EQ-ing at 96khz vs. 48khz

VST, AU, etc. plug-in Virtual Effects discussion
KVRian
1116 posts since 18 Jan, 2004 from Los Angeles, California, USA

Post Thu Nov 29, 2007 2:42 am

A couple days ago at Dreamtown Music I ended up doing some tests while the Mackie D8B was waiting for repairs. Long story short, I created two identical "draft mastering" sessions of a completed mix: one used a 48khz file, the other upsampled to 96khz. Running Pro Tools LE 7.3.1 with Digi 003 under Mac OS X 10.4.8 I found obvious differences when running several RTAS plug-ins (e.g. the URS M EQ bundle, Sonalksis SV 517 MK II and the EMI Mastering Pack) with the exact same settings. Differences between the original recording and the upsampled versions were subtle at best. Regarding the processed files, however, I found that these differences survived downsampling to 44.1khz and were easy to pick out in blind tests by either myself or the composer I was working with.

After finishing work, I came home and tried the same tests in Sonar 6PE and a demo of Adobe Audtion 2 with an Audiophile 2496. After 3-5 hours of testing (it was late) I found myself having to reconcile the fact that I did not perceive the same differences that I had at the studio, even when using the Sonalksis demo.

Has anybody had similar experiences? Is this due to differences in internal oversampling between the VST and RTAS versions or something similar? If so, I was not aware that such differences existed. I know that some plug-ins oversample more than others (if at all) but I would have thought it would be consistent between the VST and RTAS versions of a plug-in.

Is it the experience I had at the studio or the one I had at home more common? Which one indicates that something isn't working quite right?

Curious to hear others input on this. My years of experience working in this area never went far into 88khz and beyond since I often found it difficult to hear the difference in material recorded at the 96khz versus 44.1khz, but some of Massenburg's comments on digital processing recently piqued my interest.

KVRAF
1907 posts since 29 Oct, 2003

Post Thu Nov 29, 2007 3:32 am

+1

There was a thread about upsampling the whole mix and some of (us if I may say so) found that in certain hosts that allow that with little effort it's easy to mix in 44/48k and only render/export at 96(or 88k which I personally favor).

Don't have it bookmarked, but that thread is hidden somewhere in forums here. This works. The (smart) user "Kingston" outlined a tedious method of upsampling stems/channels and reimporting these into session at higher rate for hosts that don't allow effortless upsampling of the mixdown/render/export, such as Cubase (wouldn't PtLE fall into the same category?)

Btw. there were some posters in the said thread that didn't notice much improvement, while my own tests (in diff host) comparing the two exports (both finally @ 44k, i.e. the upsampled one downsampled after that) proved to me that there was a certain "fullness" present in the upsampled mix that wasn't there in the 1:1 mix @ 44.k. YMMV ... what else can I say!

Neccessary ingredient of his method is: http://www.lcscanada.com/audiomove/

Btw. since you have PtLE, can you describe what are EXACT differences between Izotope Ozone Lite (and other "Lite's") versus retail versions?

I'm eyeing mbox2pro factory / digi 003 so I ask ... and your comment regarding differences re. RTAS methods is a good one to analyse.

EDITED: I read the post below and added a comment regarding perceived improvements
Last edited by mauseoleum on Thu Nov 29, 2007 3:41 am, edited 2 times in total.

KVRian
1335 posts since 23 Sep, 2003 from ocation: cation: ation: tion: ion: on: n: :

Post Thu Nov 29, 2007 3:39 am

Funny, when I had a Firewire Audiophile I used to think that there is not much of a percievable difference between 44.1 and 96k... Thankfully now with an EMU 1616m I can hear and appreciate the difference.

Tracks recorded and/or processed at 96k sound better, because the increased resolution allows for better representation of highs and transients. This makes recordings of high freq stuff (snares, cymbals, brass) or dynamic stuff (drums, DI bass) sound percievably better - highs are less harsh and transients are more punchy.

During processing, the increased resolution allows filters to operate more smoothly in the high range and better represented transients give dynamic processors a chance to work more fluently. At 96k it's also harder for soft synths and effects to produce audible aliasing, which helps in achieving a cleaner mix.

As far as I can tell, most plugins sound different at 96k than at 44.1k - the highs will be more classy, but somewhat shy at 96k - so it's normal that a mix will sound different. If you can't hear this at home, try swapping your MAudio for something better and see if it helps. ;)
the the impotence of proofreading

KVRian

Topic Starter

1116 posts since 18 Jan, 2004 from Los Angeles, California, USA

Post Thu Nov 29, 2007 4:22 am

mauseoleum wrote: Btw. since you have PtLE, can you describe what are EXACT differences between Izotope Ozone Lite (and other "Lite's") versus retail versions?

I'm eyeing mbox2pro factory / digi 003 so I ask ... and your comment regarding differences re. RTAS methods is a good one to analyse.
I would love to help but we already had the retail version of Ozone installed so we never installed the Lite one.

As regards perceived differences, I actually found that the difference is there in LE/M-Powered versus Sonar PE. I don't know why I didn't think to do this last night, but after my last post I opened up Pro Tools M-Powered and tried the same test. The resulting wave files at 48khz and 96khz echoed the types of differences I had experienced at the studio. In Sonar, files at the two samplerates did not exhibit similar differences to each other. Furthermore, the output of the exact same plug-in with the exact same settings at the exact same sample rate was different between Sonar and Pro Tools M-Powered. I'm not even talking about how either Pro Tools attenuates or Sonar boosts the files because the output level was really different between the two. I'm talking about post normalized frequency content.

Paulie Phonic, believe me, these kinds of differences make the differences in the 1/4 or RCA outputs on an Audiophile 2496 and a EMU 1616m or even a top of the line MOTU (such as the ones that were preferred by Mary J Blige to the top of the line Pro Tools HD ones) seem very subtle. This is the sort of thing that you could tell once the files were downsampled to even an MP3, burned to CD and played in a car. :)

User avatar
KVRAF
3927 posts since 8 Mar, 2004 from Berlin, Germany

Post Thu Nov 29, 2007 4:22 am

specifically to EQs and filters there is to say, that the transition from analog (equivalent to an infinite samplerate) to digital (finite sample-rate) introduces a distortion of the frequency response in the high frequency area. this is referred to as frequency-warping. beware that this does not mean that signal frequencies are warped to other freqeuncies - it just means that the freqeuncy-response of the EQ looks "warped". this effect, however, is significant only above sampleRate/4 or so - so with 96 kHz you get rid of it in the audible range. yeah, and when nolinearities come into play, such as in distortion or dynamics-processors - the additional artifact of aliasing will occur which is even more detrimental to sound-quality. the last thing also holds for signal generation such as in synthesizers and samplers.

KVRian

Topic Starter

1116 posts since 18 Jan, 2004 from Los Angeles, California, USA

Post Thu Nov 29, 2007 4:31 am

Upon further analysis, the Sonar 6 PE processed version at 24 96 sounds more like the PT LE at 24 48 than the PT LE at 24 96. I could post audio examples if that would help others to figure this out.

KVRAF
7796 posts since 24 Feb, 2003 from Earth, USA

Post Thu Nov 29, 2007 4:44 am

Per Lichtman wrote: I could post audio examples if that would help others to figure this out.
Please do. You'd be one of the few who actually 'dares' to post examples to prove a point, and it would be appreciated that we can debate differences with examples. ;)

Devon
Simple music philosophy - Those who can, make music. Those who can't, make excuses.
Read my VST reviews at Traxmusic!

KVRian

Topic Starter

1116 posts since 18 Jan, 2004 from Los Angeles, California, USA

Post Thu Nov 29, 2007 4:48 am

DevonB wrote:
Per Lichtman wrote: I could post audio examples if that would help others to figure this out.
Please do. You'd be one of the few who actually 'dares' to post examples to prove a point, and it would be appreciated that we can debate differences with examples. ;)

Devon
http://download.yousendit.com/BBCB445D5E155F9C

KVRian

Topic Starter

1116 posts since 18 Jan, 2004 from Los Angeles, California, USA

Post Thu Nov 29, 2007 4:51 am

The above link is to a zip file containing 4 files. 2 are 48khz and 2 are 96khz. The same audio file was fed into sessions at 48khz and 96khz in Pro Tools M-Powered 7.0 and Sonar 6 PE, run through the trial version of the Sonalksis SV-517 MK II with the exact same settings and then output. Since the resulting files varied in amplitude, I brought them into a wave editor and normalized each of them to the same peak. You will notice that the greatest difference in sound is between the PT at 96 khz and all the others, making the others differences seem "subtle".

EDIT: The total zip file size is less than 1MB since the samples are all very short so I hope all interested parties download it.

KVRian

Topic Starter

1116 posts since 18 Jan, 2004 from Los Angeles, California, USA

Post Thu Nov 29, 2007 5:27 am

For anyone that doesn't have a 96khz capable soundcard (or that just wants me to back up my earlier comment about this translating past dither or downsampling), here are the same files all converted to 44.1khz at 16 bit with with the exact same pre/post filter settings and dither options.

http://download.yousendit.com/E88ABCDB2F295351[/url]

KVRian
1335 posts since 23 Sep, 2003 from ocation: cation: ation: tion: ion: on: n: :

Post Thu Nov 29, 2007 5:45 am

Per Lichtman wrote:Paulie Phonic, believe me, these kinds of differences make the differences in the 1/4 or RCA outputs on an Audiophile 2496 and a EMU 1616m or even a top of the line MOTU (such as the ones that were preferred by Mary J Blige to the top of the line Pro Tools HD ones) seem very subtle. This is the sort of thing that you could tell once the files were downsampled to even an MP3, burned to CD and played in a car. :)
Of course, I fully agree. Most probably, the frequency warping mentioned by braindoc is behind most of the perceived difference - by ear, it happens exactly in the > samplerate/4 range: above 10k @ 44.1k and nowhere @ 96k.

Mentioning the difference between an Audiophile and higher quality interfaces I meant the DACs, not the output connectors ;)
the the impotence of proofreading

KVRian

Topic Starter

1116 posts since 18 Jan, 2004 from Los Angeles, California, USA

Post Thu Nov 29, 2007 5:53 am

Paulie Phonick wrote: Mentioning the difference between an Audiophile and higher quality interfaces I meant the DACs, not the output connectors ;)
Ah, so you were ignoring the fact that I was talking about post processing on an upsampled recording? :wink:

At the studio we sometimes use Lucid converters (both DAC and ADC) and for my own work I got the first Neuman KM 184 D in the US so that I could bypass any intermediary ADC since that is done within the microphone itself. I actually prefer the sound in some cases to running a Neuman U87 or Telefunken Elam 251 E into an Avalon 737sp via a Lucid into the Digi 003. Of course if I really wanted a nice DAC/ADC, I'd get the high end MOTU.
Last edited by Per Lichtman on Mon Dec 31, 2007 12:53 am, edited 2 times in total.

KVRian

Topic Starter

1116 posts since 18 Jan, 2004 from Los Angeles, California, USA

Post Thu Nov 29, 2007 5:56 am

Still no explanation for the discrepancy between the Pro Tools and the Sonar files?

KVRian
1335 posts since 23 Sep, 2003 from ocation: cation: ation: tion: ion: on: n: :

Post Thu Nov 29, 2007 6:07 am

No I weren't - it's just that you need to hear the upsampled processed recording through something. At the studio it's a really nice setup, but at home it's an Audiophile with converters of not that high quality and somewhat colouring the sound. The difference in highs may be the same in both cases, but at home you may just not hear it.

Unless we're talking about comparing spectrum analyzer outputs - if you get different results at the studio and at home with the same material and processing it with the same plugs... well that would be weird.

As for differences between Sonar and PT - correct me if I'm wrong, but isn't PT doing 48-bit fixed point, while Sonar uses 32-bit (64?) floating point? No that it should matter much, but who knows. I can't have a look/listen to your files at the moment as I'm not at home.
the the impotence of proofreading

KVRAF
1907 posts since 29 Oct, 2003

Post Thu Nov 29, 2007 6:17 am

I would love to help but we already had the retail version of Ozone installed so we never installed the Lite one.
I see. Thanks anyway.

Return to “Effects”