FLAC is brighter than WAV

[ExtremeSkeptic]

For a week I have complained of too bright sound, I couldn't find the problem in my system, I thought my system sounded neutral. But the brightness came from somewhere. Then I tried WAV and it was very smooth. When I went back to FLAC it sounded harsh, bright and fatiguing.

I think the problem is that FLAC takes 2-3% CPU power while WAV takes 0%. I'm hearing a similar reduction in resolution as from turning on a 2nd harddrive, except the difference is much smaller.

Too bad I don't have any disk space to convert all my FLACs to WAV. :(


I thought they would sound the same, that's why I used FLAC in my system. The difference isn't huge, but it's significant enough for me. The difference between FLAC vs WAV is about 50 times smaller than 320kbps mp3 vs WAV.

Half of my albums are mp3 and I'm not bothered by them even though they have huge smearing. Lately everytime I listened to FLAC it sounded too bright and I rather chose to listen to mp3. But now when I tried WAV I heard the truth, WAV is smoother.


It's the decompression that makes it worse because it takes more CPU power = more watts = more jitter.


Here you clearly see the proof:

MP3 = 3% CPU power
http://poollogics.is-a-geek.net/pictures/forum4/blog/flac.is.bad/mp3.JPG



FLAC = 3% CPU power
http://poollogics.is-a-geek.net/pictures/forum4/blog/flac.is.bad/flac.JPG



WAV = 0% CPU power
http://poollogics.is-a-geek.net/pictures/forum4/blog/flac.is.bad/pcm.JPG

[ExtremeSkeptic]

I'm using Foobar > Kernel Streaming > EMU0404 > Nordost Valhalla > Benchmark DAC1

I have an old AMD 939 3000+ CPU because it gave smoother sound than X2 3800+. I have always had Cool & Quiet enabled because it gives smoother sound.
I'm using an Antec Phantom 350watt PSU. 1x 1GB RAM, 1x Samsung harddrive, EMU0404 soundcard, I have everything else unplugged because it gives smoother sound.


http://poollogics.is-a-geek.net/pictures/forum4/blog/cpu/cnq1000.JPG

http://poollogics.is-a-geek.net/pictures/forum4/blog/cpu/cnq1800.JPG

[Taffer]

Why on earth would you need 320kbps MP3s?

[PixyMisa]

Why on earth would you need 320kbps MP3s?
The only reason I can think of would be if you are editing the files. Since each edit can result in recompression (producing artifacts each time), you want to start out with the highest quality you can. (Uncompressed audio is best, of course.)

ExtremeSkeptic: FLAC is lossless. The audio output is bit-for-bit identical to the source WAV (assuming the WAV is really an uncompressed audio file.)

[Ducky]

wow. that is the single biggest load of unadulterated crap that I have ever seen posted. It's not right. It's not even wrong. It's just crap.

[Taffer]

The only reason I can think of would be if you are editing the files. Since each edit can result in recompression (producing artifacts each time), you want to start out with the highest quality you can. (Uncompressed audio is best, of course.)

That's true, I guess. I just assumed he wanted it for listening to. :o

[PixyMisa]

It's the decompression that makes it worse because it takes more CPU power = more watts = more jitter.
No.

[PixyMisa]

That's true, I guess. I just assumed he wanted it for listening to. :o
He probably is. But I keep my music as 256k mp3's so that if (when) I come to convert it to Shiny New Format, I won't introduce any noticeable artifacts. (I converted it all down to 128k for my iPod.)

[PixyMisa]

wow. that is the single biggest load of unadulterated crap that I have ever seen posted. It's not right. It's not even wrong. It's just crap.
You never met Franko, then? ;)

[Earthborn]

Why on earth would you need 320kbps MP3s?To fill up your MP3 player? :)

ExtremeSkeptic is a bit of an audiophile, which probably means he thinks he can hear the difference between a 320kbps MP3 and a 128kbps MP3 like he can hear the difference when his cables are wrapped in paper or having a marble ball next to the speaker.

[Ducky]

You never met Franko, then? ;)

No, but if Franko believes cpu useage = jitter I'd prefer not to. If CPU useage were to actually = jitter you could hear, then I suppose every protools/daw based studio in the country should shut down. I guess I'll call Doug McGee at Middle Earth studios in Michigan and tell him to close shop.

:rolleyes:

[Reeco]

The first thing an audiophile should do, in their pursuit of sonic heaven, is acoustically treat their listening room.

But they never do.

[MortFurd]

ExteremeSkeptic isn't an Audiophile. He's and audiophool. Or Troll who likes to pretend to be an audiophool.

He pushes a number of odd things. Amongst those things is that mp3 sounds better than wav, that wrapping his cables in (slightly) conductive toilet paper improves the sound, that placing a 24 lb chunk of marble next to his stereo makes it sound better, and that outrageously expensive power cables make a difference to the sound. Oh, and he uses his PC to drive his stereo.

[MortFurd]

The first thing an audiophile should do, in their pursuit of sonic heaven, is acoustically treat their listening room.

But they never do.
That, and get the best speakers available. ES doesn't even HAVE speakers. He uses headphones from which he removed the interconnection PC board. He claims that gives him 3dB better high frequency response - and then also admits to having no equipment with which to measure such changes, and measurements don't matter anyway.

[ExtremeSkeptic]

I have already compared 0% vs 100% CPU load in a blind test and the difference was huge. I have also done a blind test comparing 1 vs 2 harddrives running. I have the harddrive set to go to sleep mode after 3 minutes. But once 10 minutes had passed and I complained of edgy sound, then I looked at the 2nd harddrive and it was still running. The test doesn't get any better than that.

Everytime the 2nd harddrive spins down I can hear the shift to smoother sound. Everytime it spins up it sounds like the resolution is cut in half, it sounds horrible. But the difference can only be heard after the computer has been properly warmed up (huge difference), at least 2 days is needed, but about a week is the best.

[PixyMisa]

No.

Well, I suppose it's possible if you have a failing capacitor on your sound card or something. But mostly, no.

[MortFurd]

The sound cards on all of my PCs sound like crap. There's always audible noise in the background. The CPU load might change the amount and frequency content of the noise, but that isn't jitter. It's just plain old noise. I wouldn't say it sounded "brighter" I'd just say there's more noise.

But then, ES wants to impress us with the knowledge he got from not reading books. Of course, if he'd read them he might understand the terms he throws around.

[rockoon]

100% CPU usages isnt 3% CPU usage.

At 100% you might be throttled (CPU limitted) which could cause jitter. At 3% you cannot be throttled and thus there cannot have jitter related to CPU load.

Additionally, I do not think MP3 produces artifacts in the sense that decoding and then re-encoding with the same bitrate will cause artifacts. The decoded data should contain all the information necessary to produce a duplicate encoding.

MP3 is lossy in a *directed* way. It throws away specific high and low band frequencies in order to reduce the entropy of the sound data.

As a simple example from a different field: Encoding a 24-bit targa image as an 8-bit GIF image will be lossy, but after that you are free to convert back and forth with no additional data loss.

[MortFurd]

100% CPU usages isnt 3% CPU usage.

At 100% you might be throttled (CPU limitted) which could cause jitter. At 3% you cannot be throttled and thus there cannot have jitter related to CPU load.

Additionally, I do not think MP3 produces artifacts in the sense that decoding and then re-encoding with the same bitrate will cause artifacts. The decoded data should contain all the information necessary to produce a duplicate encoding.

MP3 is lossy in a *directed* way. It throws away specific high and low band frequencies in order to reduce the entropy of the sound data.

As a simple example from a different field: Encoding a 24-bit targa image as an 8-bit GIF image will be lossy, but after that you are free to convert back and forth with no additional data loss.

Unfortunately, it doesn't work that way. Each successive encode/decode degrades the audio.

[logical muse]

I'm with ES on this one. Even though FLAC is bit-for-bit identical to WAV, have any of you considered that at a quantum level, if you know the value of a bit, you can't know its position, and vice-versa?

For example, if you know where the bit is, you can't tell if it's a 1 or a 0. Conversely, if you know whether it's a 1 or a 0, then you can't tell where it is.

To make matters worse, according to Schrodinger, the bit is both a 0 and a 1 until it's observed! Of course, we have to define what we mean by 'observed'.

For CDs, is the bit observed when the laser reads it, or when the DAC (or CPU) processes it, or both? In the case of audio files on a hard drive, is the bit observed when the read head of the hard drive passes over it, or when it's transferred to RAM, or when the CPU reads it out of the RAM? What about when it passes through the ALU?

I think it's safe to say that a FLAC bit will be observed more times than a WAV bit, as it has to undergo more processing. The greater the number of observations, the less certainty we can have about its value. See the Shannon-Godel-Schrodinger hypothesis for more information.

Hence, FLAC files, whilst ostensibly the same as WAV, can never sound as good.

[rockoon]

Unfortunately, it doesn't work that way. Each successive encode/decode degrades the audio.

Yes I see that now .. I had thought MP3 used fourier transforms .. it doesn't .. it uses cosine transforms which is inherently lossy in its own right .. sigh

My mistake

[PixyMisa]

I'm with ES on this one. Even though FLAC is bit-for-bit identical to WAV, have any of you considered that at a quantum level, if you know the value of a bit, you can't know its position, and vice-versa?
That's why they use QCC (quantum correcting codes).

We know from the Pauli Exclusion Principle that you can't have two bits with the same value in the same place. So if that's what appears, we know there is an error. By adding quantum redundancy to the audio encoding, we can determine the original value no matter how many errors there might be, by taking the dot product with the next error code in sequence and recursively recalculating the probability matrix for each sample block.

What that means, of course, is the more errors you have, the longer it takes to process the file. That's why the CPU load varies between WAV, FLAC and MP3: Some of the files contain more errors than others.












This is disturbingly close to how hard disks really work.

[moopet]

I have already compared 0% vs 100% CPU load in a blind test and the difference was huge. I have also done a blind test comparing 1 vs 2 harddrives running. I have the harddrive set to go to sleep mode after 3 minutes. But once 10 minutes had passed and I complained of edgy sound, then I looked at the 2nd harddrive and it was still running. The test doesn't get any better than that.

Everytime the 2nd harddrive spins down I can hear the shift to smoother sound. Everytime it spins up it sounds like the resolution is cut in half, it sounds horrible. But the difference can only be heard after the computer has been properly warmed up (huge difference), at least 2 days is needed, but about a week is the best.

Let me get this straight. You warm your PC up for 2-7 days before you use it to listen to MP3s. And you believe that 3% CPU usage or the spinning of a second hard drive in the case will affect the quality of the decoded sound. Not electrical noise picked up in the audio circuitry, but that the digital signals are different.

That right?

[Blight]

If you're really nit-picky about the CPU usage thing, just do what I do, connect your PC to an external receiver over a digital connection (SPDIF). Set the audio-out to send PCM audio digitally (in the audio driver).

Once that's done, the audio doesn't go through a low quality D/A converter (assuming your external receiver is any good).

And yeah, on really crappy onboard audio (and even on some low-end PCI cards), hard disk usage could cause some audio leakage, although I have no idea why CPU usage, maybe if the bios is set to auto-change the CPU fan speed dynamically and when the CPU goes up, so does the fan voltage and it might "somehow" cause some noise. Surely more noise if you take the computer case noise in general into account.

[ExtremeSkeptic]

Let me get this straight. You warm your PC up for 2-7 days before you use it to listen to MP3s. And you believe that 3% CPU usage or the spinning of a second hard drive in the case will affect the quality of the decoded sound. Not electrical noise picked up in the audio circuitry, but that the digital signals are different.

That right?
Yes, I'm not hearing any electrical noise. I'm hearing edginess from jitter. The same type of jitter as a Toslink cable gives.

I have the mainboard, PSU and harddrives isolated from each other with Magix levitation feet so the reason isn't vibration. For both harddrives I'm using the same ribbon cable. So the reason for the edgier sound should be because of more noise from the harddrive entering the PSU. It sounds similar to plugging the computer into the wall instead of a power conditioner.

[PixyMisa]

Get a digital oscilloscope. Make some measurements.