|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
206.255.209.123
In Reply to: RE: Me too! posted by John Atkinson on January 26, 2015 at 11:10:56
four years ago from your native side of the pond. :)Notice that it compares a range of recordings of the same content at different resolutions. Which rules out the "it all is due to mastering differences" canard. It really doesn't matter how many studies are paraded about that don't support the ability to hear differences between Redbook and higher resolution formats. All it really takes is one or two valid ones to establish that it can_be_done.
That not every listener is able to or every piece of music benefits should be obvious to all.
Edits: 01/26/15Follow Ups:
> four years ago from your native side of the pond. :)
> Sampling Rate Discrimination
> Notice that it compares a range of recordings of the same content at
> different resolutions.Thanks very much for the linked article. I think the takeaway from this
and the other AES paper is that if a master recording is made at a high
sample rate, then any reduction of sample rate for commercial release is
going to result in a reduction in quality. So why not, then, release the
recording at the same sample rate and resolution as the master? The end
user will then know that he is getting the full sound quality.
John Atkinson
Editor, Stereophile
Edits: 01/27/15
It's funny how as audiophiles we are encouraged to use high quality recordings of classical performances to differentiate between audio components. But in this case however we are expected to believe the equipment in both recording signal paths (operating at different rates) and a DAC (also operating at different rates) in the playback path are sonically identical when presented with such challenging inputs and that the only differences audible should be attributed to the resolution of the created recordings.As someone who believes that these kinds of musical performances actually may present great challenges to audio performance I'm not really sure what I should takeaway from this kind of test. I guess my problem is I don't have a bias so I'm kind of just left hanging here without anything to takeaway.
I'll go away now.
Give me rhythm or give me death!
Edits: 01/27/15
But in this case however we are expected to believe the equipment in both recording signal paths (operating at different rates) and a DAC (also operating at different rates) in the playback path are sonically identical when presented with such challenging inputs and that the only differences audible should be attributed to the resolution of the created recordings.
No, you are not expected to believe that at all. If the ADC and DAC sounded the same operating at 44.1k and 88.2k rates, then the results would have been null. The result was that recording & playback via matched signal paths operating at different sample rates sounded different.
It's hard to believe you don't understand the point of this experiment. It's like you're arguing just to argue. Your posts are baffling.
" If the ADC and DAC sounded the same operating at 44.1k and 88.2k rates, then the results would have been null."
If that were the case then one could question (and many would including me) whether the components were capable of resolving the differences - it all seems to depend on what point one is trying to make.
" It's hard to believe you don't understand the point of this experiment. It's like you're arguing just to argue. Your posts are baffling."
LOL - as if understanding the point of an experiment requires one to accept the conclusions/suggestions others have made based on the results of it.
Give me rhythm or give me death!
Such a concept would have been unfathomable in the analog tape days or in the earlier years of digital. Today, it's a non-issue in terms of cost or deployment.
"So why not, then, release the recording at the same sample rate and resolution as the master?"
The 64,000 dollar question John.
Not sure the study proves anything other than that the converters and down sampling algorithms may impose a sonic signature. I appreciate the fact that the article itself raises those points.I would think that down sampling a higher rate recording would help negate any effects that the required filtering might have in the audible range.
So based on the results it seems the converters are pretty close but the decimation algorithm didn't work perform so well.
So though the tests prove listeners could hear a difference it does not conclusively prove the reasons for those differences.
Give me rhythm or give me death!
Edits: 01/26/15
Not sure the study proves anything other than that the converters and down sampling algorithms may impose a sonic signature.
That assertion is not supported by what you find in the study.
From page 6 under "4. Discussion", you'll find the following conclusion:
"Listeners could significantly discriminate between files recorded at different sample rates only for the orchestral excerpt, the only recording of a complex scene with different musical instruments playing in a medium concert hall. This finding provides support for theories that high-resolution formats better reproduce the details of transients and room acoustics.
That wording is pretty clear to me and makes sense that it is easier to hear differences using more demanding musical content.
"That wording is pretty clear to me and makes sense that it is easier to hear differences using more demanding musical content."
Hear differences of what? The sampling rates or the converters or the decimation algorithm or something else. One can assume the differences are due to the differing sample rates - but it's an assumption none the less.
Give me rhythm or give me death!
read the quoted conclusion from the study again!
I think perhaps you somehow missed the fact there were a number of different permutations in the comparisons done. I mention only those where no converters were used. Let's take one more look at the conclusion:
...files recorded at different sample rates
Get it this time?
Sure several files recorded at 2 different sample rates - not 2 files one converted from the other (as you seem to believe I am suggesting - the decimated file being the other, third, case). From figure 2 I see two different converters, 2 different clocks and two different recorders recording at different rates. Everything from the placement to the connectors in this set up should be suspect. Additionally from the verbiage we know a single converter operating at 2 different rates is applied at playback - we are supposed to assume identical performance at both rates during recording and playback.
The fact the results agree with your expectations is enough for you to accept the assumptions. It's too much for me.
From first hand experience - good luck with that 2 identical component idea and further had no differences or the wrong differences been heard in with this setup the first thing that would have happened is the setup would have been assumed to be wrong. Changes would then have been made until the desired results were achieved. Usually this is an honest attempt to get at the truth as flawed as it is - and it actually kind of works when the results are already known in advance and expected. Not so good when trying to present a proof of something.
And even if I did accept the veracity of these test results, it hardly matters when we are talking about mass marketed recordings to be played back on a portable player now does it?
Give me rhythm or give me death!
Sure several files recorded at 2 different sample rates - not 2 files one converted from the other
That's the whole point. Get it?
From figure 2 I see two different converters...
It's the same model converter running at a different sample rate.
2 different clocks
At the expense of stating the obvious, one needs a different clock to operate at 44.1 vs 88.2.
and two different recorders
Once again, it is the same model recorder running at the rate of the two sources.
It's too much for me
Obviously, Goob.
it hardly matters when we are talking about mass marketed recordings to be played back on a portable player now does it?
Once again, you've introduced not one, but two straw men - not all recordings found in high resolution are "mass marketed" nor do the vast majority of folks who listen to high resolution recordings use a "portable player".
"At the expense of stating the obvious, one needs a different clock to operate at 44.1 vs 88.2."
Most likely the converter used the exact same A/D circuitry running at the exact same master clock rate, and used different DSP algorithms to downsample the original data stream to either 44.1 or 88.2. In the playback, your DAC will most likely upsample whichever format back to its master clock rate and play that back. You are hearing ae the differences caused by software, as dictated by the mathematical limitations of the low-res format.
For this reason, the easiest and best way to compare formats is to use the best available software converters to convert a high res format to a lower resolution format and then back up to the original format. Any problems that you hear will be due to the limitations of the converter software or the format. If you use the best available algorithms you will discover, as I have, that the limitations come from the filters. One can try various filter settings and none of them will be transparent. One will lose something, according to the filter settings, e.g. tonality, air, transients, imaging, soundstage. Most of the filter tradeoffs hit in the downsampling process. If these are done right, then the playback filters don't matter much. If these are down wrong there will be aliasing distortion that can not be separated from the music.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
the value of upsampling and downsampling to determine whether or not a higher resolution format has merit over Redbook. If memory serves, you find Redbook severely lacking with your recording.
Why not do as this test did and simply record a single event in multiple native resolutions?
If it's a live event then it would involve running multiple converters in parallel. If it's a master tape transfer, then one could play the tape twice, and presumably there wouldn't be any tape wear involved. But if someone else did this, you have no idea what they did. There is a lot of secret sauce involved in mastering, no matter what people may claim to do or not do.
My comments were directed to people who want to KNOW what the differences are with regards to format conversions. The only way to know this is to do the format conversions oneself starting with the same recording and then playback with the identical playback chain. This has the further advantage that by experimenting with different formats and different filter parameters one can train oneself to hear all of the various artifacts involved. This training will make the various "tests" that others come up with relatively easy. I spent the better part of 100 hours doing these experiments some years ago. In the end, I concluded that it was all a waste of time. The lower resolution formats were lower resolution no matter what I did.
The 44.1 kHz format could not transparently reproduce the output of my cassette player.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
it will answer all your questions and eliminate the need for speculation.
Thanks. This answered my questions. I note that they used live acoustic music with decent venues, so it is not surprising that they got positive results, even though listeners had not been specifically trained to discriminate sample rates. However, they did not fully comply with the ITU recommendation for blind tests. [ITU-R BS.1116-1, ITU-R BS.1534-1, ITU-R BS.775-2] Had they used specifically trained listeners or used a preliminary stage to select the best scoring listeners they would likely have gotten much greater statistical significance.
The paper didn't give a specific model number for the RME ADC that was used, so I could not investigate further as to the specific ADC chips used. In all likelihood, they used a multi-bit sigma delta modulator that sampled at around 5 Mhz. The converters therefore downsampled to whatever PCM formats were chosen. As far as I know, all modern audio ADCs that output PCM downsample internally. The differences between downsampling in hardware and software will be small and will depend on details of the DSP involved. There is nothing lost in principle in following a two step process downsampling over a one step process except loss of 1 bit of resolution from the original output format. Given that the RME outputs 24 bits, this would mean restricting the precision to 23 bits, which if reduced to 44/16 would be irrelevant. In the case of reducing to 44/24 it might possibly be relevant, however the 88/24 format already has more resolution in the range 0-20 kHz because of the higher sampling rate.
One might have made the 44 kHz playbacks sound better by using different filtering in playback. Apparently, the experimenters just used the filters provided by the RME Fireface DAC.
IMO this was a good paper, great in comparison to horrible examples such as Meyer - Moran. It could have been better. One thing that I did not like is the use of second tier converters. Better results could have been achieved by taking the converters out of the picture and using file level conversions. This would allow extensive experimentation on how the different possible filter designs affect sound. It would also ensure that all of the analog equipment operated identically in all cases and the only differences being the DSP processing. Furthermore, when doing DSP processing file to file it is possible to capture that impulse response of the filters very precisely, providing much better documentation of the experimental procedure.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
All in all, I found it used a far more plausible setup than any number of contrived ones you find (like the M&M).
"That's the whole point. Get it?"
Laugh - all along."It's the same model converter running at a different sample rate."
Sure but what guarantee is there that they, the ADCs, sound/work the same? It's not unusual for professional test equipment to give different results. In fact it's not unusual to find a single piece of test equipment that gives different results when used in different locations using the same test interconnects/leads - and even easier to find when using different interconnect/cables or leads of the same manufacturers model.And the DAC, ie. the playback device, is the same unit operating at different rates. What guarantee do we have that it sounds the same at both rates? In fact the test/evidence/results would equally well prove that the DAC, sounds different at different rates - if one wanted to assume the both samples (of the dual conversion) sounded the same. No different than assuming the DAC sounds the same at two different rates - which is what we have do in order to support the printed conclusion of this test. Or that the decimation and several other factors play no part in the differences heard. We believe what we want to believe.
"At the expense of stating the obvious, one needs a different clock to operate at 44.1 vs 88.2. "
Yea but again you assume the device(s) functions equally as well at both rates.
"Once again, it is the same model recorder running at the rate of the two sources."
And again you assume the recorders are equal and that they operate equally well at both rates.
Obviously, Goob."
Sorry you need to resort to misquoting my comment in order to try to score a point!"Once again, you've introduced not one, but two straw men - not all recordings found in high resolution are "mass marketed" nor do the vast majority of folks who listen to high resolution recordings use a "portable player"."
It wasn't my intention to introduce a straw man - it was you who opened the topic on this test in this thread about a portable player.
Give me rhythm or give me death!
Edits: 01/26/15
and yet you scored twice!
"and yet you scored twice!"
LOL - but surely you've called me names many more than 2 times. How about you declare yourself the winner and move on?
Give me rhythm or give me death!
Sure several files recorded at 2 different sample rates - not 2 files one converted from the other (as you seem to believe I am suggesting - the decimated file being the other, third, case). From figure 2 I see two different converters, 2 different clocks and two different recorders recording at different rates. Everything from the placement to the connectors in this set up should be suspect. Additionally from the verbiage we know a single converter operating at 2 different rates is applied at playback - we are supposed to assume identical performance at both rates during recording and playback.
The fact the results agree with your expectations is enough for you to accept the assumptions. It's too much for me.
From first hand experience - good luck with that 2 identical component idea and further had no differences or the wrong differences been heard in with this setup the first thing that would have happened is the setup would have been assumed to be wrong. Changes would then have been made until the desired results were achieved. Usually this is an honest attempt to get at the truth as flawed as it is - and it actually kind of works when the results are already known in advance and expected. Not so good when trying to present a proof of something.
And even if I did accept the veracity of these test results, it hardly matters when we are talking about mass marketed recordings to be played back on a portable player now does it?
Give me rhythm or give me death!
One was downsampled from the 88 kHz file, the other by recording with the same model ADC as the 88 kHz file but with a different clock speed. Right?
How would you modify this test to address your concerns? And what evidence would you accept that, indeed, the perceived differences are due to the sample rate rather than the other factors you imagine?
Not sure I need to keep repeating this same thing over and over again-
"Not sure the study proves anything other than that the converters and down sampling algorithms may impose a sonic signature. I appreciate the fact that the article itself raises those points.
I would think that down sampling a higher rate recording would help negate any effects that the required filtering might have in the audible range.
So based on the results it seems the converters are pretty close but the decimation algorithm didn't work perform so well.
So though the tests prove listeners could hear a difference it does not conclusively prove the reasons for those differences."
Give me rhythm or give me death!
Every digital recording requires an analog to digital converter. A single one was used for each version.
From the mic feed, the analog signal was sent to either a 44.1 ADC or an 88.2 ADC of the same type (naturally) running at a different clock frequency.
it hardly matters when we are talking about mass marketed recordings to be played back on a portable player now does it?
Many of us choose to listen to our high resolution recordings on devices other than portable players. :)
"From the mic feed, the analog signal was sent to either a 44.1 ADC or an 88.2 ADC of the same type (naturally) running at a different clock frequency."
Yes but what guarantee do we have that both units have identical performance? Also the physical location of any component can be relevant and the connections between them and the rest of the system are different as well.
For the DAC how do we know it sounds the same at both sample inputs?
"Many of us choose to listen to our high resolution recordings on devices other than portable players. :)"
Great and a good thing. Doesn't change the fact that this test asssumes both signal paths are identical and that the down sampling algorithm has no sonic effect.
Give me rhythm or give me death!
Yes but what guarantee do we have that both units have identical performance?
It's the same model unit. You're really stretching here.
Also the physical location of any component can be relevant...
Say what?
Doesn't change the fact that this test asssumes both signal paths are identical
They are from the same mic feed.
and that the down sampling algorithm has no sonic effect.
There is NO down sampling. Only recordings sourced at different resolutions.
Get it?
It's the same model unit. You're really stretching here."
Oh?
Say what?
Also the physical location of any component can be relevant...
"They are from the same mic feed."
So? They follow different signal paths.
"There is NO down sampling. Only recordings sourced at different resolutions."
There's 3 samples in the test - 2 created by the recorders and one created by a down sampling algorithm on one of them.
Give me rhythm or give me death!
the Goober who complained:
Sure several files recorded at 2 different sample rates - not 2 files one converted from the other
might need to inform the other Goober who complained:
...and that the down sampling algorithm has no sonic effect.
that the discussion remains about the native recording results:
"To test this hypothesis, we recorded five different musical excerpts, each presented in three different formats: 44.1 kHz, 88.2 kHz and the 88.2 kHz version down-sampled to 44.1 kHz...
"Listeners could significantly discriminate between files recorded at different sample rates only for the orchestral excerpt, the only recording of a complex scene with different musical instruments playing in a medium concert hall. This finding provides support for theories that high-resolution formats better reproduce the details of transients and room acoustics."
Either you understand the concept or you don't. Have a nice life!
Lab test(your quote)-
"Listeners could significantly discriminate between files recorded at different sample rates only for the orchestral excerpt, the only recording of a complex scene with different musical instruments playing in a medium concert hall. This finding provides support for theories that high-resolution formats better reproduce the details of transients and room acoustics.
"My point -
"Not sure the study proves anything other than that the converters and down sampling algorithms may impose a sonic signature. "
You say -
"Either you understand the concept or you don't."
I understand the concept but I question whether or not this study "...provides support for theories that high-resolution formats better reproduce the details of transients and room acoustics."
There's no need for your disingenuous paraphrasing, nastiness and name calling. Either one buys into the suggestion that this test supports the theories of high resolution formats as better reproducers "..of transients and room acoustics" or they don't. You buy it - I don't (at least not completely) and I laid out my reasoning for my position.
Give me rhythm or give me death!
Not sure the study proves anything other than that the converters and down sampling algorithms may impose a sonic signature.
There was no downsampling involved
Don't you remember Goob #1's complaint? You must be Goob # 2.
They compared two native format recordings. Remember?
Is it past your bedtime?
They compared 3 "recordings". 2 recorded at the same time and 1 down sampled via a sw algorithm. The second lower resolution file was created by removing samples from the higher resolution file.
Creating a lower rate set from a higher rate file is DOWNSAMPLING. In fact it's the listening results of this decimation that cause me to question the test setup/methodology.
Please continue to state the obvious - maybe sooner or later you'll understand what you are trying to talk about.
Give me rhythm or give me death!
There were actually five recordings made in three different formats.
All of my comments have been about the native recordings. Do you understand?
You, and the author, seem to want to pick out a part of a piece of the test and hold it up as evidence of something. On the contrary the whole of the test suggests no differences due to higher resolution formats.
I have no doubt, in the special cases you rely on, that people actually heard differences. But I've made it clear throughout this thread there's no evidence proving that the format differences (as opposed to set up or equipment differences) are the reason for these differences.
You are free to believe whatever you want.
Give me rhythm or give me death!
that 44 kHz files were created both by downsampling the 88 kHz files and by recording separately at 44 kHz - all with the same recording chain/converters, etc. I don't see how you can possibly control for any other variables.
Compelling study!
"I don't see how you can possibly control for any other variables. "
Yet you call it a compelling study? We believe what we want to believe.
Give me rhythm or give me death!
"To test this hypothesis, we recorded five different musical excerpts, each presented in three different formats : 44.1 kHz, 88.2 kHz and the 88.2 kHz version down-sampled to 44.1 kHz. "
Very interesting. In my own trials, I can distinguish 24/44 from 16/44 but not 24/44 from 24/96 or 24/192. Nevertheless, I'm happy to read accounts of others who can hear the differences.
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: