[RRod]

Any of you fine people have a link to any 88/96/176/192 cymbal samples? I'm continuing to compile my list of «problems that don't exist» with 44.1/16 playback.

 

[Steve Eddy]

What's wrong with just using some high res music tracks that feature cymbals?

se

 

[esldude]

http://www.2l.no/hires/
 
This page has downloadable samples which might suffice for your purposes.
 
You also might find this page interesting.
 
http://archimago.blogspot.com/2014/06/24-bit-vs-16-bit-audio-test-part-i.html
 
If you intend to take the hirez and downsample to lower sample rates and/or bit depth you need to use good resampling software.  The best I know of is Sox for free and iZotope which costs money.  You also can resample in Audacity which in current versions uses Sox.  A third possibility is Foobar which has a Sox plug in you can use.  If you need some tips on how to do that in these softwares just ask.

 

[esldude]

Also if you have access to a good condenser mike, you can record a few keys on a keychain.  Doesn't sound like much because much of the sound of metal keys are ultrasonic.  Easy to get sound out of jangling keys with a spectrum of a few kilohertz to 35 khz or more.  That way you can be sure of minimal processing and no messing with the original sound.

 

[RRod]

What's wrong with just using some high res music tracks that feature cymbals?

se

 
I have very few and none fit the bill.
 

  http://www.2l.no/hires/
 
This page has downloadable samples which might suffice for your purposes.
 
You also might find this page interesting.
 
http://archimago.blogspot.com/2014/06/24-bit-vs-16-bit-audio-test-part-i.html
 
If you intend to take the hirez and downsample to lower sample rates and/or bit depth you need to use good resampling software.  The best I know of is Sox for free and iZotope which costs money.  You also can resample in Audacity which in current versions uses Sox.  A third possibility is Foobar which has a Sox plug in you can use.  If you need some tips on how to do that in these softwares just ask.

 
Many thanks! I use sox a good bit for looping stats out of my collection. I also use "zita-resampler" (http://kokkinizita.linuxaudio.org/linuxaudio/zita-resampler/resampler.html), which seems to work well also.
 
As far as sox, I tend to use the instructions here:
http://sox.sourceforge.net/Docs/FAQ
 
Do you have a certain flow you like to use for 96/24 -> 48/16 -> 96/24?

 

[wiinippongamer]

Any "problems that don't exist with 16/44" will be a direct result of resampling, dithering and/or DAC filtering at different sampling rates. No need to beat the dead horse.

 

[RRod]

  Any "problems that don't exist with 16/44" will be a direct result of resampling, dithering and/or DAC filtering at different sampling rates. No need to beat the dead horse.

 
I didn't say I was going to make a huge post on it here or anything. It's just compiling examples.

 

[Steve Eddy]

I have very few and none fit the bill.

Well for Pete's sake, just go over to HD Tracks and find a track that does.

se

 

[RRod]

Well for Pete's sake, just go over to HD Tracks and find a track that does.

se

 
HD Tracks doesn't support native Linux downloads, so they are on probation

 

[Roly1650]

Well for Pete's sake, just go over to HD Tracks and find a track that does.

se

Which may be upsampled 16/44 anyway with no way of knowing until after the download.

 

[Strangelove424]

  Any "problems that don't exist with 16/44" will be a direct result of resampling, dithering and/or DAC filtering at different sampling rates. No need to beat the dead horse.

 
I think RRod is just taking empirical steps to test what he can hear or not hear, or feels comfortable listening to. I commend the effort, and wish more people would take those measures before deciding what quality of music they think they must have. 
 
On a side note, I've run a few null tests between hi-res and downsampled material, and always end up with a null track with 22khz+ frequencies that's dead silent to me. I think you'd have to be a dog or bat to hear it. I can understand why people need to go through the process and test it themselves though, that's how I felt too. 

 

[esldude]

You can compare resampler specs to some extent here.
 
http://src.infinitewave.ca/
 
Zita isn't quite as good as Sox, iZotope, or a couple others though it isn't bad.
 
As far as work flow, I have tried 96/24 to 44/24, then to 44/16.  Then back to 96/24 and nulled the two files.  There is little there below 22 khz (essentially nothing there other than the noise of dither in the last bit).  I also tried other combinations.  At least using Sox you can do 96/24 to 44/16 in one step and get identical results.  So no need to do it in two steps. 
 
Here is an FFT of jangling keys which has energy well into the mid 30 khz range at levels far higher than music.  Converted 96/24 to 44/16 and back to 96/24.  Then nulled against the original.  The RMS value for residual below 20khz would be a touch below -90 db.  This with triangular dither.  Shaped dither is lower below 6 khz, and higher above.  But less easy for humans to hear.   You of course won't actually hear either anyway.  (you can click on the graphs to enlarge them of course)

This one is shaped dither for the same file.  With these residuals you need 50 db of gain to hear much of anything on playback, and it is only hiss.
 

 
I understand doing conversions yourself to hear it (or actually  not hear it) first hand.  But you can tell from the above FFT's you are wasting your time to hear a difference.

 

[RRod]

   
I think RRod is just taking empirical steps to test what he can hear or not hear, or feels comfortable listening to. I commend the effort, and wish more people would take those measures before deciding what quality of music they think they must have. 
 
On a side note, I've run a few null tests between hi-res and downsampled material, and always end up with a null track with 22khz+ frequencies that's dead silent to me. I think you'd have to be a dog or bat to hear it. I can understand why people need to go through the process and test it themselves though, that's how I felt too. 

 
I've already satisfied myself with testing, and I only tested myself anyway as a gesture of goodwill to the conspiracy theorists, as I'm a big believer in theory. Now I'm working on compiling a set of examples to offer other people who maybe aren't as convinced by theory by might be convinced by something like ABX tests. I asked for a hi-res cymbal crash because cymbals are the go-to example for people who want hi-res to sound better («oh, but the cymbals sound so much more lifelike!»), so it would be nice to have an example of them isolated that I can challenge people with. This will be perhaps useful with the Pono about to get going, as you'll have all kinds of people saying stuff like «that album I love that was made in 1956 in the back of a van sounds SO great on Pono, it must be the 192 samplez!»
 

  You can compare resampler specs to some extent here.
 
http://src.infinitewave.ca/
 
Zita isn't quite as good as Sox, iZotope, or a couple others though it isn't bad.
 
As far as work flow, I have tried 96/24 to 44/24, then to 44/16.  Then back to 96/24 and nulled the two files.  There is little there below 22 khz (essentially nothing there other than the noise of dither in the last bit).  I also tried other combinations.  At least using Sox you can do 96/24 to 44/16 in one step and get identical results.  So no need to do it in two steps. 
 
Here is an FFT of jangling keys which has energy well into the mid 30 khz range at levels far higher than music.  Converted 96/24 to 44/16 and back to 96/24.  Then nulled against the original.  The RMS value for residual below 20khz would be a touch below -90 db.  This with triangular dither.  Shaped dither is lower below 6 khz, and higher above.  But less easy for humans to hear.   You of course won't actually hear either anyway.  (you can click on the graphs to enlarge them of course)
 
This one is shaped dither for the same file.  With these residuals you need 50 db of gain to hear much of anything on playback, and it is only hiss.
 
 
 
I understand doing conversions yourself to hear it (or actually  not hear it) first hand.  But you can tell from the above FFT's you are wasting your time to hear a difference.

 
May I ask what your workflow was in sox to get these files? I find in sox i'm always getting issues with clipping and with output files that have different lengths than the original files. I'm on Linux so I just use command line, and it would be nice to write a script that will do these things automagically.

 

[castleofargh]

there was an apocalyptic fight over Arny Krueger's keychain samples if that interests you. and the files he provided at the time might very well be in a way what you are looking for.
a short summary of all this was some saying you can't hear a difference, others said you could. and then with those saying you could there was those thinking they could hear ultrasounds, and those thinking it was all nasty IMD.
it was bloody and leaked over several forums, but IMO that's a great WAR&PEACE read to get a vision of all the opinions and all the technicalities that serve to try and explain them all.

 

[RRod]

  there was an apocalyptic fight over Arny Krueger's keychain samples if that interests you. and the files he provided at the time might very well be in a way what you are looking for.
a short summary of all this was some saying you can't hear a difference, others said you could. and then with those saying you could there was those thinking they could hear ultrasounds, and those thinking it was all nasty IMD.
it was bloody and leaked over several forums, but IMO that's a great WAR&PEACE read to get a vision of all the opinions and all the technicalities that serve to try and explain them all.

 
Thanks! I'll seek it out; seems like nothing short of toilet-paper-roll direction stirs up as much hate as sample/bit-rate discussions. Thank god Mahler died before calling for keychains in the last movement of the 10th!
 
I have a recording of Ligeti's symphonic poem for 100 metronomes. I clipped out a single tick, repeated it, and filtered out above 17kHz; couldn't pass the ABX.