There is an article over at Digital Trends entitled, “Everything You Need To Know About High-Resolution Audio, The Future Of Music” written by Caleb Denison that takes a look at the emerging world of high-resolution audio. There’s a large image of three Pono players, a chart of fidelity levels and a picture of a set of Sennheiser headphones.
The author lays out the article with a little backgrounder followed by a series of questions pertaining to high-resolution audio. His outline is good but the details in each section stray pretty far from the facts.
I have to wonder whether a journalist with his background as a “professional musician, amateur chef and A/V electronics guru” actually qualifies him to explain the inner working of PCM audio and the whole subject matter at hand. Perhaps it’s just the definitive nature of his title that tweaked me this morning.
The first section “What does the term ‘high-res audio’ mean?” might seem simple but as I’ve written previously, it’s proving very difficult to nail down. The term high-resolution audio has been around for at least 15 years and is not a new term to those paying attention to this area. It’s true that HD-Audio co-existed with HiRes Audio for most of that time, but now it seems that consensus is tilting towards high-resolution audio. I wrote about the advantages of HD-Audio…but I lost that battle.
The second paragraph begins, “While high-res audio is really a very broad term that could apply to any kind of high-quality sound, it has become popular to use it to refer specifically to high-quality digital music files.” High-resolution audio can ONLY apply to digital files that have a minimum specification. When we consider high-quality analog sound the term “resolution” doesn’t enter into the picture. Analog tapes or vinyl LPs don’t have “resolution” because they store information without sampling and dividing continuously variable electronic signals into discrete parts. These formats certainly have fidelity specifications but it is incorrect to consider them “high-res audio”.
The author breezes through a brief history of audio quality (vinyl LPs still rule the top of the fidelity hill), before proceeding to a section on “How do you measure music quality?” I think he’s strayed away from his central subject and migrated over to music criticism. He would be better advised to title this section, “How do you measure the fidelity of recorded music?” Otherwise we’d be arguing about whether Bach’s “Italian Concerto” is “better” than Chopin’s “Ballade in gm”…which is a useless exercise.
This is followed by an explanation of “sampling rate”. Here’s the paragraph:
“Sampling rate is the number of times a sample (a sonic picture, if you will) is taken of an audio signal per second. The more times you sample an audio signal, the more detail you end up with. Sampling an audio signal is like shooting a video of a fast-moving object. The higher the frame rate (sampling rate) the more depth and detail you can capture and the smoother the end product is going to be. Say you’re shooting a video of a cheetah running across the savannah. At 24 frames per second, you will still be able to tell it’s a cheetah, but the details are lost in a blur. At 1000 frames per second, though, you might be able to see all of the cheetah’s whiskers, count its spots and notice its tail is slightly kinked toward the end. Again, it’s all about heightened detail.”
This explanation is not even close to being accurate. It sounds like a line from a sales brochure for the new 384 kHz DACs that are starting to come on the market. Increasing the sampling rate doesn’t provide you with “more or heightened detail” unless you define the addition of higher frequencies to mean “more detail”. The rest of the paragraph switches over to a comparison to film/video with its frames per second capture scheme. Which is where the author steps completely off the track.
Sampling audio is not analogous to using individual frames to capture motion on film. You don’t get “more depth and detail” at 1000 fps vs. 24 fps. You get a lot less strobing, the ability to do great slow motion and will need a giant reel of film! Both system are using sampling but the output is fundamentally different.
The sample rate of a PCM digital system functions according to the rules laid down by the Shannon-Nyquist theorem. We need a sample rate that is at least and at most twice the frequency of the highest component in the signal to be converted. That’s the reality. So if we present a band limited 20 Hz – 20kHz signal to our ADC, we have to have a sampling rate of 40 kHz…and no more. It’s better for the filtering if we push that out a few thousand Hz to around 44.1 kHz. And, of course, I’m an advocate of going to the next multiple out and using 96 kHz as the high-resolution sample rate.
I’ll continue tomorrow with part II.
My point in writing today’s post is less about rehashing the ideas behind PCM audio encoding and more about why readers are being confused by inaccurate articles…especially ones that tell this is everything I need to know.