Correction files

At the moment, correction files are only calculated for following FIR processors and softwares in both linear and minimal phases versions :

  • generic 1024, 2048, 4096, 6144, 8192 and 16384 taps : mono 24bits 48kHz .wav files
  • MiniDSP 2x4HD 2048 taps : mono pcm 32bits float 48kHz .bin files
  • MiniDSP OpenDRC 6144 taps : mono pcm 32bits float 48kHz .bin files
  • QSC Qsys 8192taps : mono 24bits 48kHz .json files
  • Xilica Solaro QR1 or FR1 4096 taps : mono 24bits 48kHz, .json or .txt files (ie with Rephase, just export as 32/64 bits floats lines text file)

For FIR corrections, the number of taps is the number of audio samples of the correction impulse, ie for 2048 taps, the length of the IR is 2048 samples (for any format, wav, pcm, bin, json,…).

I don’t think that I will also calculate IIR (parametric) corrections : it is very depending on the actual processor (number of parametrics, Q type,…) and FIR processors are becoming cheaper and more common (ie MiniDSP, Xilica,…). Maybe for parametrics, I’ll just export a measurement file that could be used directly into REW so to calculate IIR correction.

If you bought a plan with correction included, the correction files are found in My measurements page in the correction folder with following names and in both formats .wav (48kHz 24bits) and .bin (pcm 48kHz 32bits float, big endian):

  • …-hyblin-L : linear phase hybrid correction for left channel
  • …-hyblin-R : linear phase hybrid correction for right channel
  • …-hybmin-L : minimal phase hybrid correction for left channel
  • …-hybmin-R : minimal phase hybrid correction for right channel

To use json files for QSYS, you have to modify the json file in the correction folder : delete two first lines and first column, change the remaining column into a row with “,” between values and place brackets at both ends, ie [0,0,1,0.05,……0,0]. For Xilica, just delete the two first lines and the first column.

Important note : because correction can be positive in amplitude at certain frequencies, the software reduces the global level of the correction to avoid any overload. So when comparing before/after correction, you have to adjust level to match (with typical setup, you will generally need to reduce uncorrected level by about 3dB for a valid comparison).