|
Spectrum Analysis question
|
View this Thread in Original format
| ADSR |
I did a spectrum analysis in Soundforge of my track and a professionaly mastered track and noticed that their track, and most others, dip at about 18,500hz. Mine dips at approx 12,200hz. I'm not sure why this is, I've tried making EQ adjustments and I can't seem to get my graph looking like theirs. Any help on this subject would be greatly appreciated.
 |
|
|
| enferno |
couple things come to mind
1. what type of file compression are they using? if you are comparing your track to that of an mp3, well. aint gunna happen.
2. never try to make your track 'look' like someone elses. mix with your ears, not your eyes.
"what sounds best, is best" |
|
|
| ADSR |
| I recorded uncompressed wavs from vinyl as well as used 320kbps mp3's from anjunabeats web site for my comparison. I analyzed over a dozen tracks and found that 90 percent of them dip down between the 18,000hz and 20,000hz. I think the over all mix of my track sounds good but not great and I'm wondering if the difference in appearance of the two graphs could be one indication why. Does anyone know exactly what the graph in the spectrum analyzer represents? |
|
|
| Zombie0729 |
| not to disrespect anyone on TA but you will get far more indepth ansewrs on kvr. so my suggestion would be to post this over there. plus i am very curious on what they have to say :) |
|
|
| Nick Mimas |
| quote: | Originally posted by ADSR
Does anyone know exactly what the graph in the spectrum analyzer represents? |
Software-based audio spectrum analysers are used to study the frequency spectrum by looking at the audio output. It's basically the graphic representation of sound (I'm probably not answering your question and being Mr. Obvious) |
|
|
| Derivative |
the x axis represents the frequency. the y axis represents the level. from that you tell where the graph is at any given point the level at a particular frequency range. in most professional trance tracks the most prominant frequency is always around the kick. its nearly always bass frequency. both your track and their track have a healthy bass response. (note: just because the graph says that doesnt mean your track is as well produced as theirs. you could just as easily have poorly produced your kick and bass so that they stack up causing a peak in the level of bass frequency thus increasing the bass presence but also making it sound muddy. therefore you should not think that just because your bass and mid range spectrum analysis looks like that of a professional tune, that is necessarily on par in production terms with that of a professional tune.).
yours tails off at 12,000 hz. usually above 10,000 hz we are in hihat territory. from your graph im guessing you lack the sparkle on the hats or you simple dont have enough treble presence in your songs. it could be that your leads are upper and lower mid predominantly with no high end air. it could be that you dont have many hihats. or it could be that they arent loud enough in the mix or that there is alot of destructive phasing caused between all your treble instruments. there are so many possible answers.
note: all mp3s (even 320 kbps) will typically tail of sharply between 15 and 17,000 hz. thats just a byproduct of mp3 compression.
also note: the spectrum analysis occurs at a specific instance in your song. you should really do an analysis with the peak hold enabled. this shows you the maximum level at any given frequency over the entire track. if you do a spectrum analysis and take a snap shot right at the beginning of your tune (where for example the hihats are not in yet) then it will show a big dip after 8,000 hz. on one of my tunes it dips early up until the hihats come in then it suddenly jumps up past 10,000 hz.
also note: the spectrum analyser in sound forge is . the y axis doesnt have enough markers. its zoomed out to far and it doesnt hold peaks for more than about 3 seconds. use elemental audio systems inspector instead (its free) or if you can afford it, waves paz analyser. |
|
|
| ADSR |
| Wow, that answers alot, thanks Derivative. I will keep using the spectrum analyzer as a tool now that I know what it represents. I think I know what my problem is now. :D |
|
|
| Derivative |
just to clarify:
0 to 35 hz = sub bass territory. in reality most soundsystems even with subs roll off sharply after 20 to 30 hz with the very best ones rolling off in the 20s. you should not have much presence below 20 hz. cut it off if the graph is peaking high around this frequency range - its just eating headroom.
35 to 150 hz = bass territory. most kick drums sit between 50 and 120 hz. the kick normally has a click or a slap kind of sound right at the start caused by the rapid pitch drop. this sound resides usually in the lower and upper mids. but most of the bass part of the kick is between 50 and 120 hz. in most pro tracks, this frequency range usually peaks the highest, as near dammit to 0 dB as you can get seeing as the kick and bass is usually the driving force behind most club tunes.
150 to 3,000 hz = lower mid range territory. this is where most speech is intelligible. most vocals fit in around here as well as most of the 'telephone' quality of leads and such. in a pro tune it usually peaks in the bass frequency range and then goes flat all the way to 20,000 hz then drops off sharply.
3,000 to 8,000 hz = upper mid range territory. this is where most of the lead instruments sit as well as the definition of vocals. also, the klang part of the open hat and ride sit around here.
10,000 hz + = treble territory. this is all high frequency sound and sibilance. this is usually all hihats, the sparkle on your leads, and the air on vocals.
in a tune you basically want it loud and you want your song to have a good balance of every audible frequency. think of that spectrum analyser window as a box. the idea is to fill that box with frequencies. but you want to fill it in a way where MOST OF THE SEPARATE INSTRUMENTS IN YOUR TRACK DO *NOT* OVERLAP. note that you can fill this box easily by sloppily stacking up layer upon layer of sound.
what i will do when i run a kick drum on its own is to see what it looks like on a spectrum analyzer. a well EQed, well produced kick should be very low on the y axis on 0 to 20 hz, start to pick up sharply at 40 hz. it should peak close to 0 dB between 55 and 100 hz, it should then tail off after 200hz and maybe extend to about 2,000 to 6,000hz where the click/slap is) then drop off sharply after that. the cleanest bass drum you could have would be like an 808 kick. practically all bass and sub that tails off VERY sharply after 500 hz with only a tiny click around 600 to 2,000 hz.
similarly, with your hihats, NONE of them should be peaking anywhere near bass frequency territory or even lower mids really. they should be 0 all the way up until at least 1,500 hz then start to pick up. it should peak around 10,000 to 16,000hz and extend to 20,000 hz for some extra air. then it should drop off sharply since most audio systems arent sensitive enough to pick up sound beyond this value (additionally 16 bit 44,100 hz cds have a maximum frequency range of 22,050 hz. therefore anything really beyond that is just wasting headroom).
when most pro tunes are all in: kicks, hats, leads, pads, bass. they tend to start very low around 0 to 20 hz, pick up sharply after 35 hz, peak on around 55 to 120 hz trail off ever so slightly after that, and then run pretty much flat and fairly close to 0 dB all the way to 20,000 hz. then it should drop off sharply after that. |
|
|
| cheesy |
I find it weird that 90% of the tracks have a similar feature, it seems like each song should be fairly unique. One possibility if all those tracks were ripped from vinyl is that somewhere along the line between the vinyl and your recording software there's some hardware that causes the drop at a certan frequency (could be the cartridge, mixer, audio card, etc)
Bottom line, eferno said it best, mix with your ears, not your eyes. |
|
|
| Nick Mimas |
| quote: | Originally posted by cheesy
I find it weird that 90% of the tracks have a similar feature, it seems like each song should be fairly unique. One possibility if all those tracks were ripped from vinyl is that somewhere along the line between the vinyl and your recording software there's some hardware that causes the drop at a certan frequency (could be the cartridge, mixer, audio card, etc)
|
Yeah but after a track goes with mastering and the final mix down, each spectral analysis should be the same so each track on an album or vinyl are all the same at top quality? |
|
|
| h.vox |
| quote: | Originally posted by Derivative
(note: just because the graph says that doesnt mean your track is as well produced as theirs. you could just as easily have poorly produced your kick and bass so that they stack up causing a peak in the level of bass frequency thus increasing the bass presence but also making it sound muddy.
|
and that probably is the reason why his tracks contains more energy in bass area than in high area. however, if he does not notice that, it is most probably because that his monitor system is inadequate (just like most of the home monitor systems, including mine, are).
| quote: |
note: all mp3s (even 320 kbps) will typically tail of sharply between 15 and 17,000 hz. thats just a byproduct of mp3 compression.
|
i must correct you here - it is a byproduct of a bad encoder (xing is the first thing that comes to my mind). if you disable all filtering in lame 3.91 (and higher) the frequency response cuts off at frequencies higher than 20 k. in theory, with all filter disabled, mp3 should go as high as half of the sampling frequency (22050 Hz if you encode standard 44.1k 16bit wave). and even the encoder in cubase sx2 cuts off the frequencies above 17k.
| quote: |
also note: the spectrum analyser in sound forge is . the y axis doesnt have enough markers. its zoomed out to far and it doesnt hold peaks for more than about 3 seconds. use elemental audio systems inspector instead (its free) or if you can afford it, waves paz analyser. |
voxengo SPAN is also pretty good, and freeware.
anyway, great post. |
|
|
| h.vox |
| quote: | Originally posted by Nick Mimas
Yeah but after a track goes with mastering and the final mix down, each spectral analysis should be the same so each track on an album or vinyl are all the same at top quality? |
most definitely, no. spectrum analysis does not say anything about sound quality, it just provides you with a picture of tonal balance.
if tonal balance is different between tracks, it does not mean that tonal balance on one of the tracks is 'wrong', it just says that the atmosphere of that track is different.
the only tool used for sound quality assurance is your pair of ears. spectrum analysis is just a nifty tool which can help you find what is wrong with your mix or your master, but it should not be taken for granted. it is just like the eq - it does not matter how it looks like, because if it sounds good, it is good. |
|
|
|
|
