An Exploration Into Digital Audio File Compression

An Exploration Into Digital Audio File Compression

Written by Tom Gibbs


I first got into digital file streaming about eight years ago at the insistence of my brother, who wouldn’t shut up about it until I got somewhat involved. Of course, at that point in time, we called it “computer audio,” because, frankly, I don’t think any of us really knew what “streaming” was at all about. As I’ve mentioned here before, the late John Sunier offered me the opportunity about five years ago to review the Auralic Aries first-generation streamer, and mainly because he didn’t know what it was or what it did. I didn’t know either, so I also passed. If only we could climb into a time machine!

Just for clarity here, I’m not talking about compression as in, “the remastering of this title is totally too compressed!” As in, the dynamic range (and any of the life in the music) has been totally sucked out of it. I’m talking about file conversion algorithms that employ a selectable level of compression (or not!), and if (or how much) your files may or may not have been compressed at the point when you ripped your CDs to FLAC or whatever. Most file conversion applications employ some level of compression mainly to save file space on your music server. Remember, we’re only talking about PCM files here!

Ripping with Winamp and Exact Audio Copy

I currently work on Windows and Linux-based computers at home, and on Windows and Macs at my day job, but at the time when I first became involved in “computer audio,” it was strictly Windows – as in, Windows 7. Meaning, very simplistic. I had no concept of “ripping programs,” and started out with Winamp, which is probably less horrible than you might think, and especially at the time. It wasn’t overly complicated, and I ripped everything to FLAC – mainly because my brother told me to; I didn’t really know anything else. In my early ripping experiences, I still mostly thought that all digital sources (including the compact disc and other digital discs), weren’t particularly great-sounding, at least in their then-present incarnations (with the exception of SACD, of course). I considered digital inferior to analog sources, so it probably wasn’t worthy of “overthinking” in terms of preparation for audio playback. It would take a miracle to get the CD format sounding any better than it did, so how could a rip of a CD be expected to sound much better?

I have no idea if Winamp offered any level of selectable compression of the output file; I wasn’t savvy enough at the time for anything like that to be on my radar. And besides, the Windows 7 experience was coming to an end for me – my brother’s two sons, who were computer science students at Georgia Tech, had come home with a copy of a new bootleg program that would allow you to load any version of Windows then currently available without a license. Actually, it had a “loader” application that would generate the correct product key for whatever version you chose to install. Yeah, I know, I’ll plead the fifth on this one. I ended up installing Windows 10 Pro, skipping Windows 8 completely in the process (my current setup is completely legit; yes, Bill Gates thanks me for my support). And at that point, I moved up to Exact Audio Copy (EAC), where I actually started paying some attention to things like ripping an “exact copy” of the CD to FLAC. Compression was still off my radar at this point, however.



In my initial experiences with building a digital library, I 1) ripped CDs from my own personal library that I listened to with some frequency; 2) borrowed from my brother’s (and friends’) libraries for CDs that I didn’t happen to own but wanted; and 3) borrowed from the public libraries around me for CDs that were uncommon in my usual circles or travels. Yes, I know all about the “copyright issues,” moral implications, fair use, etc….once again, I’ll plead the fifth. I wasn’t reselling anything, it was all strictly for my own personal use, so at the time, I didn’t have any issues with it. My habits have changed significantly since then.

Moving to dBpoweramp

So, basically, what I’m getting at is that I probably had hundreds upon hundreds of CD titles that had been ripped with either Winamp or EAC, and with little concern for overall file quality and, especially, any level of compression that might have been employed. The EAC era lasted a couple of years, but when I moved into a new house four years ago and suddenly found myself with a significantly higher level of digital equipment to play with, I took the next step and bought licenses for dBpoweramp. Something else happened at this point: hard drive storage (especially solid-state drive) prices began to drop significantly, and I started reading online about “no longer needing to compress FLAC files when ripping, because with the now really low prices of storage, who cares about saving a bit of space?” So, listeners could now rip all their CDs, using dBpoweramp, as uncompressed FLACs and without a care concerning the file size. No fuss, no muss, no need to budget drive space!


At this point, I probably had about 500 ripped files in total. When I started using dBpoweramp, I suddenly decided to re-rip all my previously-ripped personal library files using the “uncompressed” settings to replace the previously compressed rips. So, over the four-year period that I’ve been living in the new house, I’ve ripped about 3,000 CDs and 400-plus SACDs and DVDs. I’ve also explored some of the dBpoweramp plug-ins – for example, the HDCD plug-in – where you can rip all your library HDCDs as 24-bit versus 16-bit versions. Now that’s a really useful algorithm, and especially in my current “disc-less” setup, which allows me to experience all my HDCDs without a disc player.

But that still leaves several hundred of my “original rips” that came from CDs, libraries, or friends, that I no longer have access to. When we moved, we changed counties, and the current county I live in doesn’t even have any music available for lending at any library locations. What’s this world coming to? I have often thought over recent years that the sound quality of some of those early rips (especially from Winamp) is probably somewhat substandard compared to rips made with dBpoweramp.

Commonly-Available File Formats

In my gigs as an audio and music reviewer, I often have access to digital files of varying formats. Usually, when I get a review assignment that involves a CD release or reissue, advance files are offered. They usually are available as downloads as WAV or higher-resolution MP3 files (320 kbps). One of the available plug-ins with dBpoweramp is a file format converter that works to absolute perfection converting WAVs to uncompressed FLACs. And I always convert WAVs to FLACs; otherwise, you have no access to any of the files’ metadata. I also have the same process with LP review releases, where digital files are also planned for release on the various streaming services. And I happen to believe that with highest-resolution (320 kbps) MP3s, there’s very little going on in terms of file content getting trimmed from the final files; it’s just getting really compressed. When I convert 320 kbps MP3s to uncompressed FLACs, I’d absolutely dare anyone to be able to double-blind tell the difference between the converted files and the actual CD files.

There was a fairly popular thread on many of the digital audio sites a few years ago where the author claimed that if you ripped your CD to 320 kbps MP3 using the LAME codec, then reconverted it to uncompressed CD quality, you would experience a significant uptick in overall sound quality. A couple of writers at Stereophile even chimed in on this. I never dove too deeply into it, but did experiment a bit, and could find no difference between the original and the LAME-processed, then reconverted MP3.

So what is all this rambling getting to? I’ve had frequent conversations with Dalibor Kasac, who’s one of the principals at Euphony Audio (manufacturer of my streaming setup). And he believes that, whenever possible, zero compression should be used. Regardless of how great the conversion algorithms are that are currently in use with most streaming software applications, there will always be losses when compression is applied. (He also believes – unflinchingly so – that DoP [DSD over PCM] is terrible; only native DSD will give you the entire picture.) I have to agree with him here; since I’ve switched to an I²S playback setup, I honestly do believe that DSD (and perhaps even PCM) sounds remarkably better when played without conversion through my PS Audio GainCell DAC. YMMV, but I stand by what my ears tell me.

An Epiphany Last Night

Like I said, I’m frequently converting WAV or MP3 files to uncompressed FLAC for playback evaluation – especially when I’m reviewing an LP of material that I’m totally unfamiliar with. I’ll convert the files for playback on my home digital system, or perhaps in my car; my process is that I can just listen to the files for an extended period of time, just to get a feel for the music, then evaluate the actual LP. I usually get LPs with a protracted period of time before the actual release date, so I generally have lots of time to play about with them before writing a review. And, surprisingly, the digital files usually sound pretty darn great over my system – in fact, I’m usually quite stunned by how very good they sound!

So, last night, I was converting some files for an upcoming LP reissue, and I suddenly had this thought – using dBpoweramp’s file converter works so effortlessly and with such good results, why not use it to convert the oldest of the old FLACs in my library to uncompressed FLACs? I didn’t have any idea what level of compression (if any) was employed by Winamp or EAC, so I started looking at files in my library – I actually opened Roon, and started looking at the album covers to help remind me of the older files in my system. It amazed me how very many I recognized as coming from sources that weren’t in my CD (or other disc) library, and here’s the real kicker: when I clicked on the FLAC files on the storage drive in Windows Explorer, and checked the Audio Properties tab, many of the earliest files were compressed as much as 35 to 55 percent!!



I currently have tons of solid-state drive (SSD) storage space available, so I made a copy of the first album I came across, Kraftwerk’s Minimum-Maximum, which is a live album I got from the Chicago Public Library. It showed the Winamp compression level as 35 percent, so I used dBpoweramp’s converter to change the compression level to zero percent. I then went downstairs for a listen; nary a tick or hiccup, and no artifacts of any kind, and the sound to me was definitely better than before, so I then went on an hours-long binge, converting everything from waaaay back that I could easily identify. Yeah, I know, you probably think I’ve totally lost it here, but it’s at the very least been really liberating to get all these early files on pretty much equal footing with more recent rips.

More Previously Unidentified Compression

After converting all the early Winamp and EAC rips, I had another thought: I have at least a hundred or more digital downloads in my system that have either been promos, gifted, or that I purchased. Could they possibly have any level of compression? A reminder – we’re only talking about PCM files here; DSD files do not have any level of compression applied.



I started looking at all my 24/96 and 24/192 files I’ve downloaded from HDtracks and Acoustic Sounds; sure enough, all of them had compression levels that averaged between 30 to 50 percent. I had downloaded the recent Beatles Abbey Road 24/96 Giles Martin remix, and it was compressed at about 40 percent. Early on, I’d downloaded Joni Mitchell’s Ladies of the Canyon, in 24/192 resolution. I always found this download very puzzling, as I expected such a high-level-resolution file to be also quite a large file. It wasn’t; it was actually smaller than many of my 16-bit files. Imagine my surprise when I checked its compression level to discover that it was a monstrous 81 percent – that explained everything! Of course, after conversion the file size literally quadrupled, but who cares? Again, I experimented, and finding no trace of any artifacts, converted all of these downloads to zero compression levels. But I have to believe that compressing a file by 81 percent – which would definitely make it much easier to download – would have to negatively impact the sound quality. This was true time and time again with every HDtracks download compression level I checked – with Rush’s classic 24/96 download of Moving Pictures, there was a whopping 68 percent compression. Of all the digital download tracks I checked in my library, those from HD Tracks consistently had the highest levels of compresssion.



Then there were the Qobuz downloads; they all had an average of 30-plus percent compression. 24/96 and 24/192 downloads from the 2L and Channel Classics labels had compression levels averaging between 40 to 50 percent. One of my favorites, Channel Classics’ Guardian Angel by Baroque violinist extraordinaire Rachel Podger was compressed for delivery by 52 percent. Again, this greatly increased the ease and speed of downloading, but seriously, 52 percent compression from an audiophile label? I’ve downloaded a number of titles from High Definition Tape Transfers (HDTT), which is one of my favorite download sites for outstanding archival transcriptions. One of my favorite titles, the Spanish music classic Espana featuring conductor Ataulfo Argenta, was 49 percent compressed. Amazing, and incredibly disappointing.



Sitting around, racking my brain to confirm that I hadn’t forgotten anything, I suddenly realized that I had the Steven Wilson Remix Blu-ray and DVD-Audio discs on hand that I’d bought of Yes and King Crimson titles that I’d recently converted to FLAC. I also realized that in the conversion setup menus, I didn’t recall seeing anything that allowed for adjusting the level of compression during the conversions. I found this really interesting: when converting from Blu-ray discs, zero compression was applied, but when converting from DVD-Audio discs, usually somewhere around 50 percent compression was applied. Again, I test-listened to everything, and felt the results were definitely an improvement.


Yeah, I know – I’m waiting for the flamethrowers to come out the day this article drops. All music player applications are designed to decompress a compressed file in order to achieve correct playback, and the transfer process is supposed to be completely seamless – but we’re talking about some serious digital processing going on to achieve that. There’s some level of compression in most digital file formats, and the transfer process between compressed file and uncompressed playback isn’t supposed to have any impact on sound quality. But I can’t believe that the transfer process involved in decompressing a file that’s been heavily compressed (like 80 percent!?!) will ultimately result in absolute fidelity. As I mentioned earlier, with the current low cost of storage – why not just maintain an entirely uncompressed digital library?

Header image from the dBpoweramp website. All other images courtesy of the author and the respective entities.

Back to Copper home page

1 of 2