What do you audiophiles think about Pono music?
- Thread starter Midpack
- Start date
All this talk is: ERD50
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
HAh-hah - so here's one theory on super-exotic-expensive speaker cables: they actually are designed to be BAD.
Yep, some of those things, in their quest to differentiate themselves and be 'special', can actually have some relatively high inductance, capacitance and/or resistance levels. Now wait a minute - isn't the purpose of a speaker cable to have low levels of all those? Sure, but then you won't 'hear' the cable! Why spend $5,000 for speaker cables you can't 'hear'?
So if you happen to have an overly-bright system/room, and you insert a magic-voodoo cable that slightly attenuates the highs... wow! You 'improved' the sound! Those $5,000 speaker cables really made a difference!
Of course, you could have done the same with a cheap pad on the tweeter, or with the tone control if your pre-amp has one, or any number of inexpensive and more controllable ways. But where's the bragging rights in that!
Yep, lot's of snake oil in high end audio. But I'm not really sure that a few extra bits of precision, or a slightly higher sample rate are snake oil.
And as skeptical as I am, I'm always a little wary to claim that someone else can't hear a difference in something. Being an amateur musician, even at my mediocre level I have found that you can become very attuned to very slight differences in sounds in an instrument.
I can recall that I would put a strip of felt under the strings at the tuning pegs of my acoustic guitar, because I could occasionally hear one of those little string segments at the head-stock 'ring'in sympathetic vibration when I hit certain chords loudly and sharply (staccato). It would drive me nuts. I don't think 99/100 people would have noticed it in an A/B test, and I bet many could not identify it in an A/B test even after they were told what to listen for. But it stood out like a sore thumb for me.
The ear/brain are complex and tricky. But there is a limit of plausibility, and you are into snake oil land. And I bet those $5,000 speaker cables would sound so much better with a Tice Clock!
-ERD50
NW-Bound
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
- Joined
- Jul 3, 2008
- Messages
- 35,712
OK, I am back after an hour or so of quality time with TaxAct, and a nap.
No, it's worse. The JBL L890 speaker is a 4-way design, with the highest crossover frequency listed as 20KHz. This means their "ultra tweeter" is put there to emit sound that nobody can hear, nor contained in regular source material such as CDs and MP3 files. However, it's ready for PONO music.
Thanks for the link.
What I found most relevant is the following paragraph.
But then, the next paragraph throws some confusion back into the mix.
Well, as I am sure that I cannot hear any of this, I am going to stick with my CDs and 320k MP3. It is depressing enough to recognize that my digitized cassettes are not worthy of the 16-bit 44.1KHz format.
No, it's worse. The JBL L890 speaker is a 4-way design, with the highest crossover frequency listed as 20KHz. This means their "ultra tweeter" is put there to emit sound that nobody can hear, nor contained in regular source material such as CDs and MP3 files. However, it's ready for PONO music.
Thanks for the link.
What I found most relevant is the following paragraph.
In 554 trials, listeners chose correctly 49.8% of the time. In other words, they were guessing. Not one listener throughout the entire test was able to identify which was 16/44.1 and which was high rate, and the 16-bit signal wasn't even dithered!
But then, the next paragraph throws some confusion back into the mix.
Another recent study investigated the possibility that ultrasonics were audible, as earlier studies had suggested. The test was constructed to maximize the possibility of detection by placing the intermodulation products where they'd be most audible. It found that the ultrasonic tones were not audible... but the intermodulation distortion products introduced by the loudspeakers could be.
Well, as I am sure that I cannot hear any of this, I am going to stick with my CDs and 320k MP3. It is depressing enough to recognize that my digitized cassettes are not worthy of the 16-bit 44.1KHz format.
I found that many mods did sound the same to me at first. But you could hear differences after listening for a while and discovering the different cues. I'm wary of listening tests that don't involve a training period and limit the decision period. That's where "golden ears", trained to listen for specific cues of particular distortions, can perform better than just grabbing guys off the street and seeing if they hear a difference in a one minute sample test.
The filter required for rolling off everything above 22 kHz has very significant effects well below 20kHz, including phase effects. Higher sample rates ease the filtering requirements and can result in less modification of the audible signal. And handle any left over >20kHz stuff without aliasing. The time attack theories might be correct, though I doubt the JBL supertweeter was phase aligned well enough to create a coherent time waveform.
I could certainly hear 20kHz as a kid, and the ultrasonic alarm systems in some stores of the day used to drive me crazy. So I might have had a shot at 23 KHz or better. 100 kHz would definitely have to be secondary effects.
The filter required for rolling off everything above 22 kHz has very significant effects well below 20kHz, including phase effects. Higher sample rates ease the filtering requirements and can result in less modification of the audible signal. And handle any left over >20kHz stuff without aliasing. The time attack theories might be correct, though I doubt the JBL supertweeter was phase aligned well enough to create a coherent time waveform.
I could certainly hear 20kHz as a kid, and the ultrasonic alarm systems in some stores of the day used to drive me crazy. So I might have had a shot at 23 KHz or better. 100 kHz would definitely have to be secondary effects.
M Paquette
Moderator Emeritus
But then, the next paragraph throws some confusion back into the mix.
Good thing there aren't intermodulation products produced in a live acoustic performance.
It takes a fair amount of volume before air starts to show nonlinear behavior that could produce intermodulation... ;-)
Another recent study investigated the possibility that ultrasonics were audible, as earlier studies had suggested. The test was constructed to maximize the possibility of detection by placing the intermodulation products where they'd be most audible. It found that the ultrasonic tones were not audible... but the intermodulation distortion products introduced by the loudspeakers could be.
[/QUOTE]Good thing there aren't intermodulation products produced in a live acoustic performance.
It takes a fair amount of volume before air starts to show nonlinear behavior that could produce intermodulation... ;-)
ERD50
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
I'm solidly with Animorph on this one. I'm definitely not claiming to be a 'golden ear', but I know that my ability to discern subtleties in music is miles above the average person who only has a more passing interest in music. I'm an amateur musician, and interested in music and music reproduction, and you just learn to pick up on things. Anyone who has become very involved in a particular field is far ahead of someone with only some acquaintance with that field.
If those listeners were not trained listeners, the results mean nothing to me. Those are the same people buying crappy systems with booming bass and shrieking tweeters pumped into double digit distortion, and think it sounds good. It's the same in any field - if you pulled people off the street to judge some craft beers, versus the guys/gals in my beer club, you'd get way different results. The guy off the street is drinking Miller Lite, and could not be relied on to detect subtle differences in brewing techniques/ingredients between two similar craft brews.
Several times, I've attended concerts with my friend and our wives, and during the intermission, me and my friend might comment on some specific sound problem. I recall once, the second acoustic guitar player was getting a huge bump in the lower mids, causing a very noticeable boom-iness when he played on the lower strings. This was not subtle to us at all, it was interfering with our enjoyment of the music. But as we talked, our wives looked at us like we were from outer space, they didn't hear anything. We were able to catch up with the performer during that intermission, complimented him on his playing, and then mentioned the boom-iness, and he agreed, and said he's trying to get the sound guy to smooth that out, and wanted to know where we were sitting, as it might be more prominent in certain areas of the hall. So it was not our imagination.
I've also found it can take time before a sound difference is apparent. See post # 4 in this thread.
I have a different philosophy on this, and I just digitized some pretty poor quality cassettes. I went with 16/44.1 for them - compression can create artifacts, and adding artifacts to something that is already distorted can magnify those issues. And lossless FLAC is only about 2x that mp3 @ ~ 720k, not really an archiving issue with cheap hard drives, and lossless means I can always create a compressed copy with no generational loss. Maybe a small point for some uses, but the drawbacks of lossless versus reasonable bit-rate mp3 are so slight, why not go lossless?
There was some great info in some of the earlier links on bit depth and dithering that I'll review and comment on tomorrow, and maybe get motivated to set up some of my own tests (spoiler alert - the info is leaning me to say that 'CD Quality' really is 'good enough', we will see if my testing agrees).
-ERD50
ls99
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
- Joined
- May 2, 2008
- Messages
- 6,509
Thanks for the link. Probably the most exhaustive elaboration I've ever read on sound reproduction. Being just a casual listener, the esoterica is interesting to me from the electronics point of view.
As an aside a few months ago on ebay I saw a SACD recording of a Sarah Brightman show sell for well over $800.- US. Rare it was supposed to be. Per the MIX discussion the listener's ability to identify the superior recording was less than chance.
HFWR
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
I'm also in the camp that knows music and plays an instrument, and I've attended hundreds if not thousands of performances, including as a stage or main system mixer. Given those parameters, I've heard both lousy mixing and lousy playing that others apparently didn't notice.
Having said that, years of loud stereos, loud guitar amps, loud machinery, etc. have resulted in tinnitus, and likely a diminished hearing range, though I haven't been tested. I started ripping music quite a few years ago, when storage was somewhat of a cost issue. I randomly looked at the rate for a few songs, and it ranged from 120kbs to 320kbs. Likely, on a good stereo in a quiet setting, I would notice the difference, but in the car, or with earbuds and an iPhone, I doubt there's much.
May have to do some listening tests to quantify that.
Having said that, years of loud stereos, loud guitar amps, loud machinery, etc. have resulted in tinnitus, and likely a diminished hearing range, though I haven't been tested. I started ripping music quite a few years ago, when storage was somewhat of a cost issue. I randomly looked at the rate for a few songs, and it ranged from 120kbs to 320kbs. Likely, on a good stereo in a quiet setting, I would notice the difference, but in the car, or with earbuds and an iPhone, I doubt there's much.
May have to do some listening tests to quantify that.
Pretty much, I’m just asking…. Please do not master the CD so it sounds best over FM, playing in a car, or on an mp3 player with headphones. That may well be the mass market, but the manufactures of car audio, mp3 players, etc can design their circuits to match the environment it will be used in.
I guess I might have mixed up Neil’s Archive project with PONO. PONO is likely an offshoot of that ever evolving project. From memory though when the Archive music was re mastered, the volume was not turned up and the audio compressed beyond recognition. I also thought he was asking other musicians to not allow their music to be master for loudness.
Last edited:
OP
OP
Midpack
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
Interesting discussion, if mostly over my head. But it's inspired me to get out several of the 400 CDs boxed up in our basement to listen to them. I've listened to nothing better than AAC and similar "resolution" MP3's for years - it'll be interesting to see if my old ears can detect and appreciate the difference.
ERD50
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
I'm going to update my earlier statement, and then hopefully get some work done on my taxes before we go out for the day...
Like I said earlier, if that test was done with average people, listening to average pop music, then it says that the average person, listening to average pop music, couldn't tell the difference. It's totally meaningless as to whether I could hear a difference, with good equipment and excellent recordings in my listening room.
When I was doing some equipment shopping, I took recordings with me that I knew to be very 'revealing', I wouldn't take pop tunes with heavy swooshy sounds all jammed on top of each other . Not that there's anything wrong with that music, it can be a 'fun' listen from time to time. But it isn't going to reveal subtle differences in equipment.
As an example, I found a site from one of the earlier links that had some interesting looking sample files and an 8-bit 16-bit blind test. I figured even through the defective analog port of my laptop (it has some noise bleeding into it from somewhere) I could tell the difference between 8 and 16, right? I've heard 8 bit recordings before, and they were not terrible, but they sure were less than anything I'd call 'hi fidelity'.
So before taking the test (you get 10 chances to identify 10 random picks), just like an ABX, you get to listen to the 8 and 16 samples as much as you want so you can accustom your ears to what each sounds like. Then I see the sample is the annoying (to me) tune 'Gangnam Style'. The snippet is all electronic whoops and swishy percussion stuff that are unidentifiable gated-noise type sounds. Both the 8-bit and 16-bit rates sounded terrible to me, and they sounded the same to me. I didn't even bother taking the test, I know I could not hear the difference with that source material.
So what does that prove - nothing useful. I guess it proves that noise bursts sound pretty much the same as 8 bit or 16 bit. But I don't care. I want to know how an acoustic guitar, going from a sharp staccato chord to a gently plucked, 'barely there' note sounded, one where you can just barely hear the players fingers caressing the strings, maybe a bit of fret noise as they bend the note and the string rubs across the fret, or a harmonic pluck. That is what differentiates the listening experience on great recordings/equipment, versus listening to pop music in your car.
BAD BLIND TEST >>> : The 16-bit v/s 8-bit Blind Listening Test
Was that the kind of music used to 'test' 16 versus 24? Of course listeners could not tell, they probably could not ID 16 versus 14, 12, 10, or even 8!
And if (and that is a big IF) 24/96 gets me closer to that, I'll consider some upgrading of my system - other than the source material, the cost is minimal, a few hundred for a new DAC, maybe an extra X MBytes of hard drive space, depending how much source material I find.
I found other samples of 8-16, not set up as a blind test, but the difference was clear, and much more revealing. Even though the source didn't sound that great (a simple male spoken word phrase that didn't seem particularly well recorded to me), they spoke a phrase, then repeated it, dropping 6db at a time. The 8 bit fell into the noise much faster, and started sounding more distorted at the lower levels. Very easy to hear he difference.
I lost those links, will have to search again later, but I was amazed at how much dithering helped. I know the basics of dithering, and was thinking it was just a 1 LSB pk-pk level of white noise added, mainly to keep that LSB from getting 'stuck' at one level or the other, smoothing it with random noise (better than non-random noise, or quantitizing artifacts). And I think understand how it can actually increase the dynamic range a little (the trade off being a little more noise?). But according to some of what I've read, but not fully digested yet, is that there are a few more 'tricks' with dithering. Apparently, dithering works about as well when you reduce the bandwidth of the noise, and move it to a range that the human ear is less sensitive (high frequencies). So you get the dynamic range improvements, with only a small amount of noise added? Hmmmm.
Now I don't know where dithering occurs. Is it in the recording? Is it done by my DAC? What about my ADC when I digitize my LPs - no dithering, and it all happens later in my DAC? I would think post processing would be better - any advances in the technique (or personal preferences) could be done later to existing recordings.
But taxes are calling.
-ERD50
Like I said earlier, if that test was done with average people, listening to average pop music, then it says that the average person, listening to average pop music, couldn't tell the difference. It's totally meaningless as to whether I could hear a difference, with good equipment and excellent recordings in my listening room.
When I was doing some equipment shopping, I took recordings with me that I knew to be very 'revealing', I wouldn't take pop tunes with heavy swooshy sounds all jammed on top of each other . Not that there's anything wrong with that music, it can be a 'fun' listen from time to time. But it isn't going to reveal subtle differences in equipment.
As an example, I found a site from one of the earlier links that had some interesting looking sample files and an 8-bit 16-bit blind test. I figured even through the defective analog port of my laptop (it has some noise bleeding into it from somewhere) I could tell the difference between 8 and 16, right? I've heard 8 bit recordings before, and they were not terrible, but they sure were less than anything I'd call 'hi fidelity'.
So before taking the test (you get 10 chances to identify 10 random picks), just like an ABX, you get to listen to the 8 and 16 samples as much as you want so you can accustom your ears to what each sounds like. Then I see the sample is the annoying (to me) tune 'Gangnam Style'. The snippet is all electronic whoops and swishy percussion stuff that are unidentifiable gated-noise type sounds. Both the 8-bit and 16-bit rates sounded terrible to me, and they sounded the same to me. I didn't even bother taking the test, I know I could not hear the difference with that source material.
So what does that prove - nothing useful. I guess it proves that noise bursts sound pretty much the same as 8 bit or 16 bit. But I don't care. I want to know how an acoustic guitar, going from a sharp staccato chord to a gently plucked, 'barely there' note sounded, one where you can just barely hear the players fingers caressing the strings, maybe a bit of fret noise as they bend the note and the string rubs across the fret, or a harmonic pluck. That is what differentiates the listening experience on great recordings/equipment, versus listening to pop music in your car.
BAD BLIND TEST >>> : The 16-bit v/s 8-bit Blind Listening Test
Was that the kind of music used to 'test' 16 versus 24? Of course listeners could not tell, they probably could not ID 16 versus 14, 12, 10, or even 8!
And if (and that is a big IF) 24/96 gets me closer to that, I'll consider some upgrading of my system - other than the source material, the cost is minimal, a few hundred for a new DAC, maybe an extra X MBytes of hard drive space, depending how much source material I find.
I found other samples of 8-16, not set up as a blind test, but the difference was clear, and much more revealing. Even though the source didn't sound that great (a simple male spoken word phrase that didn't seem particularly well recorded to me), they spoke a phrase, then repeated it, dropping 6db at a time. The 8 bit fell into the noise much faster, and started sounding more distorted at the lower levels. Very easy to hear he difference.
I lost those links, will have to search again later, but I was amazed at how much dithering helped. I know the basics of dithering, and was thinking it was just a 1 LSB pk-pk level of white noise added, mainly to keep that LSB from getting 'stuck' at one level or the other, smoothing it with random noise (better than non-random noise, or quantitizing artifacts). And I think understand how it can actually increase the dynamic range a little (the trade off being a little more noise?). But according to some of what I've read, but not fully digested yet, is that there are a few more 'tricks' with dithering. Apparently, dithering works about as well when you reduce the bandwidth of the noise, and move it to a range that the human ear is less sensitive (high frequencies). So you get the dynamic range improvements, with only a small amount of noise added? Hmmmm.
Now I don't know where dithering occurs. Is it in the recording? Is it done by my DAC? What about my ADC when I digitize my LPs - no dithering, and it all happens later in my DAC? I would think post processing would be better - any advances in the technique (or personal preferences) could be done later to existing recordings.
But taxes are calling.
-ERD50
Birdie Num Nums
Thinks s/he gets paid by the post
That is one of the primary reasons Neil Young is so into this 'sound fidelity' issue. He's got his "Old Black" Les Paul running through that Whizzer and a bunch of other gear to get his distinctive heavy rock sound, and then that vintage Martin acoustic (formerly owned by Hank Williams) that emits such a warm, gorgeous tone on Neil's softer stuff. A wide range of sounds!
NW-Bound
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
- Joined
- Jul 3, 2008
- Messages
- 35,712
I am going to make this post, then I need to get back to my TaxAct too.
We were talking about intermodulation (IM) distortions. But it occurs to me that the speakers themselves are capable of introducing plenty. Yes, and even perfectly working ones, due to the way they work.
We talked in another thread about how a speaker with fewer drivers may be better than one with many, due to less phasing difference between drivers in the crossover transition regions. So, think of a mid-bass driver that has to cover from 100Hz up to say 1KHz. When driven with two tones, one at 100Hz that requires the cone to have an excursion of a good fraction of an inch and one at 1KHz that requires a minuscule vibration, what happens with this driver? The 1KHz vibration is riding on a much larger to-and-fro movement of the speaker cone. The result is IM due to the Doppler effect, similar to a car horn sounding higher or lower depending on whether the car is approaching or receding from a listener.
The above two-tone test is very easy to experiment for myself, and I will get to that sometimes. I can excite a driver with both tones, and see if I can hear the difference compared to driving two separate drivers one with each tone. If I cannot hear the difference, I'd better be able to see the difference with a spectrum analyzer.
There are other things like taxes that must be taken care of, but the above is fun. See how an ER can entertain himself with existing toys but in a different way for so little cost?
We were talking about intermodulation (IM) distortions. But it occurs to me that the speakers themselves are capable of introducing plenty. Yes, and even perfectly working ones, due to the way they work.
We talked in another thread about how a speaker with fewer drivers may be better than one with many, due to less phasing difference between drivers in the crossover transition regions. So, think of a mid-bass driver that has to cover from 100Hz up to say 1KHz. When driven with two tones, one at 100Hz that requires the cone to have an excursion of a good fraction of an inch and one at 1KHz that requires a minuscule vibration, what happens with this driver? The 1KHz vibration is riding on a much larger to-and-fro movement of the speaker cone. The result is IM due to the Doppler effect, similar to a car horn sounding higher or lower depending on whether the car is approaching or receding from a listener.
The above two-tone test is very easy to experiment for myself, and I will get to that sometimes. I can excite a driver with both tones, and see if I can hear the difference compared to driving two separate drivers one with each tone. If I cannot hear the difference, I'd better be able to see the difference with a spectrum analyzer.
There are other things like taxes that must be taken care of, but the above is fun. See how an ER can entertain himself with existing toys but in a different way for so little cost?
walkinwood
Thinks s/he gets paid by the post
This discussion fascinates me!
I did a blind test with DW (whose senses are more discriminating than mine) with two 192 Kb/sec MP3 files ripped from the Let it Be Naked CD using the default mp3 encoder in windows media center and the LAME encoder using Exact Audio Copy. I was surprised that even I could tell the difference - but since I knew which was which, I asked DW to listen to them. She thought the LAME encoder resulted in a more "rounded" and "warmer" sound & liked it better.
Who knew that the method of ripping CDs could make a difference!
I did a blind test with DW (whose senses are more discriminating than mine) with two 192 Kb/sec MP3 files ripped from the Let it Be Naked CD using the default mp3 encoder in windows media center and the LAME encoder using Exact Audio Copy. I was surprised that even I could tell the difference - but since I knew which was which, I asked DW to listen to them. She thought the LAME encoder resulted in a more "rounded" and "warmer" sound & liked it better.
Who knew that the method of ripping CDs could make a difference!
Although I can see the Doppler effect causing problems in some situations, I think it may be kind of bogus for many. Real sounds have a time waveform. It is that time waveform that should be reproduced. And it can definitely be reproduced by a single driver without "Doppler distortion", although conceptually a single driver would seem the worst case for Doppler.
A speaker system designed for phase coherence will minimize errors in the time waveform. A speaker with an infinite number of drivers, one for each frequency if you could do such a thing (an FFT speaker!), would be a real mess to get time aligned. Even though Doppler-wise it might be ideal, it might never come to life.
Magneplanars were well known as particularly phase coherent, able to come close to reproducing actual squarewaves, unlike almost all other speakers amazingly enough. Dunlavy's were the only cone speaker I found that could match the Magneplanar's sound, with the added benefit of playing louder and deeper.
Birdie Num Nums
Thinks s/he gets paid by the post
To date three of the six (or seven) top recording companies are aboard the Pono program:
Young has noted that all three major music groups -- Warner Bros., Universal and Sony -- are participating in the PonoMusic online music store. The PonoPlayer will have a list price of $399 and be capable of storing 1,000 to 2,000 high-resolution digital albums, according to the PonoMusic statement.
Perhaps more recording companies to follow?
However, initial costs of the music to the consumer are expected to be somewhat high, I read somewhere--something like ~$25 per album. Eventually could go lower, I guess, if this Pono idea does indeed grow in popularity.
Last edited:
NW-Bound
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
- Joined
- Jul 3, 2008
- Messages
- 35,712
Yes, a speaker's job is to convert an electrical waveform into a pressure wave analogous to that excitation. If I feed the sum of a 100Hz and a 1KHz sine waves to a speaker, and measure a 100Hz sound along with a 1KHz that's frequency-modulated (FM'ed) by the 100Hz, then the speaker has introduced errors.
The question I have for myself is that how pronounced this could be, and if I cannot hear it, can I at least measure it? If it is measurable, then perhaps someone with a golden ear can hear it. And if nothing else, I can at least say that, yes, that Doppler effect exists as I surmised but it is negligible.
I have speakers with different sizes of drivers to experiment with when I get to this. I expect that a smaller driver which needs more cone excursion for the bass will have more intermodulation distortion than a larger woofer that does not have to move its cone as much. Non-linearity of the cone suspension elements, meaning the proportionality of restoring force vs. displacement (spring constant), can also add distortion at large cone movements.
Well, a large number of drivers would not be practical, and would run into other problems. It certainly creates more serious problems than the one it tries to solve.
A compromise has already been made in practice. A subwoofer would offload from the main speakers the low frequency components that require large cone excursions. This is the kind of things that I like to investigate.
Because planar speakers have a large radiating surface, they do not require as large an excursion of the radiating element compared to cone speakers for the same loudness. So, despite other drawbacks, they may have the advantage when it comes to intermodulation distortion. That's just a guess on my part, of course.
Last edited:
ERD50
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
So after spending too much time on taxes today, I got back to some of these nagging audio questions. I think I found something pretty interesting that might help me decide whether I should spend any more time considering higher bit rate audio equipment (though it really isn't that expensive anyway).
I found this site that had some free 24/96 downloads of what I'd expect to be very well recorded and revealing music:
High Resolution Music DOWNLOAD services .:. FLAC in free TEST BENCH
So I downloaded 5 tracks ( ~ 3 ~4 minutes each) - these are promos to get you to buy their full albums. My plan is to import these into Audacity sound editing program, and convert them to 16,15,14,13,12,11,10 bit versions. Then I'll put those in a playlist, set it to 'shuffle' and see if I can ID any of the various bit rates. I can make this double blind by making notes, and doing something to record which ones came up in the rotation (or maybe just add a voice announcement at the end of each file, after I've graded them).
If I can't reliably pick out 13-14 bit recordings from 16 bit, then I doubt I'd pick out 24 from 16. It must be diminishing returns.
In the process, I also found out more about dithering, and in practice it really can increase the theoretical 96db limit of 16 bit audio, without increasing the audible noise very much (which is already very low). And Audacity has the preferred dither methods included, these are used only when down-converting the bit resolution.
We will see what the listening tests tell me, but I anticipate that I should spend more time on room treatments, and just plain listening over pursuing higher bit rates.
Interesting. These compression algorithms are actually very impressive at what they do and very complex, and since they do so much, maybe it isn't too surprising that different implementations would sound different? I know the codecs for VOIP and mobile phones are optimized for voice over music (ever notice just how bad the 'hold' music sounds on a cell phone - they should avoid hold music, or use very simple stuff that can encoded reasonably well). But 192k encoding is getting up there in quality, so maybe you both are true 'Golden Ears"?
But that always takes me back to my philosophy of maintaining at least the archive at lossless. For serious listening I use lossless, for casual and where storage is limited, I can convert from the lossless to whatever format I want w/o any generational losses.
-ERD50
I found this site that had some free 24/96 downloads of what I'd expect to be very well recorded and revealing music:
High Resolution Music DOWNLOAD services .:. FLAC in free TEST BENCH
So I downloaded 5 tracks ( ~ 3 ~4 minutes each) - these are promos to get you to buy their full albums. My plan is to import these into Audacity sound editing program, and convert them to 16,15,14,13,12,11,10 bit versions. Then I'll put those in a playlist, set it to 'shuffle' and see if I can ID any of the various bit rates. I can make this double blind by making notes, and doing something to record which ones came up in the rotation (or maybe just add a voice announcement at the end of each file, after I've graded them).
If I can't reliably pick out 13-14 bit recordings from 16 bit, then I doubt I'd pick out 24 from 16. It must be diminishing returns.
In the process, I also found out more about dithering, and in practice it really can increase the theoretical 96db limit of 16 bit audio, without increasing the audible noise very much (which is already very low). And Audacity has the preferred dither methods included, these are used only when down-converting the bit resolution.
We will see what the listening tests tell me, but I anticipate that I should spend more time on room treatments, and just plain listening over pursuing higher bit rates.
Interesting. These compression algorithms are actually very impressive at what they do and very complex, and since they do so much, maybe it isn't too surprising that different implementations would sound different? I know the codecs for VOIP and mobile phones are optimized for voice over music (ever notice just how bad the 'hold' music sounds on a cell phone - they should avoid hold music, or use very simple stuff that can encoded reasonably well). But 192k encoding is getting up there in quality, so maybe you both are true 'Golden Ears"?
But that always takes me back to my philosophy of maintaining at least the archive at lossless. For serious listening I use lossless, for casual and where storage is limited, I can convert from the lossless to whatever format I want w/o any generational losses.
-ERD50
ERD50
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
A little update (which maybe should go in this other thread):
So after listening to those 24 bit downloads (with my 16 bit equipment), I decided to just use some of my 16 bit CD rips that I was familiar with as a source.
I did not see any direct way to export to lower bit rez in the Audacity program - it uses 32-bit float and tries to keep you from degrading the quality, but I posted some possible anomalies I found to the Audacity forum, and someone responded with a Nyquist prompt code segment that would convert to any number of steps (so just do the math from power-of-two bits to steps). So I did a quick test of original 16 bit and 12 bit, and along the way I had done a test of 8 bit audio.
The first amazing thing to me was that 8-bit audio sounds much better than I'd expect! Recall, 8 bit means that every sample is assigned one of 256 levels (or 255 signed I think, but close enough). The step sizes are easily seen on a computer screen, the waveform is visibly 'choppy'. Compared to a 16 bit conversion, with 65,535 steps - even zoomed in multiple times, the step sizes are not discernible on a screen with 768 pixels vertical.
Now 8-bit clearly is not 'audiophile', it is noisy and grainy, but considering the huge difference in rez, it was surprising to me that it sounded as good as it did.
So I did a quick test of 16 bit versus 12 bit (4095 steps) rez. On a quick listen, there wasn't anything that jumped out at me. When I listened to the end, as the guitar chord dies out, I could hear some gurgling/gargling noises as it jumps from one level to the next and neither level really represents the 'true' sound, and you can hear some noise. But I actually got near the speaker to hear that. So there definitely is an audible difference, but maybe still on the subtle side* ( *see below).
But if the difference between 12 and 16 bits isn't something that just smacks me in the face, I doubt I will hear anything going form 16 to 24. Recall that going from 12 to 16 is DOUBLING the available steps 4 times over again. My gut is telling me that 16 bit really is at the edge of diminishing returns.
But even if that is the case, it still makes sense to use 24 bit in the studio. It gives more room for error, and processing can end up truncating bits if it is all done at 16 bit levels. But that processing can be done in 24 bits, and still be accurate when finally converted to 16 bit for distribution.
*(from above) - I must say though, that I am pretty certain that a quick A-B comparison is not enough. I have found that small differences in sound quality affect the listening experience over time, like listening for 20 minutes or more at once. After a while, you just sense that something is missing, or there is some distortion somewhere, even if you can't describe it. It does not need to be obvious to be a problem for a listening session.
A rough parallel is a car seat. It might feel great when you first sit it in it, but take a 4 hour drive, and it can really bother you. I'd say sound is like that too. Little differences add up over time, and either feel right or wrong.
So I will do more testing in the future, going to all the bit depths between 12 and 16, but just thought I'd update at this point.
-ERD50
So after listening to those 24 bit downloads (with my 16 bit equipment), I decided to just use some of my 16 bit CD rips that I was familiar with as a source.
I did not see any direct way to export to lower bit rez in the Audacity program - it uses 32-bit float and tries to keep you from degrading the quality, but I posted some possible anomalies I found to the Audacity forum, and someone responded with a Nyquist prompt code segment that would convert to any number of steps (so just do the math from power-of-two bits to steps). So I did a quick test of original 16 bit and 12 bit, and along the way I had done a test of 8 bit audio.
The first amazing thing to me was that 8-bit audio sounds much better than I'd expect! Recall, 8 bit means that every sample is assigned one of 256 levels (or 255 signed I think, but close enough). The step sizes are easily seen on a computer screen, the waveform is visibly 'choppy'. Compared to a 16 bit conversion, with 65,535 steps - even zoomed in multiple times, the step sizes are not discernible on a screen with 768 pixels vertical.
Now 8-bit clearly is not 'audiophile', it is noisy and grainy, but considering the huge difference in rez, it was surprising to me that it sounded as good as it did.
So I did a quick test of 16 bit versus 12 bit (4095 steps) rez. On a quick listen, there wasn't anything that jumped out at me. When I listened to the end, as the guitar chord dies out, I could hear some gurgling/gargling noises as it jumps from one level to the next and neither level really represents the 'true' sound, and you can hear some noise. But I actually got near the speaker to hear that. So there definitely is an audible difference, but maybe still on the subtle side* ( *see below).
But if the difference between 12 and 16 bits isn't something that just smacks me in the face, I doubt I will hear anything going form 16 to 24. Recall that going from 12 to 16 is DOUBLING the available steps 4 times over again. My gut is telling me that 16 bit really is at the edge of diminishing returns.
But even if that is the case, it still makes sense to use 24 bit in the studio. It gives more room for error, and processing can end up truncating bits if it is all done at 16 bit levels. But that processing can be done in 24 bits, and still be accurate when finally converted to 16 bit for distribution.
*(from above) - I must say though, that I am pretty certain that a quick A-B comparison is not enough. I have found that small differences in sound quality affect the listening experience over time, like listening for 20 minutes or more at once. After a while, you just sense that something is missing, or there is some distortion somewhere, even if you can't describe it. It does not need to be obvious to be a problem for a listening session.
A rough parallel is a car seat. It might feel great when you first sit it in it, but take a 4 hour drive, and it can really bother you. I'd say sound is like that too. Little differences add up over time, and either feel right or wrong.
So I will do more testing in the future, going to all the bit depths between 12 and 16, but just thought I'd update at this point.
-ERD50
Sounds reasonable. Looked at another way, you're giving up dynamic range and leaving the resolution the same. So something that has a wide dynamic range would make the best test.
The bane of audiophiles, everything sounds fantastic until you get used to it and start hearing things you think could be improved...
The bane of audiophiles, everything sounds fantastic until you get used to it and start hearing things you think could be improved...
ERD50
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
An update to my listening tests:
A few months back I actually did some additional testing, and I just decided to document this for myself and anyone who might be interested.
Cut to the chase for those not interested in the technical details/background: I have determined, that for myself, 16 bit audio has plenty of resolution, and I will no longer put any effort into 'upgrading' to 24 bit equipment or source material (but 24 bit still makes sense in the recording studio and for post processing).
I played around with various ways to test myself to determine if 16 bit was lacking. As I've mentioned, subjective listening tests are rather tedious, you need to do them blind, and I'm convinced that some subtle differences are not apparent until you've listened for a long time (like my analogy with the car seat - it might be fine for a 15 minutes drive, but is like torture for a straight 4 hours behind the wheel).
Then I stumbled across a more objective test method. I decided that if I simply cannot hear a sound at a certain dB level below full scale, that it made no difference if my system had the resolution to reproduce that 'sound' or not. Recall that each bit of resolution provides ~ 6 dB of dynamic range. So 16 bit source theoretically provides 96dB dynamic range, but there are arguments that you lose a bit or two to dithering - or does the dithering increase the effective bit resolution? Save that argument for later...
For my first test, I used the sound program 'Audacity' to generate a 220Hz Saw-tooth wave (lots of harmonics throughout the audio range, very 'buzzy' and identifiable above background noise). With a good pair of headphones (to help block out room noise), I adjusted the sound level to LOUD (not painfully loud, but higher than I would ever listen to music for an extended time), and then used the program to attenuate the sound in specific amounts.
At -78 ~ -84 dB from that full scale LOUD reference, it was getting soft and softer (per my notes). At -87 dB, 'very soft', and by -90 dB I could barely hear it if I concentrated fully on that tone and cupped the headphones a bit (which is 'cheating' - that is effectively raising the level), and by -93 dB, I rated the sound as nothing, nada - it ceased to exist, it was pushing up daisies. That sound was dead to me.
So it would seem that for a pure tone, which I think represents a worst-case and somewhat unrealistic test, 16-bit, 96 dB range is sufficient, but maybe, just maybe a little marginal?
Then I tried again with real music ('Blake's Rag' - Bluesmen CD, Cephas & Wiggins), and setting the full scale to LOUD.
My notes say that just - 48 dB from that full scale was barely audible! I could pick up a little sound at -60 dB, but I really had to concentrate, and switching back to the full scale reference sounded ridiculously LOUD. At -66 dB, I could not detect anything at all.
I was surprised by those results, and I think they are much more meaningful than A-B listening tests. If I can't hear a sound at all at -66 dB, in isolation, at LOUD levels, with headphones on and my full powers of concentration, then I really don't think going beyond a 96 dB range is going to improve anything for me.
OK, I can hear a devil's advocate saying that the reason I couldn't hear the sound lower was because of the limited resolution. Hmmmm, maybe - but the sound wasn't masked by noise, it was just too low in volume for my ears to detect. The -60 dB that I could barely hear is still 36 dB above the limit of a 16 bit coding, so I don't think it's a coding issue.
Thoughts?
-ERD50
A few months back I actually did some additional testing, and I just decided to document this for myself and anyone who might be interested.
Cut to the chase for those not interested in the technical details/background: I have determined, that for myself, 16 bit audio has plenty of resolution, and I will no longer put any effort into 'upgrading' to 24 bit equipment or source material (but 24 bit still makes sense in the recording studio and for post processing).
I played around with various ways to test myself to determine if 16 bit was lacking. As I've mentioned, subjective listening tests are rather tedious, you need to do them blind, and I'm convinced that some subtle differences are not apparent until you've listened for a long time (like my analogy with the car seat - it might be fine for a 15 minutes drive, but is like torture for a straight 4 hours behind the wheel).
Then I stumbled across a more objective test method. I decided that if I simply cannot hear a sound at a certain dB level below full scale, that it made no difference if my system had the resolution to reproduce that 'sound' or not. Recall that each bit of resolution provides ~ 6 dB of dynamic range. So 16 bit source theoretically provides 96dB dynamic range, but there are arguments that you lose a bit or two to dithering - or does the dithering increase the effective bit resolution? Save that argument for later...
For my first test, I used the sound program 'Audacity' to generate a 220Hz Saw-tooth wave (lots of harmonics throughout the audio range, very 'buzzy' and identifiable above background noise). With a good pair of headphones (to help block out room noise), I adjusted the sound level to LOUD (not painfully loud, but higher than I would ever listen to music for an extended time), and then used the program to attenuate the sound in specific amounts.
At -78 ~ -84 dB from that full scale LOUD reference, it was getting soft and softer (per my notes). At -87 dB, 'very soft', and by -90 dB I could barely hear it if I concentrated fully on that tone and cupped the headphones a bit (which is 'cheating' - that is effectively raising the level), and by -93 dB, I rated the sound as nothing, nada - it ceased to exist, it was pushing up daisies. That sound was dead to me.
So it would seem that for a pure tone, which I think represents a worst-case and somewhat unrealistic test, 16-bit, 96 dB range is sufficient, but maybe, just maybe a little marginal?
Then I tried again with real music ('Blake's Rag' - Bluesmen CD, Cephas & Wiggins), and setting the full scale to LOUD.
My notes say that just - 48 dB from that full scale was barely audible! I could pick up a little sound at -60 dB, but I really had to concentrate, and switching back to the full scale reference sounded ridiculously LOUD. At -66 dB, I could not detect anything at all.
I was surprised by those results, and I think they are much more meaningful than A-B listening tests. If I can't hear a sound at all at -66 dB, in isolation, at LOUD levels, with headphones on and my full powers of concentration, then I really don't think going beyond a 96 dB range is going to improve anything for me.
OK, I can hear a devil's advocate saying that the reason I couldn't hear the sound lower was because of the limited resolution. Hmmmm, maybe - but the sound wasn't masked by noise, it was just too low in volume for my ears to detect. The -60 dB that I could barely hear is still 36 dB above the limit of a 16 bit coding, so I don't think it's a coding issue.
Thoughts?
-ERD50
Birdie Num Nums
Thinks s/he gets paid by the post
Beatles Catalog on Pono
Paul McCartney "reboots" Beatles catalog for Neil Young's Pono music player:
Paul McCartney "reboots" Beatles catalog for Neil Young's Pono music player:
Neil Young’s high end music player and library, Pono, has scored a coup. They’ve secured the Beatles catalog for download in what’s known as loss-less digital.
Paul McCartney himself is said to be supervising the transfer of the Beatles’ crown in the jewel recordings so they can be added to the Pono catalog which can then be purchased and downloaded into the $399 Pono player.
Paul McCartney himself is said to be supervising the transfer of the Beatles’ crown in the jewel recordings so they can be added to the Pono catalog which can then be purchased and downloaded into the $399 Pono player.
Kimo
Recycles dryer sheets
There is absolutely a very noticeable difference between SACD and CD, at least on my system. I have a hard time listening to a regular CD now. I just wish SACD had "survived" so I could have replaced all my CD's instead of just a few....I know, there are places to buy SACD's but what I mean is I wish they survived so that everything was available in SACD.
HFWR
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
Call me skeptical, but a 69yo who has been standing in front of guitar amps cranked to eleven for the past 50+ years likely cannot hear any discernible difference...
http://nypost.com/2015/01/11/do-consumers-really-care-about-digital-quality/
Sent from my iCouch using Early Retirement Forum
http://nypost.com/2015/01/11/do-consumers-really-care-about-digital-quality/
Sent from my iCouch using Early Retirement Forum
ERD50
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
My ears are still pretty good (some mild tinnitus though), but the tests I did in post 71 convinced me I won't hear a difference, and should concentrate on other things, like the room.
Well, I will never argue with what someone else says they can hear, but I am curious. It would be interesting if you would repeat the experiment I did. To my thinking, if I can't hear sounds below a certain threshold, and that threshold is well within the range of what 16 bit can reproduce, I don't see how more bits can do anything for me.
All those extra bits can do is more accurately represent things that I can't hear. If a sound is at the threshold of my hearing, I would not be able to tell a sine wave from a square wave - the harmonics would all be lower than that threshold.
-ERD50
Similar threads
- Locked
