iXBT Labs - Computer Hardware in Detail






Aspects of quality
digital recording


Today we will talk about the practice of digital sound recording. It's obvious that some day an owner of a PC will come to do it. However the process is not simple and you have to acquire some skill and experience to reach maximum quality. Here we want to give you (especially if you have the SB Live!) some recommendations which are not properly described in the papers or on the Net, in order to prevent problems connected with sound recording (such as noise and pickups).

I should note that all you will read here is a result of generalization of the experience of many people, but it still mustn't be taken for gospel.

General remarks

The following sources are used most often for digital reording:

Vinyl records

A good turntable and a preamp-corrector are of vital importance here. We write a record to a computer, remove clicks, filter infra-sound lower 16 Hz (to remove roar) and distinguish songs. Since 65-70 dB noise on the output of a player is not significant, you'd better not clean it up. And as for a background (unpleasant low-frequency tone in 50, 100, 150 etc. Hz) you better check before sampling - check the ground, terminals inside the player.


I mean a good mic and a microphone amplifier. You can read a lot on the Net, I want give you just one piece of advice. Many of you already know much about signal twisted-pair cables, but how to solder cables on the ends....

The picture show how to make a cable which won't influence the record quality (if it consists of qualitative cables). A copper braid serves a screen (it's desired that the copper be used everywhere!). Signal threads inside the screen are a twisted pair of copper multithread cables. I recommend you to buy such cable in a specialty store. Note that you should pay such careful attention to a cable only in the case of the mic, otherwise you will have to change mics and mic amps infinite times.

The principle is:

The screen is connected to a metallic base of the cap (if it's metallic) or just hangs inside a microphone case, and on input (only!) the screen is connected to the ground. In this case noise and pickup flow down into one place and do not circulate around the screen. The screen mustn't be "an antenna for the noise"!

In case of an electric guitar the principle is the same! Connect a screen of a circuit diagram (make it out of tin or foil) inside the guitar to the screen, and then connect cards or attachments with the ground on the input. Don't play the guitar near your system block - better make a longer cable.

Do not damage studio facilities! Just check how the connectors are unsoldered. A mic with a professional connector CANON has usually the ground on a pin 3 (otherwise, make sure asking the salesman)

Cables of the microphone amplifier/processor/tape-recorder are to be soldered according to the same principle (a screen are to be connected to the ground on the input), there is added one more thread of the second stereo channel. Note that you better ground the system block.

I think you would agree that a competently created cable is not that bad as compared with the Monster Cable or the Vampire; though if your 6-channel DSP-amplifier costs $1000 - better buy a decent cable.

The microphone record needs only a frequency correction (if the AFC of the microphone has a noticeable unevenness and a fall before 18-19 KHz).

As for a compact-cassette, it's not so simple. The record and artefact removal is a real art there. The matter is that a compact-cassette has a limit by 300 playbacks. After that the information can disappear, sound can get vague and broken. On the one hand, a CD with polished MP3 files is better than to mess up with a cassette since you need to clean a track with spirit, not mentioning the head. On the other hand, a restored phonogram from a decent quality cassette will offer much more details and warmness as compared with many 128 Kbit MP3-files.


First of all, you should know that if your device gives out audio data only in the range of 40-14000 Hz on a Ferro-cassette, in any case it reproduces higher frequency (up to 20-21000 Hz) but they do not correspond to the Standards and are weakened and drowned in noise. This is exactly what we are going to correct.

1. AFC approximation

Practically on any standard cassette (TDK D etc.) or an expensive one the audio record lies practically equal. If do not have the TEAC at $900 or the Nakamichi Dragon, then it doesn't matter - the AIWA or the VEGA. On such devices the quality of the reproducing signal mainly depends on the properties of the tape and less on the reproducible track. The magnetic tape on ferro cassettes won't allow the frequencies higher than 14,000 Hz, and the others will be strongly weakened.

Here you can see the AFC for the track "record-play" of the tape-recorder VEGA:

Track "record-play" depends on a tape-recorder. In order to carry out an investigation of the tract (and for the following mastering) you will need the last versions of the following programs (or at least their demo-versions):

  • Cool Edit Pro 1.2, or 2000 for creation of white noise, sine, filtering and noise removal etc.
  • Sound Forge 4.5 for editing (sampling etc.)
  • SpectraLAB 4.32 for total analysis

  • You can also use external denoisers such as the Sonic Foundry Noise Reduction 2.0, Arboretum Restoration-NR etc., though I prefer an integrated Cool Edit Pro (as well for the filtering).

Note that a commercial version of the Cool Edit (Cool Edit Pro 1.2) greatly differs from its free brother (Cool Edit 2000). For example, the Cool Edit 2000 doesn't have a function Preview, Multitrack-mode etc. Nevertheless, all necessary for our problems to be solved are included in the both versions of the Cool Edit.

The procedure of track researching should be the following:

  1. Clean a reproducing head of the tape-recorder and a clamping roller with a drive shaft, put a clean cassette (both sides should be clean!) in the tape, prepare it for recording.
  2. And then, adding the white noise* (48 KHz) into the input (Generate/Noise…/White, mono, intensity = 12 in the Cool Edit) obtain an optimal record level on your tape-recorder and make 3-minute recording (if your computer is not lower than the Celeron+64Mb) or a 60-second recording of this noise (if the machine is weaker). A magnetic tape responses irregularly on the Frequency Sweep that's why you better use white or pink noise. It will help you to reach more even and exact graph of the AFC.
  3. After that you should record the received white noise (without ejecting the cassette and turning the reproducing head) - from the cassette to the Line-In of the sound card (sampling frequency should be 48 KHz).
  4. Now normalize the received signal for -1 dB (Process/Normalize… in Sound Forge 4.5), and save.
  5. In SpectraLAB you should choose Mode/Post Process and open the received file. Make settings: 48000, 16bit, FFT Size = 65536 (16384 for a weak machine), mono, Average = linear. Mark the whole file in theView/Time Series mode, click the right button and order: Compute and Display Average Spectrum.
  6. Now we are looking at our spectrum, and notice how uneven is the AFC of our tape-recorder J. In the Cool Edit you should open the same file and raise up the weakened frequencies (FFT Filter etc.). Save (but not close) into another file and then open this other file in the SpectraLAB (the first one will be useful in case of failure of the second). In order to save spectral envelopes in SpectraLAB for comparison - click Set in Overlays.
  7. It doesn't suit you? Cancel the filtering in the Cool Edit - and do it again!

* The pink noise is preferable since it's closer to the music signal in its properties.

The Record_play tracks for good devices are very close, and you can use my results for the AFC correction of the VEGA-122C tape-recorder. So, in the file C:\Windows\cool.ini find a partition [Filters96] and record the following in the free line (one line!):

Item29=RESTORATION,3,19,0,20,426,5,845,0,1288,0,1986,0,2259,0,2855,6,3179,9,3444, 21,3583,28,3688,42,3773,48,3848,61,3925,76,3957,96,3998,100,4004,100,4012,5,4096,5, 19,0,20,426,5,845,0,1288,0,1986,0,2259,0,2855,6,3179,9,3444,21,3583,28,3688,42,3773, 48,3848,61,3925,76,3957,96,3998,100,4004,100,4012,5,4096,5,2,0,12000,1,2,0,0,1000, 100,5,-10,100,-0.5,12,24000,1,0,1,1,48000

Now in the Transform/Filters…/FFT Filter (in the Cool Edit) find the preset RESTORATION and correct the phonogram. That's all. It took me two days to get the value +/-0,5 dB from 10 to 19500 Hz for the AFC of my VEGA. Enjoy it! The AFC of the VEGA-122C before (the green line) and after (yellow line) the frequency correction:

Besides, all of you know that the AFC of the sound card itself and moreover of the external ring Line-Out-Line-In also has a definite slump in the high frequency range, and it's usually the ADC to blaim in the greater degree than the more or less linear DAC. The sampling is wan, lacking clarity and lightness in the high register. You have to solve this problem before processing the recorded material; we'll talk about it later.

2. For the SB Live! owners.

Many owners of the SB Live!1024 Value knew that the AFC of the codec SigmaTel STAC9721 is not ideal and after 4.5 KHz there starts a stepping decline:

The main contribution is brought in by the ADC of the codec. This was done in order to prevent penetration of parasitic harmonics into the audible spectrum part. On the other hand, such a cut of highs has negative consequences: with the good Hi-Fi devices it can be noticeable.

That's why I would recommend to all SB Live! owners first (before the following processing) to pass every phonogram, recorded through the Line-IN, through the next filter (the same cool.ini and the same partition):

Item36=CORRECtion,3,20,0,0,83,0,532,1,793,1,1003,2,1223,4,1713,5,2046,10,2391, 12,2569,15,2710,18,3066,24,3234,27,3398,35,3480,41,3546,47,3628,56,3726,70,3825, 89,4096,100,20,0,0,83,0,532,1,793,1,1003,2,1223,4,1713,5,2046,10,2391,12,2569,15, 2710,18,3066,24,3234,27,3398,35,3480,41,3546,47,3628,56,3726,70,3825,89,4096,100, 2,0,12000,1,2,0,0,1000,100,3,-10,100,0,14,24000,1,0,0,1,48000

After such operation the unevenness of the AFC in the ring LineOUT - LineIN for the SB Live! sound card was reduced to +/-0,1 dB!

The AFC of the SBLive sound card (the STAC9721 codec) before and after the frequency correction. The scale of the graph is 5 dB!:

Now we have the "reference" linear input! Enjoy...

If you have the MD-52A mic (professional, 20-16000 Hz) I'd recommend the following correction line:

Item40=MIC_MD52A,3,8,0,0,1723,17,2158,37,2569,53,3176,89,3316,100,3480,100,4096, 100,11,0,50,4,51,5,60,6,73,8,81,10,85,14,87,17,79,19,60,24,51,4096,50,2,1,6000,1,4, 0,0,648,31,831,57,1000,100,3,0,100,0,16,24000,2,1,0,1,48000

3. Noise elimination

I think that the Noise Reduction in the Cool Edit is the best. Sound producers praise the Arboretum Ionizer since it doesn't bring in phase artefacts in the processed material, my friends advise the Sonic Foundry Noise Reduction. Besides, there is a not bad restoration program DART Pro but we are not going to compare all them in this article.

So, after quite stressful filtering you should take 1.5-1.6 second noise before each or just any song (but only from the same side of the cassette where you have taken the phonogram from). Further:

  1. Get a profile of this 1.5 second noise (Get Profile from Selection) with the following parameters:

    Snapshots in profile = 1024…4096 (depending on the power of the computer).
    FFT Size = 4096 (I have found this figure the most optimal for noise reduction); I don't recommend you to set the maximum size of the FFT since it will cause increase of hiss in the signal.

  2. Now you are to mark out the filtered phonogram and eliminate the noise with the same Noise Reduction in the Cool Edit. The parameters can differ (the Preview button) but I can recommend you the following parameters for the tape-recorder:

  3. Precision Factor = 12 (for maximalists it will be 14, for the Preview - 8). The parameter defines the precision of measurements;
    Smoothing Amount = 0. The parameter brings in a tolerance of extracting a noise spectrum from a signal spectrum in dB. The more the worse;
    Transition Wide = 6 (for very powerful devices ($400) - 3, for the cheapest - 7…9). The parameter is similar to the previous one but it controls the process of spectrum extraction more flexible (in dB);
    Noise Reduction Level = 35…42 (with the Transition Wide = 6 depending on the level of signal recording on the cassette. If the record was exactly under 0 dB, set 37). The parameter defines the threshold level of removable noise spectrum. Usually it depends on the Transition Wide value.

Often it's necessary to renormalize phonograms for -0,2 dB (this is the standard for CD-Audio), since the noise reduction can tremendously cut down the signal energy. Now everything is ready for MP3-encoding! However, they say that normalization before compression in MP3 is very harmful, and you should do it only in case of noticeable difference between the levels of separate channels or just in case of a weak (-3…-6 dB) level of the phonogram. This is the case when it's for you to decide.

4. Compression

As for me, I'm using a coder Fraunhofer IIS integrated in the BPM Studio Pro 3.0. In my opinion, other coders from the Fraunhofer IIS can't avoid a peculiar rustle in the received file, and at all bitrates. Besides, I take carefully the LAME -f, I think that it is good only on the AFC.

Now what concerns a bitrate. Frankly speaking, all nuances remain only at 256 Kbit/s, though at 192 Kbit/s the phonogram doesn't differ much from the initial one. In order to choose the best bitrate for a certain phonogram, first I encode a fragment which is rich of high frequencies or with some familiar nuances) into all bitrates: 128, 160, 192, 224, 256. Then in the Sound Forge I connect the original fragment with the encoded one and listen. Usually the real fight occurs between 192 and 256 kbps.

For the restored phonograms 192 kbps is quite enough, though some need even more than 256 kbps. Besides, tape recording is quite difficult material for encoding. So, you can trust devices and spectrograms, but you better make a control hearing. Nevertheless, this concerns mainly the owners of a decent track anplifier-speakers (headphones).


So, with enough urge towards getting maximum from what you have at your disposal, you may reach quite decent quality of record from any source. Do not just sit with what you have got from sound card manufacturers - but seek, study and demand the Sound.

Write a comment below. No registration needed!

Article navigation:

blog comments powered by Disqus

  Most Popular Reviews More    RSS  

AMD Phenom II X4 955, Phenom II X4 960T, Phenom II X6 1075T, and Intel Pentium G2120, Core i3-3220, Core i5-3330 Processors

Comparing old, cheap solutions from AMD with new, budget offerings from Intel.
February 1, 2013 · Processor Roundups

Inno3D GeForce GTX 670 iChill, Inno3D GeForce GTX 660 Ti Graphics Cards

A couple of mid-range adapters with original cooling systems.
January 30, 2013 · Video cards: NVIDIA GPUs

Creative Sound Blaster X-Fi Surround 5.1

An external X-Fi solution in tests.
September 9, 2008 · Sound Cards

AMD FX-8350 Processor

The first worthwhile Piledriver CPU.
September 11, 2012 · Processors: AMD

Consumed Power, Energy Consumption: Ivy Bridge vs. Sandy Bridge

Trying out the new method.
September 18, 2012 · Processors: Intel
  Latest Reviews More    RSS  

i3DSpeed, September 2013

Retested all graphics cards with the new drivers.
Oct 18, 2013 · 3Digests

i3DSpeed, August 2013

Added new benchmarks: BioShock Infinite and Metro: Last Light.
Sep 06, 2013 · 3Digests

i3DSpeed, July 2013

Added the test results of NVIDIA GeForce GTX 760 and AMD Radeon HD 7730.
Aug 05, 2013 · 3Digests

Gainward GeForce GTX 650 Ti BOOST 2GB Golden Sample Graphics Card

An excellent hybrid of GeForce GTX 650 Ti and GeForce GTX 660.
Jun 24, 2013 · Video cards: NVIDIA GPUs

i3DSpeed, May 2013

Added the test results of NVIDIA GeForce GTX 770/780.
Jun 03, 2013 · 3Digests
  Latest News More    RSS  

Platform  ·  Video  ·  Multimedia  ·  Mobile  ·  Other  ||  About us & Privacy policy  ·  Twitter  ·  Facebook

Copyright © Byrds Research & Publishing, Ltd., 1997–2011. All rights reserved.