Recording- WAV, MP3, file compression and quality loss

JonSolo · Post by **JonSolo** » Wed Mar 26, 2008 2:22 pm

In a recent song post some thoughts on recording and loss came up. Instead of using the songwriter's thread, I thought I would post here, and let everyone add their comments.

In my experience I have found certain aspects to be pretty consistent. So here goes:

Anytime you go from digital to analog or vice versa you will get some loss of quality. This means if you go analog from the back of your keyboard (D/A) into an analog input on a mixing board (usually A/A) and out to your computer (A/A to A/D) then into WAV file (D/D) you have at least two to three times where your signal will get diminished. However if you went Digital out of your keyboard to Digital in on your computer there would be NO diminished signal.

Additionally, staying totally in the analog realm will diminish your signal. Every time you lower the volume you will diminish the signal for example.

Once you have a signal going to your computer, you generally should not get any degrading of it unless somehow it is processed differently. Going back and forth between two high rate formats (i.e. WAV and OGG or FLAC) will result in some loss. Anytime you go from ANY format to MP3 you WILL lose some quality. That is because WAV is a raw file and MP3 is a compressed file. That compression for size has to "steal" ticks from your file to compress it. The result is a loss of quality (usually in the lowest and highest registers).

I have a small recommendation to anyone who has this option.

Always record with 32 bit floating point on, and always save files as 24 bit audio wav. If you mixdown save it initially as a 24 bit file. As long as the khz are 44.1 or greater you are ok. A 24 bit file allows a LOT of room for mastering your song, especially if you only take up 90% or so headroom (volume) before the track peaks. This means DO NOT NORMALIZE your tracks prior to mastering.

When you master, always master in two directions: a 16/44.1 track (which is CD audio standard) and 24/48 or 24/192 track (which is DVD audio standard). When you compress to an MP3 file, it seems best to work from the CD Audio track and compress it no greater (technically LESSER) than a rate of 256bits (meaning 128bits is not going to cut it unless you have to upload to Soundclick who is going to mess with your file anyway).

I know this was long winded, and maybe there are some tricks or tips that some of you may have come across to help us out here. Please feel free to add or correct as necessary.

Jon

ellll · Post by **ellll** » Wed Mar 26, 2008 3:23 pm

Hey Jon,

Oddly, perhaps..I'm doing all and understood it,....except the mp3...which I took the lower rate to save the files need on the server....It would seem that is not now an issue...so my stuff is now better,...BUT:

I remind you that Sharp had to come in and remove some tens of items on the server...some very old, that had not been taken care of...SO we need to re-understand the protocol here, HOW SOON DO WE REMOVE?? ..but I intend to use the needed space for quality in any event...I now find my "stuff" is almost unchanged checking the expansions on some waveform engines, like Audacity...

Great work Jon...it is good to have this here, and maybe some others will add something a bit more useful than me

Best Regards!! John

chordial · Post by **chordial** » Thu Mar 27, 2008 1:15 am

I'll be trying 256 from now on, great advice.

I don't think we need to remove anything soon, but only long term.

DrWho · Post by **DrWho** » Tue Apr 01, 2008 3:28 am

Regarding converting to mp3 - how good a conversion works out depends in the encoding algorithm. mp3 is unbalanced in the sense that the effort to encode can be much more complicated than decoding - which is generic.

So as much as I dislike Apple's 128 kbit, I haven't actually noticed any audio deficiencies in any iTunes tunes - whereas I have noticed artifacts in my own encoding.

Also, the 44.1 khz sampling freq. is pretty much standard. Some software apps. do not like other rates such as 48khz.

And finally - I have not noticed any improvement w/32 bits. The default in my software is 16bits - which seems okay. Anyone else play with this setting? What does everyone use for encoding?

ellll · Post by **ellll** » Thu Apr 03, 2008 11:51 pm

Hi..anyone!!!,

How do you tell what ASIO is doing?...

To me it appears there is no way to actually "look" at it...so how do I know if it is working...just the sound?? For all I know it may have laid an egg that will blow my computer dec 31, 2020...

Regards, John

JonSolo · Post by **JonSolo** » Fri Apr 04, 2008 12:44 am

DrWho,

You are right in your comment on encoding. Actually LAME is the best encoder, when run in HIGH or SLOW settings. Either of those settings actually do a deeper scan in the audio file to make corrections of artifacts that get left behind on the POOR or FAST conversion settings.

As far as 32 bits...that is a setting for use in multi-track recording. While not making a huge difference in your hearing initially, there is all sorts of digital distortion that gets introduced that lowers the quality of the final product when you lower the resolution. If someone gives me a 16 bit track to master, almost 99 out of 100 times there is no headroom to play with. However, when I receive a 24 bit track I can can increase the overall volume, EQ, compression and almost NEVER introduce digital distortion to the track. The result? A very clean mastered recording...which has been dithered down to 16 bits for the CD final. If you want to compare what bitrates sound like, compare the old Ensoniq Mirage samples (11bit believe it or not) to modern samplers which are 16 or 24 bit. Or you can dumb down any one of your recordings. The difference between 8 bit and 16 bit is more than doubled in going from 16 to 32 bit. 16 is ok, but 24 or 32 is ultra clean and perfect when you have more than one track.

[one small edit- the reason you don't "hear" the difference in playback immediately is because your soundcard only puts out a 16 bit signal. The best way to tell is to do a project in 32 bit, and then do the same tracks in 16 bit. The clarity of the sounds in 32 bit will blow you away once you do the mixdown of the two and compare them.]

ellll,

ASIO is merely a driver. The purpose of it is to bypass all the Windows junk and allow your audio software program (which supports ASIO) to talk directly to your hardware soundcard. The result is super fast processing of effects and real time playing of software synths.

Jon

Post by **Sharp** » Fri Apr 04, 2008 1:48 am

Hi ellll .

I remind you that Sharp had to come in and remove some tens of items on the server...some very old, that had not been taken care of...SO we need to re-understand the protocol here, HOW SOON DO WE REMOVE??

There is no official time limit. All that is asked is that you respect the fellow members using the space and don't hog the server by keeping files on line that you feel nobody is listening to.

You guys are the community that makes this website what it is, so it's up to you guys to maintain the space that has been allocated. If you need more, or if you feel someone is taking advantage of the server, then I'm sure you know that you can PM us in confidence.

Regards.
Sharp.

ellll · Post by **ellll** » Fri Apr 04, 2008 2:06 am

Hello Sharp,

My thanks for your answer, for I did not want to be doing it wrong...We will all keep in mind what you mentioned about courtesy, I am sure. Thanks for the many kindnesses we all received from you for the last several years!!

ellll (John)