Why You Might Not Want to Stream Digital Music from the Net

And now for something completely different: We’re honored to present a post written by guest contributor Dave M. (“I am by profession, an audio design engineer and I design recording equipment for major recording studios and film scorers. That’s what I do.”)

Why you might not want to stream digital music from the Net:

Here is an explanation of digital audio. I hope it is not too boring. I am not talking about the legality of music downloads, that is a whole separate subject; but I’m just going to explain why downloads sound bad.

Music, (sound), is a real phenomenon made up of little waves of air pressure that reach our ears. The normal atmospheric pressure, just due to the weight of air above us and gravity, is about 14.7 pounds per square inch (psi). This varies a little due to high and low pressure areas of weather. Fast variations occur when something is vibrating or exploding which causes the air near the source of the vibration to be compressed or rarefied. This sets up a wave, not unlike the waves in the sea, which then propagates outwards from that source. The wave moves at the speed of sound, which is about 750 miles per hour and the important thing is that these waves are a real phenomenon and have as much possibility for variation as the number of air molecules in say a square inch and that is an enormously large number. These waves are not a string of numbers. If we want to convert a real world phenomenon into a bunch of numbers, then we have to somehow measure the real phenomenon and write down the results.

This is more easily explained by imagining the sea and the waves coming into the beach. If I want to record in a bunch of numbers, the representation of those natural waves, then maybe I could put a yardstick into the sand a little way from the beach and look at it from time to time and write down what I see. But it is really important how often I take the measurement and to what accuracy I read the yardstick. It is pretty obvious that if I only look at the yardstick once a day, that is too infrequent; heck I wouldn’t even catch the tides! If I read it a million times a second, that’s more than fast enough, but I need to write down a lot of numbers. (In digital electronics this is called the SAMPLING RATE.) The accuracy of the measurements is also important and that depends on your purposes and, of course, waves come in all sizes from tiny ripples in almost still water to giant dangerous tsunamis. How much accuracy is enough depends on your purposes.

In attempting to capture the waves of sound, this measurement, frequency,
in other words the sampling rate, needs to be a certain minimum. Some mathematician named Nyquist determined that it needs to be AT LEAST double the frequency of the fastest waves that are present in the incoming stream of waves.

For the purposes of listening to sound, we can consider the highest frequency that some people, (young people), can hear is about 22,000 cycles per second, called 22KHz (kilo hertz). The trouble with Mr. Nyquist’s simple theorem is that it states what is sufficient in data points to EVENTUALLY be able to figure out what the real wave looked like given maybe only a couple of data points per wave. If you want something to be able to recover the original wave shape quickly, the sampling rate should be much higher.

When CDs first came out, I think it was about 1984, we didn’t have the technology to make good enough measuring circuits (called ADCs for analog to digital converters). The trouble is CDs set a standard which persists to this day. CDs use a sampling rate of 44,100 measurements per second, which is only fast enough on paper and not fast enough to avoid major difficulties in using those limited data points later. The measurement accuracy was standardized at 1 part in 65000 (ish) because in 1984 it was a big struggle just to do that!

Hey, what’s wrong with 1 part in 65000? Sounds pretty good? Well it would be except that our hearing is logarithmic. We evolved the ability, I guess because we really needed to, of being able to hear not only very loud sounds but also pretty quiet sounds with almost equal ability. If the very loudest sound that we can measure is measured to that 1 part in 65000 resolution then the average sounds on a CD are probably only being measured to 1 part in 2000 and classical music is even worse, maybe 1 part in 500 leaving room for the crescendo!

These measurements are too coarse and we can hear the errors (called quantization distortion) even if we cannot immediately say why we don’t quite like the result. The sampling rate is too low to accurately capture the treble events.

Sure, CDs were a major improvement to most people who had badly cared for vinyl LPs and a low cost turntable, but it sure had a long way to go to catch up to the accuracy potential of LPs or even of magnetic tape.

Eventually the technology improved and us folk who work in so-called pro-audio, designing recording systems and mixing desks that studios use, always expected that there would soon be a new improved CD, maybe you could call it Super CD or something. We got it wrong. Super CD never happened because the “record stores” said NO. They weren’t going to stock two discs for that white album. They refused. And that was that. The inferior standard for CD remains to this day.

What happened next was initially beyond belief. People wanted to be able to store some songs on portable devices, and while initially there were some portable CD players that you could strap onto your belt, that was a problem because reading back the data on a CD is a tricky mechanical process and too easily disturbed by walking or running. Much better if the data could be stored in “memory chips” like in computers. OK, but storing all the data points recorded by the CD standard, 44,100 per second x 2 (for stereo) x 2 bytes per data point x say 5 minutes for a song, requires about 53 Million Bytes–that’s a large number for computer memory chips. Also, people wanted not to store ten songs, they’d like hey, how about 1000?

What to do?

Mathematicians found ways to throw some of the data away within the framework of an average person not minding too much. This so-called data compression is also needed to get music files downloaded quickly over the Internet. Remember dial-up modems? (You’re lucky if you can say no.)
The really good ones could achieve an internet connection speed of maybe 38,000 Bits per second (A Byte = 8 bits), so to download one song 53 Million Bytes = about three Hours! In a more so so 4 Mb/sec modern connection, it’s about 2 minutes.

So data compression (meaning throw away lots of data that is too subtle for the industry standard listener to complain too much about) is the online norm. Although there a few specialist internet sites that offer full fidelity downloads, the gear you have to receive the downloads is pretty much not your laptop or your Ipad or your phone and you would have to be a geek to get the specialist gear you need.

So in conclusion, digital audio was not good enough at inception and the demands of internet download speed and mass storage on your phone made it even worse.

Could digital audio be good? Sure, modern equipment could re-write those archaic standards.

Will digital audio ever be good? Maybe.

The best hope is that kids get tired of the way their parents listen to music, because kids do that, and demand something new. Once you can store 1000 songs on your Iphone or whatever, storing 10,000 isn’t really that exciting. Maybe better will get rediscovered.

Interestingly, vinyl sales are steadily rising and my local supermarket is selling record players although they probably won’t be that good; but re-discovery of the greatness of music is possible. Many people who hear for the first time a good vinyl LP played on a good turntable into a good hifi system say, “I didn’t realize.” In 2017, we still could do better. Will we?

I hope so.

Postscript: I should have reassured that the recordings made in recording studios are to a much higher standard with faster sampling rates and much higher resolution. So the digital masters are good. The trouble is that they have to be downsized to put on CD or released to the Net. If people demand better, someday, the masters will be able to provide it.

Also, all the stuff that was recorded to magnetic tape (multi-track tape machines) is very fine too, but with one important reminder: Magnetic tape has limited shelf life because it is made by gluing very small particles of iron to the flexible tape.The iron is the magnetic bit, the tape just holds them in place. Glue doesn’t last forever. After about 25 years, you can play an old tape that is in storage – once. Then you have to clean all the iron “dandruff” out of the replay machine and hope you made a good copy.

#####