Compressing CDs - FLAC compression level 0 or MP3 320kbps

Should I encode in FLAC 0 or MP3 320Kbps

  • FLAC compression level 0

    Votes: 4 40.0%
  • Mp3 320 Kbps

    Votes: 6 60.0%

  • Total voters
    10
Yes, my mastering service starts every master from 96/24. Some streaming services are now accepting 96/24 releases.
I work professionally on drum sample products. They are all recorded, produced, mixed and mastered at a minimum of 96/24, often higher.
The Roland V-Drums are 24bit/48khz audio. Higher than CD, nowhere near MP3.
I record drum tracks for people. I always record at 96/24 and send them the WAVs to mix themselves.
I have a raid system and back up my audio to WD My Passport drives which cost about $60 for 1TB.
Lovely stuff.

So when the records are mixed down, they are mastered to what for general release?

Chris, I'm not trying to denigrate your professional acumen here at all. Far from it. I think we're talking about two different things. If you're providing mixing/mastering/recording services then of course you record at high-quality. Of course you don't mix and record in MP3 because you're potentially going through several file generations. For a final product, 44.1/16 is the standard as you well know - but making a high-quality compressed MP3 of (say) 320 Kb/s has been demonstrated to be indistinguishable in all of the studies that I've read, including the one I've cited and which you've refused to read.
 
Last edited:
Yes, my mastering service starts every master from 96/24. Some streaming services are now accepting 96/24 releases.
I work professionally on drum sample products. They are all recorded, produced, mixed and mastered at a minimum of 96/24, often higher.
The Roland V-Drums are 24bit/48khz audio. Higher than CD, nowhere near MP3.
I record drum tracks for people. I always record at 96/24 and send them the WAVs to mix themselves.
I have a raid system and back up my audio to WD My Passport drives which cost about $60 for 1TB.
I don't understand the purpose of 96 000hz at all.. my hearing stops at +- 14 000hz

Can you see a difference between 16/44 and 96/24 ? It looks impossible for humans..

***
A 60$ HDD is a scary long term archival solution in my opinion.
 
You'll need to define the bitrate of the MP3, for a start...
According to my file samples.. really there is a difference between MP3 (at any bit rates) and FLAC . I need to spend time trying to hear a difference between 320CBR and 225VBR.. and even 128CBR.

But putting the MP3 bit rate asside, there I hear a difference between FLAC 0 or FLAC 7 against MP3 320CBR.

--> As if the MP3 encoding by itself produces a slightly different result and that difference is not entirely related to the bitrate but the technology by itself.
 
It's the same as the argument over cheap snares versus boutique snares.
Sorry but this is apples and oranges. Hit a snare a thousand times and measure it and you'll have a thousand different sets of data. Convert an original WAV to MP3 and back a thousand times and you'll have a thousand identical sets of data.We're talking about a measurable difference between an original and a compressed/decompressed set of 0 and 1. Since there's no original snare drum to compare to this example unfortunately doesn't fly.

Can you see a difference between 16/44 and 96/24 ? It looks impossible for humans..
There's a difference between 44.1kHz and 48kHz depending on your DAC. Because of lower quality some of them start to roll-off the heights way too early. Nyquist theorem prohibits frequencies above half the sample rate, so all DAC come with this roll-off, some do it better, some do it worse. But in general there's no need to go any higher as 44.1 for any kind of musical consumption. 48kHz just raises the Nyquist frequency into a definitely inaudible range so the roll-off definitely has no impact onto the signal covering the audible range.

What Chris is talking about is production. It's highly important to have a multiple of samples while producing audio, since lots of audio processors might add frequencies above Nyquist (e.g. any kind of non-linear amplification like distortion and exciters, bit crushing, (brickwall) limiting, even some harsh compressors), being processed further down the stream, introducing unwanted artifacts. Although todays professional grade DSP routines over-sample internally if needed.

But this is another esoteric discussion, broadly argued about in all kinds of home recording forums all over the world. Pros don't care, they just throw money at the highest grade shtuff (or better: what they perceive as the highest grade stuff).

there I hear a difference between FLAC 0 or FLAC 7 against MP3 320CBR.
Could you point your finger on the lossless compressed version? Although there is tiny differences (we're talking about 0.001-0.01dB every few samples) the majority of semi- to professionals claiming they could is (scientifically proven) wrong here, saying the MP3 is of higher quality if tested blindly.

Ask someone you trust to set up at least three (better 5) different tracks for a real blind test, mixing in some WAV to WAV tests, just for fooling your confirmation bias. In my opinion it wouldn't work outside of a laboratory (or any other controlled, unfamiliar environment) since I e.g. could just throw various mp3 decoders onto the files and compare just the diff of the data. So I could tell which is which within seconds (okay, maybe minutes) sitting in front of a Linux machine - without listening.
 
For a final product, 44.1/16 is the standard as you well know
Everything starts and fishes here at 24 bit 96khz. If I'm sending work out to partners, collaborators, it goes out at 24/96 and often comes back at 24/96. Most of the people I know are 24/96 too. If I'm offering drum samples for Roland products I convert to 24/48 or 24/44.1, depending on the product. That is the standard audio quality of Roland drums.
A lot of the streaming services now offer HQ streaming, so 96/24 is the default minimum for them. If I'm using an upload service (CD Baby type thing) they often demand nothing more than 16/44.1, so I convert my music master from 24/96 to 16/44.1.
90% of what I do is 96/24, occasionally I'm sending out CD quality.
 
I know the pitchforks are already sharp and the torches are already lit, but it's worth remembering the end use in this instance is listening on a car stereo...
That's why I mentioned the snare analogy. If the audience can't hear it, why bother?
Fact is, not EVERYONE is listening on car stereo. And by the way, a Tesla is extremely quiet with a fantastic sound system.
But it's about future proofing. We don't know what is coming next in audio. In 1980, I had absolutely no idea CD was around the corner.
If in a couple of years everyone is listening to 96/24 as standard, you are going to be sad you spent all that time creating mp3's.
 
According to my file samples.. really there is a difference between MP3 (at any bit rates) and FLAC . I need to spend time trying to hear a difference between 320CBR and 225VBR.. and even 128CBR.

But putting the MP3 bit rate asside, there I hear a difference between FLAC 0 or FLAC 7 against MP3 320CBR.

--> As if the MP3 encoding by itself produces a slightly different result and that difference is not entirely related to the bitrate but the technology by itself.
Is FLAC 0 through 7 available through the same FLAC codec? What tool are you using to test the formats? Thanks
 
That's why I mentioned the snare analogy. If the audience can't hear it, why bother?
Fact is, not EVERYONE is listening on car stereo. And by the way, a Tesla is extremely quiet with a fantastic sound system.
But it's about future proofing. We don't know what is coming next in audio. In 1980, I had absolutely no idea CD was around the corner.
If in a couple of years everyone is listening to 96/24 as standard, you are going to be sad you spent all that time creating mp3's.
Which is why 44.1/16 is still considered a decent standard, 36 years after it was formalised?

I think what this boils down to is that you're talking about studio production and audio being sent as a commercial product for others to use, whilst the intent of the thread is to ask if it makes any difference in a car stereo. Whoop-de-do you own a Tesla, bully for you - but even in controlled listening conditions the claimed benefits of lossless audio are at best, unproven and I'm yet to hear your convincing case for it being better when presented to a listener as an end product. When presented with evidence contrary to your opinion, you simply dismissed it as 'received wisdom', when it fact it was a pretty decent study into perceived listening experience and eminently relevant.

All you have offered is an unsubstantiated call to authority, telling us how great your work is and what you think you hear. You haven't engaged in any form of discussion whatsoever, haven't engaged with any of the discussion around confirmation bias, double-blind study, psychoacoustic theory, one chap going into good detail about Nyquist (which underlines this entire debate) and instead resolve into a tautology that 96/24 is better 'because it's better and I'm telling you that it's better'. Which, on a technical level it may be - but to an end consumer listening to a commercially-released product in a car, it's not going to make a damn sight bit of difference.

I'm now - goodness knows why - searching for the specifications and supported codecs for Tesla sound systems to see if they even support extreme-quality lossless and all I can find is that it 'supports FLAC' and '.WAV'. Helpful.
 
Is FLAC 0 through 7 available through the same FLAC codec? What tool are you using to test the formats? Thanks

For encoding in FLAC, I use Asunder in the Linux synaptic repository. It's unclear the difference between FLAC 0 or 7.
kz1FKNH.png


I test the format with my ears and my headphones only. I don't have a computer software to analyze the files, there was one called Spek but it can't be installed from the repository currently.
 
Whoop-de-do you own a Tesla, bully for you -

All you have offered is an unsubstantiated call to authority, telling us how great your work is and what you think you hear. You haven't engaged in any form of discussion whatsoever
Ugh, how old are you? This is classic forum bickering. Where did I say I own a Tesla? I didn't and I don't.
Where did I say 'how great' my work is? I didn't and that is completely irrelevant.
The original post asked for opinions and I gave mine. I ripped my own CD's to WAV because I can hear a difference and because I didn't want to do it twice, which I would have ended up doing.
The guy who said I was fooling myself and cited non-high end studio work in 2006 as credentials is criticising me for a call to authority?
I don't think so.
 
There are plenty of free spectrum analysers available in Linux. GStreamer might be an option if it's still being developed. Failing that, even Audacity will give you something to go on (make sure not to download the latest version!).

Which distro are you running?
 
Ugh, how old are you? This is classic forum bickering. Where did I say I own a Tesla? I didn't and I don't.
Where did I say 'how great' my work is? I didn't and that is completely irrelevant.
The original post asked for opinions and I gave mine. I ripped my own CD's to WAV because I can hear a difference and because I didn't want to do it twice, which I would have ended up doing.
The guy who said I was fooling myself and cited non-high end studio work in 2006 as credentials is criticising me for a call to authority?
I don't think so.

Have you read any of the studies yet?

Looking at the Roland technical specifications, I can't find a reference to anything higher than 44.1/16 on their module samples.


Current module.

Look, if you're not going to engage in the original discussion about subjective listening analysis and instead hark on about 96 kHz and how great it is in studio applications, fine. No problem. Do continue. But the fact is that 96 kHz audio isn't even relevant to the discussion at hand here. The discussion is about whether or not using a lossless format in a car is an improved listening experience from a high-quality MP3. The weight of evidence is against you and you're not even engaging in the discussion.

Disagree with you here. Happy to agree in most other aspects of your posting history.

EDIT: I've removed some of the more antagonistic nonsense in this post because I've had a really bad day and it's not conducive to the useful continuation of this thread...
 
Last edited:
There are plenty of free spectrum analysers available in Linux. GStreamer might be an option if it's still being developed. Failing that, even Audacity will give you something to go on (make sure not to download the latest version!).

Which distro are you running?
I use Linux mint Cinnamon 20 currently, not the latest revision which is 20.1 I think.
 
Oh man this thread is doomed. 😭
It's really not.

I'm sure there will be something out there for a Debian-based distro. In fact, here's one:


Not used it myself but next time I fire up one of the Linux boxes I'll give it a whirl. And yes, you may well see a difference on a waveform analysis. But your ears and (more importantly) brain don't work in the same way.
 
Sorry guys, I'm out for the next couple of days. Went to hospital directly after my last post, appendix. It's 20 to 1 in the morning, surgery will happen in a couple of hours. Behave! ❤️
 
Back
Top