Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Downsample before encoding or just use lowpass? (Read 9910 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Re: Downsample before encoding or just use lowpass?

Reply #25
... might be worthwhile if we don't care about anything above 12kHz...
I was thinking of the number of cases where that is done accidentally, such as whenever the speaker isn't pointed directly at you (aka off-beam), and listening to an ordinary radio (tends to filter lower than the 15khz spec).  The typical place for both of these, is in the car.  So, that seems like a good place to test drive some downsampled MP3's. 
With 24k sample rate, there's a cool trick of changing the average signal treble density (slightly) back up to what you'd expect; but, with 32k sample rate there's no extra steps, except for how nice the resampler can do. 

If goodly smaller files was the goal, then downsampling is effective.  For smaller MP3 files, the resampling cost is more than paid-off.   However, if large files is the goal, then resampling cost might not be appropriate.

Re: Downsample before encoding or just use lowpass?

Reply #26
At Max9000:
More than 20 years ago, we were listening to MP3 radios at 24kbps, or, those lucky enough to use ISDN instead of modem, could listen to some 56kbps radios.  (You would need to use a pair of ISDN channels at the same time if you would want to listen to 64kbps or more)

For some reason, you are trying to make it like we are still living at that time, and for some inconceivable reason, saying that we can go on with low quality music on a car.
I have a 64GB SDcard on my mobile phone which is enough space for all my mp3, mp4 and opus files and listening via bluetooth to the integrated car radio I can clearly hear those that are of bad quality ( very few 22Khz ones and some other low bitrate/bad quality encoding with artifacts and lowpasses between 12-14Khz).

There is literally no reason to want a 32kbps or 48kbps MP3 file sounding "not too bad" nowadays.

And obviously, there is no reason to expect to win a "smallest file" contest with MP3.
Back in 2003, with Ogg Vorbis, some tests were done with files as small as 3,2kbps.  Here you can still listen to a 46-seconds file encoded at 5.2kbps. https://sjeng.org/floggy.html
Info threads:
https://hydrogenaud.io/index.php?topic=2726.0
https://hydrogenaud.io/index.php?topic=6023.150

Official Vorbis cannot do as low, so  -q -1 at 8Khz does 12kbps "stereoish". (10 in mono), although it does sound better.

And yet, that's just history. Why would we care about them with xHE-AAC and Opus?

Re: Downsample before encoding or just use lowpass?

Reply #27
If a little MP3 sounds bad, then it was done wrong.  But, that really proves your point. 
Doing a small+good MP3 takes some extra steps and care; however, a more modern codec is easier to use in one step.
Even with extra steps and care, a similar MP3 will be ~20% larger than an AAC file.
Well, you'd only use an MP3 if you needed it.  That is still possible. 

The question was about downsampling vs lowpass, and the answer was:
downsample if the goal is smaller files.
lowpass if the goal is not smaller files.
Highpoint/max-file-size would obscure the difference, but lowpoint reveals it (lowpoint is tiny MP3).
In fact, CD-spec cripples MP3, maybe because MP3's aren't circular?  :)  

Your bluetooth solution is brilliant for using the up-to-date decoders in the phone, not the wonky-sounding early-adopter AAC decoders typically built in a car radio (old wonky AAC also available in internet radio appliance and Sony HD radio too).   My phone is an Apple with many hours spent for quickly-lost iTunes playlists.  So, either replace the phone or the car?  Yes, eventually.  But, not immediately. 

Likewise, I guess that MP3 internet radio stations are waiting for wonky-sounding, headache-inspiring AAC early-adopter internet radio appliances to die out (same problem stalled-out HD-radio--spend extra for a headache or other poorly review, didn't get popular).  If I remember right, it was but 2 years ago that I got to listen to nice sounding AAC's via Foobar2000.  The difficult-if-not-impossible to update appliances should mostly die out in 3 or 4 years more.  But, at this time, the drop in listenership might not be worth the bandwidth savings of a newer codec.  I'd like to suppose that everyone-plays-simultaneously styles like classical have already changed to a newer codec.  But, for others, there's post#13, this same thread, with the tiny+excellent 51kbit mp3 example. 

Thanks for your advice.  It is good!  Using the up-to-date (preferably convenient-to-load android phone) to get up-to-date decoder signals sent to car or home radio, is very good advice.  However, this is not easily done with my iPhone and Ford (Apple vs Microsoft).  Also feel free to post an example (for educational purpose) that competes with 51kbit mp3.  AAC can't touch that (because the example is a fluke which should have been up at almost 80kbits for clear MP3).  Probably, Vorbis can compete with that, and more easily done too. 
Anyhow, my reason for using MP3 was to illustrate that in downsampling vs lowpass, then downsampling is the clear, exceptionally clear, winner (unless you wanted maximized file size to wipe-out the difference). 

This commentary because it isn't feasible to mention all of the details in one post; and, also because [JAZ]'s counterpoise post is both good and helpful too. 

 

Re: Downsample before encoding or just use lowpass?

Reply #28
Perhaps another example is good. 
Here is Ricky Martin ft Rosie Odonnell on their CD 'Another Rosie Christmas' available for purchase:
US:  https://www.amazon.com/Another-Christmas-ODonnell-Various-Artists/dp/B00PUU81H4
UK:  https://www.amazon.co.uk/Another-Christmas-ODonnell-Various-Artists/dp/B00PUU81H4
That annotation is required in my region.  The reduced mp3 example, of one track, is presented for educational purpose. 

Perhaps consider purchasing the CD for comparison and also because it has more fun tracks.
The attached, educational purpose, stereo file is 24000 sample rate, 79kbit MP3. 
Not long ago, I was unaware of how small-size MP3's should sound and hadn't tried downsampling for MP3. 
Gosh, I had a shock.
Did the CD or the downsampled 79kbit MP3 perform better?  Hear and decide for yourself.