AIFF Vs. WAV

[Paul R]

Yep, I do believe you have nailed this one! That has to be it!

 

ROTFLMA!

 

-Paul

 

 

[Paul R]

First, please be aware that these are unduplicated and unconfirmed results. We have to find a way to sneak a duplicate run in on this, and then get some other people interested enough to duplicate the testing.

 

Are you saying there was a frequency response difference between two files known to contain identical data? ? Do you recall the magnitude (in dB and ideally, bandwidth) of the change?

 

Yes, though again, with the caveat that it is currently unconfirmed, and needs to be duplicated elsewhere.

 

The magnitude was far less than a DB, somewhere around 1/8th of a DB. I would have thought that inaudible, but apparently not.

 

don't understand the part about "speakers that were miked in a sound chamber" part. Was this comparison of two different recordings of the .aif and .wav files? or of the .aif and .wav files themselves?

 

Remember, this is in playback, all the way to analog, not in the digital domain at all. The speakers were in a special chamber, though I am not sure that the chamber was perfectly anechoic. I think it was supposed to be. Another issue could be the microphones, about which I do not know much at all.

 

We ripped (using XLD) Don't Let it Show from I Robot, and sixty seconds of pink noise, both into AIFF and WAV format. at 16/44.1.

 

Then we converted (using XLD) a copy of the AIFF files into WAV files, and a copy of the WAV files into AIFF files, for a total of eight samples.

 

We saw the spike on all the WAV samples. The WAV samples had the extra energy, whether converted from AIFF or a direct RIP.

 

The music data in each file was exactly the same.

 

Since we did this when the opportunity beckoned, let me again hasten to note that until someone else can duplicate this, perhaps under better controlled circumstances, the results are wide open to question. The more and wider the question, the better. But I think we were pretty careful and the results will duplicate.

 

 

 

 

 

[elcorso]

I cancelled my subscription to this thread, but here I'm again...

 

Hi William,

 

…5. Audiophools love small amounts of harmonic distortion. It gives that analogue tube warmth feel,…

 

Form Wikipedia:

"...Some sonic qualities are easy to explain objectively based on an analysis of the distortion characteristics of the gain device and/or the circuit topology.For example, the triode SE gain stage produces a stereotypical monotonically decaying harmonic distortion spectrum that is dominated by significant second-order harmonics making the sound seem "rich" or even "fat", while each higher order harmonic is smaller than the nearest lower order harmonic..." (Bold is mine).

 

Other descriptions says second order harmonic distortion (no also first order as measured by Paul) is 'kind' to the ear. I own an tubed amp, then it should be redundancy if I find WAV more accurate. And, or, modern tube Amps. doesn't exhibit this kind of 'second-order harmonic distortion'.

 

Hi Barry,

 

…I remember listening to a DAC with a friend who does hear differences between .aif and .wav and wondering why he didn't notice the ear-ripping treble on the DAC.

 

I own two DAC's listed on my gear bellow, and have an 'open access' to a BADA (from a friend who travels a lot). Then, in my case, I don't think this is a DAC 'sound' situation.

 

My comparison, on my previous reply to this thread, where I find accurate the Marimba sound in WAV, was against the real, unamplified and life Marimba sound. This is one ripped CD only, but I made a lot of test against a lot of musical instruments. I like and listen to life and unamplified music, like guitar, etc.

 

I didn't talk on the Marimba sound about warmth, but about harmonic content and overtones. Even the Marimba sound can't be warmth, you can't listen to it for a long time, since his overtones get you nuts after one hour. A cello and a violin (between others) can be warmth.

 

I remember, from another thread a @Miska commentary, regarding (iTunes AIFF) have some kind of compression, but I can't find the post, nor the thread.

 

I had being a Mac lover and user all my life, and my feeling about Mr. Gates is not precisely love, I only love WAV.

 

Best regards,

 

Roch

 

[bdiament]

Hi Paul,

 

Hmmm. It sounds like playback from loudspeakers was recorded.

That would make me wonder; too many variables. Perhaps the mics were "warmed up" for one recording and not another.

 

"...The magnitude was far less than a DB, somewhere around 1/8th of a DB. I would have thought that inaudible, but apparently not. ..."

 

If it is a broadband lift, I would not be surprised at audibility. I've often shown clients just how audible a 1/4 dB boost at 20 kHz can be.

 

"...We saw the spike on all the WAV samples..."

 

Wait. A spike? Do you know the bandwidth (or "Q")? A spike sounds (to me) like something very narrow. Hearing a 1/8 dB spike is something I'd have to witness (in a proper test) to believe that was the reason for perceived differences. (It may be possible however, that the ringing engendered by a narrow spike might be the culprit.)

 

Try it. Get a good parametric EQ (for example, maybe you can download a demo of ChannelStrip) and try applying 1/8 dB at 960 Hz. Then vary the bandwidth from narrow to wide.

 

***

Overall, an interesting event but as you mentioned, the test needs to be done again and I'd suggest tighter control over the process.

 

Best regards,

Barry

www.soundkeeperrecordings.com

www.barrydiamentaudio.com

 

 

[goldsdad]

@Paul

 

"First, please be aware that these are unduplicated and unconfirmed results. We have to find a way to sneak a duplicate run in on this, and then get some other people interested enough to duplicate the testing."

 

Please, please, please run these tests using Audirvana in "memory play" mode, as well as any other players you intend to use.

 

Unless Audirvana has file format decoding bug(s), there should be no format-correlated variation in output when using it as a "memory player".

 

In contrast, when using other "memory players", format-correlated variation in output is understandable (and I fully accept that my inability to hear it is irrelevant) for the reason that I tried to explain in previous posts.

 

 

 

 

 

[Paul R]

I am probably using the wrong terms. It looked like an anomaly to me, and thus I used the word "spike." The Q, if I understand the term right, was the plus/minus 10% I mentioned - about 950hz to about 970hz, which I assume is a pretty narrow band.

 

Also, it is probably not coincidental that 960hz is a harmonic of 60hz. Still, why it was there only in the WAV samples, and not in the AIFF samples is what caught our attention.

 

To be honest, none of us expected to see any differences at all, and finding one was a three beer shock, and one we are all very suspicious of. I would like to get it in a rigorous test program in a lab more suited to music than the one we used.

 

I'm actually kind of sorry I mentioned it, since it is obviously three guys taking advantage of some free lab time, not a well thought out workmanlike effort.

 

I have to stop thinking of this like a hobby, and take it more serious I guess. (grin)

 

-Paul

 

 

[thesurfingalien]

I ran an RMAA test using the original RMAA test.WAV file and a .FLAC copy of it.

 

For this test I used an E-MU 1820 with loopback cable.

 

To play both files I used Foobar2000 talking to the 1820 in ASIO, and using RMAA in record-mode for 44.1/16

 

The machine I use is an older P4, running windows XP

 

On both occasions the test results we completely equal.

 

I would like to do the same if I could find a tool to convert WAV to AIFF (not being iTunes :-)) and see how that comes out.

 

 

 

[Paul R]

User XLD on a Mac, or dbPowerAmp on a Windows PC.

It will be quite interesting to see.

 

How are you checking the playback of the files?

 

-Paul

 

 

[thesurfingalien]

I take it the User XLD on a Mac etc. was directed at me?

 

RMAA is a piece of software that does various checks on an audio signal. Please visit http://audio.rightmark.org/ for info. It will take quite long to explain how it works.

 

Basically one plays a series of test-tones, and records these. RMAA analyzes the resulting WAV file, comparing the original signal.

 

Edit1: In this case I looped back the playback to the recording channel.

 

Many sounds-cards and the likes come with a RMAA report.

 

Edit2: However, this test might not be accepted by some as it might not contain the proper method for testing :-) Another option is to have a tool that extracts the data from the audio-buffer that the hardware-driver uses to feed the hardware. By dumping this data in 2 files and compare these, one should be able to tell if the data that feeds the hardware is the same in both cases.

 

Regards,

Peter

 

[Paul R]

I meant, "use XLD on a Mac or dbPowerAmp on a PC to convert a WAV file to an AIFF file."

 

It was just in answer to your last question.

 

I would tend to just loop a digital signal into a second machine and record the input. If it is bit for bit perfect, I know I have bit perfect transmission.

 

I think it has to be converted to analog and measure the analog signal to find any differences that may exist. Staying totally in the digital domain has not revealed any differences in the files, just the opposite in fact.

 

-Paul

 

 

[thesurfingalien]

Actually the test is done with analogue loop-back :-)

 

[thesurfingalien]

/I think it has to be converted to analog and measure the analog signal to find any differences that may exist. Staying totally in the digital domain has not revealed any differences in the files, just the opposite in fact./

 

On second thoughts...

 

If one performs a digital loop-back on an audio-device that is capable of doing so, the data can be compared.

 

If that turns out to be exactly equal for both loop-back of a WAV vs. another loss-less file format, it means the audio-device actually receives the same data regardless of the (decompressed / converted) source-file used for playback.

 

In that case, conversion of the data by the DAC will result in the same output with no differences whatsoever.

 

However, if that is not the case, there must be a step within the software that alters the digital data before it gets to the hardware. At that point, audible differences could be heard.

 

Regard,

Peter

 

[Paul R]

And a full digital loop does exactly what you posit, received the data bit for bit perfect. Moreover, the files contain identical musical data before they are processed and transmitted.

 

The only possible operator I can think of that would explain it is in the software that converts a AIFF or WAV file to pure PCM, and that's a might big stretch, as there is not that much to convert.

 

It gets even worse if you consider people are convinced they hear differences in identical files produced by different ripping software, ignoring any possible errors.

 

I'm kinda thinking that Barrows or Forrest (someone) nailed it eariler in the thread, with pinning the differences down to different hardware or hardware combinations. Chips are complex, and can exhibit really odd behavior under just the right conditions.

 

Another nasty little rabbit hole. Well, they wouldn't pay us if the problems were easy, would they?

 

Oh, wait...

 

-Paul

 

 

[Teresa]

audiophile65 said, "Would love to hear your opinions and experiences between WAV and AIFF."

 

goldsdad said "In my interpretation, that certainly does not clearly imply that he or she wants only to hear from people who hear differences between WAV and AIFF."

 

My interpretation differs from yours in that I feel if one has opinions and experiences between WAV and AIFF, then by logical extension then one does not believe they sound the same otherwise it would be impossible to have opinions and experiences between WAV and AIFF as one cannot differentiate the differences between things that sound the same to them. Thus this thread was aimed at those who hear differences between WAV and AIFF not those who believe they sound the same.

 

I previously said "Also it is uncalled for attacking people who hear differences because it does not fit into someone's belief about digital dogma."

 

goldsdad said "Now, correct me if I'm wrong, but I see that statement as being aimed at me since Forrest has been so upset by my comments.

 

Please explain how a technical discussion or the way in which I presented my viewpoint is "attacking people".

 

The comment was not directed at anyone in particular but in general. Telling people that their preferences are all in their imaginations and demanding proof from people who could care less and just want to enjoy their music in the best sound possible is in bad taste. Or that it is a Apple versus Microsoft thing, when at least two of us use Mac's and actually prefer the sound of Microsoft's WAV. Sorry I find all of these demands and insults as personal attacks. I find it much easier for each listener to actually listen to what they like, with no need to justify their choices. But that's just me.

 

BTW reread the original post NO technical discussion was asked for.

 

 

[goldsdad]

Thanks for contributing your point of view to this thread.

 

 

 

 

[thesurfingalien]

Perhaps it is best if we ask the OP if he or she has a problem with the technical discussions?

 

Regards,

Peter

 

[thesurfingalien]

I am not sure if I am clear in what I wrote, but if one would loop-back and capture the data at the sound device (I can do that with my E-mu) than one can check what data ends up at the DAC-IC itself. Unfortunately, there are only so many devices that can do that, and I am not aware of on-board sound-devices that support this functionality.

 

I must say that I can not think of any hardware-issue within the computers you used for the test, neither can I see any evidence for data purposely/non-purposely being altered (the deviation you measured seems to indicate that as it is rather specific) by the audio-codec / player unless a bug in the codec is at hand, perhaps even in some versions of it. I admit that might be far-fetched, but anyway.

 

Regards,

Peter

 

[Paul R]

Yes, I agree with your assessment, and would sure love some way to be able to get the speaker/mics out of the picture, but not sure what equipment would be appropriate for that.

 

I've considered it might be a flaw in the codec, or the implementation of the codec within XLD, but have not figured out how to disprove that yet.

 

Like I said, another apparently bottomless rabbit hole, just packed with conflicting evidence. I'm personally convinced some of the folks here are really hearing a difference, and I know Karen can. I can't, but then, I have some shaky experimental evidence that there is a difference in the playback, at least in some limited situations.

 

If the damn difference would be consistent across all machines, I would be a lot happier. And if wishes were horses...

 

Maybe we can get somebody like John Atkinson interested in a better designed experiment with his much better measurement tools.

 

-Paul

 

 

 

 

 

 

[thesurfingalien]

Hi Paul,

 

I actually find your experiment not too shaky at all as you performed the test on multiple computers. Unless an error has been made during testing (for example, playing the same file twice on the computer that did not show the deviation), it shows we are talking about an anomaly.

 

In my opinion, we are talking about an anomaly to begin with as both containers should have exactly the same data representing music (according to specification, they both store the music as PCM). For what I know, Apple just created room for metadata to be included, that's all. The byte-order for the file-format is of no influence. Therefore, the conversion should be rather easy in both directions. If here is indeed a flaw somewhere in this process, we are talking about a mayor ****up programming-wise. I find it extremely hard to believe that is the case.

 

I believe the only way to really see what is going on is by a digital loop-back on the audio-device itself. Only then we can show data received by it is indeed the same.

 

Maybe I am digging in to deep, but I find it rather disturbing that, for example, Teresa is "forced" to settle for a format that apparently sounds less good while it by all means should not make any difference whatsoever.

 

As we are strictly talking digital domain here, I would say that this riddle, with some effort, is solvable.

 

Regards,

Peter

 

[Teresa]

Paul Raulerson said "As for measurements, there really is no physical phenomena that we cannot measure, including anything acoustic."

 

Really? We can only measure the most rudimentary stuff such as frequency response, noise, distortion and timing errors. No one has yet discovered how to measure soundstage, image width, image height, air between instruments, ambiance, smoothness or roughness of string tone or timbre accuracy, just to name a few unmeasurable parameters of sound.

 

At an AES conference in 1973 Dr. Matti Otala presented a paper describing a distortion that he termed Transient Intermodulation Distortion (TIM). Before that we could not measure TIM because we didn't know it existed. It's discovery greatly improved the sound of transistor amplifiers.

 

It was the discovery of how to measure digital timing errors "Jitter" that improved the realism of digital we all enjoy.

 

So many measurement parameters yet to be discovered and each new discovery will help us make better sounding components and recordings.

 

So perhaps it is something that we cannot measure yet that accounts for the differences many of us hear between WAV and AIFF even though they are bit for bit identical.

 

[thesurfingalien]

Although I am not completely sure about the way the terms you use should sound like (a lot of people use specific terms, but attach their own meaning to it), I believe most of it can be measured by measuring TIM (if I recall correctly that it is another term for "slew-rate").

 

I am not sure about this, but I seem to recall that jitter was a known and measurable phenomenon at the time digital audio was marketed. However, the effects were though to be inaudible thus unimportant.

 

As for the digital part in audio you are referring to I fear you lack fundamental knowledge about the subject. Unfortunately, this is rather common. Digital audio simply does not translate in an analogue fashion to the differences you and others hear. If something goes wrong in the digital world, it will sound like clicks, plops, silences or other rather unpleasant noises. Basically audio-garbage

 

Regards,

Peter

 

 

 

[elcorso]

Teresa you must understand that the digital (music) world is perfect!

 

Each (musical) bit is perfect, everything is bit perfect, by the way, there is an app. that his name means this (an all): BitPerfect.

 

Then you should don't care do nothing in the digital (music) world: All the app's playing digital music, all digital cables, all HD drives, all CD players, all DAC's, all interfaces, all music formats,

[Paul R]

You must be on some heavy pharmaceuticals!

 

I prescribe a dose of something silly - listened to Kermit Unpigged lately? Highly recommended!

 

-Paul

 

 

[thesurfingalien]

You beat me to it :-)

 

Oink, oink

Peter

 

 

 

[Gregor Samsa]

No,

 

Roch is right. We know how to measure already everything that could possibly have an effect on sound. Our frequency response is flat across the audible spectrum. Our TIM is inaudibly low. No abx study has ever shown that jitter is audible, so it isn't! No abx study has ever shown that a Levinson amp is better than an Aiwa receiver, so it isn't! No abx study has shown that my Turtle Beach Amigo DAC is inferior to a BADA or Ayre, so those of you who have purchased them are JUST IMAGINING THINGS! I know better than you do what you are hearing. Capacitance, inductance, resistance - other than these all wires are the same, and none of those things make any difference anyway.

 

Me, I have a hard time hearing the difference between files that have been losslessly compressed and those that have had no compression applied. For reasons that escape me now, this has not led me to the conclusion that this must be true for everyone else as well.