Detailed VolAdj Parameters

Voladj parameters can be tweaked in the config.ini file, using sets of five integers, as follows:

	voladjlow=0x1800,100,0x1000,25,60    ;default value shown
	voladjmed=0x2000,409,0x1000,27,70    ;default value shown
	voladjhigh=0x2000,3000,0xc00,30,80   ;default value_shown

The meaning of each of these is described below:

factor/second

The rate of change of the volume. Expressed as a fixed point number, with the point 16 bits in. e.g. a factor of two per second would be (2 << 16)

minimum volume

This is the smallest volume that you would you like to hear, in sample size.

headroom

After calculating the multiplier, it will be multiplied by this factor. This has the effect that large samples (especially freak samples) are less likely to cause radical volume changes to avoid clipping. Expressed as a fixed point number like factor/second. e.g. 0.75 is (3 << 14), 0.5 is (1 << 15)

real silence

If the maximum sample size for a chunk is less than this amount, then it is assumed to be silence, so the volume is gradually reduced back to normal. Expressed as a sample size.

fake silence

If the maximum sample size for a chunk is equal to this amount, then it will try to scale it to the "minimum volume" parameter above. If the size is between real silence and fake silence, it is treated as though it was fake silence.

From The files of Tony Fabris:,

From my old "Picker and Voladj for Dummies" readme file:

The first parameter is the "factor per second". This is then shifted left 16 bits and expressed as an int. So to make voladj turn up the volume so that the output is increased by a factor of two every second, you set this to (2 << 16). Weird things can happen if you set this very high or very low, but it should be OK to give it sensible values.

The second parameter is the minimum volume. It attempts to make this the quietest thing you hear. The units are sample magnitudes. If you set this to 0, then none of the other parameters should have any effect. If you set it to (1<<15)-1 then in theory everything should be the same volume, but more likely is that my log routines would get confused. Best to stick with numbers less than 30000, I think.

The third parameter is the "headroom". This attempts to reduce the number of times that a hasty volume correction is needed to prevent clipping. This is a fraction between 0 and 1, shifted left 16 bits and converted to an int. You should probably set this to 1 << 16.

The fourth parameter (real_silence) is the sample magnitude below which we assume that there is no listenable music present, so we gradually turn the volume back down.

The fifth parameter (fake_silence) is the sample magnitude that will get mapped to minvol. This must be greater than real_silence.

If your samples are between real_silence and fake_silence, then they are treated as if they were of magnitude fake_silence.

Note: Pass in the left shifted values (0x20000 for the first parameter). In an attempt to make it a bit easier to remember, sample magnitudes are just raw, and (possibly) fractional values are expressed as fixed point, with 16 bits to the right of the point.

Editor's note: In the descriptions above, notations such as "2 << 16" represent a Left Bitshift operation. For instance, "2 << 16" represents 2 left-shifted 16 bits, which works out to hexadecimal 0x20000 or decimal 131072.

You can pass either the hexadecimal values ( 0x20000 ) on the command line or their decimal equivalents ( 131072 ). When using the hex values, make sure to precede them with the "0x" prefix so that the software knows it's a hex number. Note that you can use the Windows Calculator in Scientific mode to convert numbers between decimal and hexidecimal if you need to.