[bigshot]

Whenever I see someone talking about being able to hear a difference between MP3 and FLAC without mentioning the codec or bitrate, I pretty much know they have never made any real effortt to find out for themselves. All MP3s are not the same, and doing a simple comparison test (not even a blind one) will show that.

 

[waynes world]

Quote:

Originally Posted by Brooko /img/forum/go_quote.gif
 
Here's a quick guide I wrote to using the ABX tool + ripping, tagging etc (http://www.head-fi.org/t/655879/setting-up-an-abx-test-simple-guide-to-ripping-tagging-transcoding)

 
Thanks for doing that!

 

[Dillan]

I agree with what everyone said after my post. Just saying MP3 vs FLAC in itself doesnt really make any sense. Way too many factors, way too broad. And for the person talking about ABX tests, i agree with that method too. Ive actually done it a few times the old fashioned way though, and had my back turned while a friend pressed play. Either way, I think placebo is huge in these cases, and dont think that I was defending MP3 more than anything in my post, because I was actually defending myself having all (mostly) FLAC. I just know that there's no potential harm (other than larger files) by going the all FLAC route, while on the other hand there is potential proof of harm in going all MP3. Its scientifically proven in sound science, and a lot of people feel in reality it really is a step above. Either way its a sense of security vs potential loss.

 

[Brooko]

Quote:

I agree with what everyone said after my post. Just saying MP3 vs FLAC in itself doesnt really make any sense. Way too many factors, way too broad. And for the person talking about ABX tests, i agree with that method too. Ive actually done it a few times the old fashioned way though, and had my back turned while a friend pressed play. Either way, I think placebo is huge in these cases, and dont think that I was defending MP3 more than anything in my post, because I was actually defending myself having all (mostly) FLAC. I just know that there's no potential harm (other than larger files) by going the all FLAC route, while on the other hand there is potential proof of harm in going all MP3. Its scientifically proven in sound science, and a lot of people feel in reality it really is a step above. Either way its a sense of security vs potential loss.

Agree with some of what you are saying - but why not follow the guide and take a proper abx?  The beauty is that you can set it up yourself, by yourself - and do it in your own time.
 
Your overall thoughts on lossy compression may change.
 
I'm not trying to change your mind on lossless - I archive to FLAC and use it on my desktop.  What I am trying to do is let you make informed decisions regarding good lossy encoders, and the limits of your own hearing.  Like I said - the results will probably surprise you..

 

[Dillan]

My decisions aren't misinformed if that's what you are implying, but i understand what you are saying and I would love to setup a proper abx. Only problem is, I switched over to J River a little while ago and I am not sure if they have something like that. I wouldn't mind getting Foobar again just to try that out though.

 

[Brooko]

Definitely not implying anything - just gently suggesting that the actual ABX is very easy to set-up, and just involves time.  You obviously have a Windows PC.  My guide is pretty easy to follow.  If you do ever try it, I'd be interested to hear if it changes your thoughts on lossy - especially around the 256 kbps rates 

 

[bigshot]

I just know that there's no potential harm (other than larger files) by going the all FLAC route, while on the other hand there is potential proof of harm in going all MP3. Its scientifically proven in sound science, and a lot of people feel in reality it really is a step above. Either way its a sense of security vs potential loss.

The general consensus among those in Sound Science who have done controlled tests is that past a certain point with bitrate and codec, "lossy" audio is aurally transparent from "lossless". If you can't hear a difference, it doesn't matter any more to the sound quality than whether you're wearing a green shirt to listen to music or a red one.

"Security" and "Potential Loss" have nothing to do with sound quality. Those are psychological issues. Most of us have a little bit of OCD in us. Some have more than others. Worrying about "what might be" would be more important to the latter group of people.

 

[chewy4]

Quote:

"Security" and "Potential Loss" have nothing to do with sound quality. Those are psychological issues. Most of us have a little bit of OCD in us. Some have more than others. Worrying about "what might be" would be more important to the latter group of people.

While it might have nothing to do with sound quality, it does certainly have to do with the enjoyment of music. Knowing it's all there is nice.
 
If I hear a tiny flaw in a recording I don't want to have to worry about whether it's part of the track or a result of compression. It's certainly nothing more than a psychological issue, but it's an issue nonetheless. And data storage is cheaper than getting a psychiatrist.
 
Not to mention ABX tests aren't flawless. They're great for the most part but echoic memory is extremely limited and the effects of compression can be different for different songs.

 

[bigshot]

I think differently than a lot of people, I guess... I don't want to just be confident in my decisions. I want to know exactly why I make them. When I hear an anomaly in my sound quality, I don't worry about what it MIGHT be... I grab on like a terrier until I find out EXACTLY what caused it.

I have over a year and a half's worth of music in my iTunes library... classical, jazz, pop... older recordings, newer recordings... all kinds of music. The library plays throughout my house via wifi to every room in the house. The same library serves my main listening room where I have the system I've been refining for the past thirty years.

Every time I've heard a problem with the sound, I've tracked it down. It's never due to lossy artifacting... it might be a bad recording, equalization problem, funky transducer, volume imbalance, a million things. But I've never found a single artifact in my files. I use AAC 256 VBR. I determined that setting by doing line level matched A/B comparison testing.

To be honest, I don't think an audiophile will ever achieve optimal sound quality by taking the "more numbers is better" approach. The problem never has anything at all to do with the file size in my experiience. You can't get perfect sound by relying on numbers to do the work for you. It takes proactive problem solving. In fact, if there's anything I've learned in 30 years in this hobby, that is it.

 

[xnor]

Well you gotta keep in mind that AAC is an excellent lossy codec. You probably cannot even reach AAC 256 VBR quality with MP3 at any bitrate.
I personally will take FLAC over MP3 anytime because I'm doing processing like equalization, crossfeeding .. and transcoding for portable players.

 

[flipper2gv]

In a very dense mix like death metal it's easier to spot the difference, which is contrary to popular belief. It seems to me that there is so much going on that having a lossy format will introduce artifacts because there is so little to cut.

It's really easy to notice especially in how the cymbals sound.

 

[julian67]

....But I've never found a single artifact in my files. I use AAC 256 VBR. I determined that setting by doing line level matched A/B comparison testing....

I like your approach, it makes a lot of sense, but I think you've been quite lucky...so far. AAC constrained VBR at around 256 is a very reasonable standard to settle on but no lossy format is foolproof. If your collection is of any size then you can be pretty sure it contains files with distinguishable compression artefacts or differences. It's impossible to abx each and every track so then you are back in the real world of normal listening where you are not making a side by side comparison, you are listening to music or even just casually or incidentally hearing it. No human being has enduring echoic memory. This means that even making an abx test is quite an intense experience with a surprisingly high level of difficulty (if compared to a somewhat analogous test of visual differences). Noticing a gross artefact like clipping is simple, but what are the chances of noticing in normal listening a small change in weight of bass, or a very marginal sibilance in a complex piece of music? Or a very slight change in the illusion of depth created by the stereo mix? How about several of those or similar factors being present, each by itself insignificant but together making for a degradation that is very hard to define? A side by side comparison can be revealing but in normal listening there is nothing obvious screaming "COMPRESSION!!!". What you get instead is less of the experience and enjoyment, the potential that resides in the uncompressed copy on your CD.

I did abx tests and went through the process of doing them with low quality settings so as to easily recognise the differences, then using better quality settings and repeating and so on. There are well known killer samples out there which will defeat lossless compression and, more to the point, there are other samples in my particular music collection (and probably yours too) which are equally difficult to compress. I remember about a decade ago paying over £100 for a 80GB disk drive. These days that buys me 3TB. My cheapest personal music player sounds great and has a microSDHC slot which accepts 32GB class 10 cards which cost £20 each. I wiped my lossy collection and now just use flac except in a handful of cases where I could only purchase an album in a lossy format. I can 100% guarantee that I will never hear my music collection at anything less than the best possible quality available to me. Can any rational person say the same about their mp3/aac/ogg/mpc/wma collection?

 

[bigshot]

The complexity or density of the music has absolutely nothing to do with whether a sound will artifact or not. Some sounds artifact and other sounds don't. It can be a pure clear tone or a massed chord. It either does or it doesn't. I have thousands of complex cymbal crashes in my library and not a single one of them artifacts at AAC 192 or above. I would bet five bucks that what you thought was artifacting was actually distortion built into the mix of the music. It's extremely difficult to hear artifacts when they are a part of distorted guitars. Much easier to hear them in the pure tones of acoustic instruments.
 
Lossy audio is capable of complete transparency- indistinguishable from the original CD. Try AAC 256 VBR for yourself and you'll see.

 

[bigshot]

Quote:

I like your approach, it makes a lot of sense, but I think you've been quite lucky...so far. AAC constrained VBR at around 256 is a very reasonable standard to settle on but no lossy format is foolproof. If your collection is of any size then you can be pretty sure it contains files with distinguishable compression artefacts or differences. It's impossible to abx each and every track so then you are back in the real world of normal listening where you are not making a side by side comparison, you are listening to music or even just casually or incidentally hearing it

 
I really didn't take any of this lightly. When you have a collection of music as big as I have, you don't commit to a compression setting without taking all that into account. I have over 10,000 CDs and at least that many records, dating back to the earliest days of recording. When I decided to build a music server, I spent the better part of a week compressing audio using various codecs and bitrates and comparing them with the original CD. I compared classical music, jazz, rock music, pop vocals, opera and country. I digitized acoustic recordings from 1905 to the most modern DSD recordings. I compared on my own system and the system of a friend of mine who is a sound mixer. I studied the varieties of artifacts online and learned to recognize the various types. One thing I learned is that artifacting is not subtle. It's like clipping. If you hit that line the whole thing goes splat. Even small artifacts are immediately apparent, because they sound like outer space glorps next to acoustic music. They stand out like a sore thumb.
 
I carefully inched my way up from very low bitrates to higher ones, so I could identify exactly what was being smooshed at each step. Finally at 192 CBR AAC, 256 CBR LAME MP3 and 320 Frauenhofer MP3 I couldn't detect any more artifacting. There was one stubborn CD that still artifacted very slightly at that rate, so I upped it one notch and added VBR to allow for a little bit more headroom if needed. I didn't arrive at AAC 256 VBR as my standard until I had thoroughly tested it as much as I possibly could. I could spend the rest of my life doing A/B tests on all my music, but I'm totally confident that I've allowed for a bit of overkill so I don't have to worry.
 
With the amount of music in my library, if I had ripped only as ALAC, I would have had to split the library across at least two hard drives. As AAC, it all fits on one. That makes backing up and transporting my music much easier. File size does matter.
 
I don't worry about what I can't hear.

 

[chewy4]

Quote:

I think differently than a lot of people, I guess... I don't want to just be confident in my decisions. I want to know exactly why I make them. When I hear an anomaly in my sound quality, I don't worry about what it MIGHT be... I grab on like a terrier until I find out EXACTLY what caused it.

I have over a year and a half's worth of music in my iTunes library... classical, jazz, pop... older recordings, newer recordings... all kinds of music. The library plays throughout my house via wifi to every room in the house. The same library serves my main listening room where I have the system I've been refining for the past thirty years.

Every time I've heard a problem with the sound, I've tracked it down. It's never due to lossy artifacting... it might be a bad recording, equalization problem, funky transducer, volume imbalance, a million things. But I've never found a single artifact in my files. I use AAC 256 VBR. I determined that setting by doing line level matched A/B comparison testing.

To be honest, I don't think an audiophile will ever achieve optimal sound quality by taking the "more numbers is better" approach. The problem never has anything at all to do with the file size in my experiience. You can't get perfect sound by relying on numbers to do the work for you. It takes proactive problem solving. In fact, if there's anything I've learned in 30 years in this hobby, that is it.

Well that's the thing - I too want to know what an anomaly is when I hear it. If I have a lossless file that eliminates a cause.
 
Granted you're right it's very improbable any flaw is going to be due to a compression artifacts at 256kbps+ bitrates at either LAME mp3 or aac, and I do understand why you take the stance that you do. 
 
 

In a very dense mix like death metal it's easier to spot the difference, which is contrary to popular belief. It seems to me that there is so much going on that having a lossy format will introduce artifacts because there is so little to cut.

It's really easy to notice especially in how the cymbals sound.
 

On the contrary - the  techniques that lossy compression uses work very well with complex arrangements. The artifacts are all masked by louder noises - that's a major part of how lossy compression works. Simple, high frequency stuff is what I believe might be the most noticeable.