is it possible to adjust the normalize?

thinking about it I just remembered that in Sound Forge you can choose to normalize by peak or by RMS, which could give two very different results when applied to a sample, with RMS obviously more likely to give you clipping if you do it to too high a db level. if you're not happy with the mpc results, probably best to normalize in the computer

thank you,thats what im saying! the normalize does it to one setting which is cool but still could be better.

-nocron

Yeah the whole RMS / peak thing depends on the crest factor.
If you have a low crest factor then you can get away with RMS without too much audible nastyness.
This link explains crest factor quite well. http://www.dliengineering.com/vibman/crestfactor1.htm


Basically it depends on how the MPC is computing the normalise function.
If it's doing proper peak normalising it should subtract the level of the loudest sample from the maximum loudness level, then add this number to all the sample's levels.
This should not clip any levels. This is such an easy algorithm to do they should not cock it up!

A good way to check this is to normalise a series of sine waves and observe the output spectrum. You can get spectrum analysers for computers, which use the soundcard as a digitiser. With a sine waveform you get a single frequency component only. If it is clipped, you will start to get harmonics at integer frequency multiples. For example massive clipping of a sinewave to make it look more like the top has been chopped (aka sort of square-ish) will have tons of odd order harmonics, i.e. 3x, 5x, 7x the fundamental (sinewave) frequency.

If the MPC is doing RMS normalising (which I doubt) then as mentioned above, the crest factor will affect the results. You could engineer a sample to have a very high crest factor and try normalising it then transfer it to a computer for analysis. If you could keep it in the digital domain (i.e. use the SPDIFs) then you should have a high level of accuracy. Decent wave editors allow you to measure the individual sample amplitudes. Since normalise is a linear function it should scale all samples equally. By measuring, say, positive and negative peak sample levels before and after normalisation you should highlight any odd behaviour.

Finally, might I suggest it _could_ be psychoacoustic. When you next get one of these dodgy samples, record it into a computer and check for any plateaus in level on the peaks (indicating digital clipping), if you can't see them but can hear them, then maybe it's some odd human brain behaviour. (like this... http://www.youtube.com/watch?v=QbKw0_v2clo)

i do everything in my mpc,never use computers with my tracks.(unless u count the mp)

I think you´ll get a similar effect as a rms normalise if you use the compressor and resample.

I never use the computer but to burn stuff, and the net obviously.

PrimeBeatz, its Williams. You've never heard that before? I was refering to montels nation wide dope bus. Kickin out free scripts to the needy old folks with no insurance. Spittin image.

normalising samples turns up the noise floor with everything else. so, if you think turning everything, including the noise, up to max, and then turning it back down again when you mix, is good ..... well, ok.

yo rinse,thats what was doing but it seems faster just being able to adjust the normalize to match say like a drum loop thats at the end of a song. maybe others will need it as well but i dont see it happening.

-nocron

jackymurda wrote:normalising samples turns up the noise floor with everything else. so, if you think turning everything, including the noise, up to max, and then turning it back down again when you mix, is good ..... well, ok.

Actually while I half agree, normalising will improve your signal to noise ratio...


For some reason, say you have recorded at a much lower level, normalising will increase the level of both the wanted signal as well as the noise floor as jackymurda says. However, the DAC will now be producing more output level, so you will need less gain from the MPC output amplifier and any mixer pre-amps. This is good because as you turn up the gain of the amps you amplify: the audio, the noise from the recording and the noise from the amp.
As you can see, there is an additional noise source: the noise from the amp itself. Depending on the quality of the various amps, cables and recorded noise floor, you may not notice, but from a purely technical point of view it makes sense to use the full range of the DAC to provide the "hottest" signal you can.
Also remember normalising is a linear funtion, so the recorded noise is increased as much as the recorded audio. If you then turn it down on the mixer you reduce both equally again. The trick is to have the DAC as loud as possible and every other amplifier turned down as much as possible afterwards.

It's always good to record as loud as you can without clipping, so ideally you record as close to 0dBFS as possible. This way the signal to noise ratio is good from the beginning and you utilise more of the dynamic range of the ADC (and therefore DAC). This will reduce the amount of quantisation errors, because for a given audio signal you are using more DAC codes to reproduce the dynamic range.

dave_g wrote:It's always good to record as loud as you can without clipping, so ideally you record as close to 0dBFS as possible. This way the signal to noise ratio is good from the beginning and you utilise more of the dynamic range of the ADC (and therefore DAC). This will reduce the amount of quantisation errors, because for a given audio signal you are using more DAC codes to reproduce the dynamic range.

YEAH. What HE said....

having the possibility to set a level plus an 'all samples' option would be great!