Comparing Tracks: Vinyl, YouTube & Digital Audio 16May22

On the Nest Theater and on my headphone setups I use three primary schemes to play music: streaming YouTube music videos, playing digital .flac files, and spinning vinyl LPs. Most of my HiFi setups are geared to replay digital material; only the Nest Theater has a turntable for phono LP playback. In this entry I am going to describe the process by which I directly and objectively measured and compared the "fidelity" delivered by each of these schemes as well as provide subjective listening observations. First a description of each:

Streaming YouTube Music Videos - 1st off, I am only concerned with the audio track and its fidelity within YouTube videos. YouTube uses the Opus codec to transform the audio portion of any video uploaded to it. Opus is very good, with very low latency ~ YouTube maxes out at 192kbps (44,1Khz/24bit) for audio in an HD720p or 1080p .m4a video. YouTube typically carries multiple formats behind each video (or audio only) and offers one or another format (with differing fidelity) depending on the capability of the target playback device ~ the best audio (highest bitrate) comes with the best quality video download (so know to choose best video when you can), though an audio only version is often present @ 160kbps with 48KHz/16bit encoding. A Linux program named youtube-dl allows one to get information on formats available and to download specific versions of a video.

Streaming of lower bitrate material, often using the MP3 or AAC codec as used on Pandora, Spotify and Soundcloud is not included in this comparison as I don't use them, and they've been shown to have lower fidelity many times. High bit rate streams (like 320Kbps) are much better and largely audibly indistinguishable from FLAC by humans.

 

Playing FLAC Files -  FLAC is a lossless codec often used for audio. Just as with Opus, FLAC can have a variety of levels of "quality" depending on the original material. I have FLAC files with PCM (pulse code modulation) material of 192KHz/32bit, 96Khz/24bit and 44.1KHz/16bit (CD quality) but even "higher quality" exists. I also have DSF/SACD, WAV and MP3 files but they are less relevant for this comparison... WAVs can always be made into equal quality FLAC files, DSF files are pretty rare and expensive, and MP3, a lossy scheme, will almost certainly be lower fidelity than CD or better FLAC of the same material.

Spinning Vinyl LPs - LPs have been around for a long time; they were first released in 1948 by Columbia Records. Sound is produced from them by a needle tracking in and vibrating from grooves as they are spun passed it at 33 1/3 or 45rpm. The grooves are recorded from master tapes and "pressed" into the surface of the vinyl. In recent years, vinyl LP sales have re-surged and now eclipse CD sales (but of course streaming is outstripping all other forms of music playback, as to volume "sold"). Vinyl has physical limits which make its maximum dynamic range, frequency response, channel separation, and other attributes "less" than that available from digital renderings, even CDs. Yet, many aficionados, prefer the "warmth" available from vinyl playback.

Mastering - Today it is rare for the master for a digital download, a CD, or an LP to differ, though occasionally it is true, but many think it is more widely true ~ all these types of recording/playback schemes are built from the same masters these days. This means the same master fidelity and attributes are available to creating each of these three types of playback sources. In fact, the "quality" of sound by any well delivered playback is most dependent on the initial recording and production - not the final delivery scheme (at least among these three I am comparing, and all other things being equal). Nonetheless, there are differences we hear, feel and measure between these three... the vinylphiles and the digiphiles are already lining up.

Constraints...  Physical constraints limit LPs in a number of ways: the best dynamic range is ~70dB whereas CD can exceed 90dB. Dust and/scratches on an LP's surface create noise, often audible. Even the friction of dragging the needle through the groove ultimately destroys the groove. Mechanical turntables themselves might introduce rumble or speed variations which distort their signal out. Channel separation for vinyl is about 30dB (vs 90dB for CD) and that narrow band often limits recording engineers to center bass between the channels so as to assure needle tracking. It is worth noting that the "loudness war" was being waged largely on CDs and in other digital schemes of music reproduction... there's not enough range for dynamic compression on most vinyl tracks.

Objective Measurements - Ok, here is some secret sauce. I compared the "equivalent" tracks from a YouTube Music video, a digital download, and a vinyl track by recording each in decently high resolution (96KHz/24bit) on a MacBook Pro using Audacity.

Connections & Interface Gear...

• For YouTube I extracted the 2CH PCM digital audio from Roku HDMI out (up to 192KHz/24bit) via a J-Tech Audio Extractor (which supports audio pass-through). Fiber optic SPDIF out was converted to USB for input to the Mac/Audacity by HiFiMe converter for recording.
  
• For digital download I just needed to load whatever 2CH PCM FLAC file was "equivalent" into Audacity directly.
• For Vinyl I took RCA line out directly from my Tube Box DS2 phono pre-amp into a Musou A>D sending 48KHz/16bit PCM (again through HiFiMe converter) via USB to the Mac;  Audacity did the recording.

The J-Tech Audio Extractor ~$35 (pass through audio ~ expect 96Khz/20bit or better)...

The Musou A>D Converter ~$17 (48Khz/16bit)...

The HiFiMe SPDIF fiber to USB converter ~$25 (pass through up to 96KHz/24bit)...

Material - Getting "equivalent" tracks from different sources will allow a lot of comparative analytics ~ will we really objectively see which is better: vinyl or CD, and how does YouTube hold up against these? I have acquired recent LPs of some favorite singers/songwriters that I'll use in this analysis, including Phoebe Bridgers' Punisher, Jade Bird's Different Kinds of Light, and Molly Tuttle's When You Are Ready. I have these same albums in 96KHz/24bit PCM or 44.1KHz/16bit downloads. And, all of these albums' tracks are also found on YouTube Music as pure audio (they will come down as Opus 160kbps and 48KHz). This later audio only track listening on YouTube is not exactly my practice, which usually hunts for live versions of artists' materials, but I need "equivalent" tracks for this analysis. The Opus/160Kbps/48KHz is very close to the m4a/192Kbps/44.1KHz audio I typically listen to in music videos on YT. Interesting side notes... the Jade Bird DKOL LP is a Limited Edition red 2 disc 45rpm set; Phoebe/Dead Oceans included full 96/24 digital download license with the vinyl while the others did not include digital access at all.

Measurements - What analytics and comparative measures make the most sense to determine which format might sound best? Dynamic range (the difference between the loudest and quietest bits), signal/noise ratio (comparison of what you want and noise present even when there's nothing playing), and noise floor are some of the key things I will measure to compare YT with CD with Vinyl. RMS (root mean square) is a simple measure of how loud, on average, the entire track is, in decibels. Noise Floor is a measure of all the unwanted stuff, in decibels. Peak is the loudest point (highest amplitude) on the track, in decibels. Contrast is my attempt at Signal/Noise ratio - done in Audacity by comparing two portions of the track - one before or after with "just" noise and another with most of the rest of the track, in decibels. Dynamic Range is measured in LUFSi (Loudness Units relative to Full Scale - Integrated), a fairly recent but widely used standard measure in TV, movies and music... it accounts for human "perception" of sound loudness by applying ISO standard equal loudness contours. Similar results were shown in other DR measures, with nuanced differences, so I chose just one.

I will make some general observations from the above objective measurements...

• Vinyl always has a higher noise floor and almost always a worse S/N ratio.
• I don't put much stock in the RMS as a comparison as these sources traveled through different audio chains to some extent before being measured - I could not guarantee the gain was set equally between playbacks.
• The dynamic range varied more than I expected, and not in a consistent way, especially with vinyl having a lower theoretical range. Oddly, the digital downloads often displayed the lowest dynamic range or loudest track and the vinyl DR was fine, sometimes even good.
• I did discover that YouTube does adjust for loudness ~ they may receive files that are -6dB but normalize them to -14dB or so before streaming.

While I double checked my measurements and methods perhaps the loudness anomalies have something to do with the fact that vinyl and YT streams are extractions that were recorded and the digital download was just an original FLAC file being loaded directly into Audacity. More research needed.

Subjective Listening Observations - Now this stuff is personal, no matter what any audiophile might tell you. The sound is so dependent on the system rendering the material ~ but in my case in this analysis I'll use the same system, the Nest Theater, set up equivalently for each form of playback "source." I disabled the subwoofer of the NestT for this listening analysis. I will also run two passes through the digital material - one with full Dirac equalization applied and one without any equalization. I'll do this because the LPs are not at all equalized before hitting the DRC BLHs + super-tweeters - so I wanna hear all sources rendered equally through the speakers with all their warts. I have some instant observations and I sat with some friends and we listened through the same material used to measure above and I captured our assessments. First some obvious but somewhat unexpected things:

• Not all vinyl discs are equal - there are some very bad ones with major warps and dirt and there are some better ones which are flat and clean. Standard vinyl weighs 150g and "audiophile grade" are heavier at 180g, but not all standard LPs seem this weight. These physical issues introduce noise and warble.
  
• Every LP came with the snap, crackle and pop that we all remember. Dirt in the groves, discharging static from the vinyl, etc. was typically present but only really audible before and after the track's music, even on quiet parts.
  

Now our impressions of the comparison of the source material.

• We all thought the digital material, whether streamed from YouTube or played from a .FLAC file, was more crisp sounding and clearer. Too often the vinyl mid-bass was muddled sounding. The voices sounded fine on all sources but highs were again finer on the digital.
• When YouTube streaming was done from the actual released audio track (like when I measured above), there was no appreciable difference between the YT stream and the downloaded FLAC playback. However, we all said that watching the videos attending the audio can distract from critical listening. Sometimes the FLAC files yielded a bigger sound stage than the streams and always these both were better than vinyl (L/R channel separation wasn't measured but known lower on vinyl).
  
• The vaunted visceral feel and warm sound claimed to differentiate vinyl was not apparent to us on the tracks we heard. This might be because I built the NestT on a 300B SET amp which already "warms & smooths" the sound, even on digital material.

More Action Required... I think the vinyl got a short shrift as it was playing unequalized and without a sub (as were the others for comparison) ~ but that's not how I listen most often. I have acquired an analog tone control that I will use to tame my speakers some and use to activate the subwoofer on phono playback too. Let's hope it's not good money after bad but I gotta try to improve vinyl's showing here.