|
show czech version of this page/zobraz českou verzi této stránky |
|
AUTOMATIC REMASTERER |
|
|
|
The main method of program isn't as expected clasic equalization and doesn't change frequency characteristics of sample about some multiple (number of dB) on given band on original, but it independenly on quality and frequency characteristics changes it to required state. It is not important, whether sample was few, or much distorted, result is always the same. This is proved also by fact that repeated processing doesn't lead to any other change. Program adapts sample with trimble independently on original trimble. It is very appropriate for editing sample on sound trimbles for which we've example, but we are not able to exactly revise it. Program is surely only one, which this way works with sound, since it was written mainly with my invented method nor so with other available method. If you have doubt about quality of these methods, it is nothing simple than trying of program for free of charge for 20 days on check-out. This 20 days is however only functional probation and in this time you are not permitted to use the program for commercial purposes. |
|
Program hasn't no special demands on hardware perhaps only in RAM size. Due to 32 bit processing it is suitable that RAM is minimally four times bigger than size of processed *.wav file. Mostly it is sufficient already 256 MB. Program works due to speed needs only with RAM memory. If there is memory lack it will use swap file, which cause radical reduction processing speed. However it has no influences to program functionality. |
|
Program is using for all calculation 32 bit representation of sound, which warrants dynamics of 240 dB in integer numbers and till 1667 dB in floating point numbers. This warrants after conversion back to output 16 bit format non-measurable calculation losses. Also frequencyprocessing through FFT (FAST FOURIER TRANSFORMATION) with block size 16384 samples on channel warrants high frequency accuracy. Method of noise autodetection noise requires that recordings consist of at least one place only with noise about length at least one second, otherwise it can reach the removal of also other sounds than noise so damage recordings. Unfortunately I am not conscious of that, that at present it subsists method of detection noise from sample, where there are not places with noise in clear form. Diferently from most of the programs there is not a need to search noise manually and the program finds noise. Due to 32 bit processing there is no possibility to "overburn" the signal, if peaks exceed behind borders of normalized signal, there're not otherwise displayed, but after normalization of signal they are back available distortionless. If you however sample save with overrun peaks,they will be cutted. |
