IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
Am I right that 32 kHz would suffice for FM radio w/ roll-off ~15 kHz?, [was "Tell me 32kHz sample rate is OK"]
leethedj
post Jul 13 2012, 07:27
Post #1





Group: Members
Posts: 17
Joined: 3-July 02
Member No.: 2457



I used to do a weekly radio show, and I'm in the middle of digitizing hundreds of shows recorded on VHS hifi straight from an FM tuner (from 1995 to about 2000, when we started using a minidisc recorder). I'm cutting each three hour show into three 1 hour parts, and encoding to mp3 so they can be eventually shared as a podcast. I think I would simply like someone to tell me my ears aren't crazy, and that the settings I've decided sound good enough are, in fact, reasonable.

FM radio is supposed to roll off above 15kHz, which I can see in a spectrum analysis pretty clearly. I can also see the 19kHz carrier tone for stereo. FM has limited dynamic range and limited stereo separation. Taking all that into account, I figured using LAME at V5 or V6 would be just peachy keen, and keep the bitrate reasonable since I'll be uploading and sharing a zillion hours of audio. I also included a lowpass filter at 15kHz. I settled on [-V 6 --lowpass 15] as being the smallest file size where it still sounded pretty damn good.

I was surprised to see that LAME decided a 32kHz sampling rate was all I needed. But with a max frequency of 15kHz, that's hard to argue with. It just FEELS weird. So I think I want HA moral support in my feeling that this sounds just fine, and that upping the lowpass just to get 44.1kHz sampling is useless (it kicks in around "--lowpass 15.4").

Opinions? Comments?
Go to the top of the page
+Quote Post
dhromed
post Jul 13 2012, 09:21
Post #2





Group: Members
Posts: 1286
Joined: 16-February 08
From: NL
Member No.: 51347



Why does it feel weird? Is there real information in your recordings above 16KHz?

15 * 2 = 30. So 32 gives some headroom.
Go to the top of the page
+Quote Post
greynol
post Jul 13 2012, 09:24
Post #3





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



You mean to say 32/2 = 16, so there's a little bit of headroom.


--------------------
Placebophiles: put up or shut up!
Go to the top of the page
+Quote Post
db1989
post Jul 13 2012, 09:55
Post #4





Group: Super Moderator
Posts: 5275
Joined: 23-June 06
Member No.: 32180



What exactly would be the point of forcing a sampling frequency above 32 kHz but then lowpassing below 16 kHz? IOW, why use a higher sampling rate when you need none of the higher frequencies that it would enable?

I believe -V6 defaults to 32 kHz in any case unless forced higher, e.g. by an altered lowpass, so your observation is not cause for surprise.
Go to the top of the page
+Quote Post
2Bdecided
post Jul 13 2012, 10:20
Post #5


ReplayGain developer


Group: Developer
Posts: 5059
Joined: 5-November 01
From: Yorkshire, UK
Member No.: 409



QUOTE (leethedj @ Jul 13 2012, 07:27) *
I was surprised to see that LAME decided a 32kHz sampling rate was all I needed. But with a max frequency of 15kHz, that's hard to argue with.
As far as Lame is concerned, the spectral content of the original files is irrelevant. Even if you had a pristine CD quality source, if you ask Lame to squash it into V6, the best quality is achieved by downsampling to 32kHz first.

Given that there's no useful content above 15kHz (and you're right to low pass filter it - otherwise, at some high quality settings, Lame might waste bits trying to encode the 19kHz pilot tone, depending on how loud it is in your recordings), the only downside to downsampling to 32kHz is that the temporal resolution of the mp3 encoding could be slightly poorer (the temporal smearing due to the MDCT will be longer). However, at V6 this is outweighed by the advantages of downsampling - which is why lame does what it does by default, and you shouldn't fiddle with it.

The problems you may face are more fundamental: someone somewhere may try to play it on hardware that can't cope properly with VBR, and/or can't cope properly with 32kHz sampled audio. Such things exist, but given the amount of audio, I'd still stick with VBR for the size advantage (at a given quality level).


FWIW when doing a similar thing, I settled on V5. That was a few years back, but with a mono source - and Lame may have improved since then.

Keep your lossless originals.

Cheers,
David.

P.S. how well had the VHS-HiFi tapes survived? I have a pile that I haven't dared tackle yet. I just hope they still play. They sounded great when first recorded.
Go to the top of the page
+Quote Post
leethedj
post Jul 14 2012, 01:05
Post #6





Group: Members
Posts: 17
Joined: 3-July 02
Member No.: 2457



Thanks for the reply, David! It seems to be only LAME V7 that automatically drops to 32kHz sampling. V6 still gives it a go on frequencies just under 16kHz. Lowpassing it at 15 forced the sampling drop. I'm OK with 32kHz, I just wanted to make sure I wasn't experiencing wishful hearing when I think it sounds just fine. :-) Using the lowpass at 15kHz, and therefore the 32kHz sampling rate, saves about 6 MB on a 55 minute file. Since I will, over time, be uploading and sharing HUNDREDS of hours of audio, those saved megabytes are worth it. Also, I tried V5, and the difference vs. V6 can only be described as evanescent.

On the topic of my VHS tapes, they have held up wonderfully. The only problem I'm running into is that the old RCA hifi machine that recorded them appears to have had an audio head ever so slightly out of line. That RCA died years ago, and now, even on two very good JVC machines, you still get a very low level buzz from the hifi switching noise. You can only hear it if you crank it during quiet passages, or during the news updates which had no talk bed. Otherwise, I'm surprised at how well the VHS tapes have held up, and I'm very glad I was obsessive about recording everything, and taking care of the tapes. I'm up to tape #101, with about 100 more to go. And I am saving all the uncompressed audio. Good thing hard drives are huge and blank DVDs are cheap nowadays!

Then it's on to video! *ugh*

This post has been edited by db1989: Jul 14 2012, 21:35
Reason for edit: deleting pointless full quote of above post
Go to the top of the page
+Quote Post
DigitalMan
post Jul 14 2012, 03:29
Post #7





Group: Members
Posts: 486
Joined: 27-March 02
From: California, USA
Member No.: 1631



Looks like you're pretty far into the project, otherwise I might have suggested you try V5, V6 and V7 and ABX them to see if you can hear a meaningful difference.


--------------------
Was that a 1 or a 0?
Go to the top of the page
+Quote Post
leethedj
post Jul 14 2012, 07:55
Post #8





Group: Members
Posts: 17
Joined: 3-July 02
Member No.: 2457



QUOTE (DigitalMan @ Jul 13 2012, 21:29) *
Looks like you're pretty far into the project, otherwise I might have suggested you try V5, V6 and V7 and ABX them to see if you can hear a meaningful difference.


I'm far into digitizing them to .wav files, but I'm only just now deciding what mp3 compression to use for distribution. I'm keeping the .wav's forever. :-) I did the best I could in creating a semi-direct comparison between V5 & V6, and like I said, the difference between them was barely discernible. V7 was where I could start to notice it (using some good Sony MDR-V900 headphones, hooked to a Panasonic receiver in "direct" mode, fed via digital output from the PC).

I think I have it dialed in pretty well, but I was just wondering if my hearing was failing me since I thought 32kHz sampling sounded fine. You wonder these things as you age. But I'm feeling better about it now...

Maybe I should upload a clip or two and let others see what they think.
Go to the top of the page
+Quote Post
pdq
post Jul 14 2012, 15:10
Post #9





Group: Members
Posts: 3372
Joined: 1-September 05
From: SE Pennsylvania
Member No.: 24233



You should archive the originals in one of the many lossless codecs. This way you can set up the tags the way you like, and those tags will carry over to the lossy files when you reencode.
Go to the top of the page
+Quote Post
leethedj
post Jul 14 2012, 21:24
Post #10





Group: Members
Posts: 17
Joined: 3-July 02
Member No.: 2457



QUOTE (pdq @ Jul 14 2012, 09:10) *
You should archive the originals in one of the many lossless codecs. This way you can set up the tags the way you like, and those tags will carry over to the lossy files when you reencode.



Hmmm.... that opens up another can of worms and another topic, doesn't it?

Thoughts include:
1) FLAC is probably most widespread and most future proof?
2) Metadata / tags would be nice, but I'm cutting into pieces and editing a touch before distributing, so not sure how helpful that'd be for that part.
3) How helpful would metadata be to me with the unedited lossless files?
4) WAV is just so easy, compatible, and fast to process
5) As time rolls on and hard drives get ever bigger, will I care about the saved space?

I should start a new thread, maybe in the FLAC area. ;-)
Go to the top of the page
+Quote Post
db1989
post Jul 14 2012, 21:49
Post #11





Group: Super Moderator
Posts: 5275
Joined: 23-June 06
Member No.: 32180



1) Yes
2) and 3) Well, you seem already to be quite sure that metadata would not be of much use, so perhaps we can discount that from the discussion of whether lossless compression overall is worthy of consideration.
5) Lossless compression may be especially effective at compressing audio that has relatively little high-frequency content, relatively high dynamic compression (a common concomitant of FM-based broadcasting, if I recall correctly), and possibly other traits that your sources might possess. Have you tried compressing any of your files to see? I could counter your citation of ever-expanding capacities with the equally real trend of ever-increasing processing speeds and ever-faster codecs. wink.gif So, compression may still offer something to you.
Go to the top of the page
+Quote Post
leethedj
post Jul 15 2012, 07:07
Post #12





Group: Members
Posts: 17
Joined: 3-July 02
Member No.: 2457



QUOTE (db1989 @ Jul 14 2012, 15:49) *
1) Yes
2) and 3) Well, you seem already to be quite sure that metadata would not be of much use, so perhaps we can discount that from the discussion of whether lossless compression overall is worthy of consideration.
5) Lossless compression may be especially effective at compressing audio that has relatively little high-frequency content, relatively high dynamic compression (a common concomitant of FM-based broadcasting, if I recall correctly), and possibly other traits that your sources might possess. Have you tried compressing any of your files to see? I could counter your citation of ever-expanding capacities with the equally real trend of ever-increasing processing speeds and ever-faster codecs. wink.gif So, compression may still offer something to you.


db1989: All very good points. I guess I'm hesitating because the FLAC conversion would add one more step to the workflow, and it's already rather amazing I've gotten as much done as I have! I'm still open to it.
Go to the top of the page
+Quote Post
Porcus
post Jul 15 2012, 11:52
Post #13





Group: Members
Posts: 1842
Joined: 30-November 06
Member No.: 38207



And 4):

FLAC decodes very fast (much faster than my mp3 files). It might even in some cases be faster than decoding WAV (if file reading speed caps it) -- that of course depends very much upon hardware / drives / interface / compression ratio.

FLAC encodes quite fast too, at least if you stick to the default -5 option. On the comparison at http://synthetic-soul.co.uk/comparison/los...Time&Desc=0 , then an Athlon XP 2400+ encoded FLAC -5 at 53x realtime. That could encode 350 of your 3-hour shows in 24 hours. FLAC -0 do it in ten at the cost of larger files, but still massively improve over WAV.


QUOTE
I guess I'm hesitating because the FLAC conversion would add one more step to the workflow


Could be done en bloc (then you need to have spare drive space for both your WAVs and FLACs, but hey, you wanted to have space for your WAVs and you do anyway need a backup drive). Using e.g. foobar2000, you dump them all to a playlist, right-click --> convert, pick FLAC, set the output folder, set the output filename to %filename% and leave it running.

What you SHOULD do, is tag them. Showdate, content, comment. Just like you would do about an mp3, I guess? (Tags are transferred when you convert FLACs to mp3s -- but maybe not from WAV, as the tagging schemes for that format are not widely supported.)
If you can fit all metadata into file name with a separator sign (use a character that never occurs in the text -- I suggest ¨), then you can tag-from-filename. Otherwise, there are ways to use a text file and import. (I would probably rather work with a spreadsheet first.)

This post has been edited by Porcus: Jul 15 2012, 12:02


--------------------
One day in the Year of the Fox came a time remembered well
Go to the top of the page
+Quote Post
leethedj
post Jul 16 2012, 06:52
Post #14





Group: Members
Posts: 17
Joined: 3-July 02
Member No.: 2457



I feel like I should take all the FLAC discussion to the FLAC forum.

Maybe I should upload some mp3s of said radio show with the LAME settings I've chosen and see if others think the 32kHz sample rate at V6 sounds good.
Go to the top of the page
+Quote Post
2Bdecided
post Jul 16 2012, 11:34
Post #15


ReplayGain developer


Group: Developer
Posts: 5059
Joined: 5-November 01
From: Yorkshire, UK
Member No.: 409



QUOTE (leethedj @ Jul 16 2012, 06:52) *
I feel like I should take all the FLAC discussion to the FLAC forum.

Maybe I should upload some mp3s of said radio show with the LAME settings I've chosen and see if others think the 32kHz sample rate at V6 sounds good.
It would be useful if you uploaded the lossless original too.

There's nothing wrong with keeping wave files. Especially if you're still working on them.

It's a shame fb2k didn't adopt any of the wav metadata schemes that are out there e.g. http://en.wikipedia.org/wiki/Broadcast_Wave_Format http://www.sno.phy.queensu.ca/~phil/exifto.../RIFF.html#Info There is some support out there: http://wiki.audacityteam.org/wiki/WAV#Metadata but some other software might interpret the extra chunks as audio data or strip them.

I have all my CDs as flacs, but most things I've digitised myself stay as waves because the next time I go back to them will be to edit them or process them. Finished things get converted to flac, but in progress things, and multitracks, never do. The folder and filename is all the metadata they'll ever have.

It sounds like conversion to FLAC when you've finished working on the files, as a precursor to mp3 encoding, might make sense for you. It'll save space, and it'll let you put the metadata in your lossless archive for automatic transfer to mp3 (or any other lossy format you might use in the future).

Cheers,
David.
Go to the top of the page
+Quote Post
db1989
post Jul 16 2012, 11:41
Post #16





Group: Super Moderator
Posts: 5275
Joined: 23-June 06
Member No.: 32180



QUOTE (2Bdecided @ Jul 16 2012, 11:34) *
It's a shame fb2k didn't adopt any of the wav metadata schemes that are out there
foobar2000 gained the ability to tag WAVs in 1.1.9, though you’d have to ask Peter or another developer for the specifics; I don’t recall any having been provided yet.
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 25th July 2014 - 12:53