Topic: Handling 48kHz

[morningdew]

Just in case there are any other brain dead morons like myself who also tape, I'll pass along this newb lesson and also ask a few questions.

I recorded my first show in 48kHz.  I normalized it, cut it up and exported to seperate wavs using Sound Forge 8.0.  I then burned it to a CD.  I have a 30 minute ride to work and I'm listening to what I recorded and going insane because there is some noise in my recording.  I could only pick it up on the highs, not the lows, but it's bad.  At times it sounds just like when you have your volume way up on a speaker that can't handle it and you get that vibration or distorion.  At other times it sounds like this hiss.  Almost like the drummer is hitting a cymbal throughout the entire recording.

I'm having a major panic attack because this is also the first show I've used my hyper caps on and I'm worried I bought a bum pair.  Finally, it occurs to me that I never noticed how Sound Forge handled the file being that it was recorded at 48kHz.  I get home and check out the properties of the wavs it created and sure enough they are 48kHz.  I didn't realize that CDs could be burned with wavs recorded in 48kHz, but they can and my software did it no problem.  Not only did it burn but the discs (or coasters now) are playable and even listenable but has something that is obviously off to your ear.

I re-did the show in 44.1kHz and all is good now.  The mics are just fine...I'm the one who is stupid.

My questions:

When dealing with 48kHz.
1.  Should I normalize first and then resample to 44.1kHz or should I resample to 44.1kHz first and then normalize?
2. I have the choice to apply anti-alias filtering when resampling.  Do I apply it or not?
3. Is there any other common editing technique that is important to do before or after resampling?

Thanks.

[TNJazz]

Actually what your burning program did was resample the file to 44.1kHz on the fly.  This is why it sounds off to your ear.  CD burning programs are not known for high quality sample rate conversion.

[eric.B]

I have allways heard to do ALL post processing of a file in its original (highest) bit/sample rate, then lastly perform resampling, then dither(change in bit depth)..

I would imagine any normalization, eq etc would be more accurate(less artifacts) when performed *before* any processing to produce file to be burned in the redbook standard.

[Roving Sign]

I usually save resampling for the final step - do all the editing in the file's original format...

[Steve J]

My questions:

When dealing with 48kHz.
1.  Should I normalize first and then resample to 44.1kHz or should I resample to 44.1kHz first and then normalize?
2. I have the choice to apply anti-alias filtering when resampling.  Do I apply it or not?
3. Is there any other common editing technique that is important to do before or after resampling?

Thanks.

1.  All editing first (fades, normalization, etc)
2.  Resample (use anti-alias'ing in SF, highest level)
3.  Bit Depth Change (ONLY IF you're going 24>16 bit; otherwise, not necessary)

[admkrk]

make sure you resample before splitting tracks!  everything else first. you'll end up w/ all kinds of sbes if you don't. trust me, i'm on my 3rd in a row i forgot/thought i did. too much like work this way.

[live2496]

I don't think this question got answered.

2. I have the choice to apply anti-alias filtering when resampling.  Do I apply it or not?

When going from a higher sampling rate to a lower, an anti-alias filter should be used.

The reason for this is. At 48kHz, you can have a frequency present up to 24kHz. When resampling to 44.1kHz, the maximum frequency you can reproduce is close to 22.05 kHz.  Any frequencies present in the original recording from 22kHz to 24kHz can come back into the resampled file as an alias frequency in the range of 1 hz to 2kHz. The filter removes this prior to resampling.

Gordon

[SparkE!]

I don't think this question got answered.

2. I have the choice to apply anti-alias filtering when resampling.  Do I apply it or not?

When going from a higher sampling rate to a lower, an anti-alias filter should be used.

The reason for this is. At 48kHz, you can have a frequency present up to 24kHz. When resampling to 44.1kHz, the maximum frequency you can reproduce is close to 22.05 kHz.  Any frequencies present in the original recording from 22kHz to 24kHz can come back into the resampled file as an alias frequency in the range of 1 hz to 2kHz. The filter removes this prior to resampling.

Gordon

Gordon, I think that the spectrum folds back on itself when you resample, so that frequencies that are originally between 22.05 and 24 kHz end up in the range between 22.05 and 18.1 kHz.  The anti-aliasing filter will be used to pre-attenuate the products in the 22.05 kHz to 24 kHz band prior to doing the sample rate conversioin.

[live2496]

Gordon, I think that the spectrum folds back on itself when you resample, so that frequencies that are originally between 22.05 and 24 kHz end up in the range between 22.05 and 18.1 kHz.  The anti-aliasing filter will be used to pre-attenuate the products in the 22.05 kHz to 24 kHz band prior to doing the sample rate conversioin.

Thanks for correcting that.

[SparkE!]

Gordon, I think that the spectrum folds back on itself when you resample, so that frequencies that are originally between 22.05 and 24 kHz end up in the range between 22.05 and 18.1 kHz.  The anti-aliasing filter will be used to pre-attenuate the products in the 22.05 kHz to 24 kHz band prior to doing the sample rate conversioin.

Thanks for correcting that.

No problem.  It was a picky point anyway.  You answered the important part and that is that you should use the anti-aliasing filter unless you already bandlimited the signal with an analog anti-aliasing filter when you first recorded it at the higher sampling rate.  In that case the digital anti-aliasing filter only adds round-off error to the samples and you get no S/N benefit from running it.  In fact if there is nothing to filter in the 20.05 to 24 kHz band, you actually hurt your S/N by running it.  Most people do not run an antialiasing filter, except the fixed width filter that goes righ in front of the sample and hold on the input to the A/D converter.  Most of those filters will not provide much attenuation between 20.05 and 24 kHz.  They usually kick in above 24 kHz, so you will need to run the digital anti-aliasing filter.  It's pretty rare that someone would eq the signal ahead of the A/D to remove any signal below 24 kHz, so it's pretty rare for anyone not to get some benefit from running the digital anti-aliasing filter prior to doing the sample rate conversion.

[Roving Sign]

 

Always appreciate your posts...I feel like Im listening to someone speak spanish...I can understand most of it...but dont ask me to speak it!!!

[SparkE!]

Well it's nice to be appreciated.  Whenever I start geek talking around my wife, she just tunes me out until she can change to subject to something important, like the cute thing that the dog did today.  (She never ignores the dogs... Should I be worried?)

[pfife]

make sure you resample before splitting tracks!  everything else first. you'll end up w/ all kinds of sbes if you don't. trust me, i'm on my 3rd in a row i forgot/thought i did. too much like work this way.

I don't know about Soundforge 8, but in version 6 it did not cut on the sector boundaries.   I'd suggest using cdwav.... Its really easy to use, and cuts on the sector bounds.

A final thing - save your regionlists/cue sheets with your archived files!  You'll be glad you did.

[Roving Sign]

make sure you resample before splitting tracks!  everything else first. you'll end up w/ all kinds of sbes if you don't. trust me, i'm on my 3rd in a row i forgot/thought i did. too much like work this way.

I don't know about Soundforge 8, but in version 6 it did not cut on the sector boundaries.   I'd suggest using cdwav.... Its really easy to use, and cuts on the sector bounds.

A final thing - save your regionlists/cue sheets with your archived files!  You'll be glad you did.

I think he's right though - even if you use CDwave...if you reopen the cut files - and then re-sample them - you can get SBEs...

[mhibbs]

make sure you resample before splitting tracks!  everything else first. you'll end up w/ all kinds of sbes if you don't. trust me, i'm on my 3rd in a row i forgot/thought i did. too much like work this way.

I don't know about Soundforge 8, but in version 6 it did not cut on the sector boundaries.   I'd suggest using cdwav.... Its really easy to use, and cuts on the sector bounds.

A final thing - save your regionlists/cue sheets with your archived files!  You'll be glad you did.

Yeah, but even if you cut the 48khz file in CDWav THEN resample the individual files you'll end up w/ SBEs.  The cuesheet route is definitely the way to go....resample the entire file, then use the cue sheet to cut it up via CDWav and it should be fine.

mitch

[admkrk]

i don't understand about the q sheet. is that a record of were you want splits to be and then select the times in cd wave? that would save a lot of time if i could figure it out.

sometimes i'll cut into discs in wavelab and then just put a split at the first and last point i can in cd wave and delete them. all the indevidual tracks i do in cd wave.

[mhibbs]

Once you set all of your tracks in CDWav, do File>Save Cuesheet.  The result is a cuesheet that tells CDWav where to cut the tracks in the large source file.  If you want to cut the file into tracks of 48khz files, then you can do so by File>Save like you've done in the past (keeping the big source file intact).  Now you can goto Wavelab, etc and resample the original large source file down to 44.1.  Once that is done, open the large 44.1 file in CDWav, do File>Load cue sheet and load the cuesheet you saved when you set the tracks in the 48khz file.  You can now do File>Save again and cut the 44.1 file into tracks of 44.1 files.  Make sense?

[willndmb]

1.  All editing first (fades, normalization, etc)
2.  Resample (use anti-alias'ing in SF, highest level)
3.  Bit Depth Change (ONLY IF you're going 24>16 bit; otherwise, not necessary)

always forget this myself
resample then dither

[admkrk]

Once you set all of your tracks in CDWav, do File>Save Cuesheet.  The result is a cuesheet that tells CDWav where to cut the tracks in the large source file.  If you want to cut the file into tracks of 48khz files, then you can do so by File>Save like you've done in the past (keeping the big source file intact).  Now you can goto Wavelab, etc and resample the original large source file down to 44.1.  Once that is done, open the large 44.1 file in CDWav, do File>Load cue sheet and load the cuesheet you saved when you set the tracks in the 48khz file.  You can now do File>Save again and cut the 44.1 file into tracks of 44.1 files.  Make sense?

so like when i forget to resample first, all i would have needed to do was go back and resample then load the q sheet and not have to go through retracking it all over again?  i've got to try this!  +T to ya bud  sounds like a good habit for absent mind(ed) tapers like me.

[pfife]

make sure you resample before splitting tracks!  everything else first. you'll end up w/ all kinds of sbes if you don't. trust me, i'm on my 3rd in a row i forgot/thought i did. too much like work this way.

I don't know about Soundforge 8, but in version 6 it did not cut on the sector boundaries.   I'd suggest using cdwav.... Its really easy to use, and cuts on the sector bounds.

A final thing - save your regionlists/cue sheets with your archived files!  You'll be glad you did.

Yeah, but even if you cut the 48khz file in CDWav THEN resample the individual files you'll end up w/ SBEs.  The cuesheet route is definitely the way to go....resample the entire file, then use the cue sheet to cut it up via CDWav and it should be fine.

mitch

This is exactly what I do too, both when I used to resample and now that I dither.

[pfife]

Once you set all of your tracks in CDWav, do File>Save Cuesheet.  The result is a cuesheet that tells CDWav where to cut the tracks in the large source file.  If you want to cut the file into tracks of 48khz files, then you can do so by File>Save like you've done in the past (keeping the big source file intact).  Now you can goto Wavelab, etc and resample the original large source file down to 44.1.  Once that is done, open the large 44.1 file in CDWav, do File>Load cue sheet and load the cuesheet you saved when you set the tracks in the 48khz file.  You can now do File>Save again and cut the 44.1 file into tracks of 44.1 files.  Make sense?

so like when i forget to resample first, all i would have needed to do was go back and resample then load the q sheet and not have to go through retracking it all over again?  i've got to try this!  +T to ya bud  sounds like a good habit for absent mind(ed) tapers like me.

exactly correct.  It works very nicely.   What I do is store a cue sheet on my archived dvds w/ my original 24/44.1 file.   if I need a 16bit version, open the one on the DVD in wavelab, dither, and save onto the hard-drive.  Open the cue file in a text editor, and change the location of the input file.... then load it... unselect the first track and last track for exporting (those are the small tracks to keep out SBE's), and then save (ie export individual tracks).

I was very hesitant to change to CDWav for tracking, but I think it rules. 

[shruggy1987]

some more advice on processing:
before you do any post production (normalization, adding EQ, compression, etc.) it is wise to increase bit-depth to max (in Sf that is 64) and upsample to 192kHz.  then apply the post production and go down to 16/44.1, or whatever your desired final format is.  it takes a long time (depends on your system), but it is time well spent. 

[morningdew]

What does increasing the bit depth and up sampling before doing any post production work do for you?

If you recorded at 16 bit / 48 kHz you can't really make it any better can you?

Thanks.

[shruggy1987]

What does increasing the bit depth and up sampling before doing any post production work do for you?

If you recorded at 16 bit / 48 kHz you can't really make it any better can you?

Thanks.

the sound quality itself cannot improve over what you have originally, but the higher sample rate/bit-depth allows the processing (equalization/compression) to be as accurate and detailed as possible.

[SparkE!]

Yup, the higher bit depth and sampling rate for processing doesn't make things better than their original accuracy at 16 bits and 48 kHz, but it does help you to keep from making things any worse.