Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Jon Hodgson]

squeegybug wrote on Thu, 22 December 2005 10:25

Jon Hodgson wrote However what is NOT possible is for this additional accuracy to survive a SRC unscathed. Now it is perfectly possible for a converter design to have certain inadequacies which mean that it is better to take its output at 96kHz and then downconvert it using an additional unit or software (that's actually what is going on inside the converter anyway), but this would be a result of the physical implementation of that circuit or the internal processing, and not some magical circumvention of Nyquist. And of course that slight, but obvious, degradation from the SRC is apparent between the original 96 and the 96-resampled-to-44.1 files that I posted, and Ivo reports similar observations. I find it interesting that our different converters and SRC still both produce superior (subjective) results at the higher rate.  I guess the mathematical hunt here is for any common ground that will explain this... or perhaps this would actually prove to be the case with all converters, except those with obvious deficiencies at high or low rates.  Sounds like a difficult project. Steve

What sample rate conversion are you using? Also why did you also dither down to 16 bits?

[maxdimario]

We need real samples of real 44.1 and real 96K with no compression, on the same converter, the same clock, and same signal path....HELP!

the amount of technical observations one can make about a system, increases with the amount of knowledge at his disposal...that does not mean that the technical explanations are to the benefit of the system!

only that which noticeably improves the output of the system is worth discussing and elaborating.

These arguments will never ever end unless some samples are posted...

[Jon Hodgson]

squeegybug wrote on Thu, 22 December 2005 09:38

So that is where I would expect the discussion to focus -- as the OP asked, why do the higher rates sound better?  But I would also think this is a question that has surely been asked since the beginning of analog to digital conversion?

But you can't start by asking that question whilst giving two identical rate files as your evidence, it's a non-sequitur.

[squeegybug]

Jon Hodgson wrote

What sample rate conversion are you using? Also why did you also dither down to 16 bits?

I used the integral AD converters and DSP in my Akai DPS24 digital mixer/recorder.  I dithered the files simultaneously with SRC, in order to obtain a representative CD production sample, and see how the entire process affected the originals.

Got to step away for a while, back later...

Steve

[Ivo]

Jon Hodgson wrote on Thu, 22 December 2005 12:30

But you can't start by asking that question whilst giving two identical rate files as your evidence, it's a non-sequitur.

For sure you can ... If we discuss whether for example a green tea tastes better if we use boiling water or just 90 degrees C water when preparing it, we will certainly not taste it in these original temperatures to judge, but we will wait till it reaches a drinkable temperature (the same in both cases) and then we may perceive some differences.

The same with our discussion - it is quite irrelevant how the sample sounds in 96k because nobody is usually able to listen to this sample rate with present common home devices. We have to cool it down somehow and then serve. And ask - does the sound originally prepared in 96k taste better than the sound prepared in 44k ? This is the point. If the difference remains only when listening in 96 k and then disappears, then it has no practical meaning.
If, after cooling down, some sound differences can be clearly heard (and I say yes, they are definitely there),it has a real meaning and the differences are directly caused by the original "temperature". Whether 44k "tea-kettle" is well designed in this case or not is another question.
I think it is quite simple in practice, but of course, the nets of theoretical reasoning can spread mist around anything.
As for the "performance differences" - yes they are naturally there, but try for example to listen first to both samples in 44k (they are really a bit different performance-wise) and then to other two samples originally recorded in 96k. Whatever tones come, these two have generally slightly different sound flavour as such (clarity, depth, intimacy, what is played does not matter). And this is the point we are talking about. At least I hear a difference, irrespective of "different performances" (which are not that much different IMHO).

[kraster]

Ivo wrote on Thu, 22 December 2005 14:39

The same with our discussion - it is quite irrelevant how the sample sounds in 96k because nobody is usually able to listen to this sample rate with present common home devices. We have to cool it down somehow and then serve. And ask - does the sound originally prepared in 96k taste better than the sound prepared in 44k ? This is the point. If the difference remains only when listening in 96 k and then disappears, then it has no practical meaning.

I think it's very relevant how we sample sound at 96k or other rates and then present it for analysis. The thread title is "Why recording in 96k through Lavry Blue sounds clearly better ?". Therefore we must hear the sounds at 96k. Otherwise the thread title should be "Why recording in 96k and then downsampling to 44.1khz through Lavry Blue sounds clearly better ?"

I listened to the samples and the most obvious thing to me was differences in the timbre of the instruments as a result of a different performance. There are a lot of factors at play here: variances in performance, the downsampling process and your own knowledge of what sample rate was being used. All of these have a bearing on the sound.

I don't mean to be disrespectful and I truly appreciate the effort you have put in.

Best regards,

Karl Odlum

[squeegybug]

kraster wrote

I think it's very relevant how we sample sound at 96k or other rates and then present it for analysis. The thread title is "Why recording in 96k through Lavry Blue sounds clearly better ?". Therefore we must hear the sounds at 96k. Otherwise the thread title should be "Why recording in 96k and then downsampling to 44.1khz through Lavry Blue sounds clearly better ?" I listened to the samples and the most obvious thing to me was differences in the timbre of the instruments as a result of a different performance. There are a lot of factors at play here: variances in performance, the downsampling process and your own knowledge of what sample rate was being used. All of these have a bearing on the sound.

I can add nothing to Ivo’s superb analogy, it is valid and in fact the exact reason I also resampled to the standard CD resolution.

I understand your point about analytical data submitted for review.  If it was indeed possible for the resampling process to “improve” the original audio, then I would agree that only raw information would be acceptable for discussion.  However, of course we know this is not reasonable – the SRC could (and does) slightly degrade the audio resolution, but in no way can it enhance it.

I will certainly submit my original raw 48k data alongside the raw 96k file previously posted.  To me, the relative differences remain as in the resampled 44.1 clips.

Would that I, or Ivo, or any musician, could actually increase stereo imaging or add more depth or smooth out harsh tonal transitions of actual physical instruments (or voices) – only from the simple notion of just “playing differently”!  Of course the styles or presentations can change – but, the overall organic structure of these instruments preclude creating something that is physically not there.

In my opinion, the opposite result is actually the point here – I contend that there is already further detail available in the music than what is able to be captured by the 44.1 (or 48) resolution.

In my case, it was quite welcome to hear the comparable ease of the performance in the higher sampled version.  I can accept that as a fact, given my own 40+ years experiences with my own instruments and voice.  While I understand the reason for this forum is to explore technical explanations of physical phenomena, let’s not dismiss the submissions of evidence for the wrong reasons here.

Steve

[kraster]

The SRC can add stuff to the original that may be construed as more euphonic. People like to add distortion to a signal to make it more musical. This may or may not be the case in this SRC. But it could feasibly be a part of the equasion. Without having the original 96k file it's impossible to tell. Also, in order to compare two signals accurately they have to be identical. Not two different performances. I could say , for example, that I prefer the performances in the 96k tests.

There are many factors at play hear. Governed by not just what we hear but what kind of emotional response that test like this will bring about. This is what makes subjective analysis so prone to Bias.

I really don't want to knock any effort that has been put in by anybody. It is much appreciated. And forgive me if I sound ungrateful.

Best regards,

Karl

[Jon Hodgson]

squeegybug wrote on Thu, 22 December 2005 17:26

kraster wrote I think it's very relevant how we sample sound at 96k or other rates and then present it for analysis. The thread title is "Why recording in 96k through Lavry Blue sounds clearly better ?". Therefore we must hear the sounds at 96k. Otherwise the thread title should be "Why recording in 96k and then downsampling to 44.1khz through Lavry Blue sounds clearly better ?" I listened to the samples and the most obvious thing to me was differences in the timbre of the instruments as a result of a different performance. There are a lot of factors at play here: variances in performance, the downsampling process and your own knowledge of what sample rate was being used. All of these have a bearing on the sound. I can add nothing to Ivo’s superb analogy, it is valid and in fact the exact reason I also resampled to the standard CD resolution. I understand your point about analytical data submitted for review.  If it was indeed possible for the resampling process to “improve” the original audio, then I would agree that only raw information would be acceptable for discussion.  However, of course we know this is not reasonable – the SRC could (and does) slightly degrade the audio resolution, but in no way can it enhance it. I will certainly submit my original raw 48k data alongside the raw 96k file previously posted.  To me, the relative differences remain as in the resampled 44.1 clips. Would that I, or Ivo, or any musician, could actually increase stereo imaging or add more depth or smooth out harsh tonal transitions of actual physical instruments (or voices) – only from the simple notion of just “playing differently”!  Of course the styles or presentations can change – but, the overall organic structure of these instruments preclude creating something that is physically not there. In my opinion, the opposite result is actually the point here – I contend that there is already further detail available in the music than what is able to be captured by the 44.1 (or 48) resolution. In my case, it was quite welcome to hear the comparable ease of the performance in the higher sampled version.  I can accept that as a fact, given my own 40+ years experiences with my own instruments and voice.  While I understand the reason for this forum is to explore technical explanations of physical phenomena, let’s not dismiss the submissions of evidence for the wrong reasons here. Steve

I'm not dismissing the evidence, I am using facts to filter out what might or might not be the cause.

Fourier and Shannon-Nyquist are not theories, they are theorems, they are mathematical equations, every bit as deterministic as 1+1 = 2.

What many people don't understand, when they think they are challenging these theorems, saying things like "they require perfect filters to work and you can't have perfect filters so they can't work", is that they DO NOT require perfect filters, because their "job" is not to create or recreate waveforms perfectly, but rather to tell you how to do it, and what happens when you get something wrong.

So for example if your anti-aliasing filter doesn't completely cut out everything above half your sampling rate, then Nyquist still correctly predicts what you will get.

Now, given that we know Fourier and Nyquist-Shannon to be correct, then all things being equal, the only difference between a perfect 48 kHz sample and a perfect 96kHz sample would be bandwidth, the second sample would have an extra 24 kHz of bandwidth available.

Now, if you put this through a "perfect" sample rate converter, then you will lose that extra bandwidth, and what you end up with is the same as if you originally sampled at 48kHz.

If everything is working correctly, then there is no way for any of the extra information resulting from a higher sample rate(call it imaging, clarity, whatever you like), to translate through the Sample Rate Conversion process.

Therefore if there is something coming through, then something is working incorrectly from a technical point of view (artistically it might be great, but then so is valve distortion on a guitar).

I'm kindof busy the next couple of days, but I'll try to set up some tests for you guys to try, I'm interested to see the results.

[squeegybug]

Jon Hodgson wrote

I'm not dismissing the evidence, I am using facts to filter out what might or might not be the cause. Now, if you put this through a "perfect" sample rate converter, then you will lose that extra bandwidth, and what you end up with is the same as if you originally sampled at 48kHz. If everything is working correctly, then there is no way for any of the extra information resulting from a higher sample rate(call it imaging, clarity, whatever you like), to translate through the Sample Rate Conversion process.

I did not intend to imply that you were not addressing the evidence, Jon, I agree you are as best as Ivo and I can present it.  Only that the typical responses to these types of live performance comparisons inevitably leads to the ubiquitous “performances were different, therefore no knowledge can be obtained”, as often voiced by many other participants.

And Karl, I don’t take any offense from your observations, quite the opposite, I welcome any input to this discussion.  And your opinions are of course as valid as mine... a distinction being that I had the experience of my actual performance of a song I have done for 25 years as a reference, and so can appreciate the details of the differences in reproduction from a different perspective.

Jon, it’s helpful and practical information regarding the bandwidth differences and sampling definitions.  I can certainly agree that it is likely the signals are just not arriving in the same form, after coming through the two different “pipelines” – not that the 96k is necessarily “better” or has more information, but that more likely the 44.1/48 is “worse”, or not as accurate.  Or at least, not as esthetically pleasing.  The fact that this appears (to my ears at least) to happen with my system as well as Ivo’s indicates that it is perhaps not strictly equipment (mal)function, since we both use different products.  And as I’ve mentioned, to my senses the “general” relative differences of the raw audio remain even after SRC.

But of course, this is just a topic for discussion.  As in selecting so much of our audio equipment, the personal preferences from experience will win.  It’s always useful to question how we can define those preferences and try to quantify them so we can apply the lessons in creating and using our tools.  I have not previously seen these types of posts that attempt to display the differences between conversion rates using actual audio files, so I guess we’ll all continue to learn how to describe them.

Steve

[danlavry]

Try the following:

Here is a little experiment:
Start with a 48KHz material. Call it Audio1.
Then, upsample it to 96KHz. Call it Audio2.
Then, downsample it to 48KHz. Call it Audio3.
Now compare Audio1 to Audio3.
Do you hear a differance?
If there is a differance, which one do you prefer?
(Keep it all at 24 bits. No dither, no processing)...

Regards
Dan Lavry
www.lavryengineering.com

[squeegybug]

Good idea Dan, thank you.  I'll try that soon as I get a chance.

I'm also considering this approach:  Play back my original 96k clip through my PMC monitors and record that with a fixed mic and preamp gain, at both 48k and 96k.

Now of course we know that the music through a speaker will not approximate the live recording, but could yield information strictly about the converter behaviors... if these new "broadcast" 48 and 96 recordings sound significantly different from each other

At least they will be "identical perfomances".  What do you think about that procedure, and do you have suggestions?

Steve

[squeegybug]

danlavry wrote

Here is a little experiment

I did this test tonight.  Audio1 as original raw 48k stereo tracks, Audio3 the up-then-downsampled version.  All at 24 bit, no dither.

I cannot reliably hear a difference between the two short recordings in blind A-B listening on speakers and headphones; but when I do detect any slight perceived (or imagined) differences I preferred the original.  Perhaps the tiniest bit of more openness on leading "ss-es" and overall treble frequencies in the original.

I compared the signals by reversing polarity on one stereo set relative to the other.  And then also with the two left channels and two right channels.  The signals do not null in any cases.

When the two stereo samples are routed to the main bus simultaneously, there is a noticeable "phase-shifted" sound.

If wanted, I can post the clips as soon as I can write them out, I don't use a computer in the studio.  But they are not remarkable at all in any differences that I can tell.

It seems the Akai SRC is doing a "reasonable" job here.  Although it is not transparent, since the signals are different.  It certainly does not add any euphonic qualities or better imaging.

Steve


Edit:  I also just ran this SRC test with the 96k example clips -- downsample to 44.1, then back up to 96.  Same type of results as above; almost indiscernable audio difference, but the reversed-polarity signals do not cancel.

Conclusion:  My SRC does not "improve" the audio.  But doesn't destroy it either.

[Yannick Willox]

I really whish the 96K version sounds better, because then I'd have a good reason (besides the commercial one) to record all CD/SACD projects in 96K.

In these examples 96mon1 sounds better than 44mon1, but 44mon2 sounds better than 96mon2. With the vio samples it is the other way around. The 96k drum samples is clearly played a lot better (more defined/enthousiastic attacks, as if the player knows, hey this is a good take, or, hey, this needs to be great cause it's 96K ...)

Anyway, after almost 15 years of mobile recording, and a lot of soundchecks (where you NEVER get the same performance twice, but still have to decide which is the better setup), I have to conclude in this case I cannot find any difference.

Changing a mic cable would have a greater impact on recording quality than 96/44K with this converter, IMHO.

[Ivo]

Yannick Willox wrote on Fri, 23 December 2005 15:53

Changing a mic cable would have a greater impact on recording quality than 96/44K with this converter, IMHO.

Can you, please, specify ? There have been already quite a few discussions about "esoteric" cables. I use standard decent professional cables ... ("cordial") Do you have a suggestion for a clearly better alternative ?

Although I have already finished my short experimental period I will do one more thing: I will play two identical things on monochord (the only instrument which position is absolutely fixed towards microphones) and post the original 44 and 96k sample. Everybody is free to make any personal conclusion. Hopefully today or tomorrow ...