Processing clips

[Robert Bass]

Hey Ed, if that thing can run at 7.5", and if you're ever gonna junk it, I wouldn't mind getting my greasy paws on it. ;D

Too late! I already bought it. ;D

R

 

[kevtronics]

Question for Goran...

Were the Omnia-6 and DSP-X clips you uploaded processed using 50 uS or 75 uS pre-emphasis?

 

[gtomas]

Were the Omnia-6 and DSP-X clips you uploaded processed using 50 uS or 75 uS pre-emphasis?

50 µs. That's what we use here in Europe


Regards,
Goran Tomas

 

[kevtronics]

50 µs. That's what we use here in Europe

OK, I noticed they sound quite bright compared to American standards.

I also noticed that you have the channels reversed -- at least compared to the uncompressed sample music tracks that you have uploaded.

 

[gtomas]

OK, I noticed they sound quite bright compared to American standards.

Really? I'd thought this market is not that bright, compared to some others... It would be interesting if you can make some recordings of the station where you live?

I also noticed that you have the channels reversed -- at least compared to the uncompressed sample music tracks that you have uploaded.

Quite possibly as I just hooked the processors to the PC for the recording (those who were not recorded off-air).


Regards,
Goran Tomas

 

[kevtronics]

Really? I'd thought this market is not that bright, compared to some others... It would be interesting if you can make some recordings of the station where you live?

The DSP-X Tina Arena clip sounds really really bright. Most stations here don't push the high end that much because today's music is already treble-happy right on the CD, and that's not such a good mix with 75 uS clipping. (Kelly Clarkson's "Because Of You" is a killer -- her vocals during the beginning get horribly clipped by every station I've heard it on!)

 

[kevtronics]

Quite possibly as I just hooked the processors to the PC for the recording (those who were not recorded off-air).

BTW, you also set the recording level on the Omnia clips a bit too high, causing some clipping in the WAV files when the peaks hit full scale (0 dBfs).

 

[gtomas]

Quite possibly as I just hooked the processors to the PC for the recording (those who were not recorded off-air).

BTW, you also set the recording level on the Omnia clips a bit too high, causing some clipping in the WAV files when the peaks hit full scale (0 dBfs).

No I haven't, I first recorded the clips and normalized the files afterwards to 0dBfs. But some software (Adobe Audition and Sound Forge are both examples) can indicate peaks with heavily peak processed audio as OVER, although they aren't....


Regards,
Goran Tomas

 

[FFoti1]

No I haven't, I first recorded the clips and normalized the files afterwards to 0dBfs. But some software (Adobe Audition and Sound Forge are both examples) can indicate peaks with heavily peak processed audio as OVER, although they aren't....


Regards,
Goran Tomas

All the more reason why this practice is not a good idea. Who knows *exactly* what the normalizing function added to the audio? Taking audio, that has already been significantly processed, and adding further dynamics functions to it, is only asking for more trouble.

This further reduces a fair evaluation of the audio clip.

-Frank Foti

 

[rrsounds]

No I haven't, I first recorded the clips and normalized the files afterwards to 0dBfs. But some software (Adobe Audition and Sound Forge are both examples) can indicate peaks with heavily peak processed audio as OVER, although they aren't....


Regards,
Goran Tomas

All the more reason why this practice is not a good idea. Who knows *exactly* what the normalizing function added to the audio? Taking audio, that has already been significantly processed, and adding further dynamics functions to it, is only asking for more trouble.

This further reduces a fair evaluation of the audio clip.

-Frank Foti

Is the normalizing function a single, one-time level adjustment for the entire piece, or is it dynamic? If the former, no big deal; if the latter, all bets are off.

David P. Reaves, III
TransLanTech Sound, LLC
Home of the Award-winning "Ariane Sequel" Digital Audio Leveler

 

[gtomas]

All the more reason why this practice is not a good idea. Who knows *exactly* what the normalizing function added to the audio? Taking audio, that has already been significantly processed, and adding further dynamics functions to it, is only asking for more trouble.

This further reduces a fair evaluation of the audio clip.

If it would make any difference, I can just as easily make the files without normalizing. But I get a feeling you're against this practice in general, regardless of the peculiarities, so I'm pretty sure it wouldn't matter for you one way or the other...

What does normalizing do? It scans for the peak value in the file, calculates the headroom and then scales (multiples) all the samples with the coeficient that will put the highest peak(s) to 0dbfs. It's not a dynamic function of any sort and does not affect dynamics at all. And it doesn't cause any clipping, the sometimes OVER indication in some software is just the inaccuracy of the metering and it's ballistics. Afterall it's a fixed gain change and affects dynamics just as much as does the output level control on Omnia, or output level of any other processor...


Regards,
Goran Tomas

 

[gtomas]

Is the normalizing function a single, one-time level adjustment for the entire piece, or is it dynamic? If the former, no big deal; if the latter, all bets are off.

Of course it's static. It's a fixed digital level change for the entire clip -> no big deal.


Regards,
Goran Tomas

 

[FFoti1]

If it would make any difference, I can just as easily make the files without normalizing. But I get a feeling you're against this practice in general, regardless of the peculiarities, so I'm pretty sure it wouldn't matter for you one way or the other...

What does normalizing do? It scans for the peak value in the file, calculates the headroom and then scales (multiples) all the samples with the coeficient that will put the highest peak(s) to 0dbfs. It's not a dynamic function of any sort and does not affect dynamics at all. And it doesn't cause any clipping, the sometimes OVER indication in some software is just the inaccuracy of the metering and it's ballistics. Afterall it's a fixed gain change and affects dynamics just as much as does the output level control on Omnia, or output level of any other processor...

Yes, I'm against this practice in general, and for reasons shared many times over.

Also, you're putting blind faith in the normalizing function, without any proof that it's is doing as we're lead to believe. I've done plenty of 'normalizing' using various S/W packages and have found some that are not what they claim. Adding another variable to the mix creates more problems.

-Frank Foti

 

[kevtronics]

Also, you're putting blind faith in the normalizing function, without any proof that it's is doing as we're lead to believe. I've done plenty of 'normalizing' using various S/W packages and have found some that are not what they claim. Adding another variable to the mix creates more problems.

I've found that I get more accurate results in Cool Edit / Adobe Audition if I scan the audio file to find the highest peak value, and then manually amplify or reduce the gain to the desired peak level. The automatic normalize feature is often off the mark by a few tenths of a dB.

I also prefer to normalize things to at least -1 dBfs, if not even lower, just to keep some digital headroom and prevent any chance of unwanted clipping. This is especially important if you're preparing audio that will go into a lossy codec, since that often introduces peak overshoot in both the encoding and decoding/playback processes.

p.s. You can also argue that normalizing 16-bit audio adds quantization noise and raises the noise floor, but that's the nature of the beast when you're dealing with any 16-bit audio, and I'd rather have an unnoticeable increase in background noise level than a very noticeable amount of clipping!

 

[Robert Bass]

Personally, I don't normalize any recordings I have made, from a broadcast signal. I set the levels to around -3 dBfs and record as-is. This removes the added variables that can occur with normalizing.

R

 

[kevtronics]

I've uploaded some new processing clips, based on the three sample songs that have been provided (Tina Arena, Styx, and Elton John) as well as two other examples of songs that often sound bad when run through "major-market" FM audio processing (Sarah McLachan and LeAnn Rimes).

All were processed by the wideband audio processor built into John Burnill's Sonos I (version 1.0.27), which is basically a scaled-down version of the MBL4 multiband processing which is also the heart of the Inovonics Omega FM hardware box processor.

Using some innovative settings (many of which totally go against Burnill's own recommendations!), I have been able to eliminate virtually all traces of "pumping" and other wideband processing artifacts, while still preserving competitive loudness and brightness. In fact, it often sounds cleaner than multiband processing, because it isn't boosting the bass and treble only to have the limiters smash it back down to a flat response!

 

[gtomas]

I've found that I get more accurate results in Cool Edit / Adobe Audition if I scan the audio file to find the highest peak value, and then manually amplify or reduce the gain to the desired peak level. The automatic normalize feature is often off the mark by a few tenths of a dB.

Didn't notice that, I'll have to check your claims. Since the software uses the same tools (built-in statistics, etc) that you're using to do it manually, there's really no reason why it would be off, but I'll check it out.

I also prefer to normalize things to at least -1 dBfs, if not even lower, just to keep some digital headroom and prevent any chance of unwanted clipping. This is especially important if you're preparing audio that will go into a lossy codec, since that often introduces peak overshoot in both the encoding and decoding/playback processes.

I agree with you on that and I've made the same recommendation myself. I'd advise leaving 3dB of headroom as in tests I've done I've noticed as much as 2.8dB of overshoot with some types of codecs. But these recordings you're talking about are not coded, they are in WAV...

p.s. You can also argue that normalizing 16-bit audio adds quantization noise and raises the noise floor, but that's the nature of the beast when you're dealing with any 16-bit audio, and I'd rather have an unnoticeable increase in background noise level than a very noticeable amount of clipping!

Generally yes, not necessarily. If the internal operations in the software are done with higher wordlength, any quantization noise added will be way below the output -96dB dynamic range (and even lower than that with dithering). But, for the sake of the discussion, even if it doesn't do so and you add some noise, it will certainly be below 90dB. With the average RMS level of these processed samples around -10dBfs, we are talking of S/N ratio of 80dB. Is this audible in any practical situation as the audio level within the songs rarely goes low enough to hear the noise 80dB below even if you're in the completely dead acoustic chamber and not in a normal environment where the background noise is much higher? Wouldn't the high levels (in hi-fi terms) of distortion produced by clipping in the processor and the S/N reduction due to processing itself be much more limiting factors to the listening experience?

Why are we so nit-picking on this??


Regards,
Goran Tomas

 

[Robert Bass]

Why are we so nit-picking on this??

Excellent question! Other than a difference in loudness, I personally cannot distinguish any other difference between normalizing vs not normalizing. I think the only practical reason to normalize audio, is if you're cutting a CD using various different sources. The only other reason I'd normalize, is if the original recording was so loud, that it pegs the VU meters to the end of the scale and sounds distorted, when using an analog broadcast console.

R

 

[karasak]

But if don't normalize, i have situation with some very old cd's that peak level is too low so, processor agc don't react, and processor gate mute the sound.


Kreso Croatia

 

[fugazi]

Guys,

When is someone going to upload a Omnia One soundclip?
I'm sooo curious.
I am considering a test drive via mpx. So clips come in handy anyway.

Regards,

Evert