Right way to compare 16/44 and 24/96?

[Jeff In San Diego]

Ok, I just purchased and downloaded an enjoyable classical selection from the Linn download site at 24/196. (It is Susanna Yoko Henkel with Tchaikovsky violin concerto if it matters -- passionate playing and laser accurate pitch on a fretless instrument - what more could a guy ask for?)

 

So far I've just listened to the original 24/196 downloaded file downsampled on the fly by Audirvana into my 24/96 capable USB DAC. (I don't have any connection that would let me pipe it in there at 24/196)

 

I'm new to this Hi-Res stuff, and want to take a stab at learning the difference in sound between 16/44 and 24/96...but I want to make sure I'm not fooling myself by doing it wrong.

 

So finally getting to my question...how best to down-res this file into the two formats given the equipment in my sig and not spending any money for additional hardware or software?

 

Would creating a 16/44 and a 24/96 version of this using "Max" provide a reasonable basis for some non-scientific a-b comparison? Any settings I need to be careful of to avoid fooling myself?

 

And finally, if the difference isn't immediately apparent to me, what should I listen for to train myself to hear the difference?

 

thanks,

jp

 

[bottlerocket]

I don't think Max will down-sample. I did a similar thing using dbPoweramp, but it is Windows only. I run it under vmware on my Mac. I kept the 24/192 originals and created 24/96 copies as that is the maximum my current DAC will take.

 

[The Computer Audiophile]

"...how best to down-res this file...?"

 

Hi JP - Don't do it.

 

There are a few professional software packages that do this and some professional hardware that does this, but there is no consensus what works better. Reference Recordings has even used two identical D to A and A to D converters to accomplish this task as good as possible. RR converted a 24/176.4 album from digital to analog with a DAC and sent that signal into the analog to digital converter ADC to be captured at 24/96 for HDtracks download site.

 

Based on my conversations with people who do this for a living it's not a trivial process and all methods provide different results.

 

[Jeff In San Diego]

I knew it did a lot of formats so I ass-umed it would do some other things as well.

 

I didn't find anything laying around that could fully do the conversion I wanted, but I did manage to find "Switch", which was able to do half of it. I was able to make 44.1 and 96k versions, but both at 24 bits. hrrrrmphf

 

[Jeff In San Diego]

...one has to admit though that this sounds like the folks who sell them asking us not to compare so that we won't discover we can't hear a difference.

 

Given all the disclaimers about how difficult it is to do this well perhaps it doesn't mean anything at all, but I was able to hear a difference between my poorly and partially converted down-res attempts.

 

I used an app called "Switch" and got as far as making two files, one at 44.1 and one at 96...but both at 24 bits depth.

 

I CAN hear a difference between the two and prefer the bigger one. I can't begin to put words to the type of difference I hear. It is like trying to use English to describe the difference between two glasses of Merlot. One can sound learned and impressive, but I wouldn't call it a successful communication.

 

Best description I can come up with is the bigger one sounds "better". Perhaps another way to describe it would be to say that Ms. Henkel sounds like she is playing a more expensive violin in the bigger file.

 

Then again, maybe it has more to do with the quality (or lack thereof) of the down-rez-ing in one vs. the other. Who knows.

 

I do know that if there IS a tangible difference between the different formats, the Hi-Res download sites would find it to be a VERY powerful marketing tool to give away a 44.1 file whenever a customer buys one of the higher res format files.

 

Until I have a good opportunity to do an apples to applesauce comparison and convince myself there is a palpable difference, I'm not likely to spend much money on Hi-Res.

 

[Miska]

Reference Recordings has even used two identical D to A and A to D converters to accomplish this task as good as possible. RR converted a 24/176.4 album from digital to analog with a DAC and sent that signal into the analog to digital converter ADC to be captured at 24/96 for HDtracks download site.

 

I would call that worst possible method, since most ADCs and DACs are "oversampling" with constrained quality plus there's added analog noise and distortion from two analog stages.

 

Practically all ADCs downsample from higher internal rate, just like DACs go to opposite direction.

 

IMO, best way is to choose good quality software converter and let it handle the task, and not trust on (processing power-) constrained ADC chips doing the same...

 

When done in software with unlimited amount of CPU time spent, the results are technically best. But naturally some may still prefer other ways sonically, due to different added "colorations".

 

 

[Miska]

I'm new to this Hi-Res stuff, and want to take a stab at learning the difference in sound between 16/44 and 24/96...but I want to make sure I'm not fooling myself by doing it wrong.

 

There is no clear and simple way to subjectively compare differences between two materials, since inevitably you end up having software and/or hardware included in the comparison.

 

So finally getting to my question...how best to down-res this file into the two formats given the equipment in my sig and not spending any money for additional hardware or software?

 

IMO, best quality converter you can have for free given your current computer is the one in OS-X CoreAudio if you are running 10.5 or newer. Unless there's some converter I don't know about.

 

 

[diogenis]

Ok,

Max can convert amplitude (24, 16, whatever) but not frequency (It can't do 96->48)

It also crashes on my snow leopard... So I am not really sure what it can do

 

XLD @ http://www.macupdate.com/app/mac/23430/x-lossless-decoder

can change amplitude in aiffs, wavs and pcm files and frequency in ALACs, but not in one go (maybe you can do it in 2 steps. My bet is that it uses some low level routines from Apple core to do this) . It can also output multiple files from one input file.

 

Sample Manager does all that and in style, with automated workflows that can help you if you convert routinely, but it costs $$.

http://www.audiofile-engineering.com/samplemanager/

 

So, why do you want to mess with all these?

 

 

[blueixus]

Korg's Audiogate will down (and up) sample, it is free except that when you 'export' your new file you have to send a 'tweet', but it is good software.

 

It may be better to take 96kHz files to 48kHz not 44.1kHz as the maths is much easier.

 

[sq225917]

There's no point rushing into it head first choosing sample rate convertors without actually knowing their performance first. Fortunately someone has already done that for you.

 

http://src.infinitewave.ca/

 

You'll find the 'Audacity Best' SRC is almost perfect in technical terms though it does introduce pre and post ringing, but then most SRC do. The good news is that Audacity is free and works on Mac and Windoze so makes an ideal starting point. The link doesn't include the very last SRC's but it does include 'most' of them. In technical real terms it is unlikely that much improvement has been made in SRC apart from conversion speed and memory footprint in recent times- the maths is too well understood for that, though some apodizing filters do exist now that work and measure well.

 

Of course you aren't just hearing the difference in the files, but also the differences in the way they are handled by the hardware, none integer mutiples of a 44.1 source may well be handled differently on chip, by the dac, compared to 44.1 source, the same holds true for none integer downsampled files from hi-res source.

 

Having done these tests many times with various SRC and dacs I think I can hear a difference between 44.1 material and higher res 176.4+ source material, but I've never bothered putting it to the test in a blind AB. I settle for sourcing as much 24/96 as I can and take any 176+ stuff when I really want the music.

 

best of luck.

 

 

 

[Miska]

It may be better to take 96kHz files to 48kHz not 44.1kHz as the maths is much easier.

 

Easier for the software developer or the CPU?

 

Easier for the developer yes, not for the CPU.

 

[The Offenbach Kid]

Hi jp,

 

I'm not sure why you bought the 24/192 version when your DAC max's out at 24/96, but that's another question.

 

It involves spending some more money on the same piece of music, but if you really want to do some serious comparisons between 16/44.1 and 24/96 why not go back to Linn and buy both versions. Hey, you could even download the MP3 version and get the full set. (only joking).

 

Actually I'm only partly joking. Downloading the MP3 version as well could actually help you to hear the differences in the qualities between the file resolutions i.e. if you can hear and understand the differences between the MP3 and 16/44.1 files, you might be able to apply this learning/understanding to hearing the differences between the 16/44.1 and 24/96 files.

 

Doing it this way means you aren't introducing any other variables into the equation.

 

Just a thought.

 

Regards,

 

Neil

 

[scotty]

just change the setting in the MIDI -- restart iTunes and play that file at different sample rates - not sure how iTunes resamples the files - but pretty sure it does?

 

of course this makes it harder to do a A/B comparison as you have to back out of iTunes, make the change, then restart iTunes.

 

[Miska]

I'm not sure why you bought the 24/192 version when your DAC max's out at 24/96, but that's another question.

 

I would do the same, since it makes the investment future proof. No need to repurchase once hardware is changed to something 192/24 capable... I guess for many the purchased songs outlast most pieces of hardware or software.

 

Meanwhile, playing back a downsampled version is good option as long as the algorithm for doing it is good.

 

Doing it this way means you aren't introducing any other variables into the equation.

 

Except then comparing downsampling algorithm used by the vendor... Who knows if they would use the best one? Which one is the best in first place? Who should be the one deciding it?

 

 

[dsnyder]

Hello,

 

The last time that I played with this, I used the SoX command-line utility. It is the same tool that cPlay uses. For example:

 

$ sox orig.flac -b 16 -V test-cd.flac rate -v -M 44.1k dither -f gesemann

$ sox orig.flac -b 24 -V test-24.flac rate -v -M 44.1k

$ sox orig.flac -b 16 -V test-16.flac dither -S

 

These command-line options require a fairly recent version of SoX (I'm using v14.3.0). These will create three versions of the original file. The "test-cd.flac" is standard redbook format, but I think it's more interesting to compare the test-16.flac and test-24.flac: a 16-bit version with the original sampling frequency vs. a 24-bit version resampled at 44.1kHz. The last time, I consistently preferred the 24-bit version with the low sampling frequency vs. the 16-bit version with the original frequency. I was also not able to consistently tell the 24-bit 44.1kHz version from the original. This is fun though. Perhaps I'll give it another try.

 

-- David

 

 

[The Offenbach Kid]

Hi Miska

 

I also would do the same, and have already done so as I bought the 24/192 version a few months ago. I was just trying to respond to jp's request in that he wants to do this test without having to buy anymore hardware or software.

 

I also discounted what the vendor was using as a downsampling algorithm (rightly or wrongly) because I saw it as outside of my control and a decision already taken by somebody who knows far more about these things than I do.

 

Regardless (?) of the algorithm used, Linn sell MP3, 16/44.1, 24/96 and 24/192 versions of this piece of music. That is what they sell and that is their offer to the marketplace.

 

Do you not think that it is an option for jp to buy these and listen to them and to try to hear the differences between them? It's the music that jp will hear, not the algorithm's. And as a one-off exercise it's much cheaper doing it this way than buying new hardware or software.

 

And anyway, it doesn't matter what we think does it. This is the very reason why this site exists. We who are lovers of music can use this site, ask questions, search the interesting threads etc. It's up to jp to decide the best course of action for himself, and if along the way we've been able to help, then it's part of why we exist.

 

Neil.

 

[dsnyder]

However...I suppose the best way to compare formats is to actually purchase a CD (or 16-bit, 44.1kHz files) and a 24-bit, higher sampling frequency version of the same tracks and compare them directly. After all, the goal is to decide if you prefer high resolution music enough on your system to justify the difference in price.

 

[eggers]

However...I suppose the best way to compare formats is to actually purchase a CD (or 16-bit, 44.1kHz files) and a 24-bit, higher sampling frequency version of the same tracks and compare them directly. After all, the goal is to decide if you prefer high resolution music enough on your system to justify the difference in price.

 

It's a bad idea to compare the merits of 16/44.1 vs. 24/XXx by listening to a CD vs. a high-res download, as they are mostly mastered differently.

 

Mostly the high-res downloads/DVD-Audio/Blu-ray have less compression and less admissions towards mono-compatability. That is why they usually sound better than the CDs, not because the high-res material is inherently better.

 

Downsampling of material can be performed on-the-fly in OS X but setting the output device to whatever combination you want in Audio MIDI Setup. If one setting sounds better than another it's the DAC doing that, not the material.

 

Downsampling high-res audio is trivial in software. There are many software and hardware devices that can do this with no issues. That RR uses a DAC feeding a ADC to downsample I find hilarious.

 

[bottlerocket]

JP, another suggestion is to download these samples that were set up with comparison of different sample rates in mind

 

http://www.soundkeeperrecordings.com/format.htm

 

They're free.

 

 

 

[Manelus]

eClassical gives for free mp3 and 16bit downloads when you buy a 24bit download.

I purchased (among many others) Suzuki’s Bach cantatas vol. 22 (that I already have in CD version) for comparison.

24 sounds better, no surprise (fuller bass, better timbre,…).

Then I compared 16 bit downloaded one with my CD rip (EAC, secure). Guess what?

Downloaded version sounds way better than ripped version.

I took both on a pendrive and performed blind test at a friend’s house (similar hardware stuff than mine). He listened blind… same results.

 

 

[dsnyder]

Quote: It's a bad idea to compare the merits of 16/44.1 vs. 24/XXx by listening to a CD vs. a high-res download, as they are mostly mastered differently.

 

Ok...I went back and re-read JP's original question in which he was asking for help on how to train his ears to hear the difference between redbook and higher resolution recordings. For that purpose, I agree with your statement above.

 

I guess I was responding more to this statement of JP's: Until I have a good opportunity to do an apples to applesauce comparison and convince myself there is a palpable difference, I'm not likely to spend much money on Hi-Res. From this statement, the goal of the exercise seems to be to decide if commercially available high resolution recordings are worth the extra spend vs. redbook releases of the same. If so, the difference in mastering for the higher resolution releases is part of the value add (assuming that it's beneficial) and should be included in the comparisons. Well, that's the thinking behind my suggestion anyway. Cheers.

 

-- David

 

 

[Jeff In San Diego]

...have mirrored my thoughts on this pretty well. I've found myself on both sides of the above mentioned fences, and currently sit firmly atop some of them.

 

All of the responses have been useful as I wade slowly into this icy water of Hi-Res. The process is laying bare some of my incorrect assumptions and also inviting me to further analyze my own reasons for wanting to do this comparison.

 

I guess ultimately for me it comes down not so much to the cost of different downloads etc. as to the poor range of selection in Hi-Res. A decision about how much time to spend trying to find some morsel of interest among the different Hi-Res sites vs. just shopping on Amazon for CD's from a huge available catalog.

 

I think it is something many of us are struggling with...balancing the interest in audiophile and its associated learning curve and enjoyable tweaking with our love of music. Right now, because of the very limited selection of Hi-Res, those two have a maddeningly small intersection.

 

I'm trying to keep that distinction clear in my mind, and separately enjoy each of those often disparate hobbies. Sometimes though, I get them confused. This thread is ultimately all about that dichotomy, the intersection, and the resulting confusion of goals.

 

thanks for the thought provoking discussion and ideas etc.

jp

 

[Jeff In San Diego]

I will check that out tonight while I watch American Idol (with the TV sound OFF)

 

 

jp

 

[eggers]

Here is a procedure to test your hearing, perception of sound, and your DAC. Take your best high-res audio track and create the following derivatives in an audio editor:

 

1: Down-sample to 16 bit, 44.1 kHz (CD quality)

2: Take the down-sampled file from above and up-sample it to the original resolution.

3: Take the original file.

 

Now do an ABX test of these three files. If you can hear no difference then you are a human – congratulations.

 

If you can repeatedly identify 1 from 2 and 3, and find the sound of 1 less good than 2 and 3, then your DAC sounds more to your liking when fed higher bit-depths and sample-rates.

 

If you can repeatedly identify 2 from 3, and find the sound of 2 less good than 3 then your name is Clark Joseph Kent and I bow before you.

 

[elcorso]

Max & AudioGate can downsample and upsample, the easy way on Max is to do the conversion to FLAC first. But I don't see the reason to do it.

 

I believe in Barry Diament explanation (and achieve to the same conclusion) that everybody should listen to computer music tracks in the downloaded resolution. You can downsample, with no problems, if your DAC, or computer music player app. doesn't reach the higher sample rate or "resolution". But this is only, as always, my personal taste.

 

If somebody is having crashes under Max, they should try the "Unstable Max Version 0.9.2" (for Snow Leopard), that is in the Max download page.

 

The other trick under Max is to quit the app, after a conversion, and then reload again for a new conversion, since the "Temporary files" fouls the app.

 

By the way, last night I was comparing Louise Rogers "Black Coffee", Chesky 96/24 download, versus the 44.1/16 ripped from the CD. The difference was not huge in % terms, but huge in "audiophile" terms.

My conclusion: If you have the CD, do not download the hirez tracks, because this recording is so wonderful, maybe you'll be happy with the ripped CD tracks.

 

Happy listening,

 

Roch