Quality of digital recordings

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I know that we all thrive for the best quality of our music. I have music in CD, GDCD, DVD-A, SACD formats - the albums that I like I always tryed to buy in the best resolution.
Recently I started to use the foobar2000 with the spectrum visualizer on bottom. I realized why some recordings always sounded good to me and others dull. Some hi-res formats sounded dull too, so I did plug in the analog outs of the DVD-A/SACD player into my new E-MU1820m and capture snippets. Played in foobar. Well... Beside some albums that are looking and sounding great, I was no so surprised to find that the albumes that sounded bad... have reduced bandwidth or reduced content of high-frequency sounds, even in "hi-res" formats that I do own. This is just a frequency observation, dynamics usually mirror the bandwidth performance.
Samples below. First some "good" ones.
1. Police - The Classics (SACD).
An externally hosted image should be here but it was not working when we last tested it.

Nice looking, nice sounding. Constant high bandwidth.
2. Pink Floyd - TOSOTM (SACD). Below is a clip from the beginning of "Time":
An externally hosted image should be here but it was not working when we last tested it.

It's rather an exception for Pink Floyd music, their cymbals don't pass 18kHz:
An externally hosted image should be here but it was not working when we last tested it.

I would say that is marginally hi-res. Not so much difference from the CD.
3. Metallica - The Black Album DVD-A
An externally hosted image should be here but it was not working when we last tested it.

This track is nice sounding, nice looking. But other tracks barely hit 17kHz. I am confused.
4. Genesis - Genesis (SACD)
An externally hosted image should be here but it was not working when we last tested it.

Marginally hi-res. Content of hi-freq sounds is present but limited in level.
5. Tom Petty - Damn the Torpedos (CD). Love the album, but...
An externally hosted image should be here but it was not working when we last tested it.

That's all you get.
6. Eagles (CD) Love the music. Not the quality.
An externally hosted image should be here but it was not working when we last tested it.

7. Rush - Moving Pictures (CD)
An externally hosted image should be here but it was not working when we last tested it.

That's all the master tapes had?
8. Yes - 90125 (CD)
An externally hosted image should be here but it was not working when we last tested it.

Nice sounding, maybe a little "relaxed". Spectrum shows a slight taper-off after 17kHz.
9. Something "modern". Lady GaGa - Born This Way(CD)
An externally hosted image should be here but it was not working when we last tested it.

I can not only hear, but see, the dynamic range compression. Content of hig-frequency is mostly... white noise.
10. Lady Antebellum - Own The Night (CD)
An externally hosted image should be here but it was not working when we last tested it.

This is a rare capture, most of the time you will see signals below 15kHz. Sounds good, not compressed, relaxed, but in the end dull...
 
Last edited:
I have seen in a couple instances in the multiway forum where people have commented that adding a super tweeter brought the music"alive" , but I'm sure they wouldn't make much of a difference if there's no musical info present. Maybe the engineers are just bandwidth limiting for the I-pod generation?
 
Bandwidth is indeed an issue, but compromised dynamic range is even more pernicious. Play around with Keith Howard's DynamicsAnalysis freeware. You'll soon see that gross compression and limiting is the rule rather than the exception. And that many recordings praised in the audiophile press as "detailed and dynamic" are the worst offenders.
 
Some older recordings sound a little dull (indeed like The Eagles) but that's probably the result of the limited bandwidth of the recording equipment of those days.
But it's the heavily compressed "loudness war" CDs of today (actually it already got bad back in the 1990s, e.g. Oasis albums) that I try to steer clear of as much as I can nowadays. They just hurt my ears and can give me tinitus.

The first image is the tell tale look of a waveform of a loudness war track. The signal almost completely fills the available headroom indicating almost no more dynamics present and normalizing nearly up to 0 dB.
The spectrum analysis shows the bandwidth has been sharply cut off at around 17 kHz. Originally it was an MP3 of 192 kbps, this may have something to do with the bandwidth limitation.

The other track doesn't sound compressed and it shows. It looks much more "ragged" because the dynamics are still there. Even after normalizing to 100% (i.e. largest peak at 0 dB) the waveform doesn't look different. The original recordings must have been made on analogue tape (late 70s early 80s) hence the slow roll off of the frequency. Luckily the dynamics were left intact when the recordings were remastered for CD in 1996.

Loudness war music all start well recorded but at some point in time during production the decision is made to compress all living soul from the dynamics, IMHO destroying the music. I agree with Bob Dylan that loudness war music is much like static. I can only hope music companies stop this asap.

Oh yes, the waveform in the first image is Amy Winehouse "You know I'm no Good". The second image is from The Blues Brothers Complete "Expressway to your heart".
 

Attachments

  • Image1.gif
    Image1.gif
    94.2 KB · Views: 89
  • Image2.gif
    Image2.gif
    125.7 KB · Views: 92
You folks know/remember that CDs are brickwalled at 22K, right?

I'm also not convinced that "dull" sounding older recordings are due recording limitations but rather deliberate choices by the mastering engineers. I suspect it more likely that current recordings highs are boosted along with the compression.
 
Brickwall or not, you can easily pick out 70s pop recordings, IMO they have the most recognisable sound "signature" of all recordings out there. But I'm talking about recordings that were remastered in the earlier years of digital. Nowadays it's likely remasters are "treated" with all the gadgetry that's available.
Example: get some old AC/DC-CDs (e.g. "Higway to hell") and compare the sound to the ca. 2003-remasters.
 
Your findings confirm what I've suspected, even some of your examples are songs I use as references. The Police had absolutely brilliant production- the peak of the vinyl era and analog studio recording. Pink Floyd's The Wall is also brilliantly mastered. And of course, my fave genre is late 70's prog rock- I suspect in part because that was the high point of analog recording.

I built a set of Short Thor speakers a couple years ago and was absolutely astounded at what I heard. Several tracks on Yes (the album) have very faint "snicking" sounds, which I figured out are from the engineer punching tracks in and out. In fact, and again this is what I love about that era, mastering was all hands on deck, everyone turning the right knobs at the right time- band members, roadies, engineers, everyone.

I absolutely despise the high compression trend.
 
What happened to the pics in the first post? This is the first time I've seen this thread, and they're all a - sign inside a circle for me.

I'm doubtful that frequency response is a major problem. It annoys me at an intellectual level that mp3's are usually cut off below 15k, but I can't hear past 12k on a good day anyway. A look at the response of any 15IPS reel-to-reel recorder or mixer from the -60-s or '70's is probably pretty flat to 20k and only down a little bit at 30k. I'd think of LP's as more limited, on the other hand there were was discrete quad LP's in the '70's using supersonic carriers and sidebands for the back channels, much like stereo multiplex FM.

There are people who claim "LP's sound better" who can't say exactly why they feel that way, but I'm pretty much convinced it's because they they're hearing the dynamics on older LP's that don't exist in recently produced or recently remastered recordings.

I remember as early as the late '70's, the tenor from little I read of the hifi magazines was "wait till we have digital as a consumer format - the dynamics are gonna BLOW YOUR SOCKS OFF!" We got digital with the capabilities, but the dynamics sure went to hell in a handbasket. Now we have "all the bits flipping all the time" except when it's at one rail or the other.

Woops, sorry. I just wanted to add my voice to the chorus saying how terrible everything is nowadays.
 
It annoys me at an intellectual level that mp3's are usually cut off below 15k

Mp3 encoders are far better than they use to be, and I don't think it's the norm for the good quality ones to use a cut off anywhere near 15k

but when they do it is because the quality would be far worse with the sounds above, this usually applies to low bitrates, like 128kbps and below

I did a quick test the LAME encoder, at 320kbps they use a 20323 Hz - 20903 Hz transition band.

256kbps: 19548 Hz - 20129 Hz

192kbps: 18774 Hz - 19355 Hz

128kbps: 16452 Hz - 17032 Hz

112kbps: 15115 Hz - 15648 Hz

MP3 gets a lot of flak it doesn't deserve probably because of the old days when you could identify each encoder by the audible artefacts it created, now using variable bitrates, psycho acoustic modelling, blind listening tests and loads of CPU, you can get as low as 180kbps with no audible difference to the source file.
 
Mp3 encoders are far better than they use to be

They are a lot better, LAME has improved significantly over the last 5 years although there hasn't been much improvement in the quality of constant bitrate MP3's. I've never really liked any kind of psychoacoustic audio compression scheme because they always seem to introduce knocking sounds as well as watery and metallic sounds into the final output. I've noticed significant differences in quality between codecs such as MP3 Musepack and WMA at bitrates at bitrates as high as 256kb/s. The treble cut-off doesn't necessarily mean the codec will have better quality than a codec with a lower one, sometimes it can be a negative thing at medium bitrates as the extra information required to encode the high end will result in a loss of temporal resolution (especially in MP3) resulting in smeared transients (Xing used to suffer particularly badly from this problem).

Remember, it's not only the high frequencies that are lost in the compression process. Just look at a spectrogram and you'll see the dropouts.
 
They are a lot better, ...Just look at a spectrogram and you'll see the dropouts.

Well, they may have improved ever so slightly, but I can still pick out an mp3 any day of the week. No matter what bitrate. The noise floor in general and stillness in transitions and gaps are filled with a smeared droning of sorts. Even though you may not have 20-20k hearing you can still easily pick out the high and low cutoffs as well, because it affects the sound in the entire audio band indirectly. The sounds you normally would not notice have been smoothed away and replaced with tiny smooth shortcuts so to speak, causing a sort of smooth noise that is not present on lossless formats.

Good quality .ogg Vorbis on the other hand, I have difficulties telling apart from CD quality wav sometimes. But it was always a far superior algorithm in my eyes.

Regarding the issues mentioned about different sound engineers having different sound:

Great sound engineers will not apply compression, or use compression in such a way you cannot detect it by listening to the finished product. They listen to the tracks on a multitude of different systems in different circumstances and settings, and adjust the mastering accordingly.

Good sound engineers always listen to the tracks on several different setups and headphones, when they use a peak limiter to take care of the short peaks and the compression set slow and easy, so the drums will not fade out flutes or strings for example, and transitions will appear to be dynamic even though they are not for the trained ear.

I know about several allegedly pro sound engineers who rely on aggressive compression settings and just 1 (one) setup for listening, because they delude themselves that it's the best one, and often even apply EQ to the system they listen to, to "smooth it out". Studio monitors are not always better and more neutral than commercial hi-fi equipment, everything you can buy for money in this world is made to perform a specific task according to the design ideals/subjectivity and compromises of their creator. Studio equipment NOT an exception. Professional self deluding amateurs!


There is also a big difference on lossless formats, not between the formats, but how they are ripped from the medium. A poorly ripped flac is not much better than a 320kbps mp3.
 
Last edited:
Well, they may have improved ever so slightly, but I can still pick out an mp3 any day of the week. No matter what bitrate.

What encoder did you use?
And what were your results in the ABX blind test? if you didn't ABX-Blind why should I trust your results?
What was your sample?

Why does Hydrogen Audio, who critique Vorbis and LAME constantly, recommend using LAME with the V0 (Variable bitrate averaging 245kbps) setting and claim it to be transparent (except in rare occasions), yet you claim to be able to 'easily' hear the differences between the original and 320kbps, Vorbis on the other hand can go up to 500kbps can you hear anything different at these bitrates?
 
I admit I have mostly constant bitrate mp3's, if this really matters at 256 and 320kbps. I have no idea which encoder was used, but last I tried the lameenc.exe or whatever it's called on foobar I could still pick out mp3's.

I have a lot of the same tunes as Flac, .ogg Vorbis and mp3, they have different playlists on foobar (tabbed ofcourse), sometimes I get the wrong playlist, most of the time I have no problem spotting the mp3 playlist, but it's hard to notice the ogg. Sometimes my wife will put on something while on my own computer, and after a while I'll ask why the sound sucks. Even my wife notices it on occassion. Some of the flac files I have from friends are also completely useless, guess I should just delete them, but it would make gaps in my 18500 song collection (flac only).

Like i said, an mp3 I am fairly confident I'll spot in blind tests, a bad rip is also easy to spot whatever the file format, be it .ogg, flac, ape, aac whatever.
 
Sorry about the bump.

The difference between a good quality mp3 and a cd quality flac is about as obvious as the difference between the integrated dac and analog out from my tv compared to the soundcard on my pc. I will notice it every single time, but it may take a few minutes if I am not listening for it, which translates to a perfect blind test in my opinion. The best blind test is when you're not aware you're being tested!

I still state that mp3's have their own sound/noise because of the algorithm used, you do not listen for the high and low cutoff, but the sound replacing the upper and lower frequencies. You do not listen for the details when the music is going full tilt, but when it should be quiet and calm and there is a constant noise that do not belong.

Edit:
I see no point in 500kbps ogg, most cd ripped flac of sensible quality is between 650-980kbps. Mp3's for my car, .ogg for our phones, flac to use at home.
 
Last edited:
Administrator
Joined 2004
Paid Member
I took a look at a lot of recordings that have a pleasant (to me) tonal balance and those that don't. Many of the pleasant ones were old, the unpleasant fairly recent.

What I found was a lot of energy in the upper-mids to high end in the unpleasant recordings. They sounded harsh, bright and fatiguing. Now wonder, with all that energy up at 5K-8K. Ouch. I don't know how that got there, if it was mixing, mastering, bad gear. But it sounds bad. More agreeable recordings have a more natural roll-off at the top.
 
Mp3 encoders are far better than they use to be, and I don't think it's the norm for the good quality ones to use a cut off anywhere near 15k

[snip] ...

MP3 gets a lot of flak it doesn't deserve probably because of the old days when you could identify each encoder by the audible artefacts it created, now using variable bitrates, psycho acoustic modelling, blind listening tests and loads of CPU, you can get as low as 180kbps with no audible difference to the source file.

The cutoff frequency is determined by the codec ... There is nothing an encoder can do about it. If the codec and bitrate call for 15k it will be no more than 15k (for example).

Maybe a nitpick but for clarity it seems you use "mp3" where "codec" should be. An mp3 is an mp3 ... Its a patent-protected codec that has not changed since the 90's when it was introduced; devepment goes back earlier. Psycho-acoustic modeling and blind listening were instrumental in its formulation ... Its not new.

If modern desktop software does a better job of encoding mp3 its mostly due to licensing and refinement of consumer hardware and software ... the (expensive) tools were available to professionals a long time ago and are used commercially for mp3 from the onset. VBR had to wait for the players' chipsets to support playback but its part of the mp3 spec ... you could have used it a decade ago on the desktop but no portable or car stero mp3 player could play it then.

Newer codecs however do the things you mention and use the same development strategies pioneered by Freuhoff. A compressed file might be an mp3 but it also might be another format. Most of the improvements you mention are found in the more modern codecs ... an mp3 is an mp3.

Look at Satellite radio ... Sounds OK to most people ... Bitrate is well under 128kbps for music; its been a while since I studied the spec but I seem to recall 80 and 64kbps ... talk channels are lower still. Different codec than mp3 but still lossy low-bitrate compression developed with psycho-acoustics, blind listening, etc..

A well-encoded mp3 doesnt sound any better or any worse today to my ears than one from a decade ago, but there are lots of poorly encoded mp3s out there. Other lossy formats developed since do sound better though.
 
Last edited:
Administrator
Joined 2004
Paid Member
... you could have used it a decade ago on the desktop but no portable or car stero mp3 player could play it then.
Indeed. I was fiddling with VBR MP3 about 12 years ago. Most software players then would play it OK, but the counter never worked right - like it couldn't understand the length of the song. There were no off the shelf car players at the time. I don't remember about portables.

OGG, WMV and others are nice, but once you get up to about 240KBs, they all sound similar. The phasey sounds of MP3 are gone by that time.

While some with good ears can tell the difference in comparison tests, most have no idea it's MP3 if you don't tell them. Tell them, and they hate it. ;)
 
There are people who claim "LP's sound better" who can't say exactly why they feel that way, but I'm pretty much convinced it's because they they're hearing the dynamics on older LP's that don't exist in recently produced or recently remastered recordings.

Few years ago, I converted some LPs to CDA. (CD). cause the parts for turntable were running out from the market here.

They did sound good ( although not very same). but later in couple of years the owner company published the re-mastered CDs. I was expecting lot from them. when bought and heard, they had know presence and was flat with no details sure it was not about treble bass or mid ( missing the dynamics). and felt as they were not worth buying. so probably what you said is true.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.