatgbt said that RCA interconnect was good enough.
Now, there is a twist to this story that might make those doubters think twice. A digital cable carries packets of information that are rechecked to assure that the streamer is recieving correct information. There is the timing concern, though. But my Moon 280D has an asynchronous DAC with a clock as part of the DAC. Any information sent by my transport, whether it is clocked by the transport or not, will go through the Moon's asynchronous DAC's clock. So ;there shouldn't be a timing problem. Should there?
Can anyone make a case that I should buy a "better" coax cable?
- ...
- 278 posts total
Lol @sls883 |
@audphile1 @devinplombier @richardbrand Okay, now that we're all friends and we've thrown one-ups-man-ship (Richard, you can correct the spelling on that) out the window, Richard, would you please explain to us what is the difference between a 16 bit word length and a 24 bit word length. I know that a bit is an on/off switch and that sixteen of them gives us 2 to the 16th power possibilities. But what do they store in that word or packet or whatever you want to call it, and why do they need so many other possiblities from 2 to the 24th power? I can certainly hear the difference, but I have no idea why. |
I'll try, but it will be digital 101 First I want you to count to 15, just using the fingers (digits) on your left hand. The least significant finger will be your little finger. For zero, none of your fingers is up (on). For one, your little finger is up (on). For two, it gets a bit trickier. Stick your ring finger up, and your little finger down. For three, both your ring finger and your little finger are up. For four, only your middle finger is up. Repeat counting like this until you get to 15 (all four fingers up). With me so far? Now keep counting using just the fingers (no thumbs) on both hands. You should get to 255, which is the maximum you can count in with 8 fingers, or one 8-bit byte. This number frequently appears in IPv4 addresses such as 192.255.255.1 because these are written for us to read more easily as four bytes. Internally they are just 32 bit whole numbers. What about including toes? Ok, excluding your big toes, you can get to 256 x 256 less one, or 65535 for just 16 digits or bits. Your forbears thought they were doing well to count to 20! And the Romans really struggled, because they had not invented zero and had really awkward ways of writing numbers, to boot. Now let’s go back to your graph paper, where you can draw a squiggle representing sound pressure on the y axis and time on the x axis. Let the graph paper be divided into millimetre squares. We are going to map the sound pressure on to 16 bits, so the graph paper needs to be 65.535 metres high. We are also going to use a millimetre for each time sample, so for CD quality every second will need over 44 metres of graph paper going sideways, per channel. To convert your squiggle to Pulse Code Modulation (PCM), you just need to put a dot on the closest grid point on the graph paper. The y readout is an exact whole number that can be represented as 16-bits, that can change about 44,100 times a second In this model, silence is centred on 32,768. Note that the coding is linear – the tiniest wriggle is represented at quiet times and at peak levels. Error detection Let’s add a spare bit to each 16-bit sample. Set that bit so there are always an odd number of bits. On read back, if there are an even number of bits, you have detected a single bit error, though not multiple bit errors. This spare bit is known as a parity bit. Wait, there’s more. Arrange 16 words into an array and add a vertical parity bit per column. Now you can find out exactly which single bit in in error, and therefore you can correct it. You can also discover when you have two bits in error. There are very clever ways of adding data redundancy to detect and correct errors. The Reed Solomon Cross-Interleave coding used for CDs allows 4,000 consecutive wrong bits to be corrected. Upping the ante Higher time resolution is simple. You want 192-kHz? Use four times the width graph paper. More sound pressure resolution? Try 24 bits because modern computers work on bytes. You have 256 times as much resolution, so your graph paper needs to be over 17-kilometres high. But the principle is the same. So what’s the basic problem with PCM? Even with just 16 bits, there are pressure level changes where lots of bits switch off simultaneously and one more important bit switches on. The important bit should contribute tens of thousands more, but how can this level of accuracy be guaranteed? It is the monotonicity problem and it is why 24-bits is a myth even with the best dacs. Enter Direct Stream Digital (DSD) Instead of recording sound pressure as an absolute number, why not sample much more frequently (at least 64 times as often as CD for DSD64 at 2.8-MHz) and just record if the sound pressure is now different (higher or lower) compared with the previous dot on the graph paper. Of course, you need 2.8-kilometres of graph paper per second! I have not gone into how digital noise is filtered out, but removing MHz noise is much simpler than the steep filters used by dacs which do not oversample.
|
- 278 posts total