Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Interpolation of Clipping and Bit-Rate Questions. (Read 12515 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Interpolation of Clipping and Bit-Rate Questions.

I've been testing out the trial of Adobe Audition and have noticed that it is a bit more detailed in analyzing and manipulating peaks and "clips" than Audacity. My immediate example is a needle-drop of Krisiun's "The Great Execution." (One track--1st illustration below.) The CD and iTunes Matched AAC files are a complete mess in that most of the entire recording is clipped or squashed or whatever. (Same track--second illustration below.) (Krisiun is a Brazilian "death metal" band, so there's quite a bit of sound anyway!)

LP Rip @ 16/44.1


iTunes Matched AAC file



If you look at the needledrop in the first illustration you'll notice five places (bottom/right channel) where the sound clipped for a small fraction of a second. I was thinking of zooming in and doing a simple "auto heal" on those as I'm very pleased with the sound of this album when burned to Redbook CD-R. You'll also notice that, if you look at the bottom-right of both illustrations, that Audition is working at 16-bit for the vinyl rip (recorded at 16/44) and 32-bit float for the iTunes AAC file.

Some questions: will working on those few clipped samples and exporting the results at the same bit-rate reduce the quality of the original recording? Why does Audition indicate that it is working at 32-bit float for a 16/44 AAC file while it remains at 16-bit for the FLAC?


The Loudness War is over. Now it's a hopeless occupation.

Interpolation of Clipping and Bit-Rate Questions.

Reply #1
Why does Audition indicate that it is working at 32-bit float for a 16/44 AAC file while it remains at 16-bit for the FLAC?


Whatever AAC decoder its using almost certainly operates on 32 bit floating point values, so the output PCM is also 32 bit floating point.  FLAC, on the other hand, is a lossless format, so the samples it outputs are identical to whatever were input into the encoder.

Interpolation of Clipping and Bit-Rate Questions.

Reply #2
...almost certainly operates on 32 bit floating point values, so the output PCM is also 32 bit floating point.  FLAC, on the other hand, is a lossless format, so the samples it outputs are identical to whatever were input into the encoder.


Thanks. That's good to know.

I'm pretty new to most of this stuff. I guess my remaining questions (to whomever) would be "would editing/making interpolative corrections in 16-bit mode and exporting corrected files result in a theoretically lesser-quality file than the original files? Is there some sort of destructive resampling going on or anything I should be aware of?" (I'm content with 16/44 for playback purposes though I do have a DVD-A player. I just see it as a bit pointless.)

I was thinking of using this software to augment my use of VinyStudio. I can't seem to as easily make such corrections with Audacity.
The Loudness War is over. Now it's a hopeless occupation.

Interpolation of Clipping and Bit-Rate Questions.

Reply #3
To clarify, you're bringing 32-bit float files in and editing them in 16-bit mode?

Interpolation of Clipping and Bit-Rate Questions.

Reply #4
To clarify, you're bringing 32-bit float files in and editing them in 16-bit mode?


No. (Sorry for the confusion.) I record albums with VinylStudio @ 16/44.1 and was just doing some tweaking/manipulation in Audition.
The Loudness War is over. Now it's a hopeless occupation.

 

Interpolation of Clipping and Bit-Rate Questions.

Reply #5
Working in 32 bit floating point, even if the original recording is 16 bit, is much more accurate. The quantization errors, which you cannot avoid, regardless of format, are much smaller. Doing transforms in 16 bit will often produce distortion and noise which are visible in the spectral view and the frequency analysis graph (this is much more true in simple generated waveforms, not often visible in the massive visual confusion of real music). So, the theoretical results are perhaps an order of magnitude "higher quality" when the standard is comparison to the starting waveform. However, if you dither the transform, there will be only extra noise, no distortion.

On the subjective side, will you be able to hear a difference? Again, with digitally generated, or well captured and recorded, simple signal (e.g. a pure sine wave) definite audible differences can exist. With music, especially something you've recorded from an LP, something audible is rather unlikely.

Interpolation of Clipping and Bit-Rate Questions.

Reply #6
Working in 32 bit floating point, even if the original recording is 16 bit, is much more accurate...


Luckily for me I just realized that VinylStudio will export a copy of my existing 16/44.1 FLAC saves as 32-bit float WAV or AIFF (at whatever bit-depth.) I guess if I do it that way; importing it then into Audition, it would be far "better" academically. Would that be feasible in your opinion, Andy? At 32-float/44.1 then just export from Audition at 16/44.1 again?

The reason I'm worried about all of this is I want to archive some decent vinyls as appropriately as possible. I'm no professional, but (for the asking price of Adobe Audition) I'd really like to get professional results. I prefer CD if all things are equal, but many of my even recent LPs (such as this one which was released in 2011) just aren't maxed and clipped like what those "professionals" ended up putting on the CD. (I wish some of you guys here, who know a hell of a lot more about this stuff than me, were working for these recording studios! CDs would get the treatment they deserve.)
The Loudness War is over. Now it's a hopeless occupation.

Interpolation of Clipping and Bit-Rate Questions.

Reply #7
These are the settings I used in converting the sample-type back to 16-bit FLAC from 32-bit/44.1 AIFF. Hopefully someone can correct me if I should change the dither-type or turn on noise-shaping?

The Loudness War is over. Now it's a hopeless occupation.

Interpolation of Clipping and Bit-Rate Questions.

Reply #8
dither bit depth 0.5 bits
p.d.f. shaped triangular
noise shaping C1

These are the CoolEdit settings I use for LPs, primally because these settings add the minimum amount of noise that works well. Ten times as much noise added probably would make no difference in the sound, especially from an LP. When I write "works well" I come from my experiments with simple tones. Actual music recorded from an LP has so much disk background noise that ignoring dither is an option unlikely to be detected in the final result.


Noise shaping, like dither itself,  is always a good idea, never a detriment. Lack of noise shaping might add audible hiss to some music, but all LPs are so noisy that it probably makes no practical difference, so the advantage is just theoretical.

You can do some easy tests to see the Audition result (in Spectral View) by generating a few seconds of silence in a 32 bit space, then converting it to 16 bit with different dither settings. All that will exist in the final result is the dither itself.

You ask about importing into Audition. I don't know if "importing" is the correct term, especially for wav. You simply open the file, no modifications take place in the process. When you are completely done with everything, convert to 16 bit and save. Don't go through the process multiple times. If you don't do all work at one time just save the floating point format, then re-open the next time you want to work on it.

Interpolation of Clipping and Bit-Rate Questions.

Reply #9
dither bit depth 0.5 bits
p.d.f. shaped triangular
noise shaping C1


Thanks for your insight, Andy.

I know Cool Edit was bought by Adobe and is what Audition is based off of. I think the only setting presented as yours is the "triangular shaped." Under "Noise Shaping" I find: None, High Pass, Light Slope, Neutral (Light,) Neutral (Heavy,) U-Shaped (Shallow, Medium, or Deep,) Weighted (Light, or Heavy.)  There's also in some instances "Adaptive Mode" which can be defined as Constant or Dynamic (Light, Medium, or Extreme.)



I'm thinking this should still be beta software until Adobe actually publishes more comprehensive help on it. Supposedly it's coming this month.
The Loudness War is over. Now it's a hopeless occupation.

Interpolation of Clipping and Bit-Rate Questions.

Reply #10
What does the recording side look like (hardware, resolution, sampling rate)? While 16/44 is adequate as an end format, recording should be done in 24/48 at least, or 24/88 - 24/96 when using lesser A/D converters. (The only exceptions exist in bottom of the barrel type hardware like an old Realtek ALC262 with its digital filter that includes a highpass and is identical for all the sample rates, so highpass cutoff shifts upward.)

I'd guess that Audition can record in 24 bit without having to jump through any hoops (Audacity is limited to 16 bit via DSound or MME, so as a Windows user you need to compile it with ASIO support). In Vista/7, it is also useful to use some sound API that gives complete control over the sound device, like WASAPI in exclusive mode or ASIO. Otherwise you have to mess with sample rate setting.

Anyway, Audition has a top-notch resampler, so that can be used for resampling to 16/44 after all the editing is done.

EDIT: My, those are some pretty advanced dithering options. (I'd be happy with some generic Gaussian p.d.f., ATH-based dither already.) Documentation is definitely needed there.

Interpolation of Clipping and Bit-Rate Questions.

Reply #11
Assuming he has a preamp that let's him set the level correctly there is no real advantage to 24 bit other then that it let's you be marginally more sloppy with your levels.

Interpolation of Clipping and Bit-Rate Questions.

Reply #12
In this instance the recording was done with a Pro-Ject Debut III USB turntable with on-board preamp. The USB out is limited to 16/44 due to the USB 1.1 protocol. (I have no idea why it couldn't be USB 2.0...especially when Pro-Ject's literature claims it's a "premium" A/D converter.) In the future I plan to pick up a Tascam US-144MKII USB Audio Interface to record most LPs directly from the preamp's RCA-outs.
The Loudness War is over. Now it's a hopeless occupation.

Interpolation of Clipping and Bit-Rate Questions.

Reply #13
I doubt it is any more "premium" than your average entry-level (CMedia or PCM29xx based) USB interface. Looks like the Debut III USB has also been around for 5 years already, so you wouldn't expect bleeding edge technology either.

What kind of cartridge are you using? The stock OMB5e isn't much to write home about, AFAIK...

Another thing to watch out for: Correct speed. I was decidedly displeased when I found that my little Technics (an entry-level model, SL-BD20) played about 0.5% fast and, unlike the next model up the food chain, has no speed adjustments. This is distinctly audible when compared to a CD version of the same disc. I haven't digitized another record since (I wasn't 100% pleased with the cartridge or the rumble level either). The next one will be a direct drive with a strobe, or quartz direct drive. Let's see what the used market will supply. This analog stuff has a much less steep diminishing returns curve.

Maybe one really doesn't need more than 16 bits (a phonopre/cartridge already contribute a good bit of noise, but I think I looked at spectra once and didn't considering dithering sufficient)... but given that the OP wants "pro" quality and modern technology easily pulls it off, using 24/48 - 24/96 seems like a no-brainer. Then nobody will be able to complain about the A/D side at least. (I'd entrust such a job to a basic Audigy SE, though in 24/96 and accounting for a 1-sample offset between channels. Unlike analog technology, digital can be inexpensive and totally decent.)

Interpolation of Clipping and Bit-Rate Questions.

Reply #14
An Audigy SE is out for me...I do my stuff on a MacBook Pro. The cartridge is indeed the stock one (Ortofon Red) that comes pre-fitted from Pro-Ject.

I believe may "downgrade" to the newer Debut III Carbon and sell this one before I upgrade the cartridge. You mentioned the speed and that's one of the reasons I got into Pro-Ject TTs; they offer decent upgradability. However I found out that the Speedbox doesn't work with the USB model, and that was part of my long-term strategy. I haven't noticed any speed problems thus far...but then again I never really looked into it either.

I've thought about buying a quartz-driven TT like a Technics. It would certainly make cleaning a record far easier

When I get the proper equipment (between the preamp and the MBP) I may begin recording at 24/88.2 and archiving that way. I have a SACD/DVD-A player in my stereo system but I never really listen to anything at that resolution. So I guess I'd just be dithering for playback copies anyway.

Digital is my preference over vinyl. Sometimes I find records (such as this one) that aren't badly clipped like the CDs and would rather listen to a digital copy of it. Some records are nothing more that the CD's master turned down a bit for vinyl but still clipped. So it's just a crapshoot for me. It's really more of a hobby to keep me collecting.
The Loudness War is over. Now it's a hopeless occupation.

Interpolation of Clipping and Bit-Rate Questions.

Reply #15
I've thought about buying a quartz-driven TT like a Technics. It would certainly make cleaning a record far easier

I'll be looking for a used one - any decent middle-class model of 25 years ago is likely to be as least as good as any affordable new one.

Not accounting for inflation (40-odd %), a Dual CS505-4 costs twice as much as its predecessor in 1990, while being somewhat more cheaply made. Those used to be a common middle-class semi-automatic belt drive TT, not even anything too fancy (if solid). Now they're 500€. That's what (no more) mass production does. The venerable Technics SL-1200/1210 used to be somewhat more expensive but ran out 2 years ago, certainly not due to lack of demand (I'm guessing they kept production up for as long as they could, but eventually parts ran out and people retired, and there wasn't enough money in it to restart development).

Even basically decent new models seem to have their share of problems here and there (e.g. AT-LP120-USB: complaints about bad QC and audible motor noise).

The only big problem when buying a (used) record player is shipping. Most people don't know how to package one correctly, others just shun the effort. As a result, ruined dust covers, trashed pickups, bent tonearms and even more severe damage are seen on a regular basis. Ironically, it's probably easier to get a lightweight plastic cheapie in one piece than a fancy Thorens or something.
Digital is my preference over vinyl. Sometimes I find records (such as this one) that aren't badly clipped like the CDs and would rather listen to a digital copy of it.

A common position these days, and one I'd agree with. I just wish one had the option of obtaining a properly mastered CD, too...

Be sure to contribute to the DR Database if you can (at least an offline scanner for Mac does exist, though I'm not sure what the pitfalls are - the old Windows version of the offline meter had some issues with metadata).

Interpolation of Clipping and Bit-Rate Questions.

Reply #16
Maybe one really doesn't need more than 16 bits (a phonopre/cartridge already contribute a good bit of noise, but I think I looked at spectra once and didn't considering dithering sufficient)...


What do you mean by this?  What does the spectrum have to do with the number of bits required?

Interpolation of Clipping and Bit-Rate Questions.

Reply #17
...

Be sure to contribute to the DR Database if you can (at least an offline scanner for Mac does exist, though I'm not sure what the pitfalls are - the old Windows version of the offline meter had some issues with metadata).


I wasn't even aware of the database. I'll try the scanner out and see about contributing.

There's not a lot of affordable new Technics TTs anymore...but I certainly wouldn't mind having one (or a good used one too.)

Anyway...heres what I settled on for my settings and just called it a day.



I'm beginning to wonder if any of this was even necessary as Audition may just make these corrections at 32 bit floating point anyway. Perhaps these settings really only exist to facilitate the generation of a Redbook file from, say, an existing 24-bit file or whatever. IDK.
The Loudness War is over. Now it's a hopeless occupation.

Interpolation of Clipping and Bit-Rate Questions.

Reply #18
Are you sure about Audition doing something at 32 bit with 16 bit files?  With Cool Edit, if I open a 16 bit file, all operations are done in 16 bit. This is preferable. It gives me control of what is done.

I can convert the data to 32 bit after opening, or have the program automatically convert upon opening, but then I no longer have a 16 bit file. It stays 32 bit until explicitly converted to something else.

Interpolation of Clipping and Bit-Rate Questions.

Reply #19
Are you sure...


Not at all    I was just thinking I read that somewhere but I could be completely full of s___.  (I haven't been feeling all that well lately so I may not be thinking too clearly.)

I'm hoping that promised documentation from Adobe is coming any day now.
The Loudness War is over. Now it's a hopeless occupation.

Interpolation of Clipping and Bit-Rate Questions.

Reply #20
Are you sure about Audition doing something at 32 bit with 16 bit files?


Update: I found this link that seems to indicate that it does. Audition Help

It seems to be noted under "Understanding Bit Depth."
The Loudness War is over. Now it's a hopeless occupation.

Interpolation of Clipping and Bit-Rate Questions.

Reply #21
The documentation is awkward.
What is "best audio quality"? The quality Audition assigns to everything it does or what it uses when the user in some manner requests "best" or what it uses  when the input is 32 bit?
What did the writer mean by "transforms"? by "in 32 bit"? by "saves"? Converts the entire input to 32 bit and only reconverts to the original bit depth when the file is put away? Converts that part of the data it uses for any particular operation to 32 bit and reconverts to the original bit depth as it "saves" the result? Nothing there is specific enough to assure one and only one reasonable interpretation: poor documentation.

Interpolation of Clipping and Bit-Rate Questions.

Reply #22
The "32 bit" in their table seems to be 32 bit integer, not floating point, no?

Interpolation of Clipping and Bit-Rate Questions.

Reply #23
I guess it is too generic. I suppose I was just being a bit too naïve and optimistic.

I agree that the documentation is very poor at best. There's supposed to be a book published later this month. Amazon pre-order ...I wonder if that will be better detailed.

For the asking price of this software, it wouldn't be too much to ask for better documentation. I mean it is geared towards professionals of which I am decidedly not. I can't even seem to get an answer to simple questions based on my limited usage.

...but anyway: I am going to continue to export files from VinylStudio at 32-bit floating point and work with those in Audition (saving corrections as such and finalizing back to 16/44.1.) This way I can be reasonably assured that it's getting done properly. That's not a problem for me at all as I like and always use VinylStudio for my rips.
The Loudness War is over. Now it's a hopeless occupation.

Interpolation of Clipping and Bit-Rate Questions.

Reply #24
Another thing to watch out for: Correct speed. I was decidedly displeased when I found that my little Technics (an entry-level model, SL-BD20) played about 0.5% fast and, unlike the next model up the food chain, has no speed adjustments.

Guess what - turns out I was wrong. I got a PM stating that it very much is possible, and lo and behold, the service manual now available on the interwebs (unlike the last time I tried) shows how to do it. I was looking at the wrong holes earlier, no wonder I only found bare PCB underneath.

I'm now at -0.1..-0.2% rather than +1.5% (!) before, and pitch has unsurprisingly normalised. Basically you can do it with a small Phillips (size 0) jeweller's screwdriver, a small object to place on the platter as an indicator that whizzes by at a clearly defined location, and a stopwatch. A fluorescent light and a strobe disc in theory are more comfortable, but here it meant that I had to drag the player into the cellar. Back upstairs, it was 1% slow. Hence the stopwatch method.