POLL Lossy Streaming Sound Quality

Spoiler

A was kinda wonky, I'm getting that uncanny valley effect. B was kinda dead. I think both are lossy, just different levels of lossy. EDIT: re-read, it is lossy lol.
ZX2 > Andros, Pietus > PortaPro

 

Spoiler

Yep! I made sure to get the exact same version of the song from the same release of the album up on both streaming platforms just to eliminate as many variables as possible. This one's from No Need to Argue: The Complete Sessions (1994-1995).

 

Spoiler

For completeness, here are the L & R spectra of the difference file:



And the frequency analysis:



Linear frequency scale on the former to highlight the higher frequencies, log scale on the latter because tradition - and because it highlights quite a big difference in the bass content, 20-200 Hz, more than I would have expected just from eyeballing the 2 individual files. Note also the peak centred on ~3k (also more than I would have thought) and the 15-20k peak spotted by @roderickvd .

Now we need to see which files have the greater/lesser magnitudes in the bass & upper-mid areas. But no matter how I fool with Audacity's spectrogram display settings, I can't discern contrasts for differences this small.

 

Is it time to stop with the spoilers yet?

Spoiler

Massively magnifying the vertical scales of the frequency-content plots and comparing side-by-side, it's file B that has more of everything - in both peaks and troughs. As the screenshot in the post above shows, the magnitude of the difference varies with frequency.

Funny, to me A had the subjectively slammier bass but it's B that has the (fractionally) higher bass SPLs. Could this suggest (gasp) that it's not all about frequency content?

OTOH A did seem to me to have more distortion in the bass, and randomly using the analyzer on a few different bass frequencies does indicate 2-3% higher for A than B. Maybe coincidence at these levels though.

 

Honestly the spoilers are kinda silly at this point and I'm just having fun reading along trying to figure out how things correlate to what I'm hearing. This is all great fun albeit above my brain capacity hahaha

 

Definitely not all about frequency content (and related bitrate) but also the *way* the codecs process the audio in and out. What I like about this poll is that more bitrate in my ears did not equate to more better.

 

The funny thing to me was that when taking a bird's eye view of the waveforms it was actually Spotify very high that was pushing the quieter bits upwards, per here (just that on closer inspection it did look like YTM high quality was actually more compressed on an instantaneous level (this is a different song I wanted to look into ergo why it looks very much NOT like the Cranberries):

YouTube Music - top; Spotify - bottom


YouTube Music RMS


Spotify RMS


Plus visualisation from someone else on Discord who's using a program I'm unfamiliar with (think this is Fabfilter Pro Q?) where Spotify here (RED) seems slightly louder overall compared to YouTube Music (GREY) with some more notable discrepancies in the very low bass and above around 5kHz.

 

Spoiler

I listened to the clips on my desktop system in two ways. (1) my Dell Desktop's DAC, a LEPAI LP-2020TI amplifier, B&W 600DM Series 3 speakers; (2) the same desktop, an iFi Audio Uno DAC/amp, Hifiman HE-400se headphones. In both cases, I found clip B (my favorite) to have clearer highs and less muddy bass. My preference was stronger with the the speaker setup. With the headphone setup, the lossy nature of the clips seemed to be emphasized; I still preferred clip B but wondered whether the clip A "sound" would be more pleasant for long-time listening.

 

Spoiler: spoiler

B sounded smoother and the soundstage was better. However, both sound substantially worse than most post-2010 recordings I listen.

Update: Just listened the original from Tidal and it sounds much, much better than both files you posted. Clearer, soundstage deeper and wider, more pleasant overall. However I will accept that just Tidal app sounds better than foobar2000.

 

Given I once upon a time found myself preferring JRMC to foobar2000 (I have not revisited this in a long time and do not plan to go down that rabbithole again, please) I can buy that haha. Would be curious to get your thoughts on how both files perform on something like PlayPcmWin if you happen to be on a Windows machine, but at the end of the day it's just a bit of fun.

Spoiler

both files here are just lossy streams ripped via loopback so of course neither is going to be nearly as nice or refiend as a proper lossless copy

Thanks for the input! I've had folks who've waffled on preferences here, good fun to see the spread in choice. I should maybe try and somehow integrate lossless into it as a baseline, though that might be too easy for most of the folks on here...

 

Spoiler: spoiler

I did not get back to foobar and as you suggested switched to PlayPcmWin. But this time I listened 4 different variants - A, B both with Windows 11 upsampler and Equalizer APO; then Tidal but one with Windows 11 upsampler another with Sox upsampler, both going to Equalizer APO. I was unable to listen A and B with Sox upsampler because it crashed on those files. Listened twice each.
What is the main problem with this recording, which is on all 4 variants: the voice is not perfectly recorded, there is a harshness in the microphone, the distortion is not in the mix but it was there when microphone captured it, it will sound as such even on $100k system.
The second problem is that the background guitar is placed right behind the voice. Maybe she is playing? It's like the voice and guitar are recorded from the same mics but guitar is slightly delayed.
I still preferred B to A, and the main difference in A the guitar behind the voice was muffled, like it was wrapped in a blanket. B was better, the guitar was there.
Tidal both Win 11 and Sox much improved that, the guitar was very clear in both cases. They also improved the voice. It still sounded harsh and unnatural, but I had a clear feeling that I hear just the mic deficiency, the recording captured everything possible in this situation. The most improvement came from that pesky guitar, it separated from the voice a bit and I can concentrate and hear just the guitar.
I think I preferred Sox to Win 11 upsampler, but it was within placebo delta, I am just very happy that I am able to bypass Windows 11 upsampler and have the full control on the chain.

However if you want to hear how the rock music can sound - right now I am listening the live concert 2013 by Dead Can Dance "In Concert". It is full HiRes 24/96 and there are clearly many microphones on stage, and the best onces. When the same sound is captured by several mics it creates some substantial air effect, as if they are playing in a large cathedral. What's funny, part of the percussion is captured with very close mics and then has a direct sound, no air. It is as if part of the percussion was recorded separately in a treated studio and overlapped over. But only part of the percussion - some of it is super airy. It's a funky feeling, as if the mastering engineer is playing with me. And there are many changes in sound from track to track.
I kind of feel I like Sox more than Win 11 but cannot say why, it's probably still placebo. I was also trying turning NOS and OS in the DAC several times. Each time I initially liked the NOS sound but after some time was getting annoyed with it and switched the OS on.