|
vinyl rip -> normalization
|
View this Thread in Original format
| brinboston |
wut's up everyone....
I've been lurking around the forum for a long time and have learned quite a bit, so thank you all for the useful info i've picked up over the years....
I'm in the process of converting my vinyl to CDs. I know you should normalize your tracks, but I am not sure which option to use in sound forge. Do you guys normalize using the 'peak @ 0db' or do u use the other options (I think it is average RMS and they have several options for you there).
Also, if when I recorded my signals are not very high.... would I first normalize at peak 0db and THEN normalize using the average RMS method?
Your input would be greatly appreciated... thanks
- brian |
|
|
| Zild |
| I try to get it as close to 0db when I record it without clipping and then from there I normalize using the music preset. |
|
|
| Derivative |
you should really never normalise.
why?
because normalising increases the noisefloor in proportion to the peak level of the signal. which is always bad.
you should increase the gain of the recording up to -0.1 dB. as close as you can get without clipping really.
you should really be using dB meters whilst mixing to keep the gain between tracks fairly constant anyway and you should know roughly the peak level of each track in your set. enough that any song doesnt suddenly come in whole decibels louder than any other.
even still, you can apply gradual gain increases using a wave editor like soundforge or by using a sequencer with automated gain control (all of them should be able to do this to be honest). increase the gain gradually on incoming tracks that are too quiet for instance. as long as the peak waveform and peak level is constant throughout and in proportion across all tracks, then you can gain up to -0.1 dB and everything is dandy.
then all you need to worry about is premastering the result and dithering it down to 16 bit PCM wav for the CD master. assuming you recorded at 32 bit float or 24 bit. that or you could pay an engineer to do it for you. but its charged per song so i dont know how much a set would cost. metropolis charge £75 per track. |
|
|
| brinboston |
thanks for the replies...
in response to your post, derivative.... i understand using the dB meters to mix to keep levels constant.... however, at the moment i'm ripping one song at a time. the problem with this is, even when i level a song as close to 0dB as possible at the loundest point in the song as can be seen from the meters on my mixer, it does not reach 0dB in the levels on my recording software (audacity), so i have adjust gains even higher (because digital outs are independent from master volume). I am using a 24-bit 96khz digital connection through the digital out on my mixer into my m-audio audiophile interface. shouldn't my recording be still extremely high quality even if i normalize afterwards? and the damn meters on top of audacity are too small so i can't really see how close to 0dB i am getting.
on a side note, maybe u guys can help me out with this. I use audacity to record but sound forge to make any edits or processes. sound forge seems to eat up a lot of CPU and for some reason, when I record in sound forge through SPDIF, the time is not moving in real time. by this, i mean when 1 actual minute has passes on my vinyl, on my computer screen only 40 seconds have passed. the result is that when its rendered it is both sped and pitched up. any idea why this is happening? i am running my audiophile through ASIO drivers and my laptop is a centrino 1.6... which is plenty fast. No problems when recording at 16-bit analog 48khz.
thanks in advance for all ur help,
- brian |
|
|
| Derivative |
your dB meters are inaccurate. i think voxengo span has a set of meters. failing that, waves PAZ analyer has a series of meters that can be used and they are pretty accurate. if you use izotope ozone, those meters are also spot on. some of the meters in fl studio for instance are waaaay off. im not sure about audacity though. soundforge is usually pretty spot on. it always plays back several dB hotter than fl studio...
although to be honest, if you are normalising and you dont have much headroom anyway then you probably wont notice the difference. but thats not as important as the procedure, which should be squeaky clean at all stages of A/D D/A conversion.
with regards to speeding up. check that soundforge is set to record from a source using the correct sample rate. if you import a 44.1 khz render into a project that is designated at 96 khz then soundforge will speed it up to allow it to be played at that sample rate. |
|
|
| brinboston |
will try that out now.... thanks for the quick reply.
regards,
- brian |
|
|
| brinboston |
i somewhat fixed the problem in sound forge and yes, its meters are much more accurate than audacity. however....
i played around with the settings on my audiophile as well as sound forge... when my audiophile is set to record at a depth of 24-bits, it can record at a max of 96khz. when i alter the settings in sound forge, the depth setting is also at 24-bits.... however, when i start recording at 96khz the running time on sound forge is still noticeably slower than real time. this problem goes away once i drop the sampling rate to 48khz (still at 24-bits).... any thoughts?
am i correct in thinking that the difference in sound quality is not that noticeable between 48khz and 96khz at 24-bit recording? should i just continue ripping my vinyl at the 48khz setting?
thanks again. help is greatly appreciated...
- brian |
|
|
| Derivative |
| quote: | | am i correct in thinking that the difference in sound quality is not that noticeable between 48khz and 96khz at 24-bit recording? should i just continue ripping my vinyl at the 48khz setting? |
its sort of hard to explain since i need to explain the difference between digital copies of analogue sound sources.
if you look at a sine wave that is generated from analogue (i.e. from an analogue monosynth) it is a continous wave.
a digial copy of that would look for the vertical height of the sine wave and take loads and loads of snapshots of it over time to build up that sine wave in discrete intervals.
the horizontal distance between each dot is determined by the sample rate. obviously, the higher the sample rate, the more snapshots are taken in the same time, and the more continous it is horizontally and the more accurate it is to the original analogue source.
the vertical distance between each discrete sample is the bit depth. the manner in which an analogue sound source is 'digitised' is called quantisation.
to cut a potentially very long story short, there is a noticeable difference between a 96,000hz recording and a 48,000hz recording. at least, its audible in my productions. this is all digital at 24 bit/96 khz - nothing i use is analogue. more than likely, all your vinyl is actually a digital recording put back onto an analogue playback medium. there cant be a record released in the past 5 to 10 years that doesnt have a single analogue to digital conversion. most productions are in 24/96 at every stage before being dithered down to 16/44.1 for cd, although some arent. to keep things simple, its best to keep everything in 24/96 throughout to cover for any eventuality. unless you need to put your mixes onto CD in which case you still need to keep everything at 24/96 up until the very last stage where you dither it down to PCM wave. but that would take another thread to explain how to do that.
a 48,000hz recording has a theoretical frequency range of 0 to 24,000hz. therefore, why would you need to record in 24/96 you say? the limit of human hearing is only about 20 to 20,000hz to 22,000hz! this is adequately covered!
you are of course correct. however, downsampling a 24 bit, 96,000hz sound source to 24 bit, 44,100hz will introduce quantisation error.
because, you now have half the horizontal discrete samples to express the same sound signal. this results in very very low level cyclical noise in your recording. most people wont have a clue what to listen for, and it is difficult at normal listening levels but once you notice it, it can be quite evident and it can be annoying. the more samples and bits you throw away, the more evident the noise is although it is very small at this bitrate/sample rate.
interesting fact: bitcrushing effects are basically distortion effects that achieve their result by truncating bitrates to the point where the audible signal just becomes a mush of quantisation noise. its fairly unpleasant sounding.
but back on topic. the obvious solution then, is to keep everything in 24/96. ideally the resulting mix would be burned to dvd-audio since it can playback at 24/96. if not, then it needs to be dithered down to 16 bit/44.1khz PCM wave in order to be burned to CD. you can simply render the result to PCM wave but of course, you would simply be truncating the bitrate as well as the samplerate, introducing more quantisation error.
to hide this quantisation error, you need to use a filter process called dithering before you truncate a 24/96 recording to 16/44.1
there is a thread in the production forum on doing this - ill try to find it and link to it here. dithering is something you attempt at your own risk since if done wrong, can easily make your recording sound worse than if it were not dithered at all. it is essentially a process in which you add very low level random noise to your recording to cover up the very low level cyclical noise caused by quantisation error.
------------------------------
bear in mind that this is absolute audiophile stuff. my sister doesnt know what quantisation noise is, what it sounds like at high amplitude and can find no perceivable difference between a 192kbps mp3 and a 24/96 wave recording. for 99% of the human race (the remaining 1% being audiophiles, engineers, producers and so forth) alot of this is practically a non issue.
but if you are going to make it your demo cd in an official capacity...thus representative of the work you do - i am a firm believer that whatever you are doing, you should always endevour to make it absolutely the best it can possibly be.
dithering is something that i would leave to a qualified engineer to do though unless you know what you are doing and want to give it a go - an engineer however will more likely be in possession of the right tools whereas you probably are not - they will have a very good set of ears, a very very good set of monitors that very few mortals can afford, and a range of dithering alogrythms which they know how to use.
if your demo release is pretty serious then it may even be worthwhile to get a professional engineer to do this for you although i expect to do an entire cd's worth of material, the cost would be high. hope this helped in some way. or at the very least, proved interesting. |
|
|
| PersianMafia |
| awsome buddy! that was a great read! btw, what books/sites/universtiy majors :p can you recommend about this stuff. it seems pretty interesting. |
|
|
| brinboston |
wow, talk about an informative post, derivative....
I actually know which thread you are talking about that was (i think) posted in the production forum regarding dithering. had to read it several times and still do not understand it completely...
I can understand the concept of wanting to keep everything at 24/96 for as long as possible before dithering down to CD quality.... however, I am quite an amateur producer at the moment and the recordings will simply be burnt on CD to be spun. I still can't figure out why my sound forge isn't running in real time at 24/96...
so, from what i could extract from your post, would you simply suggest recording at 16/44.1, so that i don't have to worry about quantization errors from dithering down to 16 when i burn the files as .wavs on a cd?
thank you for all your help...
-brian |
|
|
| Zild |
| quote: | Originally posted by PersianMafia
awsome buddy! that was a great read! btw, what books/sites/universtiy majors :p can you recommend about this stuff. it seems pretty interesting. |
If you want to get a degree get something that is in demand. Then you can pay for a studio and the leisure time to learn how to operate it. |
|
|
| Derivative |
| quote: | I can understand the concept of wanting to keep everything at 24/96 for as long as possible before dithering down to CD quality.... however, I am quite an amateur producer at the moment and the recordings will simply be burnt on CD to be spun. I still can't figure out why my sound forge isn't running in real time at 24/96...
so, from what i could extract from your post, would you simply suggest recording at 16/44.1, so that i don't have to worry about quantization errors from dithering down to 16 when i burn the files as .wavs on a cd? |
actually i just realised something. ive had my producer's hat on this whole thread.
its probably NOT a good idea to dither AT ALL. because theres a very good chance that the songs you are using (especially if they from a CD source,) have already been dithered.
dithering twice is a really big no no. i guess the safest bet then is to render straight to 16 bit/44.1 khz. if you are rendering from vinyl then theres a good probability that the recording is 24/96 and you'll just have to live with the quantisation error. im not sure of the processes used in vinyl production. on CD it will almost certainly be dithered 16/44.1 so its fine to render it at this bit depth and sample rate. problems only arise when you start downsampling and truncating bitrate.
that or start hyping up the benefits of dvd-audio.
| quote: | | awsome buddy! that was a great read! btw, what books/sites/universtiy majors can you recommend about this stuff. it seems pretty interesting. |
well i learnt it mostly on my own initiative - i stopped maths and physics at GCSE and have no formal engineering qualifications - alot of this is just peripheral to music production - which i do alot of in my spare time.
course, it helped an awful lot that the izotope guys wrote up some really informative and easy to read articles on mastering and dithering which you can get on the izotope website.
alot of this is actually pretty simple in principle and it makes complete sense if you break it down and get over all the jargon. it takes a while to sink in but any questions about dither should probably go into the dithering thread in the production forum. its a very simple process in reality although it makes sense to think about dither filters in terms of images because you can see what it is doing. i direct you to the dithering guide from the izotope guys since they explain it very well. |
|
|
|
|
