Analyzed: John Hiatt's Dirty Jeans and Mudslide Hymns at 24 bit / 96 kHz

<cite>I setup the Aurender S10 music server to shuffle between 44.1 and 96k tracks while I tried to identify them as 44.1 or 96k. This testing was a disaster as I selected the incorrect version half the time and really confused myself the longer and harder I tried to listen for differences.</cite><br />

<br />

If you cannot hear the difference, then there <em>is</em> no difference. The <em>only</em> way to overcome expectation bias is by blind randomization – the ABX.

Chris, as per the spectrograms I can see (not listen to) your Lion is not full Burn-in, the music tracks, or both... (forgive me, as always joking).<br />

<br />

For me also is hard to get an conclusion from similar SQ music tracks, but I prefer to take an spectrogram after the listen session, in order do not get a prejudge that can influence the session, even if I'm not an Audacity fan, is free and handy when you find an obvious problem on some recording.<br />

<br />

Thanks for the links and review!<br />

<br />

Barry, What do you mean with <i>"...even if upsampled because the playback filter is different..."</i>? The recording was made on PCM at a low sample rate, DSD, or analogue?<br />

<br />

Or, simply upsampled from 16/44 Redbook, and the is playback filter in the DAC that gave you more extension in highs?<br />

<br />

Kind regards,<br />

<br />

Roch

Hi Chris<br />

<br />

Did I just see that you are using the TAD CR1! Did that replace your Verity speakers and your new reference speakers or is it an item in for review? Exciting stuff. Any pictures?

Hi Hiro - I purchased a pair of TAD CR1s :~) They are a tremendous loudspeaker for pleasurable listening and as a reference tool. <br />

<br />

the Meitner

I am a big John Hiatt fan too. I'd love to see them get this right, but this looks not so good. I was looking forward to buying the DVD and am still waiting for someone to give me an excuse to do so.<br />

<br />

A question I have is does the spectrum show you anything about the bit depth?<br />

I have no problem with 24/48 music if characterized that way.<br />

<br />

I don't understand why, if you had a plan to market 24/96, the spectrum would not look perfect. Obviously they don't follow this website. <br />

<br />

Personally I think the Hiatt engineers need to answer to this. Its great that CA can reveal this and demand better. And the musicians themselves need to hold their engineers accountable. Come on, it's your music John, get it right. Neil Young will laugh at you at a party. I am pretty sure of that.<br />

<br />

<br />

<br />

<br />

This is from the Bring the Family DVDa rip.<br />

<br />

<IMG SRC="http://www.computeraudiophile.com/files/hiatt.jpg"><br />

<br />

Hi Roch,<br />

<br />

<i>"...Barry, What do you mean with "...even if upsampled because the playback filter is different..."? The recording was made on PCM at a low sample rate, DSD, or analogue?..."</i><br />

<br />

If I start with say, a 16/44.1 recording and play that back on my DAC, I'm using the DAC's filter for 44.1 recordings (at 22.05 kHz with a slope of x).<br />

<br />

If I take that same recording and convert its sample rate to say, 96 k and play <i>that</i> thought my DAC, I'm using the DAC's filter for 96 k recordings (at 48 kHz with a slope less steep than x and with a "knee" -- where the roll off begins -- that is more than an octave higher in frequency and so if further away from the top of the audible spectrum.<br />

<br />

If all else is equal (a big "if"), I can see how the gentler filtering, much further up the spectrum, will do less damage to the audio than a steep filter just above the audible spectrum.<br />

<br />

This is why I said that even if the recording is upsampled and not genuine 96k, I can see there being an audible <i>difference</i> (even if not necessarily an improvement but certainly a <i>difference</i>) between the 96k result and the lower sample rate original.<br />

<br />

To be clear, I am not advocating this and I am not suggesting this is what I would expect when a product is sold as 24/96. I'm just saying I'm not surprised if an audible <i>difference</i> is heard.<br />

<br />

Best regards,<br />

Barry<br />

www.soundkeeperrecordings.com<br />

www.barrydiamentaudio.com<br />

what if you take the native 16/44.1 use the DAC to upsample the native 16/44.1 rather upsampling the data to 96.<br />

<br />

I have heard people claiming that they like ripping to higher sample rates than native. I have tried listening to native at native DAC sampling, same music ripped or reformatted to higher sampling rates and using native sampling and upsampling via the DAC's upsampling.<br />

<br />

My own personal preference has and continues to be to listen at native. I understand the theory as to why it might sound better or could sound better, but to my ears, resampling, either the file or via the DAC has always added an less than natural listening experience. <br />

<br />

In fact, after many fatiguing hours of listening, I choose to re-rip my disc collection to native.

Hi Priaptor,<br />

<br />

<i>"...what if you take the native 16/44.1 use the DAC to upsample the native 16/44.1 rather upsampling the data to 96..."</i><br />

<br />

If I was going to upsample, I would do it using the best <i>off-line</i> algorithm I could find (which in my experience, is currently -and easily- iZotope's 64-bit SRC). I've never heard on-the-fly sample rate conversion that does not (to my ears) do sonic damage.<br />

<br />

Even most of the better off-line algorithms will (to my ears) brighten and harden the sound. In my experience, iZotope's is an exception that creates results that sound more like the unconverted original than any of the few dozen other SRC algorithms I've tested.<br />

<br />

To be clear, I'm not talking about a DAC with an oversampling filter, I'm talking about converting the rate of the data to a higher rate (which I believe is what you're talking about too).<br />

<br />

I with you with regard to native. As much as I like iZotope's SRC (and use it in my daily work), I never apply it for just listening. I always listen to files on my music server at the sample rate at which they were recorded. (And I load them into the server at the native rate at which they were recorded.)<br />

<br />

As to some folks <i>liking</i> to use SRC when they rip, that is a personal call and I'd be the last to call them "wrong". If it is a sound that pleases them, that is all their system needs to do. However, if one is seeking the sound of the <i>source</i> (not everyone is), in my view, native is the way to go. (Keeping in mind that I wouldn't do it with what I consider far and away the very best SRC algorithm extant. And most folks are using other algorithms.)<br />

<br />

All my rips are native, to raw PCM (in my case, my preferred format: .aif).<br />

<br />

Just my perspective of course.<br />

<br />

Best regards,<br />

Barry<br />

www.soundkeeperrecordings.com<br />

www.barrydiamentaudio.com<br />

I can to the same conclusion, after many painful hours of ripping one way.<br />

<br />

My server is all native, both in AIFF and in FLAC, the latter for storage and backup. I too prefer AIFF.

Hi Barry,<br />

<br />

Thanks for your detailed explanation, since I'm 'learning Audacity', there are things I can't still understand.<br />

<br />

What I found, since some computer music player app's allows you upsample, and, when I do this (from a 16/44 rip), the recording has to be very, very good, to allow the DAC achieve this 'high ceiling' that could benefit (a little) the SQ, but, if it is a 'normal' (bad) recording, it sound (to my ears) similar to high jitter.<br />

<br />

Kind regards,<br />

<br />

Roch

ChrIs did I read right that you just changed every component in your system? what do you do with all your other expensive toys LoL

So Chris, if you ABX yourself and it becomes a mess, meaning you get it wrong 50% of the time, what do you think of this? There is in fact a difference in the 44khz vs 48khz versions. Yet you apparently cannot discern it. So do you conclude.....well what do you conclude? What would you gain by actual 96 khz files? I agree if marketed as such they should in fact be 96 khz, but what if that also wasn't discernible blind?<br />

<br />

I posted an alternative variation on blind testing, which is subjectively benign compared to ABX testing in a recent post. Apparently it wasn't of interest to anyone as not one person has made a comment on it after a few days. <br />

<br />

I just wonder if this pursuit of higher and higher sample rates isn't chasing a ghost if the lower rates of say 48 khz are well done. <br />

<br />

What would it take to convince you that well done 48khz/24 bit is truly enough? (I am not declaring I think it is enough).

Why do we only (mostly) looking just at the frequency spectrum, when talking about HiRes Audio. Sure, real HiRes Recordings offer better open space and easy flow of the music, but my experience does also indicate, that we should not forget to look at the dynamic of the music, well known as loudness war. My feeling is that I have more recordings that are suffering from the loudness war, then suffering from the “limitation” of the recorded bandwidth.<br />

<br />

Beside being an audio hardware engineer for nearly 30 years, and making music for over 35 years, I have also a small record studio and doing mainly tutti (the complete orchestra at a time, no overdubs) classical and jazz recordings and be an active member of “pleasurize music foundation” to do at least a little bit against the loudness war and do only a very limited compression on percussion instruments, but the rest leave it as it is.<br />

<br />

Attached are two examples of a good dynamic range recording “Yuri Honing Trio” with 16 dB loudness dynamic and a compressed recording “Donald Fagen – Morph The Cat” with only 10 dB loudness dynamic. For those who are inexperienced with that figures, should read a little bit at the pleasurize music foundation, but I can tell “only 2 dB” difference in numbers, gives you a huge difference in listening.<br />

<br />

Best Regards<br />

<br />

Juergen<br />

<br />

Chris: Congratulation on your new speakers.

Hi esldude,<br />

<br />

<i>"...What would it take to convince you that well done 48khz/24 bit is truly enough? (I am not declaring I think it is enough)...."</i><br />

<br />

I'm not Chris but I have an answer for that one: <br />

<br />

It would be hearing the same signal encoded at 24/48 and at higher rates. It would be in comparing those recordings with the input signal. And lastly, it would be in not hearing the higher rate recordings (particularly 24/192) get <i>significantly</i> closer to the sound of the original.<br />

<br />

The rate just below the one that ceases to get closer to the sound of the input signal is "enough".<br />

<br />

Just my perspective of course.<br />

<br />

Best regards,<br />

Barry<br />

www.soundkeeperrecordings.com<br />

www.barrydiamentaudio.com<br />

Great write up. Thank you. And I loved that you mentioned what has become my pet peeve:<br />

<br />

<i> "... I could have sworn the difference between the 44.1 and 96 versions was huge and readily apparent.... To my dismay I was unable to duplicate this experience in any subsequent listening session."<i> <br />

<br />

This has been my uncomfortable experience for 30 years -- both in my personal listening for pleasure, my reviewing, and on a couple of occasions when I ran single-blind tests for audiophile friends. It's also why, like a broken record, I keep saying: Use your ears, but never trust them completely. Our ears are connected to our brains, and our brains will always process "the facts" in ways we can't be aware of.<br />

<br />

That said, there are many reasons why our audio systems sound different from one day (or one hour) to the next -- reasons that don't involve our "ear/brain systems" fooling us. I also think Barry's explanation of DACs sounding different at different sampling rates is very reasonable. But if you couldn't hear a difference between the 48k and 96k files when listening critically in a blind, randomized session, then I'd say your DAC is doing it's job. <br />

<br />

I just ordered the Apple Lossless files from Hiatt's website. Downloading now.<br />

<br />

Thanks again - Mark B

I agree with you Juergen.<br />

<br />

I guess it is because almost no recordings come close to using even 16 bits that you hear little about it. <br />

<br />

Once a few years ago I heard a very popular CD and commented I needed to use a dynamic range expander on it. I happened to have an older pro DBX expander. It would do up to 50% expansion, and I declared I could take this recording and have 4.5 dB dynamic range rather than just 3 dB. I was only joking, but later analyzing it with some software, it only had 5.5 dB on most of the tracks without any expansion. I was shaking my head and figuring the band must have wanted it that way though I don't know for sure who made that decision.

"It would be hearing the same signal encoded at 24/48 and at higher rates. It would be in comparing those recordings with the input signal. And lastly, it would be in not hearing the higher rate recordings (particularly 24/192) get significantly closer to the sound of the original."<br />

<br />

"The first principle is that you must not fool yourself - and you are the easiest person to fool. by Richard P. Feynman."<br />

<br />

Well, that is the obvious answer Barry, and I have no problem with it. The real question of course is how do you decide when you do or don't hear a difference. The following quote is well known and very much something to keep in mind. <br />

<br />

"The first principle is that you must not fool yourself - and you are the easiest person to fool. by Richard P. Feynman."<br />

<br />

I have yet to find a convincing argument that makes blind testing null and void. In some fashion it has to work. Either you can reliably hear something or you don't. It is even possible to develop the edges of thresholds. I have the same problems as others who don't like it. Some things sound so obvious, yet under scrutiny they turn out to be maybe not obvious maybe not real. But your gut level experience makes it hard to accept. I see an explosion in money, time and effort spent by people who refuse to accept blind testing at all chasing what I think isn't even there. I think it has it own problems, but in principle it will work. I think if one nails down what really works in finding the limits of perception then all this effort can yield much better results with everyone better off. <br />

<br />

I also think plenty of reasonable differences are overlooked that could be solved because of all this ghost chasing going on. <br />

Hi esldude,<br />

<br />

<i>"...The real question of course is how do you decide when you do or don't hear a difference..."</i><br />

<br />

For me, it is direct comparison against the input, sometimes with the aid of my assistant who will do the switching while I listen, not knowing which is which.<br />

<br />

So far, blind or sighted, the results have been identical, every time: 24/192 (on the system I'm using) is hard to tell from the mic input. At lesser rates, the differences are not subtle.<br />

<br />

Best regards,<br />

Barry<br />

www.soundkeeperrecordings.com<br />

www.barrydiamentaudio.com<br />

Concerning Dynamic Range, the book from Bob Katz “Mastering Audio – The Art and Science” is a very good source.<br />

<br />

For me personally, a loudness dynamic of about 13 dB is a good and practicable choice for 16 Bit CD medium. When I master for CD, this is my roughly goal. When it is for more popular music, I tend toward 12 dB and when it is for more “serious” listening, I tend toward 14 dB.<br />

<br />

Attached I have a list of some “typical” tracks, that you can hear on HiFi shows (I know, typical will differ from room to room, but just take this as an example). You will see that an average of 62 “typical” HiFi show tracks gives you a value of 12 dB loudness dynamic.<br />

<br />

I personally also do like progressive metal and progressive rock a lot and it is a shame, that most of the modern albums are mastered way too compressed to be loud. Not being a progressive metal album, but also mentioned on the former blog is Death Magnetic with DR of 5.<br />

<br />

Best Regards<br />

<br />

Juergen

What I get from Chris his story, is that he not only couldn't differentiate between the two, but that he also couldn't select the proper versions (selected A while thinking he selected B). A decent means of ABX ?<br />

<br />

NO<br />

<br />

Because one can't be conclusive by any means this way.<br />

<br />

I can imagine that Chris lets the thread go along thinking what "we" think, but it wouldn't be quite right.<br />

<br />

Am I wrong Chris ?<br />

Regards,<br />

Peter

Dear Everybody,<br />

Chris mentioned that his package contained a CD and a DVD-Video disc. As some of you might remember the DVD-Video specification does allow for 24/96 Stereo Sound with no picture and in the past a few discs like this were offered by companies like Chesky and Classic Records. The benefit of this disc is that it will play on any DVD-Video player and should work just as well on any Blu-Ray Player. As both DVDs and Blu-Ray is working with 48 Khz base sampling frequencies it may be entirely possible that one of these machines does give a better sound in playback compared to the standard CD with its 44,1 Khz sampling frequency. Thus the idea may have been to give "Joe Sixpack" a better sound when he uses his one universal player and I think that is quite laudable. It is rather unlikely anybody thought of the tiny community of computer audiophiles who would rip the disc to play it back on their computers.<br />

<br />

Jan

Hi Jan - I think I understand what you're trying to say but I don't really understand how the community of computer audiophiles and Joe Sixpack really matter in terms of what was discussed in the article. The album was released as 24/96 but the source was clearly 24/48.<br />

<br />

Maybe I missed your point. It's entirely possible :~)<br />

<br />

agentsmith - <i>"Chris did I read right that you just changed every component in your system? what do you do with all your other expensive toys LoL"</i><br />

<br />

Hi AgentSmith - Please stop by and pick up the DAC of your choice. Only joking :~)<br />

<br />

 <br />

<br />

esldude - <i>"So Chris, if you ABX yourself and it becomes a mess, meaning you get it wrong 50% of the time, what do you think of this? There is in fact a difference in the 44khz vs 48khz versions. Yet you apparently cannot discern it. So do you conclude.....well what do you conclude?"</i><br />

<br />

I conclude I could not tell the difference between the 44.1 and 96k versions of the new John Hiatt album Dirty Jeans and Mudslide Hymns using the components specified below, in my room, on the days I listened to the album.<br />

<br />

 <br />

<br />

<i>"What would you gain by actual 96 khz files?"</i><br />

<br />

The only way to know is to have the files. <br />

<br />

 <br />

<br />

"<i>"I agree if marketed as such they should in fact be 96 khz, but what if that also wasn't discernible blind?"</i><br />

<br />

Then I would have to conclude I could not tell the difference between the 44.1 and 96k versions of the new John Hiatt album Dirty Jeans and Mudslide Hymns using the components specified below, in my room, on the days I listened to the album.<br />

<br />

 <br />

<br />

<i>"I just wonder if this pursuit of higher and higher sample rates isn't chasing a ghost if the lower rates of say 48 khz are well done. What would it take to convince you that well done 48khz/24 bit is truly enough? (I am not declaring I think it is enough)."</i><br />

<br />

If a recording is well done and the engineers doing the recording suggest a higher rate would not make the specific recording sound better then I'd be cool with whatever that rate is. Heck it could be 32k for all I care, as long as the aforementioned condition is met.<br />

<br />

 <br />

<br />

PeterSt - <i>"What I get from Chris his story, is that he not only couldn't differentiate between the two, but that he also couldn't select the proper versions (selected A while thinking he selected B). A decent means of ABX ?"</i><br />

<br />

Hi Peter - I conducted my listening tests as nearly every other listener would conduct such a test on his home system. No matter what testing methodology was used the only conclusion that may be logically reached is that I could not tell the difference between the 44.1 and 96k versions of the new John Hiatt album Dirty Jeans and Mudslide Hymns using the components specified below, in my room, on the days I listened to the album.<br />

<br />

<br />