bdiament
bdiament's picture
Offline
Last seen: 8 months 2 weeks ago
Joined: Dec 1 2006 - 3:52am
16/44 vs. 24/96 Format Comparison
ethanwiner
ethanwiner's picture
Offline
Last seen: Never ago
Joined: Sep 1 2005 - 2:26pm

Hi Barry,

Nice test, thanks for making that available for download. I love stuff like this!


Quote:
most folks have no means of making a valid comparison


No kidding.


Quote:
Often, a high resolution version of a record is mastered at a different session, sometimes by a different engineer.


That too. Especially that. Likewise for people who would like to compare CDs to SACD or to vinyl or to 1970s 8-tracks for that matter. If the sources are not identical, then of course you'll hear a difference.

Okay, I'll stick my neck out and go first. I hear no difference between your two files.

I did notice that the files do not null exactly, but that makes sense since I had to up-convert the 16/44 file, which included re-sampling and anti-aliasing in Sound Forge. I also noticed that the 24/96 file starts half a dozen samples later than the 16/44 file, unless an offset was added as part of the up-conversion. None the less, the nulled level did not change when I slid the data backward in time to compensate and make them align.

I went back and forth many times being careful not to move my head even a quarter inch, and in my reflection-free listening environment I couldn't hear any difference at all.

--Ethan

bdiament
bdiament's picture
Offline
Last seen: 8 months 2 weeks ago
Joined: Dec 1 2006 - 3:52am

Hi Ethan,

Have you heard differences between 16/44 and 24/96 on other occasions?
Or have those been the type of compares I mentioned, where you hear mastering differences that inhibit any reasonable format comparison?

All I can suggest is listening for things like focus of the space in which the musicians are playing. Also, instrumental definition, particularly the bass (!) but it should be evident on all the instruments.

I'm very proud of how the 16/44 version came out but to my ears, the 24/96 just contains more of the information from the microphones.

Or perhaps your upconversion is making it sound indistinguishable from the 24/96? If so, this would be the first time I've heard of this. (Also, I find the SRC in Sound Forge, like many other SRC algorithms, to not be very transparent. This alone should have created a somewhat brighter and harder tonality.)

The only thing that comes to mind if you didn't hear a difference is that perhaps the 24/96 was being processed in some fashion as well (?). I can't say for sure.

Anyway, thanks for listening.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

ethanwiner
ethanwiner's picture
Offline
Last seen: Never ago
Joined: Sep 1 2005 - 2:26pm

Barry,

LOL, I'm surprised nobody else has chimed in. I guess I have more courage than I thought.

Had I prepared this test I would have faked it for sure. I'd have padded a 16-bit file to 24 bits just to make the size bigger, then down-rezed the 24 bit version to 16 and back. Or some such trickery. I wonder which version most people would say they prefer. Better, prepare two such tests, and invite 20 people to take each, and compare the results. I'll bet $500 that most people would pick what they thought was the 24/96 version.

Actually, I did do a 24/16 comparison a few years ago, but I swear I did not do anything tricky. Link here:

http://www.ethanwiner.com/BitsTest.html

Over the years I've received 65 emails, and only two people got them all right. Which is just about exactly what chance guessing would yield.


Quote:
All I can suggest is listening for things like focus of the space in which the musicians are playing. Also, instrumental definition, particularly the bass (!) but it should be evident on all the


I listened to everything carefully and heard no change. I'm not saying there is no difference, and as I said the files did not null so obviously there is some difference. Also, I'll be 59 next month, so that's a factor. Then again, my hearing acuity - if not sheer extended HF response - has only gotten better over the years.


Quote:
Or perhaps your upconversion is making it sound indistinguishable from the 24/96? ... if you didn't hear a difference is that perhaps the 24/96 was being processed in some fashion as well (?).


Initially I played both files as is in Sound Forge through my M-Audio Delta 66 card. I heard a tiny click a few times as I switched from one to the other, which happens when the card switches formats. So I'm pretty sure each played in its native format.

Have you tried letting someone else switch while you listen, to make a sort of single-blind test for yourself? Do you happen to live within driving distance of me? I'd love to get together with you in person some and fool around with stuff like this.

--Ethan

bdiament
bdiament's picture
Offline
Last seen: 8 months 2 weeks ago
Joined: Dec 1 2006 - 3:52am

Hi Ethan,


Quote:
...I'm surprised nobody else has chimed in. I guess I have more courage than I thought.

I'm glad you listened. I've posted the link elsewhere and received a great deal of feedback.


Quote:
Had I prepared this test I would have faked it for sure. I'd have padded a 16-bit file to 24 bits just to make the size bigger, then down-rezed the 24 bit version to 16 and back. Or some such trickery. I wonder which version most people would say they prefer. Better, prepare two such tests, and invite 20 people to take each, and compare the results. I'll bet $500 that most people would pick what they thought was the 24/96 version.

My purpose with this example has been to provide the files *without* trickery. Tricks are what I complain about in the comparisons I've heard - where different levels, different masterings, etc. etc. cloud one's being able to hear the formats as they are, in a fair comparison.


Quote:
I listened to everything carefully and heard no change. I'm not saying there is no difference, and as I said the files did not null so obviously there is some difference. Also, I'll be 59 next month, so that's a factor. Then again, my hearing acuity - if not sheer extended HF response - has only gotten better over the years...

...Have you tried letting someone else switch while you listen, to make a sort of single-blind test for yourself?

I have always double checked after doing comparisons (whether with components, accessories, mastering changes, engineeering decisions or format comparisons like this), with the help of an assistant to choose which to play at a given moment, so I just sit and listen and don't know what is happening at the controls.

With a fair test, like this example, a recording I made and mastered myself, I can tell which I'm listening to without an A/B.


Quote:
Do you happen to live within driving distance of me? I'd love to get together with you in person some and fool around with stuff like this.

I'm in NY, west of the Hudson, not super far away but a drive that would require an opening in my schedule, which unfortunately for me (as far as meeting soon - and fortunately as far as work) is increasingly harder to find.
Sounds like a fun idea when such an opportunity avails itself though. Thank you.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

Elk
Elk's picture
Offline
Last seen: 3 years 7 months ago
Joined: Dec 26 2006 - 6:32am

When I first saw this posted elsewhere I downloaded the files and burned them to a DVD-A using Disk-Welder. Played back on my main system (using a Grace M902 as DAC) I can hear the difference, especially in the highs.

OTOH, I can't hear any difference when I compare them on computer speakers. I need to play them back on my main system or my my Adam studio monitors.

bifcake
bifcake's picture
Offline
Last seen: Never ago
Joined: Nov 27 2005 - 2:27am

I played them on my computer using my headphones. I heard a difference, but it was very subtle. The 24bit file was a bit richer and fuller. The highs were slightly less harsh, there was slightly more background detail and slightly more envelopment.

ethanwiner
ethanwiner's picture
Offline
Last seen: Never ago
Joined: Sep 1 2005 - 2:26pm


Quote:
My purpose with this example has been to provide the files *without* trickery.


Of course, I was mostly kidding.


Quote:
I'm in NY, west of the Hudson, not super far away but a drive that would require an opening in my schedule


That's not far, so if you ever get that opening please let me know. Will you be at AES this year? If so, please stop by the RealTraps booth and say Hi.

--Ethan

Elk
Elk's picture
Offline
Last seen: 3 years 7 months ago
Joined: Dec 26 2006 - 6:32am

Please let us know if you two ever get together. I, for one, would rally enjoy hearing what the two of you learn from each other!

Barry, on a related question, do you find that you hear any improvement if you record in 24/88.2 and the downrez to 44.1 as opposed to recording originally in 44.1? I've tried and can't hear any improvement when the original is recorded at a higher sampling rate.

I like the extra bits for all the extra headroom however. I love setting levels leaving a good amount of headroom and not needing to worry about it.

bdiament
bdiament's picture
Offline
Last seen: 8 months 2 weeks ago
Joined: Dec 1 2006 - 3:52am

Hi Ethan,


Quote:
..Will you be at AES this year? If so, please stop by the RealTraps booth and say Hi...

I'm not sure if I'll have the time but there is a chance and if I'm there, I'll be sure to stop by.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

bdiament
bdiament's picture
Offline
Last seen: 8 months 2 weeks ago
Joined: Dec 1 2006 - 3:52am


Quote:
...Barry, on a related question, do you find that you hear any improvement if you record in 24/88.2 and the downrez to 44.1 as opposed to recording originally in 44.1? I've tried and can't hear any improvement when the original is recorded at a higher sampling rate...

Hi Elk,

Actually, I do all my recording at 96k and yes, I find there are distinct advantages, even if I'm going to downsample a copy to 44.1k for CD.

I use 96k instead of 88.2k because I use the 96 for the 24/96 DVD release and because with the better sample rate converters nowadays, there is no need to limit oneself to integer conversion --96 to 44 will be just as transparent as 88 to 44.

When I'm mastering a recording done by someone else and know I'll be applying some kind of processing (like EQ), even if the original is at 44 and the target is a CD, I've found it advantageous to do the processing at 96 (and 32-bit float), so I'll save a copy of the original as 32-float, then SRC (sample rate convert) to 96 for any work I need to do. At the other end, I'll SRC to 44, then as the last step, dither to 16. Keep in mind that the advantages offered by these steps is only realized with the better SRC and dithering algorithms I've heard. I would not bother doing this with most of the current SRC and dither algorithms out there, which to my ears, are riddled with artifacts.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

bdiament
bdiament's picture
Offline
Last seen: 8 months 2 weeks ago
Joined: Dec 1 2006 - 3:52am

The Format Comparison page has been moved here.

ethanwiner
ethanwiner's picture
Offline
Last seen: Never ago
Joined: Sep 1 2005 - 2:26pm

Barry,

I meant to say this three times now, but kept forgetting: I think your sample recordings sound fabulous. Both versions.

--Ethan

bdiament
bdiament's picture
Offline
Last seen: 8 months 2 weeks ago
Joined: Dec 1 2006 - 3:52am


Quote:
Barry,

I meant to say this three times now, but kept forgetting: I think your sample recordings sound fabulous. Both versions.

--Ethan

Hi Ethan,

Thank you very much. That means a lot and is appreciated.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

Elk
Elk's picture
Offline
Last seen: 3 years 7 months ago
Joined: Dec 26 2006 - 6:32am

Thanks, Barry!

What do you use for SRC and what other SRC do you find sounds good?

bdiament
bdiament's picture
Offline
Last seen: 8 months 2 weeks ago
Joined: Dec 1 2006 - 3:52am


Quote:
Thanks, Barry!

What do you use for SRC and what other SRC do you find sounds good?

Hi Elk,

I have been using the SRC in Peak Pro XT 5 (at its best, which is slowest, level - doing several tracks as a batch process over night). I was involved with testing this after sending in complaints about the previous version. Peak 5's SRC was the first I'd heard that I would refer to as anything like "transparent".

Recently, I've been experimenting with another SRC algorithm from iZotope, which I believe may well be better -i.e. more transparent- than the one in Peak. (iZotope also makes a very interesting dither algorithm, MBit, which may become my new favorite over Pow-R. More testing to be done before I reach my personal conclusions.)

Most of the other SRC algorithms I've heard leave "boing"y artifacts on sine wave sweeps and harden the sound of music (I would guess from generation of spurious harmonics but that's just a guess). Those seem to do a little (emphasis on "little") bit less damage with integer conversions, which might explain why some prefer 88.2 to 44.1 conversions over 96 to 44.1 conversions.

Hope this helps.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

audiodb
audiodb's picture
Offline
Last seen: Never ago
Joined: Sep 7 2007 - 7:23pm

A short affirmation of this:

"At many format shootouts I've attended, I hear level differences, EQ differences, etc."

I was comparing a half speed mastered vinyl to CD a couple of weeks ago and found the left and right channels reversed between them.

Elk
Elk's picture
Offline
Last seen: 3 years 7 months ago
Joined: Dec 26 2006 - 6:32am

Thanks, Barry!

I am also impressed with iZotope's mBit dithering. I have Ozone 3 (their mastering suite) and originally turned it off as it was part of their "loudness maximizer". I don't like these limiter/compressor type processors. It wasn't until using the program a bit that I realized that the dither was separate even though it is on the same software page.

I believe that you are correct that you are hearing spurious tones generated by various SRC's. I have seen the spectra of a 1kHz sine wave converted from 96kHz to 44.1kHz using various SRC's. The best leave only the 1kHz tone (such as Weiss). Many generate all sorts of extra tones. Many also leave ghost reflections from the application of their filters. So much for the idea that digital is perfect.

Thank you again for your thoughts.

Now to figure out what went well and what did not in last night's recording session...

trevort
trevort's picture
Offline
Last seen: 13 years 2 months ago
Joined: Aug 21 2007 - 8:05am

Thanks for the fun -

I would actually really enjoy having the wool pulled over my eyes with this test, as Ethan suggests, especially if you were to give us 10 links to identically sized files, numbered 1-10, and ask us to give our A or B guesses, but I can appreciate that's effort that is a bit at side purposes to your post.

Here are some notes between the two samples:

1) First bar after the bass enters. The second bass note (the long note, not counting the little repeat after he lifts his bow) is stronger than the first, and is more pronounced in sr001-01-2496.wav (B). The blooming bass note integrates better with the (string against fretboard?) buzz

2) Opening melody, doubling banjo and mandolin (?): resulting harmonics richer with (B)

3) Interestingly, I've felt that picked instruments are less revealing than bowed instruments or voice, but I couldn't find any points in the voice part that suggested a difference between the two recorded versions.

Now, I did find the difference subtle, and as discussed elsewhere in this forum, its not easy to be sure if what you are hearing is really a difference in the physical sound, or simply a perceived difference. I will be amused but not terribly surprised if I learn that the two files actually contain identical music!

bdiament
bdiament's picture
Offline
Last seen: 8 months 2 weeks ago
Joined: Dec 1 2006 - 3:52am


Quote:
...I will be amused but not terribly surprised if I learn that the two files actually contain identical music!

Hi ttt,

Glad you had some fun with the samples.
The *do* contain identical music, the same performace -- just not identical sound. ;-}

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

trevort
trevort's picture
Offline
Last seen: 13 years 2 months ago
Joined: Aug 21 2007 - 8:05am

gotcha!
my semantic error.
I guess you understood my meaning - the sounds were close enough that I was vulnerable to thinking I heard differences when their were none.

Incidentally, I have done sound format comparisons before, and found it easier to distinguish different samples when the identical source music was
a) live, unprocessed recording
b) accoustic instruments, not electronic
c) voice or bowed string instruments

There was quite a bit of (electronic?) reverb, I think, which made it a little harder to get my head into the primary sound. But as a music-enjoying listener, the sound was pleasant and plenty clear.

bdiament
bdiament's picture
Offline
Last seen: 8 months 2 weeks ago
Joined: Dec 1 2006 - 3:52am


Quote:
gotcha!
my semantic error.
I guess you understood my meaning - the sounds were close enough that I was vulnerable to thinking I heard differences when their were none.

Incidentally, I have done sound format comparisons before, and found it easier to distinguish different samples when the identical source music was
a) live, unprocessed recording
b) accoustic instruments, not electronic
c) voice or bowed string instruments

There was quite a bit of (electronic?) reverb, I think, which made it a little harder to get my head into the primary sound. But as a music-enjoying listener, the sound was pleasant and plenty clear.

Hi ttt,

This might be another "gotcha".

The samples in this comparison do indeed represent a "live, unprocessed recording". More so than most in fact. It was recorded with only two microphones (one per playback channel), with no mixing or overdubbing. The signal went from the mics to my mic-preamp/A-D converter and from there, via FireWire, directly to hard disk.

The samples feature "acoustic instruments, not electronic" and also contain "voice or bowed string instruments", in this case, in some sections of the track (for example the very beginning) the acoustic bass was played with a bow.

The reverb you hear is the real sound of the church in which the recording took place. Nothing from the microphone outputs was added to, taken away from or processed in any fashion.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

Elk
Elk's picture
Offline
Last seen: 3 years 7 months ago
Joined: Dec 26 2006 - 6:32am

Barry, just for my own learning purposes - how were the mics set up? (ORTF, X/Y, Blumlein, spaced omnis, etc).

I keep struggling with mic placement, have read lots of stuff but have been finding the good practical examples and my own experience are a lot more helpful than written guidelines.

I'm currently working on recording a brass quintet and am having a bit of a problem getting the balance I between the instruments - ideally I want to capture how they control balance as performers and not the balance by how I record.

bdiament
bdiament's picture
Offline
Last seen: 8 months 2 weeks ago
Joined: Dec 1 2006 - 3:52am

Hi Elk,

After a great deal of experimentation, I arrived at the mic setup I used for "Lift", which is a matched pair of Earthworks omnis, spaced about 15" apart and separated by a baffle I built myself.

If you want to read a summary of the experiments, check out Recording in Stereo (Part 1) and Recording in Stereo (Part 2).

Hope this helps.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

Elk
Elk's picture
Offline
Last seen: 3 years 7 months ago
Joined: Dec 26 2006 - 6:32am

Fun reading. I particularly like "ORTF omnis result in 'fat mono'". I've tried near-coincident omnis and this is a great description.

Thanks!

trevort
trevort's picture
Offline
Last seen: 13 years 2 months ago
Joined: Aug 21 2007 - 8:05am

Indeed -- the news that this was a live recording is worth the price of admission alone!

Goes to show I made the unfounded assumption that you laid on some electronic reverb... and upon listening carefully couldn't distinguish that sound as real or fake.

I did hear the most difference bewteen your samples in the bowed bass, so that will be my consolation prize.

But thanks for exposing may hearing prejudice -- I will listen more carefully to echoes in future, see if I can detect a difference.

bdiament
bdiament's picture
Offline
Last seen: 8 months 2 weeks ago
Joined: Dec 1 2006 - 3:52am


Quote:
Indeed -- the news that this was a live recording is worth the price of admission alone!

Goes to show I made the unfounded assumption that you laid on some electronic reverb... and upon listening carefully couldn't distinguish that sound as real or fake.

I did hear the most difference bewteen your samples in the bowed bass, so that will be my consolation prize.

But thanks for exposing may hearing prejudice -- I will listen more carefully to echoes in future, see if I can detect a difference.

Hi ttt,

I record "live" for Soundkeeper, meaning everyone plays in real-time, with no overdubs, mixes or processing. There is no audience present except for the microphones and myself.

It isn't surprising that with a recording of pop music, we might *expect* any room sound to be electronically generated. To my knowledge, there are only a handful of pop records where one can hear a real acoustic.

Reminds me of one of my earliest experiments in recording this way. The recording was done after hours when I worked at Atlantic Studios. I was allowed to conduct my recording experiment (one of my first that used only two microphones - bypassing the control room, console, etc.) and the next day, played the results for some of my colleagues.
The chief engineer asked "What did you use for the room sound?"
My response: "The room."

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

Log in or register to post comments
-->
  • X