Records and Tapes to Digital

[Brownian Motion]

I know that some folks here do their own conversions of analog to digital, but, if your needs are limited, here is an alternative:

http://reclaimmedia.com/index.html

I've used them recently, and I was satisfied with the results.

Edited December 3, 2007 by Brownian Motion

[AllenLowe]

well, as long as we're advertising - I do transfers from LP to CD; using an audiophile turntable and high-quality A to D converters and will also give it a run through CEDAR to eliminate about 80-90% of surface noise; also will re-eq and can usually improve original sources - will do for $20 per LP plus shipping - I've done transfer work for Sony, Ryko, NPR, Micahel Feinstein, Shout and Rhino, and numerous small labels including Venus (in Japan), Global Village and others - my price is little high than these guys but CEDAR is much superior to the program-based de-clickers -

Edited December 3, 2007 by AllenLowe

[robviti]

anybody have an opinion about this part of their process?

Trimming & Amplification

The digital sound files are then trimmed to remove the silent "leader space" at the beginning and end of each side, so that only the full audio program remains. They are then amplified to the maximum possible volume to fully exploit the CDs 16-bit dynamic range. The amplified 24-bit sound files are then down-sampled to 16 bit stereo, the resolution of an audio CD. At this point, we have two fully-processed sound files, one for each of the record's sides.

and again:

Do you do normalization?

Sure. Every side of every record is amplified to just max-out a CD or MP3 file's dynamic range. This always comes out perfect because all records are originally mastered so that each cut has the same volume as its neighbors.

[AllenLowe]

the first part is ok (the 24 bit to 16 bit) but I would run from normalization, which sucks - they are doing this to keep the service inexpensive and reduce the need for human attention - in normalization the program "evens" out the level peaks and valleys by just basically looking at the wave form - since it is not really "hearing" the music it is simply doing this based on a digital mathematical estimate - so a quiet passage may get artificially boosted to match an average - in reality a piano at -3 dbs sounds much differnt than a sax at -3 dbs - so it is apples and oranges, and the program does no know this - this has to be done by ear, by a person, to be done correctly -

Edited December 3, 2007 by AllenLowe

[Brownian Motion]

well, as long as we're advertising - I do transfers from LP to CD; using an audiophile turntable and high-quality A to D converters and will also give it a run through CEDAR to eliminate about 80-90% of surface noise; also will re-eq and can usually improve original sources - will do for $20 per LP plus shipping - I've done transfer work for Sony, Ryko, NPR, Micahel Feinstein, Shout and Rhino, and numerous small labels including Venus (in Japan), Global Village and others - my price is little high than these guys but CEDAR is much superior to the program-based de-clickers -

I wish I'd known; I probably would have paid the extra. I think I put a feeler out on the board about a year ago, looking for someone to do this or to suggest someone who would. Must have been too subtle a feeler.

[AllenLowe]

I used to do that work for a living, but have cut back due to a day job; LPs, however, are pretty easy to do; just shoot me an email if you ever need anything -

[Daniel A]

in normalization the program "evens" out the level peaks and valleys by just basically looking at the wave form - since it is not really "hearing" the music it is simply doing this based on a digital mathematical estimate - so a quiet passage may get artificially boosted to match an average - in reality a piano at -3 dbs sounds much differnt than a sax at -3 dbs - so it is apples and oranges, and the program does no know this - this has to be done by ear, by a person, to be done correctly -

They could be doing some sort of RMS normalization which is slightly better in this regard, as it takes into account the sound levels as perceived by the human ear, but another drawback is that they will have to record everything with a generous headroom to avoid the risk of clipping, and then normalize, as they will probably not play through the whole LP beforehand to note the peaks. This way the dynamic range - of both the analog and digital parts of the chain - is not used optimally.

Do you do normalization?

Sure. Every side of every record is amplified to just max-out a CD or MP3 file's dynamic range. This always comes out perfect because all records are originally mastered so that each cut has the same volume as its neighbors.

The last part (italics by me) is false.

Edited December 4, 2007 by Daniel A

[couw]

I think they mean the same by amplification and normalisation, as in applying a constant amount of gain (=turning volume knob), maximising volume without distorting the signal. As they are applying it to sides of an album and not to single songs and as long as there are no distinct differences between the audio level of the 2 sides, this should work out fine and not affect the dynamics.

[couw]

Do you do normalization?

Sure. Every side of every record is amplified to just max-out a CD or MP3 file's dynamic range. This always comes out perfect because all records are originally mastered so that each cut has the same volume as its neighbors.

The last part (italics by me) is false.

yes, that part is weird

[Dan Gould]

I stay away from all software driven removal of pops and clicks. The results are a hundred times better if you do it manually in a program like Goldwave. I don't have a great TT, my soundcard was standard for our "media" PC (or however Gateway described it), yet I end up with LP transfers that have pleased the ear of that well-known lover of great sounding stereos, Jazzbo.

[jazzbo]

Dat's a fact Jack! Er. . . I mean Dan.

I'm sure Allen does amazing work for that double sawbuck!

[robviti]

... I don't have a great TT, ...

apparently, neither do they:

The record is played on one of an array of direct-drive Numark TT200 turntables.

[Dan Gould]

... I don't have a great TT, ...

apparently, neither do they:

The record is played on one of an array of direct-drive Numark TT200 turntables.

I don't doubt that puts mine to shame.

[AllenLowe]

"applying a constant amount of gain" - re normalisation -

Couw, I don't think this is correct, at least relative to one program I use, which basically just takes a print of the wave form and tries to make it "consistent" - and I have a feeling they are using a program to do this, not their ears - though I could be wrong -

turntables are important - I use a VPI and it really makes a big difference -

[Christiern]

I once used software to remove the surface noise--it did too good a job, so I sent the tape to Phil Schaap, who instinctively reinstated the noise and added bonus tick tracks.

[couw]

"applying a constant amount of gain" - re normalisation -

Couw, I don't think this is correct, at least relative to one program I use, which basically just takes a print of the wave form and tries to make it "consistent" - and I have a feeling they are using a program to do this, not their ears - though I could be wrong -

What you describe, would be a type of (dynamic range) compression, right? Applying different amounts of gain for different parts of the song.

I have seen the term "normalisation" used for that as well, particularly in audio programmes.

I'm no expert on these terms, but I have always used them in the sense wikipedia does.

Anyone know?

[AllenLowe]

on the program I use - Steinberg Wavelab - it does not apply compression, though that is another way to normalize a wave file - I suspect with these guys it is basically a method of trying to make sure levels are consistent - and a lazy way out, as to do this correctly and in a way that is sonically best, one needs to listen to the LP, find a peak passage, and set levels accordingly (though somewhat risky, this assumes that the LP will be internally consistent, and that other passages will not overload - and in 15 years of doing this I've never had a problem with an LP transfer ) -

[Daniel A]

I always want to keep the difference between the levels, track for track, for each side of the original LP. So before recording I play the LP side, set my software (Cool Edit as it is) on "peak hold", and when finished adjust the recording level accordingly (for example up 2.5 dB if the peak of that side is at -3 dB). My aim is to have the peak value of the side as close to 0 dB as possible without clipping.

[erhodes]

on the program I use - Steinberg Wavelab - it does not apply compression, though that is another way to normalize a wave file - I suspect with these guys it is basically a method of trying to make sure levels are consistent - and a lazy way out, as to do this correctly and in a way that is sonically best, one needs to listen to the LP, find a peak passage, and set levels accordingly (though somewhat risky, this assumes that the LP will be internally consistent, and that other passages will not overload - and in 15 years of doing this I've never had a problem with an LP transfer ) -

I use both Wavelab and Sound Forge to normalize, though I generally use the Sound Forge utility because, even though it is not a real time editor, the interface is easier to read and I seem to have more control over levels. My understanding is that all that a normalization utility does is apply a constant amount of gain to the selected waveform so that the highest peak value becomes whatever you set when you configure the utility. E.g., if I set it at -1 db the utility will apply sufficient gain to the selection such that the highest peak value is -1 db. This can be a useful thing to do but not for equalizing real or apparent volume between two selections. In other words, just because two selections are both normalized to -1 db does not mean they will have the same apparent loudness. As Allen points out, it depends on the instruments, whether the passage was soft or loud to begin with, the internal dynamic range of the passage and a million other considerations.

The problem I have had with Wave Lab is that my version will not allow me to normalize to fractions of a db, e.g. -.1 db, which is what I can do in Sound Forge. The problem I have had in Sound Forge is that the utility is not always accurate and sometimes applies to much gain, so that the highest peaks are clipped. I've resorted to using the analysis tool to find the highest peak and simply applying sufficient gain to take it to -.1 db. The analysis tool seems more accurate than the analysis done by the normalizaton utility, though I can't figure why this should be true.

Normalization has been useful to me as a final step in maximizing gain after a file has been edited and after one is satisfied that the internal dynamic contrasts are what they should be. It is all but useless as a tool for equalizing apparent loudness.

I should add, per Allen's comment about listening to the lp and finding a (the) peak passage, that I have virtually no control over the absolute volume of the initial analogue to digital transfer...something about the driver for my sound card. All of my lp's...in fact all of my analogue sources...generate wav files with peak values several db's below 0. So I edit and then maximize/normalize as a final step.

[Dan Gould]

Like Ed, I always err on the side of caution and keep my inputs low when dumping an album into the PC. Then after editing out pops, I "normalize" or "maximize" each track. Obviously it helps to get rid of the pops because otherwise they will inhibit the amount of gain applied to the music.

[Chalupa]

http://www.numark.com/tti