Digital Audio on VHS – The Technics SV-P100 [video]
youtube.com
youtube.com
IIRC, WBRS, the radio station at Brandeis University, used one such system to record live performances at the studio in two channels of digital audio starting in the late 1980s. I also worked with a studio in Taiwan in 1997 that used analog video cassettes to record 16 channels of digital audio, which were then mixed down to 2 channels and burned to a CD-ROM.
One of the limitations that we were very aware of at the time was being able to extract the audio from the tapes at a later date. No home users had gear that could play back the tapes, and it seemed that very few studios even had access to these specialized pieces of machinery.
In the early 1990s, I also worked in a 24-track London recording studio, which still used very thick analog tape for multitracking, but would output demos or test mixes to the then-emerging DAT standard (https://www.newworldencyclopedia.org/entry/Digital_Audio_Tap...).
The second part is finding a good sample rate that you can easily clock on VHS. That’s where the this stems from although they technically could have gone for a different sample rate at a lower bit depth. 44100 ended up being a good trade off between different requirements.
Anyone saying the 44.1 kHz sampling rate was because of DRM [0] and that they should have used a more sensible rate like 48 kHz is just mistaken. DAT was the first commercial product to have 48 kHz sampling, and that wasn't available until 1987.
[0] The Red Book CD Digital Audio specification obviously existed and was available, otherwise other equipment makers couldn't create compatible players.
Can you elaborate? How could it be a form of DRM? Did you mean that 44100 Hz was incompatible with existing designs using 48000 Hz, or that 44100 Hz created technical challenge for replication due to its smaller frequency margin?
Limiting frequency response at 22 kHz ensures that you get a high quality material, but not the highest quality.
Record labels still have the leverage to extract some pennies in the future by selling a higher quality record and this makes them feel comfortable.
In other words, 20 kHz is good enough.
44.1 tracks tend to be "normalized" at the highest portion of the spectrum trying to squeeze the juice right at the studio.
Tracks mastered for 48+ kHz are usually more relaxed in this regard, they just let the sound flow.
Both sound exceptionally well, but tracks with higher sample rate sound a bit more natural, a bit softer if you prefer. Higher sample rate tends to add a pleasant edge to some kinds of music: live performances, jazz, minimal techno.
If you happen to live in a part of the world where crickets are active in the summer then take a closer look. Their audio spectrum goes way beyond 20 kHz and if you try to listen carefully then you may spot the point where it goes beyond 20 kHz limit.
Not everybody is physically capable to experience this. Personally I can hear a bit of those frequencies, so it's amusing to realize that a little bug can be such a masterful audio source. Nature is beautiful. And this beauty is a selling point when it comes to music.
Could you explain this to me? What does it mean to “normalize something at the highest portion of the spectrum”?
Assume I have a fairly solid background in signal processing theory and understand the fundamentals of music production / mixing / mastering.
Whereas being mastered for 27khz would allow some of those higher frequencies to remain as is and sound closer to the natural sound you'd hear.
I may be wrong but as far as I know, this is because those higher frequencies are still picked up and interpreted by us whether we actually hear them or not.
Personally, i've never really noticed the difference between 44.1kHz and 48kHz, but I do notice the difference between listening to audio on headphones with different cutoffs.
20-20000hz headphones sound noticably less good than 18-22000hz or even 24000hz. It makes even more of a difference than price point usually does. To the point where it's the only thing i really look for when looking for headphones for music.
That’s the Nyquist frequency. The cutoff is wherever you like, but typically at 20kHz.
> …higher frequencies are normalized down around 20-22khz, blending them with those higher frequencies and distorting and changing them slightly.
Can you explain what you mean by “normalized down”? It sounds like you are describing some kind of processing done to music, something I have never heard of before, something different from aliasing.
I want to assume that you’re commenting in good faith, but are you just rewording the original comment? I don’t think the original comment makes sense, and rewording it isn’t going to change anything.
> I may be wrong but as far as I know, this is because those higher frequencies are still picked up and interpreted by us whether we actually hear them or not.
This is wrong, or at least wishful thinking—under ordinary listening conditions with typical program material, you simply can’t tell the difference between music which has those frequencies and music which doesn’t. That what experiments are testing when they test if you can “hear” something.
> Personally, i've never really noticed the difference between 44.1kHz and 48kHz, but I do notice the difference between listening to audio on headphones with different cutoffs.
Different headphones sound different, this is a known fact. There are simply too many differences between different headphones to draw conclusions here.
When I’m making music, I try to listen to it on several headphones before I release it. Each set of headphones I have imparts a unique sound on the music. Same thing applies to loudspeakers.
Let’s say you want to resample from 32kHz to 48kHz with a filter that has N taps. The sample rate ratio is 3:4, so you need 4N coefficients total. Basically, what you have is four N-tap filters, where filter K has a delay of K/4 samples. With higher values for N, you get lower ripple and lower aliasing.
Resampling 44.1kHz to 48kHz, the ratio is 147:160 so you need 160N coefficients. This wasn’t a problem for very long, but it was inconvenient in the 1980s.
I think it's more likely that 44100 was chosen for other reasons--either because it a multiple of the number of active NTSC or PAL lines per second, or because it has a convenient factorization (14x14x15x15), or both.
44.1kHz is a great sample rate for distribution. 48kHz is a little more convenient for creating sounds because it often uses less CPU power to process 48kHz signals (because the transition band is twice as large).
I remember buying a vinyl LP that had been DIGITALLY MASTERED™ and expecting all kinds of magic, but in fact it sounded a bit thin and dead.
The early systems were 14-bit, and that's not quite good enough for smooth audio.
If you're interested in hi-fi, photography, new/classic/weird/historical technology, I'd recommend checking out the rest of Techmoan's catalog.
[1] https://www.youtube.com/channel/UCy0tKL1T7wFoYcxCe0xjN6Q
Was this a common thing, or am I just misremembering this? Maybe just a hack a college station would have done to use commodity hardware/tapes and get 6 hours of recording time per tape?
I also understood that it was a term of our (short) license that we had proof of what was said in air, should anyone make a complaint etc. since all our shows were played live.
I recall one of the problems is the fire hazard of keeping 8 weeks or so of tapes.
https://en.wikipedia.org/wiki/ArVid
(Data storage on VHS)