Author Topic: highest possible quality resample filter (plugin)  (Read 2176 times)

iFriend

  • Jr. Member
  • **
  • Posts: 26
Hello,

instead of using the internal resample function (Player -> resample to), I 'd like to use a high-quality resample plugin. Please, do you have any ideas?

iFriend

  • Jr. Member
  • **
  • Posts: 26
guys? Any Idea please? E.g. like foobar, which has DSP items for resampling...?

iFriend

  • Jr. Member
  • **
  • Posts: 26
sorry to get on your nerves, but I still don't have a solution for this issue. Please note that I'm NOT a freak of up-sampling or so, but I need the 48kHz for technical reasons of my rig (it does not accept any other in the config that I need). Atm, I use the in-build up-sampling but there are really quite some differences in quality, so I'd like to use the best wavailable. Meanwhile I will have a look if I find a VST plugin that does the job, as I discovered today the MB supports VSTs. If you know of any HQ VSTs that do upsampling pls let me know!

Thank you in advance!

P.S. For winamp and foobar there are resample plugins that use SSRC: out_ds_ssrc.dll and foo_dsp_ssrc.dll . I tried to load both into MB but no success. Somewhere I read that winamp plugins are supported but seems not, or am I missing smt pls?
Last Edit: September 03, 2020, 04:41:24 PM by iFriend

ikonomov

  • Jr. Member
  • **
  • Posts: 71
Did you ever find any solution?  Also did you do any testing to establish the quality of the internal resampler used by MB?

iFriend

  • Jr. Member
  • **
  • Posts: 26
Nope, did not find any solution. And I did not do any testing (if you mean with other means than by ear), but by my ears it's a clear difference...

ikonomov

  • Jr. Member
  • **
  • Posts: 71
Not trying to start an argument, but unless Steven went out of his way to choose a bad resampler I think it would not be easy to hear a clear difference.  There are definitely resamplers out there that result in bad quality, but given that latency performance is irrelevant for a music player I highly doubt this is what we have.  Still, I can agree that it can definitely be auditable and significant in quality degradation, which is why I hope we can get a little more insight about the resampler used.
Edit: Hopefully somebody knowledgeable here or in the new thread I started https://getmusicbee.com/forum/index.php?topic=36624.0 will reply to give some information about what kind of resampler is being used.

iFriend

  • Jr. Member
  • **
  • Posts: 26
thank you, ikonomov. I fully understand your perspective. I personally learned that "audible" depends on so many factors that it's difficult to tell. In my case, being a musician, I invested quite some money in good stuff, e.g. headphones for around 800 EUR, etc. I personally definitely hear a difference. There is also a reason I suspect why some people program high quality resamplers. Anyway, again, thanks for commenting, etc. Btw, I use RME too :)

ikonomov

  • Jr. Member
  • **
  • Posts: 71
hiccup replied in the other thread that I posted stating that likely BASS's built-in resampler is the one used by MusicBee.  It looks to be a popular audio engine, likely with a quality resampler, but somebody created a resample library on GitHub https://github.com/aidan-g/BASS_SOX specifically to be used with BASS, so many not.  You said you can definitely hear a difference.  Can you provide a little more detail about the source material that you use to compare and what sample rate do you use to upsample to.  Since you said you use RME I assume you use ASIO driver for the output, but can you confirm that this is what you use when you do the comparison.  I just did a bit of A/B comparison between 44.1 kHz file with and without being upsampled to 96 kHz and I can't hear a difference.  I use AKG K712 Pro headphones together with RME Babyface and an external O2 headphone amp.  I tested it with many different sound files.  Theoretically if the resampler is "good enough" and not actually bad (some examples here of both https://lastique.github.io/src_test/) there could actually be a potential benefits with the filter used when upsampling.

iFriend

  • Jr. Member
  • **
  • Posts: 26
Thank you ikonomov! This whole thing is quiet some time ago. Atm, I'm not able to remember any details about the material, nor do I have resources to repeat any tests (my health is not the best). But I will answer as much as I can: yes, I always use ASIO (RME ADI 2 DAC FS), using AKG reference headphones. I definitely upsampled from 44.1 to 48, nothing else, because that's what I was interested in at that time.
Just as a hint, you may want to look into DAC upsampling tech on youtube. There are quite some details as far as I remember. Still, as these things are also sometimes very subjective, I may just have heard "a ghost", so as long as you are ok with it maybe just enjoy!
I just think that it would be good if MB could offer the possibility to include hi-q upsampling plugins (e.g. SSRC).

ikonomov

  • Jr. Member
  • **
  • Posts: 71
I agree, it would be nice to have some indication as to the quality of the resampler used.  I tried to find tests of the BASS resampler that somebody might have done, but I couldn't find any.  It is very likely that it is good enough as I highly doubt that it can be worse quality than the one Windows 10 uses, and that has already been instrument tested couple of times and shown to be good enough.  So a solution would be to allow sampling plugins as you suggested or maybe include a high quality resampler as an option in the settings.  SOX comes to mind since somebody already did a plugin for it https://github.com/aidan-g/BASS_SOX and it is highly regarded resampler for quality when configured that way.  As for your hearing tests, among many factors that can contribute to this maybe it is possible that you have heard sound degradation because you were upsampling from 44.1 to 48kHz.  The general consensus seems to be that it is best to upsample with an integer factor of 2 or 4.  I couldn't hear any difference at 96kHz which also has interpolation, but maybe interpolation works better at 96kHz.  If we are stuck with BASS resampler hopefully one of those days somebody will care enough to do some instrument tests or maybe people more knowledgeable about the BASS resampler can provide some information.

phred

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 8160
 If we are stuck with BASS resampler hopefully one of those days somebody will care enough to do some instrument tests or maybe people more knowledgeable about the BASS resampler can provide some information.
This is something you should take up with the BASS developer.
Download the latest MusicBee v3.5 patch from here.
Unzip into your MusicBee directory and overwrite existing files.

----------
Check out the MusicBee Wiki.
How to post screenshots is here.

ikonomov

  • Jr. Member
  • **
  • Posts: 71
If we are stuck with BASS resampler hopefully one of those days somebody will care enough to do some instrument tests or maybe people more knowledgeable about the BASS resampler can provide some information.
This is something you should take up with the BASS developer.

Thank you for the reply phred.  According to this post here http://www.un4seen.com/forum/?topic=17860.msg125464#msg125464 the resampling by bassasio.dll can have either BASSASIO or BASSmix handle resampling stating that having BASSmix handle it offers higher quality via the BASS_ATTRIB_SRC attribute.  Do you know if this was done in MB?  As I understand it the BASS library provides configuration options for the resampler which MB hasn't made configurable through the settings, do you know how it has been configured?  Once we know this then I can ask in the BASS forum and hopefully we can gain a little more insight.

hiccup

  • Hero Member
  • *****
  • Posts: 6344
I agree, it would be nice to have some indication as to the quality of the resampler used.
The only relevant indication on that (to me) would be reproducible blind tests.
And that seems to be a very big problem when it concerns up-sampling.
Even a sensible notion such as that up-sampling would best be done in factors of natural numbers is sometimes disputed by some people with solid technical backgrounds who say it really doesn't make any discernible difference.

So my advice to anyone that has some feelings of insecurity about the matter of up-sampling (in relation to listening and enjoying music, which is what any music player such as MusicBee is intended for):

1. Either don't use up-sampling.  (you can then be confident that MusicBee will output a bit-perfect signal)
2. Or use your ears, and decide for yourself if you are able to hear differences, and then decide on the best approach for you and your equipment and circumstances. And that will be strictly personal.


edit:
To the respected forum members participating in this thread:
To test various upsamplers yourself, you don't need to wait for further explanations about the way how MusicBee uses upsampling, nor what upsampling plugins do work, or which ones don't.
There are plenty ways to try out different up-sampling methods and algorithms, without any need to (solely) use MusicBee for testing them.
Last Edit: May 27, 2022, 05:32:17 PM by hiccup

ikonomov

  • Jr. Member
  • **
  • Posts: 71
I agree, it would be nice to have some indication as to the quality of the resampler used.
The only relevant indication on that (to me) would be reproducible blind tests.
And that seems to be a very big problem when it concerns up-sampling.
Even a sensible notion such as that up-sampling would best be done in factors of natural numbers is sometimes disputed by some people with solid technical backgrounds who say it really doesn't make any discernible difference.

So my advice to anyone that has some feelings of insecurity about the matter of up-sampling (in relation to listening and enjoying music, which is what any music player such as MusicBee is intended for):

1. Either don't use up-sampling.  (you can then be confident that MusicBee will output a bit-perfect signal)
2. Or use your ears, and decide for yourself if you are able to hear differences, and then decide on the best approach for you and your equipment and circumstances. And that will be strictly personal.


edit:
To the respected forum members participating in this thread:
To test various upsamplers yourself, you don't need to wait for further explanations about the way how MusicBee uses upsampling, nor what upsampling plugins do work, or which ones don't.
There are plenty ways to try out different up-sampling methods and algorithms, without any need to (solely) use MusicBee for testing them.

Be that as it may, your response has the potential to leave us in the dark when it comes to the question of upsampling.  I just don't see what's the problem of having a bit of information about how it has been configured in MB.  Upsampling can be a useful tool depending on people's setup and there is no reason to defeat its purpose or the importance of its quality by talking about bit-perfect signal chain or using our ears when iFriend has already confirmed that he hears sound degradation.

sveakul

  • Hero Member
  • *****
  • Posts: 2033
Before this ends up being drowned in discussion about what resampler is best, how do they affect sound, etc., which are all discussed ad nauseum elsewhere on the web, I do think that the original "root" question is fine to ask on its own just out of curiousity, which is, when the "resample to" box is checked in MB's player preferences, by what method does MusicBee perform the resampling?  And I politely present that now to Steven.

Like hiccup I perform no resampling (WASAPI Exclusive) so personally have no horse in the race.