apopsisdj
I'm more than 100% sure that something is wrong with the gain in virtualdj.
In earlier posts i wrote that i hear distortion with Auto gain "Always 0db" and i had to reduce the gain about 2-3 db or more to clear the sound.
Now i wanted to do some tests for various compressed formats and bitrates, as there is a topic about this in general forum, and i used FFT graphics to find out frequency responces.
So i generated some "pink noise" and "white noise" samples (for those who understand what i'm talking about...)
I was VERY SUPRISED how virtual dj altered the sound adding "harmonics" to the original signal. It was not only visible but very very audible too.
What i found is: At every gain setting different than the original level (even for 0.1...) of the song, distortion and harmonics was added.
So double click the gain knob of a song (default skin).
Lets say it say -3 at the display.
By altering this setting either manually either auto (Auto gain) you get a terrible distortion.
Its not so audible for regular songs but for a tone like white noise it is very easy to see and HEAR...
I disabled Auto-gain and use manual gain in the hw mixer, like the old days...
until a fix came up..
And i dont understand !
HOW virtualdj is measuring the gain ? Are these really db? This in not "Peak" level,
either "RMS" level, what is it.
I don't trust it.
Soundforge, wavelab etc, report different values for every song i tested.
Vdj has "unique" -way off-, different results...
Example: a song mastered at peak -0.5 db, and has RMS power -12db (wavelab)
Virtualdj reports gain: -3.3 db It can be right.
I hope the Team will look at this.
Sound quality is directly affected.
In earlier posts i wrote that i hear distortion with Auto gain "Always 0db" and i had to reduce the gain about 2-3 db or more to clear the sound.
Now i wanted to do some tests for various compressed formats and bitrates, as there is a topic about this in general forum, and i used FFT graphics to find out frequency responces.
So i generated some "pink noise" and "white noise" samples (for those who understand what i'm talking about...)
I was VERY SUPRISED how virtual dj altered the sound adding "harmonics" to the original signal. It was not only visible but very very audible too.
What i found is: At every gain setting different than the original level (even for 0.1...) of the song, distortion and harmonics was added.
So double click the gain knob of a song (default skin).
Lets say it say -3 at the display.
By altering this setting either manually either auto (Auto gain) you get a terrible distortion.
Its not so audible for regular songs but for a tone like white noise it is very easy to see and HEAR...
I disabled Auto-gain and use manual gain in the hw mixer, like the old days...
until a fix came up..
And i dont understand !
HOW virtualdj is measuring the gain ? Are these really db? This in not "Peak" level,
either "RMS" level, what is it.
I don't trust it.
Soundforge, wavelab etc, report different values for every song i tested.
Vdj has "unique" -way off-, different results...
Example: a song mastered at peak -0.5 db, and has RMS power -12db (wavelab)
Virtualdj reports gain: -3.3 db It can be right.
I hope the Team will look at this.
Sound quality is directly affected.
Posté Tue 19 Jul 05 @ 3:41 am
djbambi
That's why I run all my tracks through mp3gain first to normalize them, and I don't use VDJ's gain auto-gain feature.
The least VDJ has to do, the better the sound is: I don't use autogain, I don't use time-stretch, I don't use volumes/EQ since I have an external mixer, and I don't use effects since I have an effect section on my mixer. Result: the sound coming out of my mixer (through my XP10s) is pristine. When comparing side by side (xp10 on mixer channel 1, and turntable on mixer channel 2), I cannot tell the difference between the ripped file and the original vinyl. However, as soon as I turn on some of VDJ's sound-processing features, slight (or less slight) audible differences appear.
The least VDJ has to do, the better the sound is: I don't use autogain, I don't use time-stretch, I don't use volumes/EQ since I have an external mixer, and I don't use effects since I have an effect section on my mixer. Result: the sound coming out of my mixer (through my XP10s) is pristine. When comparing side by side (xp10 on mixer channel 1, and turntable on mixer channel 2), I cannot tell the difference between the ripped file and the original vinyl. However, as soon as I turn on some of VDJ's sound-processing features, slight (or less slight) audible differences appear.
Posté Tue 19 Jul 05 @ 9:38 am
lopez453
sometimes I encounter problems when using mp3gain. Some tracks get distorted in vdj
even when i reanalyze them or use different autogain settings.
maybe it's about 3% of songs but when I play it the sound distorted.
I use maximize at 95db. but I think even if using other values (the distortion dissapears), these few songs will play much louder than others .
even when i reanalyze them or use different autogain settings.
maybe it's about 3% of songs but when I play it the sound distorted.
I use maximize at 95db. but I think even if using other values (the distortion dissapears), these few songs will play much louder than others .
Posté Tue 19 Jul 05 @ 10:04 am
djbambi
"or use different auto-gain settings": if you mp3gain all your files, you should not use the VDJ autogain at all. The VDJ autogain is the faulty one. To check, try the same files with winamp or WMP and notice that their volume is similar.
Posté Tue 19 Jul 05 @ 12:00 pm
lopez453
so you think let vdj autogain to always 0db.
When autogain is off. It actually sets the level for me, or is it my misunderstanding and
it only "recalibrates" the 0 value to the "scanned value"
so when
- autogain is off
- a track is normalized in mp3gain
- vdj analysed this file to be a -6 db
the displayed gain is something like analysed gain+gain correction via knob.
Therefore if I see a gain of -6 it means that VDJ makes nothing to the gain?
Anyway i get distortion in vdj.
Maybe a can provide the extracts of song somewhere ...
just have to find that songs in my db.
When autogain is off. It actually sets the level for me, or is it my misunderstanding and
it only "recalibrates" the 0 value to the "scanned value"
so when
- autogain is off
- a track is normalized in mp3gain
- vdj analysed this file to be a -6 db
the displayed gain is something like analysed gain+gain correction via knob.
Therefore if I see a gain of -6 it means that VDJ makes nothing to the gain?
Anyway i get distortion in vdj.
Maybe a can provide the extracts of song somewhere ...
just have to find that songs in my db.
Posté Tue 19 Jul 05 @ 12:09 pm
Chromolli02
Must i reanalyse my mp3s after using mp3gain and what is the best db for mp3gain???????
Posté Tue 19 Jul 05 @ 1:56 pm
clax
Does mp3gain adapts the whole song's level or does it change the general level ?
In the past I have heard a lot of songs which had a general level and after a break the sounds goes down at the next beat so I'm a little affraid to use softwares like mp3gain.
In the past I have heard a lot of songs which had a general level and after a break the sounds goes down at the next beat so I'm a little affraid to use softwares like mp3gain.
Posté Tue 19 Jul 05 @ 4:51 pm
apopsisdj
Instead of talking about mp3gain, i think its better to ask for a fix :)
Auto-gain to 0db and "always match" in vdj was very usefull to me until now, and easy to use.
PS. if the team need the samples i made i can send them
Auto-gain to 0db and "always match" in vdj was very usefull to me until now, and easy to use.
PS. if the team need the samples i made i can send them
Posté Tue 19 Jul 05 @ 4:59 pm
clax
You're right; as I use VirtualDJ with low power speakers I didn't really noticed this problem but it seems to be a real problem.
Posté Tue 19 Jul 05 @ 5:01 pm
apopsisdj
Its easy to hear it even with laptop speakers with the samples i used.
Its like a "modulation" or wah effect aded mostly at the mid frequencies..
Needs some correction/filtering
Its like a "modulation" or wah effect aded mostly at the mid frequencies..
Needs some correction/filtering
Posté Tue 19 Jul 05 @ 5:07 pm
clax
Can you upload your samples with a specific name 'test' or something like that ?
Do you get the problem with all music or only with these samples ?
Do you get the problem with all music or only with these samples ?
Posté Tue 19 Jul 05 @ 6:07 pm
lopez453
@clax - it losslesly (without repacking the mp3) adjust the volume of the whole song (not for its parts) in the 1.5 db steps
it plays somehow the song for itself and decides how loud it sound to the human ear, but sometimes itsounds like i have a incompatible ears on my head :-)
it plays somehow the song for itself and decides how loud it sound to the human ear, but sometimes itsounds like i have a incompatible ears on my head :-)
Posté Tue 19 Jul 05 @ 7:19 pm
frd1963
I have experienced similar sound 'coloring' while using the 'limiter' feature of VDJ. I haven't noticed it happen since I disabled that feature.
Posté Tue 19 Jul 05 @ 7:47 pm
bagpuss
Personally I wouldn't let any "mp3gain" touch my mp3's, extract the tracks as they are from the source (e.g CD or vinyl), this is usually a good level, and it maintains the best possible quality, if you 'quiten' any files, they can become distorted once they are turned up loud, and increasing the gain is not a good idea either, obviously for the same reason...
The solution for now is set VDJ up as good as possible, it can be good, but not perfect (for all tracks), once your settings are good (i.e limiter off, auto gain on or off), then set gain at or below -3.5, this is effectively a safe level, further adjustment should be done manually at the mixer, and NOT in VDJ.., if things are done this way, then you shouldn't have any real problems..
The gain implementation in VDJ needs improvement, both the analyzer, and adjusting.., but it's not such an emergancy.., rather a quite needed improvement, especially for those that perform proffesionaly.
Bagpuss.
The solution for now is set VDJ up as good as possible, it can be good, but not perfect (for all tracks), once your settings are good (i.e limiter off, auto gain on or off), then set gain at or below -3.5, this is effectively a safe level, further adjustment should be done manually at the mixer, and NOT in VDJ.., if things are done this way, then you shouldn't have any real problems..
The gain implementation in VDJ needs improvement, both the analyzer, and adjusting.., but it's not such an emergancy.., rather a quite needed improvement, especially for those that perform proffesionaly.
Bagpuss.
Posté Tue 19 Jul 05 @ 9:44 pm
Rune (DJ-In-Norway)
Dev Team is aware of the problem, and working on it :)
Posté Wed 20 Jul 05 @ 10:06 am
apopsisdj
@Bugpuss, i totally agree
@Dj in Norway,
Thanks again for the good news.
I love it when i found this kind of bugs... and i love it more when they fixed.!
@Dj in Norway,
Thanks again for the good news.
I love it when i found this kind of bugs... and i love it more when they fixed.!
Posté Wed 20 Jul 05 @ 12:45 pm
Rune (DJ-In-Norway)
and a great thanx to you Apopsis, for doing extensive tests and giving constructive feedback, making the software progress and better :)
Posté Wed 20 Jul 05 @ 3:07 pm
Mike Marble
Last week I "cheated" while mixing, and re-did a mix I screwed up. So basically, I hit cue an "Track A", throw crossfader back to "Track A", I rewind, reset EQs, and afterwards I continued my session. Afterwards I loaded the created WAV (the complete mix) into CoolEdit, to fix that portion.
What stroke me was the fact, that the waveform of "Track A" looked completely different the second time I played it. When I screwed up, the whole waveform had spikes all over it, and the second (and I even recorded it with VDJ a third) time, the waveform looked original. By original I mean: It looks just like it looks when I directly open the mp3 in CoolEdit.
For VDJs sake: I had a really hard time, telling those two versions apart. From hearing them, I´d say they were identical, but well, the waveform says otherwise.
@Apopsis: Just a hunch, but is this what you also saw on your testfiles?
What stroke me was the fact, that the waveform of "Track A" looked completely different the second time I played it. When I screwed up, the whole waveform had spikes all over it, and the second (and I even recorded it with VDJ a third) time, the waveform looked original. By original I mean: It looks just like it looks when I directly open the mp3 in CoolEdit.
For VDJs sake: I had a really hard time, telling those two versions apart. From hearing them, I´d say they were identical, but well, the waveform says otherwise.
@Apopsis: Just a hunch, but is this what you also saw on your testfiles?
Posté Thu 21 Jul 05 @ 4:13 pm
apopsisdj
Dj in Norway, no problem. It is because i love this software..
Mike it is like the "underwater" and modulated sound you wrote in your other post.
One of the samples i created in wavelab called "Sweep 16Hz-22000Hz".wav
Imagine the fft analysis as a "smooth" single -moving- wave from lows to highs (the original)
and ...many and "dirty" waves (harmonics/modulation) the one played in vdj.
Mike it is like the "underwater" and modulated sound you wrote in your other post.
One of the samples i created in wavelab called "Sweep 16Hz-22000Hz".wav
Imagine the fft analysis as a "smooth" single -moving- wave from lows to highs (the original)
and ...many and "dirty" waves (harmonics/modulation) the one played in vdj.
Posté Fri 22 Jul 05 @ 4:25 am
Mike Marble
Thanks apopsis. Sounds like we all are on to something here....
Posté Fri 22 Jul 05 @ 12:04 pm
