Does Analog Music Really Sound Better Than Digital?

Everyone has that friend who loves records and insists that analog music is superior to digital. But is that true?

Hosted by: Hank Green (he/him)

Correction:
2:06 Although this graphic can be found in many seemingly reputable sources, we’ve ultimately determined that it’s misleading. Digital recordings can produce continuous sound waves with no information lost if the sample rate is sufficiently high. The Nyquist-Shannon theorem states that if the digital sample rate is at least twice the highest frequency of the signal, the output is a perfect copy—not an approximation—of the original analog sound wave. SciShow appreciates the many comments on this and regrets the error.

----------
Support us for $8/month on Patreon and keep SciShow going!
Join our SciShow email list to get the latest news and highlights:
----------
Huge thanks go to the following Patreon supporters for helping us keep SciShow free for everyone forever: Odditeas , Garrett Galloway, DrakoEsper , Kenny Wilson, J. Copen, Friso, Lyndsay Brown, Jeremy Mattern, Jaap Westera, Rizwan Kassim, Harrison Mills, Christoph Schwanke, Jeffrey Mckishen, Eric Jensen, Chris Mackey, Adam Brainard, Ash, You too can be a nice person, Piya Shedden, charles george, Alex Hackman, Kevin Knupp, Chris Peters, Kevin Bealer, Jason A Saslow
----------
Looking for SciShow elsewhere on the internet?

#SciShow #science #education #learning #complexly
----------
Sources:

[sgtproper]

I'm an audio engineer and in my opinion, in modern music at least, the most difference that you can make to being able to enjoy your music fully is in the last step of the chain - aka the DAC, your speakers and any processing device you may have in between. It's astonishing how different different speaker, tweeter and woofer designs, materials, and other adjustment sound with the same track. Also don't be afraid to use an EQ in your room, or measure it and use adaptive EQ if you want to get the best possible experience. For me, vinyl is more about the ritual of listening and not skipping; and I like the "craftsmanship" aspect of it

[silverXnoise]

I’m an audio engineer, and the question of what sounds better is individually subjective. As to whether analog music is more accurate, linear, or truer fidelity? No, in most cases it is not. Compressed digital formats may be worse, but it’s nothing inherent with their being digital, and analog media has its own long list of caveats.

[BillySugger]

2:05 No Hank, digital sound is _not_ staircase shaped. Experienced electronics engineer here. There is no approximation and there is no information lost. It’s true that the sound is sampled, but those samples technically have no width in time. They are discrete impulses, not staircase steps. And when those samples are sufficiently frequent to capture all the frequencies contained in the original sound, (just over two samples for the period of the highest frequency in the signal), and are played back in a competently designed player with a reconstruction filter that meets the Nyqvist criteria, the result is a _perfect_ analogue signal reproducing the original sound. There are no steps in the resulting signal at all; it’s technically _perfect!_ This is testable; just look at the playback signal on a good oscilloscope, there are no steps. So please don’t perpetuate this terrible and unfounded myth about digital sampling being stepped or staircased - it just isn’t true. Technically speaking, it is absolutely superior in fidelity to every analogue recording medium ever implemented, ever. Period!

[frank_lotion_]

Real audiophiles know that since the sound wave travels through the air, you have to pump in pristine air direct from the Italian Alps into your listening room to help mitigate random molecular compression (RMC) that you get from sub-optimal air, which tends to muddy up those higher frequencies

[albertoporras]

Sorry Hank, but you're perpetuating a widespread but fundamental misunderstanding of digital recording. The sample rate defines the bandwidth of the signal that can be captured, the number of bits defines the signal to noise ratio present in the recording. So by increasing the sample rate you can make the bandwidth arbitrarily high, and by increasing bit depth you can make the noise floor arbitrarily low. No analogue recording technology has infinitely high bandwidth or infinitely low noise. The bandwidth and noise floor of good quality digital recording standards and equipment exceed the the performance of analogue recordings by some margin, and they are more "faithfull" to the original sound. Of course, incompetent use of the equipment or techniques can produce bad digital recordings, just as they can produce bad analogue recordings. People who claim that analogue is better are either just exhibiting a preference for the sound of the added analogue recording artifacts and distortions, or have allowed the spurious arguments of the "analogue is better" movement to unduly influence their subjective perceptions.

[jeanmichel]

No sampling does NOT create loss of information. The samples are NOT an approximation of the original signal. According to Shannon's theorem after filtering there is only one curve that can relate the samples: the exact original signal. Not an approximate signal.

[Conor_]

I think Dankpods put it best

"It's not about sound quality, it's about VIBE quality."

[SquintyGears]

Don't use that visualization for sample to sound conversion. That's not what it does at all. Mathematically we know sound is a sinwave so it fits a sine to the discrete datapoints. It is not a derivative "area under the graph". That would totally leave a ton of noise on the signal. Sine fitting at regular sample frequency does not.

That visual is used so much and is so wrong it reinforces People's perception that digital storage of analog signals is bad. I really would have expected scishow not to fall into this pitfall because it's quite literally the most common error made in this music format discussion... This deserves a reupload...

[pedroff_]

The 44 kHz or whereabouts sampling rate is not coincidentally a bit over twice the theoretical 20 kHz upper limit on frequencies we humans can hear. Through clever mathematical shenanigans, one can use discrete samples to recreate perfectly all frequencies below half the sampling rate. As the world isn't perfect, the actual sampling rate is slightly higher than that to give leeway for the flawed methods we use.
For anyone wanting to look deeper into this, it has to do with Fourier Transforms and with the Nyquist Limit

EDIT: I'm a moron and mistook double for half. Twice.

[HippoOnABicycle]

For me, vinyl records are like candles: It's not about them serving a practical purpose, it's about the coziness. A LED light does an objectively better job at lighting a room than a candle does, and is also better for the environment, but turning on the lights just isn't as cozy as lighting a candle.

[zAAmpie]

I think all engineers with some background in signal theory cringed a bit at that explanation of signal discretization. Spend enough time with Fourier and Nyquist/Shannon and you know that sampling can produce a perfect reproduction without aliasing up to a given frequency. As for what people find more pleasing, I think that's subjective. Same argument why people like tube amplifiers etc. They are provably worse and introduce distortion, but people still find that distortion pleasing, even if it's not the artist's intent.

[MonguzTea]

That is not how digital audio works. There is no stairstep pattern, samples describe 2 points of a sine wave. Every sound can be described with sine waves.

[nathan]

The best explanation of digital audio and why it's as perfect a representation of audio as you'll ever get can be found if you search YouTube for "Digital Show & Tell" given by Monty Montgomery. He shows how there ARE *NO* STAIR STEPS in digital audio and that even a 20kHz waveform can be perfectly reconstructed at even a 44.1kHz CD sample rate. He also explains why a 24 bit sample depth is pointless for consumers.

[andrewlecouteurbisson]

As others have pointed out, a signal with a specific maximum frequency can be sampled in time, without loss of fidelity. Its actually the sampling of the signal level into discrete voltage steps that introduces noise. For 16-bit (CD) audio the signal to noise ratio is 96dB which might be audible under rather contrived conditions. 24-bit audio theoretically gives you 146dB which exceeds that capability of human hearing. Techniques like noise shaping can substantially improve these figures. Suffice to say that under normal conditions you will not detect the noise in either case.

[Blackmark]

"But there's gonna be a tiny amount of information lost no matter what."
That just CANNOT be true. The digital sampling can be much higher than air's ability to reflect the difference between digital and analog. It boils down to how air vibrates and our ears pick it up, not the technology.

[Tuings]

As an audio engineer. I would say. If the original recording has an analogue workflow. Then i
Understand the argument for Vinyl(even if i dont partake myself). If there is a digital conversation somewhere in the chain. Then a High quality FLAC/WAV will be of better quality than the vinyl. This is assuming that your Dac is of sufficient quality.

[yasunakaikumi]

This video pretty much didn't show the Nyquist-Shannon theorem at all when mentioned about the part about there's a lost, which incorrect about PCM, and also saying steps on representing digital audio sample? it's 2024, and there's a video about it from XiphOrg channel that you guys could have watch before making this video... digital audio is easy to research why the heck you guys skipped the very fundamental.

[LePedant]

This is why Open Source projects like FLAC are so important. We get lossless compression, that is free for everyone to use.

Not only does it allow people to rip their own CD's and retain the full 16bit 44.1KHz sampling rate. It allows CD quality audio to be streamed with relatively low data.

[goaty.]

technically speaking, vinyls are the one that need to compromise on quality, since the little needle can only change direction so fast, meaning that vinyl needs to be mastered to account for this, this can can actually make vinyl sound different

[bartpappas]

How can you create a video on this topic and not address the Nyquist-Shannon Sampling Theorem. Basically all the information you gave about the digital waveforms being "discreet" is misleading at best.

The data stored in digital media is discreet, much like 2 set of linear coordinates. There is a theorem of geometry that 2 sets of linear coordinates uniquely define the infinite number of points of a line.

The same is true with sinusoidal waveforms or any linear combination of them. The Nyquist-Shannon Sampling theorem similarly defines a waveform by the minimal number of points needed to define that form completely, NOT just approximately.