STAGES REQUIRED FOR GETTING HIGH QUALITY .MP3 FILE*

A. Getting initial sound file

1. Particulars of getting initial sound file from CD-ROM drive
2. Particulars of getting initial sound file from analog sources

B. Processing initial sound file**

1. DC offset removal and frequency response limiting to 20 Hz-20 kHz
2. Declicking and denoising
3. Frequency range equalization and special processing
4. Channels balancing, normalizing and sound form density increasing

C. Encoding into MP3


A. Getting initial sound file

1. Particulars of getting initial sound file from CD-ROM drive

o The CD's surface is of decisive importance for qualitative ripping. It doesn't matter if the scratches are visual or not (previously mentioned test CD "Europe. The Best 1983-1991" has no visual scratches at all, but the 17th track can be hardly read). The point is if they interfere with the laser beam. The CD can be full of visual scratches, but the quality of files received from it will be excellent. And vice versa. So, first, you should clean the CD's surface off dust and fingerprints by CD Cleaning Kit or soft cloth moistened in cleaning solution (all cleaning solutions for tape- and video recorder heads are of the same composition, so you can use any of them). Avoid detergents and solvents. After damp cleaning, breathe on the CD's surface and then wipe to dry the generated condensate layer. This little trick will let you get rid of stains and spots forever. Now you've got a CD with absolutely clean surface.

o If you wish to work off scratches on your CD with the help of restoring solutions, keep in mind that you can get as a result visually smooth surface of the CD which won't be actually read any better. The purpose of these solutions is to prevent further destruction of the CD caused by damage of the upper protective layer. If you rent the CD or just play it from time to time, this problem is not yours.

o You shouldn't use without extreme need kits for cleaning CD-ROM drive's lenses: having no direct contact with CD's surface, the lenses are not subject to direct action (unlike heads of tape- and videorecorders). Exclusion is work in conditions of excess dust and smoke (however, that could be hardly imagined).

o Forget about any ripping speed other than x1 (150 kbps). Only this will allow you to get sound file of high quality. Existing opinion that brand name (not pirated) CDs can be spinned at speeds up to x32 is absolutely wrong. You can't know if there is any failure, where it occurs, how long it lasts, what sort it is, etc. And in this case you will have to monitor the total file's length, that means setting of high ripping speed to be senseless. You can set ripping speed in ripper's menu. You may also turn on "spin up disk before reading" option (if available).

o The best results go to setting of ripping method to "unbuffered burst". Names may change, but the idea is that the CD is ripped sector by sector at minimum speed, without buffering. As for "jitter correction" mode, it is included for old CD-ROMs with unstable focusing. This mode is not necessary for modern drives, but you can experiment with it because tandem "particular CD - optics and mechanics of particular drive" is always different and the result here is not predictable.

o You should be careful while using Normalize option which is available in some rippers for peak level normalizing. The point is that tracks with different EQ at the same volume level sound with different loudness. This is caused by some peculiarities of the human ear. Perhaps, better results can be achieved with RMS normalization available in some sound editors.

o If you nevertheless got file of inadequate quality (usually with clicks and crackles), you can improve the situation using sound editors. Such defects as speech glitches, missed or duplicated segments are not subject to correction. In this case you should try either to change ripper's settings or to change the ripper itself.



2. Particulars of getting initial sound file from analog sources

To get qualitative sound file from analog sources through the line input of your sound card you should follow the following rules:

o Always choose 16 bit recording resolution. This standard is accepted in recording industry not for nothing: it provides maximum reasonable quality at minimum necessary costs. If you reduce resolution up to 8 bit - you will get sharp audible decrease of signal/noise ratio. If you increase resolution up to 20 or even 24 bit (modern professional sound cards support such resolutions) - you will get files of huge sizes at theoretically improved but practically imperceptible signal dynamics (signal/noise ratio better than 90 dB has no sense, because there always exists natural background premise noise which is much heavier).

o Sampling rate should be chosen taking into account desirable audio range: at 44,1 kHz the upper limit of frequency range will be 22,05 kHz (that is more than the corresponding parameter of most Hi-Fi systems which is usually 20 kHz). Thereafter, such setting has sense only if qualitative sound sources are available: vinyl record player, reel tape recorder at speed more than 19 cm/s, cassette tape recorder with Cr or Metal tape. But if you digitize speech, recordings from cassette tape recorder with Normal tape or reel tape recorder at speed less than 19 cm/s, programs of FM radio or satellite audio broadcasting, there is some sense to set 32 kHz sampling rate. The audio range of your sound file will exactly match audio range of these sources. Higher sampling rate will lead to senseless increase of file size, the less one - to distinct decrease of sound quality.

o Always choose recording level with maximum peak at -3 dB. The worst thing you can have while digitizing is overload. Distortion caused by overload can not be remedied in principle. In case if the volume level of your sound file is low, you can easily increase it up to maximum with the help of Normalize option or Waves UltraMaximazer plug-in.

o Always monitor noise and pickups level. This is the common rule for all cases when wires are used. There mutual location and connection quality strongly affect recording and sometimes may noticeably reduce signal/noise ratio and add crackles. Always use for power supply of your PC and recording sources line filters like APS SurgeArrest with the proper ground.

o It is highly recommended to apply so called "wet playing" while recording from vinyl records. In early 80-ies this method generated a real boom among Hi-Fi fans. It consists in the following: first, you should wipe the record with watered soft cloth (some antistatics may be added) to remove usual dirt and dust. Then, upon having lowered pickup on the record and started playing, with the help of pipette you should drop some water directly before the pickup stylus. There will be formed something like water path on the record's surface, where the stylus is sliding. In this case at the expense of friction reduction and water layer formation, the number of clicks and crackles will be in order of magnitude less, all kinds of distortions will be decreased and signal/noise ratio will be significantly improved. However you should control the stylus to be always in water and use the pipette as necessary.

o While recording from tape recorder, do not forget to clean its heads and capstans from dirt and tape powder beforehand. Otherwise you will get dull or "floating" sound.



B. Processing initial sound file

1. DC offset removal and frequency response limiting to 20 Hz-20 kHz

o DC offset is a shift of sound form from the baseline. The shift itself means nothing, but if you are going to edit sound file, first thing you should do is to remove it. Otherwise any further file processing will be done incorrectly and result in distortion. DC offset removal can be done by Remove DC offset command of the sound editor.

o Limiting frequency response to 20 Hz-20 kHz (i.e. to audible range) is necessary for reduction of distortion appearing during file processing and its subsequent DA conversion. This can be done by applying Paragraphic Equalizer of the sound editor.



2. Declicking and denoising

o For declicking may be used any considered plug-ins and appropriate options of sound editors. The result in every specific case will be different and depends upon type of clicks and particulars of sound form. Be careful - unsuccessful settings may result in sound peaks distortion.

o For denoising the best choice would be Sonic Foundry Noise Reduction DirectX plug-in. Other plug-ins and appropriate options of sound editors either intended only for broadband noise and tape hiss reduction or do not work in real-time. Be careful - unsuccessful settings may result in unpleasant artifacts and sound distortion. Don't be too enthusiastic about noise reduction degree (except for Sonic Foundry Noise Reduction plug-in where you can freely apply 100 dB noise reduction to get highest signal dynamics. Sound quality in this case will remain practically the same). Final noise reduction in pauses can be done with Noise Gate option available in most sound editors.



3. Frequency range equalization and special processing

o Frequency range equalization of sound file should be done very carefully. This editing field is fraught with appearance of artifacts and distortion. The best results while processing go for Waves Paragraphic EQ plug-ins. For bass enhancement (if the sound file has normal low frequencies) can be used Waves MaxxBass plug-in.

o Appearance of Steinberg FreeFilter plug-in kicked up a row in the world of PC sound editing. This unique plug-in is an intellectual equalizer able to remember reference phonogram and apply its frequency characteristics to your own sound file. Thus you can mix into one file different music pieces or match bad CD to your favorite sound pattern.

o For stereo enhancement and sound space processing you can apply Waves StereoImager+ plug-in. The best effect so far produces its Spatial Enhancer preset.

o For realization of recent fancy QSound technology you can apply QTools plug-in. With its help you can get effects of 3D sound, Pseudo Stereo, Expanded Stereo, etc. However it is necessary to note that yet becoming nice the sound acquires some affectation.



4. Channels balancing, normalizing and sound form density increasing

o Channels balancing is done by increasing or decreasing volume in one of the channels by Volume option of the sound editor. To apply this you should first play the file and mark average difference between channels in dB as per volume meter, then set this value in Volume option.

o For normalization, increase of sound density and audible resolution you can apply Waves L1-Ultramaximizer+ plug-in. This is the most effective plug-in for such operation. A similar Steinberg LoudnessMaximazer plug-in is ineffective. However, incompetent use of Waves L1-Ultramaximizer+ may result in sharp sound distortion as well. This operation should be applied only at the final sound processing stage.



C. Encoding into MP3

o While encoding into MP3 the optimum should be 128 kbps bitrate. Arguments of audiophiles like Tord Janssen (creator of BladeEnc) about differences in sound quality at 128 kbps and 256 kbps are absolutely abstract. The main point of MP3 technology is removal of excess information, that presumes definite losses and compromises. Simply speaking you will hardly or easy, but always tell original WAV file from its MP3 version both at 128 kbps and 256 kbps.

o While encoding at bitrates higher than 128 kbps you can assign Stereo mode (it's not available in most Fraunhofer based encoders). At standard 128 kbps encoding is usually used Joint Stereo mode. The difference is the following: while encoding in Joint Stereo some frequencies in both channels are mixed together. This method of stereo reproduction has been well known to Hi-Fi fans since 60-ies. But at that time it was applied only to low frequencies due to opinion that stereo effect is not apparent at such frequencies. Now this method is widely used in MP3 encoding, but is applied to some set of frequencies. As a matter of fact the true audiophiles always encode only in Stereo mode.

o Do not apply Checksum option (if available). The checksum data is needed for error correction when streaming the .mp3 file in real-time over Internet (as done by Internet radio stations). It slightly lowers sound quality since the checksum data also needs to fit in the specified bitrate and is not needed for normal use.

o You should fill in ID tag of your ready MP3 file. Then, while playing it you will have visual information about Artist and Title. To fill-in ID tag you can use the appropriate option of WinAmp MP3 Player.



* All operations on sound files should be done using headphones only. Monitor loudspeakers may be used only at final stage for expert listening.

** It is recommended to perform sound file processing exactly as per specified order of steps (1-4).