if digital compression is evil, what about "normalizing"?

Not quite! My cautions are because to avoid truncation, a quality editor must use dither to avoid it. Yes, truncation is the worst, but repeated dither is almost as bad. In higher-bit, 24-bit and above, most software will not dither. You shouldn't dither until that last step of going to 16-bit, and then only do it ONCE!.

Yup, it does help maintain the resolution, but you still don't want to do it too many times on a 16-bit file. Why? Because layered dither makes the music sound dull and thin, and the dither noise becomes more audible each time it is applied. If the 16-bit file was noise shaped, and the majority of CDs produced in the last decade were, depending on the curve, you may get away with it a bit more. BTW, Cool Edit let's the user define how much dither the program will use in processing.

I have been keeping up with that website of Bob Katz for years now. He may have updated it a couple of times in the last couple of years, but when he wrote it originally in the late 90s, there weren't any programs out there that did 32-bit, except, maybe Cool Edit Pro.

Dithering IS necessary, except, possibly for working on 16-bit files that have already been dithered once.

Yes.

Correct! To the initial poster, monkboughtlunch, there is nothing to get crazy about here. I'm not saying you shouldn't ever process in 16-bit, gosh, i'm working on some stuff in 16-bit right now! I even do it sometimes. After all, there's no practical reason to convert a 16-bit file to 24 or 32-bit just for normalizing if you are just making a CD-R for yourself or the car. Besides, you'll have to dither to get back to 16-bit. BUT, if you are going to say, do an LP-to CD-R transfer, and plan to do something to the files, yeah, do it in 32-bit. Your sound will come out smoother. If you use some noise shaping with that dither, you can tame the effects of the dither. But, again, I DO process some LP-to CD-R stuff in 16-bit, particularly if it is going to be a quickie, or if I don't care that much about the album at all. Otherwise, it's 32-bit all the way.

BTW, the main reason Steve does not use high bit-depths on his CD projects is exactly because of dither. Dither does alter the sound a bit. So does noise shaping. But, if you are going to use high bits, when you come down, you, as the uh, producer, have to make some decisions about what how much dither, and what noise shaping you will use to get as close to the sound of your 24 or 32-bit file in 16-bit. Dither: you can't live with it, you can't live without it. It is a neccessary evil.

As far as sampling rates go, i'm with you, in that if your final destination is going to be 44.1, why not just record there and save some time, hard drive space, and change in sound? Sample rate reduction will affect your sound in a very bad way if it is not done well. In fact, 88.2 is a choice of many people because it divides evenly into 44.1. 96kHz was more of a marketing thing. The number just sounded better, I guess. Who knows. But, I suggest using the sampling rate your card uses internally, then converting with your audio editor software, unless you have a card that does transparent sample rate conversion, like my Turtle Beach Santa Cruz card. The internal sample rate conversion is so good that I can record straight at 44.1 with confidence. Otherwise, I use Cool Edit.

I hate to say it, but Sound Forge 5.x and 6.x is so lousy at dither/sampling rate conversion that I will not use it in SF. Last year, someone posted a link to a lab test of all the "pro" audio editors on the market on the Syntrillium forum, and SF came out so bad in the dither test that the testers wouldn't even bother to comment on it.

 

what is "dither"?

 

Simply, dither is a random analog white noise applied to a digital signal in order to smooth out the digital distortion and quantinization noise caused by mathematical calculations or bit truncation. It introduces a bit of hiss into the data but the effect is preferable to not having it.

 

1.2 vs. 2.1?

Hi Grant,

I'm sure 2.1 is fine for LP-to-CD projects! My question is if you find it a big advantage (worth the $$$ for an upgrade) compared to Cool Edit pro 1.2a. Are the signal processing / click removal functions better, for instance?

Thanks,
Joe

 

The upgrade is free. There are so many things that are different in 2.1 from what you know. Oh, it looks the same, cosmetically, but it has TONS of new features, especially with multitrack! It now has a mixer with real-time processing...

Why not just download it? You can run the demo on the same drive as CEP 1.2.

The processing was already great. All they did was include a spectral decay option in the NR.

 

That being the case, what might be recommended for us Sound Forge users--process in 16bit or 32bit and not dither? What audible signature is there non-dithered sound?

 

The funny thing is that it works exactly like JPEG graphics. The JPEG algorithm performs a Fourier transform on images to store them in the frequency domain instead of the spatial domain... and of course, sound is stored in the frequency domain as well.

The number of similarities between digital audio and digital image processing are really surprising to me. I suppose it all boils down to producing the most pleasing result given the limitations of your sampling rate.

 

Gross digital distortion.

I'm not saying you can't use it, but it does not sound good. Try it yourself.

Ron, I thought you once mentioned that you have Cool Edit. All you have to do is transfer your 32-bit files from SF 5.0 when you are ready, and do the dithering in CEP. If you run CE2k or CEP 1.2, be careful! SF5.0, 6.0, , and the newer CEP 2.x uses the standard 32-bit floating point type of file. The older Cool Edit uses their propriety 32-bit version which is incompatable with SF.

 

No, for better or worse, all I've got is SF 5.0. And to be honest, I've been real happy with it. This dithering problem does concern me a little, though.

 

Not to say that 5.x is a bad program, it is the same as 6.x under the hood. I do use it, or the DX-plugs quite a bit, but not as much as CEP. SF 6.x is VERY fast!

If SF had a way to configure the dither as in Cool Edit and other editors, I would use it more.

 

On a related note:

I have a cassette-to-.wav transfer that I recorded with Cool Edit 2000 to a PC hard drive via an M-audio 24/96 sound card, at 48 kHz/32-bit.

I can open it fine in Coole Edit, but no matter what I do so far, Sound Forge 6.0 refuses to open it properly (I've tried opening it as a Raw file, setting the sampling freq to 48,000 and the bit depth to 32, 'little Endian').
It opens in SF 6.0 as a big blue block of noise (and sounds that way too!)

Suggestions for fixing this are welcome! I want to keep the file in a higher sampling/bit domain while I do some work on it, but i don't want to have to stay in Cool Edit exclusively.

 

Easy! Open the file in Cool Edit.

Choose "Save As" and click options.

Select 32-bit Normalized Float (type 3)

Save file (s)

Open In Sound Forge 5.x or 6.x. They will not work in versions earlier than 5.x..

 

Worked like a charm. You da man! Thanks.

 

Hello krabapple.

How come you record at 32-bit if the card can only provide 24?

Are there true 32-bit cards?

Thanks in advance.

 

Most audio software can either record at 32-bits or convert to 32-bits. It is actually 24-bit with 8-bit mantissa to give an unlimited dynamic range. You could boost the file to crazy levels without clipping!

 

Interesting, so the extra bits are put at the top, leaving the original numbers the same.

Do current filters/enhancers work on 32-bit samples? Do they NEED 32-bit samples?

I'd guess, then, that for Lp transfers, which won't be massaged, there's no use in adding extra headroom so it is a waste of space and time to go beyond the native 24 bits of the card?

TIA.

 

Most quality audio software does indeed work at 32-bits at best. Some only work at 24-bit, but that's fine also. Even if the file being worked on is 16-bit, the software still does it's internal calculations in 32- or even 64-bit. In Sound Forge 5.x, you can disable this and work in what they call the direct mode. But you will then deal with truncation and no dither.

IF you do not plan to do anything to your LP transfers, yes, anything more than 16-bits would be a waste of time. But, you still have to deal with the sampling rate of your card. If it's internal rate is 48kHz, you are usually better off recording that way. If your card does good sample rate conversion, then you could safely record at 44.1.

You just have to find what works for you and your situation.

 

Hi Grant,

Going from CEP 2.0 to 2.1 is free, but going from CEP 1.2 to 2.0 costs $80. That's what I'm wondering about...

Thanks,
Joe

 

Oh yeah! I forgot!

 

Grant,
I had formerly used Cool Edit 2000 with a pop/click removal plugin demo, until the demos ran out. IIRC the pop/click removal plugin had a simple button-push option to allow yout o hear *just* the removed material. Now I'm trying Cool Edit Pro's demo, and its pop/click removal function seems to have a much more complicated means of listening to the removed stuff, involving cutting, pasting, and inverting. There's no simple option I can see for doing the same thing.

Is my memory of CE 200 off, or is there some way to access the simpler way of doing this, from CEP2? I can't chekc msyelf, because my demo privelages for CE 2000 have expired.