Who's done a 88/96k vs 44/48k listening test?

[Zep Dude]

I'd be interested in your results -especially if they were well constructed, blind tests. Did you consistently choose the higher rate? Did you find better imaging or clarity (not just in the high frequencies) with the higher rates?

Thanks

[ekuehnl]

Personally, I didn't need to do a blind test. I put a 48kHz rig up against 96kHz, recorded guitar, bass and vocals, and the difference was beyond obvious. A guy off the street could pick the 96. Both high frequency details and imaging are vastly improved.

[Zep Dude]

Ekuehnl, what were the details of this test? Were the converters and gear all the same except the sample rate? Which gear did you use -was this with the new PTHD? Did you play once and split the signal into the two systems or was it two different performances? Thanks

[Nika]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:<HR>Originally posted by ekuehnl:
Personally, I didn't need to do a blind test. I put a 48kHz rig up against 96kHz, recorded guitar, bass and vocals, and the difference was beyond obvious. A guy off the street could pick the 96. Both high frequency details and imaging are vastly improved.<HR></BLOCKQUOTE>

Yes, I'm most amused by this. You didn't need to do anything scientific like a blind test because your unscientific and uncontrolled listening test was so inconclusive that it left no question for you? The entire notion that you didn't need a blind test and the method used totally invalidates any results you have.

Were you using the same converters to do the test? What converters were they? How was the material split between the two?

All that your test indicates so far is that on some unknown converter (assuming you used the same converter between the two tests) there is a difference so long as only the things are held in check between the two that you held in check.

Please tell us, what converters were you using? Because the fact that they don't sound the same at 48 and 96 is a strong indicator of how bad those converters are. Any good converter should sound the same between the two different sample rates. So please, which ones were you using? Or were they really not the same? I'd like to avoid those converters in my own rig if they were really as bad as you say they are..

Nika.

[Park Seward]

This is one way to measure the difference between two items.

Double Stimulus Continuous Quality Scale (DSCQS) Method Instructions

Dear subjects, Thank you for participating in this test. In the DSCQS tests, a series of two pairs of sequences will be displayed on the monitor. Your task is to evaluate the QUALITY of BOTH test sequences in each pair. You do that by marking one point on each of the following rating scales:

=======A===B==========
Excellent
Good
Fair
Poor
Bad

The rating scale on the left is for grading the first sequence (A), the rating scale on the right is for grading the second sequence (B). Please use an horizontal mark (no crosses or other signs that could make your vote ambiguous). Each evaluation must reflect your opinion of the global quality of the whole sequence. Therefore, only vote after the end of the second sequence and base your evaluation on the entire duration of each sequence. Do not hesitate to rate a sequence either at the top or bottom of the scale, if that is how you believe it should be rated. A voting form will be distributed before each test session. On this form will be a series of rating scales, like the ones above, one scale pair for each sequence. The scale pairs are numbered sequentially. Use scale pair 1 for the first sequence pair, scale pair 2 for the second sequence pair and so on. The first sequence you will be see will be announced by the message "A" (standing for 2 seconds) then the second sequence you will be see will be announced by the message "B" (standing for 2 seconds) Then the two sequences will be shown again and the messages will change into "A*" and "B*". Finally you will see the message "VOTE N". Look at the scoring sheet and check for the correct number. Then mark the box corresponding to the quality level you have chosen. During these tests do not talk with other assessors or comment on the sequences you have seen. Before recording your vote, always check to be sure you are using the correct scale on the score sheet. Finally, it is important that you keep your concentration throughout the test session. Now try this evaluation procedure in a practice session. You will see a series of pairs of sequences using the exact same timing as will be used during an actual test session. This will allow you to become familiar with the timing of the test and to practice using the rating scales. If you have any questions, please ask them now.

[Zep Dude]

Hey, you know what, I just realized last night that I am actually fully equiped to run my own listening test. Turns out my Apogee PSX-100SE can switch quite readily between 48 and 96k on the AD side and simply pass the signal live right back out of it's DAC. Although I can't record, I can do a live listening test. And so I have!

This wouldn't exactly be accepted for submission by a scientific journal, but I'm quite satisfied with the results.

I had someone consecutively sing, play acoustic guitar, play djembe and hit a cymbal while I blindly switched the Apogee between 48 and 96k.

My equipment consisted of two mics (Neuman KM184 and a AT 4047) fed to Mastering Lab mic pres and direct to the Apogee. The Apogee was set to AD convert at the desired sample rate (48k and 96K) and then send the signal directly out it's own DA. From there the signal went to a McIntosh 2200 power amp via Kimber KCAG cable and then to DAS monitor 8's.

Not having anyone in the control room at the moment to switch for me I employed a highly scientific method of randomization. This consisted of me closing my eyes and pressing repeatedly the "2x" button (which toggles between 48 and 96k) while reciting multiplication tables in my head until I had no idea where the button was set and furthermore felt my sixth grade teacher was about the enter the room and give a pop quize.

I then simply listened for about 20-30 seconds (still with eyes closed) and toggled the "2x" button and listened again. I decided if this sounded better or worse and then switched back one more time to confirm my assessment. Then I opened my eyes and checked the result.

I tested only two times for each instrument, primarily because I felt the differences were evident enough that I was not just "shooting in the dark" with my decisions. Each time without fail I chose the 96K rate as sounding more open and clear in the high end. The dead giveaway was that there simply was more "air" with the 96k setting. Of the different instuments, predictably the cymbal hit was the most obvious -I knew immediately upon switching that I had gone to 96k from 48k. It sounded like a high frequency filter had been removed (pretty close to the reality). The djembe I didn't expect to hear because it's not really a high endy instrument. However there was more of a sense of "air" around the instrument and clueing into that I was able to pick the 96k both times. This was the same for acoustic guitar and vocal -more open in the highs. I would classify the differences as subtle but noticable and with the cymbal quite noticable.

Everything was mic'd from about 3 feet away in a wide stereo pattern. I wanted to blend in a good amount of room sound in order to capture any acoustic interactions. I did not sense any shift or tightening of the stereo image however given that these were all mono point sources this may have been asking too much.

Also, one point about listening tests. I find that about 20-35 seconds of exposure is a good amount of time before switching. It allows you to become very familiar with the sonic qualities of "A" without yet being fatigued or losing focus. The switch then to "B" presents itself at a time where familiarity and focus are optimal. Tests that drag on too long or require too much switching become meaningless because one simply looses focus and sensitivity. I also believe (in some cases) to pre exposing myself to A and B. This is simply because sometimes one needs to figure out where the differences might occur and therefore where you should be listening.

The only thing missing from this test was the ability to switch between the actual source (signal comming from the preamps) and the converted sounds. This would have allowed me to hear which sample rate sounded closer to the original (it's possible that the 96k rate was hyping the high end). I may run another test comparing 96k to source just to see if there was any additional highs being added. Although the unit was set to 24 bit without any UV22 or soft limit it is remotely possible that something other than a shifting of the high frequency filter was happening.

Perhaps, Nika, you might know if there is any noise shaping or concentration of high frequency energy that might occur with the 96k setting?

[EJolson]

Great post! And great idea. Thanks Zep!

I am gonna go try this with my PSX-100 too! Will report back...

[McGriffy]

It's perhaps a bit surprising that people can so obviously hear the difference though this has been reported enough that I believe it by now. BUT, can you still hear any difference after downsampling to 44.1? There are, of course, DVD-A's and someday it may even be possible for me to make one, but for now it's Redbook all the way. Can you still tell on a CD? (Zep, I realize that you don't have the gear to test this yet but many people out there do.) Everyone report your results. A few highly scientific tests would be nice but a whole bunch of informal ones might also say a lot.

DMcG

[Renie]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:<HR>Originally posted by McGriffy:
It's perhaps a bit surprising that people can so obviously hear the difference though this has been reported enough that I believe it by now.
DMcG<HR></BLOCKQUOTE>

I wonder how many people wouldn't dare say otherwise for fear of being seen to have inferior hearing.

Nika you said---

"Please tell us, what converters were you using? Because the fact that they don't sound the same at 48 and 96 is a strong indicator of how bad those converters are. Any good converter should sound the same between the two different sample rates. So please, which ones were you using? Or were they really not the same? I'd like to avoid those converters in my own rig if they were really as bad as you say they are.." ------

This has surprised and confused me!!
So if the 192 is any good it should sound the same whether at 48 or 96???
What's the point of 96 and above then?
Is 48 the lowest sample rate where difference is audible in good convertors?


Thanks

Renie

[Nika]

[QUOTE]Originally posted by Renie:
This has surprised and confused me!!
So if the 192 is any good it should sound the same whether at 48 or 96???

Yes. That's the idea. There are many who say that there is no reason that a 96kS/s or 192kS/s converter should be able to outperform a 48kS/s converter. In this list of people I'm including the designer of the converters for the Sony Oxford console, engineers at Digi, myself, and many distinguished people. It's not to say that there isn't a reason to WORK at 96kS/s, but to SAMPLE at 96kS/s is the conversation at hand.

As Digi has alluded through Frederick Umminger and I believe Chris Townsend, the only potential benefit that Digidesign sees for 96kS/s or higher has to do with SOME plugins whose processing can be advanced from running at faster rates. For additional information on this subject I'd highly recommend the following threads:
http://www.musicplayer .com/ubb/ultimatebb.php?ubb=get_topic;f=3;t=002469

The one above has to do with converter design and where the differences lie in different converters. The one below has to do with benefits of higher sampling frequencies (or the lack thereof):
http://www.musi cplayer.com/cgi-bin/ultimatebb.cgi?ubb=get_topic&f=3&t=000822&p=


What's the point of 96 and above then?

Yeah. That's a good question. Certainly there are individual converters where there are differences, but conceptually one does not have an advantage over the other sonicly if both are well designed. The advantages kind of come in that you can design cheaper converters that will perform well if you can do it at 96kS/s. But there's no reason you can't get the same results with a well designed 48kS/s converter.

Is 48 the lowest sample rate where difference is audible in good convertors?

One has to design a converter that has a very steep filter that rolls off nothing below 20kHz and rolls of entirely above the Nyquist frequency. The more room you have to do this in the easier it will be to make it. People indicate to me that it's POSSIBLE to do at 48kS/s and even 44.1kS/s, though not necessarily easy. One manufacturer said that he thinks 60kS/s would be about ideal.

Check out the first thread above to learn why sample frequency isn't all that we care about in a converter's design.

Nika.