What specs would digital music have to have to be indistinguishable from analog?

Factual Questions
41

Here is an analogy.

In 2020, Apple rolled out the Retina Display. It was so named because the resolution was comparable to what the human retina can resolve at typical viewing distance.

Suppose we wanted to test that claim. We would take a high-resolution photo of a painting and digitize it at least the resolution and color discernment of the retina display. Then we would hang the original painting next to its image displayed on the retina display. Now we take an analog photo of both of them at the average resolution of human vision. We zoom in and see if we can tell the difference. We are just interested in whether the Retina Display really shows an image with a resolution, dynamic range, color discernment that is indistinguishable from real-world viewing based on the parameters of an average human eye.

The difference between this and my experiment is that you must assess the final product using a method that requires recapturing an image, which is similar to the process used to capture the image to display it on the retina display to start with. However, with audio, you can capture the signal before it becomes audible; there is no need to mic a speaker and repeat the audio capture process.

42

I mean I bet one could digitally imitate your turntable so you could not tell them apart. Don’t know that anyone has tried precisely what you are describing, though.

43

I’m not an audiophile… but-
But this is the big problem - every device, especially analog, has a response curve. Microphone, amp, speakers, storage medium… Other than listening to the original sound at the source, there is no guarantee that an analog recording or playback is any more or less faithful than digital.

One thing I recall reading was that hearing deteriorates with age, and some young people - those who haven’t abused their headphones - can hear amazingly high, 25kHz to 30kHz. By our twenties, we’re back to nominal human frequencies. (I recall a friend was trying this out with his son, a young teenager, and the boy could definitely hear 22kHz.)

What I read - CD’s allegedly got the bad rap because the earlier releases, if not DDD, tended to be rushed to release cheaply, simply using the original vinyl mastering tapes as input. Because the needles on cheap stereos tended to be less responsive at higher frequencies, these (analog?) masters had the higher frequencies boosted. As a result, the CD’s faithfully reproduced a more “tinny” sound versus the “deep, rich” tone of vinyl. Later releases corrected this issue.

The second CD I bought was a DDD recording of Beethoven piano music. On my crappy Radio Shack stereo, you could hear the pedals squeak in the quieter parts as the piano was playing. My friend was listening to this and it convinced him to go out and buy a CD player himself, even though they were $800 at Radio Shack at the time… IIRC, Glenn was humming or moaning along with the variations on his CDs.

BTW, the first CD I bought was Born in the USA.

44

The trouble with using wave comparison for audio reproduction comparison is that doesn’t measure what matters here.

Our ears don’t really respond to the wave amplitude in time. They are for the most part actively controlled frequency analysers. They don’t see a lot in the way of phase anomalies. The ear-brain needs to be considered as a unit. There is active control of the hair bundles in response to sound that both modifies the response and informs the brain’s knowledge of the sound.

There is some ability to detect absolute time offsets for frequencies below about 1kHz. This likely fits with the distance between our ears and allows both amplitude and time difference to inform stereo location. Changes to frequency response with angle (the HRTF, head related transfer function) inform this as well.

But the system is insensitive to a huge set of phase anomalies. Comparing waveforms is ruined by even tiny phase anomalies. (Then there the stupid arguments about absolute phase. Something is almost always inaudible outside of extreme events. )

So you need to be very clear that your comparison is not claiming terrible results for things that just don’t matter and yet be relatively insensitive to things that matter greatly.

One area digital streams were shown wanting is the propensity to create distortion products that were not harmonically related to the source signal.
This was an interesting surprise.
In the past distortion was neatly measured by feeding a system a signal at a fixed frequency and looking for all the harmonics generated by the chain. Add up the energy in all the harmonics and you have total harmonic distortion, THD. It didn’t take long to realise that second harmonics were almost inaudible and if they were audible they added a nice warmth to the sound. Third harmonics in small doses could add a pleasant richness. Given these are the octave and a fifth above the octave this is perhaps not a surprise. It is a staple of studio production adding just the right amount of harmonics. Doing so in an amplitude related manner even more so.

Higher harmonics are less desirable and a metric that weights the harmonic by its order has been used to better relate measurements to audibility.

Then digital came along and could generate spurious products at heterodyne frequencies relative to the sample clock. These aren’t harmonics and turned out to be very objectionable artefacts audible at levels well below what we were seeing with harmonic distortion.

So we could measure systems to the best of our ability and miss things that matter.

In the end there is no go/nogo metric. Measurements must be informed by our understanding of how the ear brain system functions.

The interesting studies were done during the development of perceptual compression systems. So MP3 and its successors. They performed many double blind trials to determine the limits of perception to tune the compression schemes. A criticism of the work was that it most used classical music. But the results have been very interesting. They could identify a large amount of signal content that was simply inaudible. There are interesting tells, where if you know what to listen to, are give aways. An identifiable splashy edge to cymbals and high hats. And the easy one, loss of the faint ambient fade out of the sound, replaced by a sudden cut to silence.

Better compression systems at reasonable bit rates are impossible to pick.

Then you can get into other weirdness like the precedence (aka Hass) effect.

For anyone that cares about the reproduced sound, they know that the single biggest factor is the room. The sound field you hear is dominated by the room, and effort spent getting that right (if there is a perfect right) pays dividends way above effort spent on the equipment. Stereophonic sound is so compromised relative to live sound that there is no point worrying past a certain point. It is intrinsically impossible to reproduce the original sound field. Getting to an acceptable compromise that is a good experience in and of itself is the best we can hope for.

45

To add one useful trick that can be used to evaluate audio devices.

If you have a reproducing device (this is usually done with amplifiers) you feed your original signal into the device, and then attenuate it back to the level of the input signal. Then feed both the original signal and the new signal into a double blind comparing switch, and get humans to try to detect which is which in a statistically meaningful manner.

This could be used to evaluate a ADC/DAC pair. There will be a very very tiny delay down the digital path, but should be imperceptible. However it is little tells like this that can confound such tests.

Bob Carver had an open challenge to people who claimed to be able to tell the difference between power amplifiers and ascribed magical properties to the high end stuff. He would take one of his own amplifier designs, and add nothing more than a simple frequency response shaping circuit to it that exactly matched any anomalous response inherent in the other amplifier. (This needs to be done in the context of expected speaker load as some amplifiers are so poorly designed that this affects their frequency response.) The challenge was that nobody could then tell the difference between them. The implication being that all of the other claims about magical performance was inaudible and that those high end amplifiers with eye watering prices were nothing more than snake oil. He was pretty much right.

46

The defensible position is that to some, vinyl subjectively “sounds better”, and the technical justification that’s frequently used is that the inherent limitations of vinyl require the music to be equalized differently. Indeed when CDs first appeared and I got my first CD player, I loved the sound but my subjective impression was that high frequencies tended to be more pronounced than on vinyl, so it sounded comparatively harsher. Again, this is extremely subjective and some of it, as already mentioned, may have been due to sloppy CD mastering. I do have a turntable and small vinyl collection but it’s mostly for nostalgia.

47

And the response to that is always the same: digitally record the record output. If you like all the extra noise and the special EQ and stuff, then leave it in the digital version.

48

An even simpler response is to buy an equalizer, or an amp that has one built in.

But for us Old Farts none of that is a substitute for the genuine sound of the good ol’ days. :wink:

49

This was a common complaint. The standard answer was that in their haste to put out CDs the record companies simply digitised the master tape used to cut the vinyl master.
These were often created with an ear for compensating some of the inherent problems in vinyl. This is somewhat reasonable. Depending on the type of music and production values.

I well remember taking a new CD of The Police, Synchronicity into my favourite HiFi store and we played it through an extremely high end system. It sounded terrible. Really strident and ear bleeding highs. But that doesn’t represent the recording. Pull down a modern mastering and it sounds fine.

There was all manner of finger pointing, and an entire industry of snake oil appeared. One that has never fully gone away. Green pens anyone? Or magical stones. Or insane interconnects.

50

There has to be an upper limit to the amount of ‘detail’ that any original sound can contain, because sound is transmitted by variations in air pressure and air is made of discrete physical particles.

So whereas recording sound at too low a sample rate risks the possibility that something interesting happened in between samples, that cannot be interpolated from the adjacent samples, there has to be a point where that gap is too small for anything interesting to fall down the gap - if your sample rate is sufficiently high, then any theoretically-missed pressure change events would have to be happening so fast (and therefore with such energy) that would be turning the air into plasma.

51

I don’t believe it’s possible to get proper audio unless your cables are made from either oxygen-free copper or oxygen-enriched copper (I forget which) and of course the actual connectors must be pure gold. :smiley:

I will say, though, that one time a friend gave me a whole bunch of A/V cables and other stuff, among which was an RCA type cable intended for a subwoofer. In the course of whatever playing around I was doing, I tried using it to feed composite video to a video projector, and the results were terrible. Clearly, this cable in some way was optimized for low frequencies and video being the opposite extreme, it just couldn’t handle it. I’ve also experienced comm errors with inferior HDMI cables but not with good quality ones.

But yeah, in general, almost all beliefs about interconnects are bullshit.

52

And funnily enough, this is exactly what Nyquist and Shannon provided us with.
If you sample at greater than double the bandwidth you are guaranteed that nothing is missing. Simple mathematical fact.
The amplitude of the pressure change is reflected in the signal to noise aka dynamic range. That defines the bit depth needed.

Eventually you get to Shannon’s pivotal work on signaling in a noisy channel.

Despite the handwringing and special pleading of the audiofools, this exactly covers the ground. One still hears arguments about stair step waveforms and stuff that falls between the cracks, or claims that phase information is lost. All wrong.

Once you have defined your channel characteristics it is just implementation.

53

I can only provide a friend of a friend story here, but back in the 90s I had a friend who knew a guy who’s day job was doing field recordings for movies, sound libraries and stuff like that. He initially had a portable rig that did (I assume) 44.1KHz. He recorded birds with a stereo mic and came home and played it for his cat. His cat didn’t care.

He later got a rig that would record at a higher rate (I assume 96KHz). He did a similar recording with the same mic, and came home and played it for the cat. The cat was instantly trying to find the birds this time. It sounded the same to the guy who made both recordings. Since the cat can hear higher frequencies than we do, Nyquist seems to pass this test with flying colors.

I can offer a first hand account of this, though. Mastering counts for a lot. I was the person in my bands managing our vinyl releases in the age of digital. I’d send the tracks off to the mastering guy with instructions on the order and how to gap the tracks, and he’d send back two stereo tracks (one for side A and side B), that I’d send to the record press along with sleeve/label art, etc.

I’ve listened to those tracks digitally before sending them off to be converted to records. If I didn’t at least check that the song order was correct before doing so, I’d be an idiot. If I listened to them for sound quality, I wouldn’t be much better. They seem to be kind of heavy on the high end of the spectrum, if not piercing in the way that CDs that lacked a proper remastering sounded. Conversely, those same stereo tracks sound great on vinyl.

And we just released the digital tracks as they were at final mix down without mastering. They sounded fine, so no level matching or anything. From what I understand these days, you can just send your side A and side B digital tracks in order and volume matched to the pressing plant. If anything’s necessary to adjust, they’ll remaster it for you. If I was still in the business of buying 500 copies of the same record, I’d probably still send it off to one of the mastering guys I know. Heck, I’m still considering sending the most recent tracks I’ve recorded off to one, but telling him to not worry about vinyl. Mastering is a relatively inexpensive step in the recording process, but they can fill things out nicely, Sometimes a second set of ears manning a different set of compressors is worth far more than $30 or so a track.

54

I remember a thing that was going around where the expert would tune your audio system by sending you a pack of (enormously expensive) foil stickers that you would apply to your equipment, then there would be some sort of phone consultation where they would ‘cure’ your audio problems by playing various tones, over the phone to you.
And people believed this worked, apparently.

Here’s a variant of it: Quantum Sticker - Telos Audio Design

Every Quantum X2 stickers contains patented formula infra-red powder (which undergoes 36 hours of cryogenic process at -196oC), patented formula negative ion powder, titanium dioxide photocatalyst materials, silica wafer insulation coating material, pentoxide antistatic agent material, calcium oxide CAO desiccant material, zironia filter material and alumina structural ceramic material.

Infra-red powder and negative ion powder eh?

55

Mastering is one of the interesting things that has grown organically since the advent of digital. Used to be that your master tape was the final mix down and the mixing engineer knew what was needed and wanted. The guy driving the cutting lathe was the final arbiter of what happened, so there was arguably a final human touch before the stylus hit the wax.

There were stories of CD plants that would do all manner of indignities to what they were given. Including recording a digital master to tape and back again to digital. This isn’t as insane as it sounds but certainly messed with intent.

Somehow the role of mastering engineer became something in its own right. Which is remarkable. Sadly it is in their hands that the loudness wars are sometimes fought. But they do what the customer asks for, so it isn’t their fault.

56

best way to nullify any argument is " … whatever error is produced in the digital domain, the error added by your speakers (or headphones) will be 100s or 1000s of times more severe".

so … somewhat in a “black-and-white” nutshell:

Digital Domain: error = 1 (order of magnitude)
amplifying the converted, analogue electric signal: error = 10 to 100 that
converting electric energy into kinetic (via a “speaker/headphone”) error is 100 to 1000 that

case in point Dig.Domain is the least of your problem.

which makes sense: more compressed music is where “everything is louder than everything else” … and louder music is being perceived ase “more gooder” by the brain

57

This brings up a real issue in terminology. Compression of dynamic range is orthogonal to compression for data size reduction. There is quite literally zero in common.

Yet there are legions of commentators that still confuse the two with no idea. Heck, I remember hearing an entire radio broadcast with commentators missing the difference, yet there should have been people in the room who knew better.

A reasonable bit rate mp3 is under most circumstances indistinguishable from the original. Down at 128kb/s you can start to hear artifacts, but even here, it is better than most people are used to hearing any music. 256kb/s and you are well on the way to needing controlled conditions. Above 300kb/s and you will really struggle.

But, and this is key, there is essentially no dynamic range compression occurring. Not in the manner that has wrecked contemporary popular music - with the loudness wars.

To be fair - “everything louder than everything else” is a couple (closer to three) decades older than the loudness wars. (It pre-dates Motörhead.) The loudness wars are about filling every second and the entire frequency spectrum with full level sound. That and things like the use of gated compressors to make the sound “pump”. Add autotune and you have the fresh hell of popular music of today.

Everything louder than everything else, was an acknowledgement that each band member always wanted his instrument turned up louder than that of the other band members. A conversation mixing engineers dealt with every day.

58

Wowwww. I suppose that’s one way to add tape compression. A studio would probably charge you for rental of the tape for the same service.

Even today the pressing plant will fix stuff that’s obviously not going to work in a vinyl groove. For example, I’ve got a track I’m working on with two different kick drums panned hard left and right. If I sent it off to get pressed (I don’t plan to), they’d move at least the low end frequencies of the kicks to the center.

Yeah, in the era of CDs, they kind of needed to make the position of mastering more important. Digital allowed the track to be louder, since you weren’t worried about bouncing the needle out of the groove. At the same time you had to be a lot more worried about peaks, because digital doesn’t deal well with going into the red. Analog tape would just get you a little more distortion if you peaked hard (and that could even sound pleasant), digital would generally give you awful noise.

Hehe, I’ve actually asked to have my instrument turned down on two recordings. I think since I was the guy footing the bill and watching over the mixing, they wanted to please me (or avoid having me ask for more me in the mix). I’m so proud of myself.

59

It wasn’t a brain thing, it’s more like how everybody liked disco in the 70’s, and now they don’t. Popular taste is a function of popular availability. I suspect this is also why people who grew up around vinyl believe it has a superior sound. It’s just what they’re accustomed to enjoying.

That study was in 2009, I’d be very interested to see it run again now that low-bitrate mp3s haven’t been a dominant format for quite some time now.

60

Has anyone mentioned, too, that sound waves in air attenuate with distance, particularly the high frequencies - so the positions of listener, speakers, microphones used for recording and digitizing, etc. - all matter. The actual sound waves change depnding on where you sit. When you add in the environment, and reflective surfaces, this becomes even more important. Studios may have sound absorbing walls, but your basement rec room doesn’t. Fancy concert halls have been designed for acoustics and many come with large moveable panels that can aborb noise.