We Need To Talk About Ones And Zeroes
Several well-respected audiophiles in this forum have stated that the sound quality of hi-res streamed audio equals or betters the sound quality of traditional digital sources.
These are folks who have spent decades assembling highly desirable systems and whose listening skills are beyond reproach. I for one tend to respect their opinions.
Tidal is headquartered in NYC, NY from Norwegian origins. Qobuz is headquartered in Paris, France. Both services are hosted on Amazon Web Services (AWS), the cloud infrastructure services giant that commands roughly one third of the world's entire cloud services market.
AWS server farms are any audiophile's nightmare. Tens of thousands of multi-CPU servers and industrial-grade switches crammed in crowded racks, miles of ordinary cabling coursing among tens of thousands of buzzing switched-mode power supplies and noisy cooling fans. Industrial HVAC plants humming 24/7.
This, I think, demonstrates without a doubt that audio files digitally converted to packets of ones and zeroes successfully travel thousands of miles through AWS' digital sewer, only to arrive in our homes completely unscathed and ready to deliver sound quality that, by many prominent audiophiles' account, rivals or exceeds that of $5,000 CD transports.
This also demonstrates that digital transmission protocols just work flawlessly over noise-saturated industrial-grade lines and equipment chosen for raw performance and cost-effectiveness.
This also puts in perspective the importance of improvements deployed in the home, which is to say in the last ten feet of our streamed music's multi-thousand mile journey.
No worries, I am not about to argue that a $100 streamer has to sound the same as a $30,000 one because "it's all ones and zeroes".
But it would be nice to agree on a shared-understanding baseline, because without it intelligent discourse becomes difficult. The sooner everyone gets on the same page, which is to say that our systems' digital chains process nothing less and nothing more than packets of ones and zeroes, the sooner we can move on to genuinely thought-provoking stuff like, why don't all streamers sound the same? Why do cables make a difference? Wouldn't that be more interesting?
Data center equipment is selected for high throughput, low latency and other factors including manageability and cost. Low noise isn’t one of them, and reasonably so since noise has no effect on digital data transmission. As a result switching mode power supplies and cooling fans are the norm, so these are noisy environments.
That’s generally true, but I’m not sure what you mean in this context? |
@devinplombier - I usually stay out of these conversations because they always seem to go sideways and inevitably lose the point of the original post. But today I have stepped into the fray.... So, here is my take on things... I agree with your overriding idea for the post. The topic is full of assumptions, misconceptions and lack of understanding. Even after my 40 years experience in computers, networking, servers, and software in multinational manufacturing companies, I’m still guilty of them at times.
Correct. Digital files, whether complete files or data being streamed should always match between source and destination. If it doesn’t then the data is corrupted and dies. Files are unusable and streaming data packets are unable to be converted to an analog signal.
You are correct about the confusion. So, let’s discuss the noise that travels over transmission lines. There is noise, and bottle necks, and delays in transmissions, and even packets taking completely different paths across the Internet to get to the destination. (Welcome to TCPIP and the original design of the ARPANET, the core of the Internet and many wide area (WLAN) modern networks.) So, what is the affect of all the general "noise". Typically it only will cause delays in packet transmission and receipt. The protocols in place are designed to assure the data comes through "unscathed" (and yes, I’m using that word on purpose), even it if requires a lot of time, possibly to the point of timing out and the connection getting dropped. And there is one of the rubs, speed of data flow. More on that in a minute. But first, it needs to be noted that the "noise" present across any transmission/receiver leg ends at that receiving hardware. This would include the legs from the original server to network switch, network switch to firewall, firewall to router, router to your ISP, your ISP to home modem, home modem to network switch, or network switch to streamer. Why is this true? It is because the data always travels in packets, not in a start to finish steady stream like an analog signal. So each leg starts off with a complete packet and that packet is clean and isolated from the previous leg. Now back to the rub of the data flow speed. This is where somethings can go wrong. In the case of a sequencial "streaming" of data, whether being video or audio, eventially the data packets need to be sequenced and converted to a continous analog signal. (BTW, I’m trying to keep this very simple...) And here is where time comes into play. If there are delays in receipt due to noise or packets coming into the buffer out of order, the data won’t stay ahead of the requests from the processor. And thus you get Jitter.
One misconception in the statement above. "super-noisy industrial grade switches". These industrial grade switches are designed to route the data packets in high volumes and with minimal requests for packet resends. The onboard software control is significant and monitored (i.e. level 3 managed switches). They have large data transfer backplanes and these days are running at 10 GB upwards of 800 GB ethernet speeds. Your home networking equipment typically maxes out a 1 GB. Any noise inside the circuitry is negligible in this discussion. Then, to the question of how can the last 10 or 20 feet in you house have an impact. First, the hardware is cheap and relatively slow. Enterprise level network switches, like what AWS is using, can run up into the $10K plus range. Routers and firewalls used by corporations and ISP’s in the same range, or even higher. Your home modem and network switches, a couple of hundred bucks. In fact, they are actually noiser than Enterprise level equipment. Second, power supplies. Typically, home equipment use power supplies that are barely enough to run the equipment. Start loading it down with constant data flows, and they start to fade out, speeds drop, induced noise goes up, packet resend requests rise and things go boom on the receiving end. Third, shielding. Enterprise level equipment is usually in metal cabinets. Power supplies are shielded internally. Processors and buffer memory as well. Most home modems/routers/ and switches, not so much.
Honestly, power supplies outstanding, I have no idea what some of these companies do to create an "audiophile" switch. I have seen a few that have been opened or completely torn down, and other than a different paint job, they seem to be typical home equipment. Just slap a different power supply in, call in audiophile, and charge a $400 markup. I don’t know. All I know is that my Enterprise level Cisco switch in my basement will out perform it. Thanks if you read this all the way though. I know I written a a novel. Time for bed. - Jeff |
It’s wearisome to answer Ethernet experts with their thinking that their knowledge/profession make them totally knowledgeable in digital High-end audio. Their “I’m right until you prove me wrong” requires much effort to educate. Then another one comes along. The digital data rides on an analog signal that can pick up noise. High-end computing ≠ High-end audio where we care how it “sounds” |
