An urgent message about digital noise reduction

[guygee]

Originally Posted by rerem

Actually-a big peeve of mine is torrents put out with major clipping. Some of these were recent vintage,from DAT so analog hiss is no factor,but somewhere along the line the waves got major clipping. A few clips are fixable-but if the whole thing is maxed-there's no hope. I actually have run into that more than a botched NR job. Almost as annoying-but at least fixable are those where the whole show needs a 300% volume boost-and I've DL'd plenty of those. The way I try to do it-up what is first rate. If something seems scarce-yet what I have really is flawed...odds are a better version exists,and often it appears on a bt site. Later-after I do the next puter,with a better soundcard-I'll hunt through my analogs and see what still has not appeared online. Already several have shown up-and in better quality. A few probably won't appear until I digitize them.

rerem - I think that this is a very good point to make. I have run into many recordings that are ruined by clipping, and really no way to recover the music, only some "peak-rounding tools" that yield far from perfect results.

There are several different causes of clipping. In analog recordings, the pre-amp (like any amplifier) has a linear range relative to input level, and beyond that a non-linear range where distortion is introduced. A good piece of analog recording equipment should have a VU meter that calibrates 0 dB to the top of the linear range. If recording levels go just a little above the calibrated optimum, distortion is introduced. Low levels of distortion might actually sound pleasant to some people's ears, adding a kind of warmth, but if the levels get too high then the sound becomes nasty and the recording is ruined in those parts. One might refer to this as "analog clipping". I think "analog clipping" is also referred to as "overdrive" or "over-saturation", and it leaves the recorded waveform rounded but distorted compared to the actual waveform of the input. I do not know of any "cure" for this, maybe someone out there can point to some tools that can help recover from this type of problem?

More pertinent to our discussion here is digital clipping, where 0 dB represents the maximum level that can be represented with any given n-bit digital recording.

Digital clipping occurs when you have exceed the maximum sound level that can be represented (in binary, either 16 1's for 16 bit formats or 24 1's for 24 bit formats). Digital clipping completely flattens the waveform, which introduces some very nasty distortion, so it absolutely must be be avoided.

Digital clipping can occur in the digital recording process, in A/D'ing an analog recording, or in the process of "remastering" a digital recording.

One example is A/D'ing an analog cassette - if your preamp levels are set too high you can introduce clipping. One of my A/D units, a Motu 828, has a warning light that stays lit if I clip even a single sample. It is a 24bit-48 kb/s unit, and I figure that I am just "wasting bits (not using the full dynamic range) unless I set my input level "just high enough" so that it is at the maximum level that will not introduce clipping. In practice, this means I set my levels and A/D, then tweak my input levels up and re-A/D, repeating until clipping occurs, then tweak the levels back down "just enough" and do the final A/D conversion so that no clipping occurs. This is the only way I can figure to take advantage of the full dynamic range that 24-bit offers. If the recording is clean and well-balanced (no drop-outs, no warble, no speed correction needed, etc), then the only only other operations I will perform is to remove DC bias from the untracked (large) wav file, and normalize the entire show at once "by peak" (not compression!).

I think that what some people do not realize, when they are attempting to remaster a show, is that many operations increase the levels and can potentially can cause digital clipping. For example, any kind of active filtering (which can include some NR algorithms), dynamic equalization, and poorly-implemented compression are some examples (to add to the discussion, maybe somebody can point out other operations that change levels?). If you plan to perform these types of operations, leave yourself some "digital headroom". Most decent audio remastering software will have a utility to check for digital clipping, so this should be run after any other operation is performed, and at the end of your remastering effort before tracking, just to be sure.

As a related aside, I hate it when folks normalize "by track" instead of "by show". The relative dynamics between the tracks from a show should be preserved, because some songs are meant to be "soft and sweet" while other songs are meant to be loud. If you normalize by track, you are subverting the true intentions of the artist or band when they performed the show, and if you didn't document exactly what you did, there is no going back, so your version of the show is in a sense "polluted" forever.

[guygee]

(Note to self ) Even the operation "Remove DC Bias" can cause digital clipping. I always remove it because I figure it is an artifact introduced by the recording gear, but since it shifts the whole waveform up or down, it can clip off "wavetops" in the direction it is shifting the waveform.

And using the "replaygain" tags in flac files and editing them to your liking can eliminate the need for a final normalization, but you still have to find the right levels when A/D'ing from analog sources.

[Lou]

Maybe the shows that have digital clipping were recorded with the volume too high to begin with? In which case there's nothing you can do about that, that's how the source is.

With analog cassette>digital transfer, however, there should not be digital clipping other than maybe an occasional pop, whistle or clap.

[guygee]

Originally Posted by rerem

[...]The way I try to do it-up what is first rate. If something seems scarce-yet what I have really is flawed...odds are a better version exists,and often it appears on a bt site. Later-after I do the next puter,with a better soundcard-I'll hunt through my analogs and see what still has not appeared online. Already several have shown up-and in better quality. A few probably won't appear until I digitize them. [...]

rerem - (You probably already know this, but for the sake of others who are new...)When you build your next computer, try and go with a soundcard with a "Stand-Alone" A/D convertor. Prices have gotten pretty reasonable on these. You do not want the analog signal entering your computer case, there is all kinds of nasty noise inside of there, and it will show up in your recordings. Doing the A/D outside, in a shielded seperate box, eliminates that problem.

Like you, I have a lot of cassette recordings, but all of unverified lineage. I keep waiting for better sources to come out, but cassette tapes do not last forever, they become increasingly demagnetized, so hiss increases with age.
Many of these shows are not circulating, so it makes me think maybe I should get on with the job. If better sources emerge, people could always just trash my version. But since it is a lot of work to do a good job on even one show, first I think I would 1) Re-listen and start with the best sounding shows that don't circulate in digital form, 2) "google hard" and try to trade-up to get multiple copies and see if anything better is out there. I would probably want multiple copies anyways to use as patch material if my copy has any drop-outs, cuts or other flaws that were introduced somewhere after the master recording was made.

[guygee]

Originally Posted by Lou

Maybe the shows that have digital clipping were recorded with the volume too high to begin with? In which case there's nothing you can do about that, that's how the source is.

With analog cassette>digital transfer, however, there should not be digital clipping other than maybe an occasional pop, whistle or clap.

I am curious, what is the manufacturer/model of your A/D unit? I have both a Motu 828 24/48 and a M-Audio Duo 24/96. Both of of these units have input level controls that you have to set yourself (e.g. you have to twist the knobs to the correct settings). I can guarantee you, from personal experience, that if you set the input levels too high on these units clipping does occur, the Motu unit even has that warning led that tells you that you have clipped. If I set my input levels to full setting, massive clipping will occur on anything but the weakest input signal (any such analog recording would be very poor, very much lacking in dynamic range).

Maybe your unit has some sort of "input level limiting" feature that prevents clipping? I do not think I would want to use such a feature, because the peak amplitude values are different for each analog recording, so such a feature could easily result in unintentional compression. I would rather have control myself, and I do not want to compress any shows.

I would be interested to know if there are any other methods to prevent clipping when A/D'ing a cassette, but I cannot think of any other possible ways other than what I mentioned above. With my A/D units, setting levels is just like what a taper has to go through, except the taper gets only one chance to get it right, while I can tweak to my heart's content until I get it just right.

[freezer]

RE: this thread and the current seed on TTD of the Entwistle "rough mixes" from 7/18/79...

Just so it's known, I mastered the Entwistle 7/18/79 rough mixes from a cassette which is now the furthest back the material can be traced in that format. (I have no idea if the 2" master multitrack reels still exist. OR if any of the multitracks were re-recorded.)

The Rough Mix reel for 7/18/79 was destroyed soon after my cassette 1G copy was recorded. (Any remaining tapes from that bankruptcy sale were sold as blank reels after being bulk erased.)

The 1G cassette was recorded too "hot" and does have some flaws, so I made 3 different digital transfers and sent the discs out to three different seeders for offering on three different sites.

Each transfer is slightly different.

However I used no digital noise reduction on the set offered at TTD.

Originally Posted by guygee

If the seeder A/D'ed from an analog source, or from a rare/uncirculating digital source, and applied NR processing, then I think it would be fair if the seeder would offer, upon request, to distribute the raw source in the original form (hopefully 24 bit/high sampling rate) to any person claiming the experience to do a better job, or to anyone offering to archive the files on a publicly accessible site.

Your opinion only. Others may not agree.

Your mileage may vary.

[Lou]

Originally Posted by guygee

I am curious, what is the manufacturer/model of your A/D unit? I have both a Motu 828 24/48 and a M-Audio Duo 24/96. Both of of these units have input level controls that you have to set yourself (e.g. you have to twist the knobs to the correct settings). I can guarantee you, from personal experience, that if you set the input levels too high on these units clipping does occur, the Motu unit even has that warning led that tells you that you have clipped. If I set my input levels to full setting, massive clipping will occur on anything but the weakest input signal (any such analog recording would be very poor, very much lacking in dynamic range).

Maybe your unit has some sort of "input level limiting" feature that prevents clipping? I do not think I would want to use such a feature, because the peak amplitude values are different for each analog recording, so such a feature could easily result in unintentional compression. I would rather have control myself, and I do not want to compress any shows.

I would be interested to know if there are any other methods to prevent clipping when A/D'ing a cassette, but I cannot think of any other possible ways other than what I mentioned above. With my A/D units, setting levels is just like what a taper has to go through, except the taper gets only one chance to get it right, while I can tweak to my heart's content until I get it just right.

I actually don't have any taping gear, I go to at most 2 concerts a year, it's just not worth spending hundreds of dollars to buy something when the likelihood is high nowadays that I'll be able to trade for it.

Converting analog cassettes to CDR however, I have a Sony RCD-W500C standalone burner which has a VU meter, letting you know immmediately if it's clipping. I always edit everything I transfer with Soundforge so I'll see the clipping there as well.

So I'm not talking from experience, but it's just something to consider that maybe the taper set the levels too high, so it's clipped not because some bozo edited it and made it that way, but because that's how it was taped.

[guygee]

Originally Posted by Lou

I actually don't have any taping gear, I go to at most 2 concerts a year, it's just not worth spending hundreds of dollars to buy something when the likelihood is high nowadays that I'll be able to trade for it.

I am considering getting some gear, despite the cost, because I have become really pissed-off at myself for not having taped some of the really good shows I've seen in my lifetime that will never circulate, like John Hartford solo, tap-dancing on a piece of plywood in a Pittsburgh pizza parlor (late '70's), or like Gregg Allman, surprisingly burly, who showed up at a tiny bar around here (EC FL coast) and jammed with some blues guys for awhile. A couple of years ago, I tried to trade for the 1986-07-03 Petty-Dylan-Dead show that my brother and I saw together so I could make a really nice birthday gift for him; all I could find after months of searching was 3 copies, no lineage, all pretty high-gen (yes, NR was applied to my "remaster"). I sent out about 50 offers, so I guess my list just wasn't good enough, despite being fairly large. I've never seen the show torrented either. Too bad for me I didn't have the Master's of all those shows I'd seen in my life...

Originally Posted by Lou

Converting analog cassettes to CDR however, I have a Sony RCD-W500C standalone burner which has a VU meter, letting you know immmediately if it's clipping. I always edit everything I transfer with Soundforge so I'll see the clipping there as well.

So I'm not talking from experience, but it's just something to consider that maybe the taper set the levels too high, so it's clipped not because some bozo edited it and made it that way, but because that's how it was taped.

I'm just starting to research the newest technology in taping gear, so I am not talking from experience either, but I know with older shows tapers had to set their levels, and some of them set them too high, you can hear the results in their recordings, so I agree with you here.

But the clipping can also come from someone is careless (with your unit or mine) and doesn't watch the preset levels and/or the VU meter every second, and then clipping can occur (major clipping). At least you know you can check for clipping in Soundforge, newbies might not. And even if the newbie finds the clipping, will he go back and start from scratch, or will he use that little "peak rounding" tool to fake it and think he just did a "good job"?

The point I was trying to make earlier is that, at that one moment when the show hits its highest peak, you want your VU meter on your Sony RCD-W500C to "almost, but not quite" touch zero dB. For 16-bit units that holds true (well it also depends on the noise floor level of the cassette) but for 24-bit, maybe it is overkill since 24-bit has more dynamic range than even the cleanest cassette (127 dB vs. 90 dB sound about right?)

[guygee]

Originally Posted by freezer

[...]
The Rough Mix reel for 7/18/79 was destroyed soon after my cassette 1G copy was recorded. (Any remaining tapes from that bankruptcy sale were sold as blank reels after being bulk erased.)
[...]

I went through a time when I bought a lot of unknown reels on ebay, sold as blanks. Maybe I have these? Then, just like on "CSI Miami", I could use special "magnetic ghost recovery software" and recover these shows?

Life is too short...

Thanks for the very nice seed, freezer.

[freezer]

Originally Posted by guygee

I went through a time when I bought a lot of unknown reels on ebay, sold as blanks. Maybe I have these? Then, just like on "CSI Miami", I could use special "magnetic ghost recovery software" and recover these shows?

Life is too short...

Thanks for the very nice seed, freezer.

Life's too short to chase after the same show a dozen times looking for the 'best sounding version'.

Better to waste your time recording shows with a realistic mono cassette recorder from PBS Lite Jazz in the Morning.

And you're welcome. 8 more to come at here at TTD and more on other torrent sites.

And I'm breaking out the Sound Forge this evening for a work-out on Who Biloxi 1982-12-01.

[Lou]

I was basically referring to say someone taping with a DAT machine who set the levels too high. If that happens you'll see digital clipping.

You should never see digital clipping though from an analog source...even if it's distorted on the tape, it doesn't have to be distorted on the digital transfer too.

[Five]

clipping can also be intruduced thru an ANA > ANA transfer.

best thing to do is record a bit at a safe level which is too low, zoom in on the peaks and see what clipping exists on the master. then crank it up slowly, each time checking if the peaks are still okay. back it up a couple db from the top and run the whole tape.

and watch out. vu meters look cool and everything but they're averaging meters and won't always show momentary spikes. give a gander in a wav editor to really see how good your transfer is. Also, compare with your ears. Something is always lost, but you can minimize that by cross-checking frequently.

[el c]

A couple cool links re: clipping and dynamic range issues. Important for the fixer-uppers out there and an interesting history of professional CD engineering.
http://www.mindspring.com/~mrichter/...s/dynamics.htm
http://www.cdmasteringservices.com/dynamicrange.htm

from experiance i have found if you master at 48 khz 24 bit you will not get as many digital artifacts from noise reduction !

as for what to use i reccomend waves x-noise !

[tomska]

To add my opinion, i often use soundforge to clean or boost a recording this is for my soul benefit,i will never corrupt the origional source which can be passed around the pool,as is.The saying "a little knowledge is dangerous"rings true.I consider myself to be quite good at remastering but who am i to make such an arrogant claim? It's all down to horses for courses, i weld lumps of metal together for a living, thus i'm no studio engineer,and i never will be,but there are people out there who think otherwise.I say leave the trickery to the experts.Never corrupt the master.