Live's resampling quality aka Sample Rate Conversion

[Machinesworking]

EVERY DAW has this problem.

Exactly. Only a 25,000+ Pro Tools HD setup seems to match the 2-500$+ stand alone applications in SR conversion tests, so in that sense you have to ask yourself how important it really is? If to you it's important enough to spend a couple hundred on, then by all means do.

Honestly, if you were to use 8 bit samples from a cell phone of you playing acoustic guitar in a commercial release, it's my guess you would probably apply EQ and noise filters to it pretty heavily you know?

[Timur]

In theory it doesn't even matter if I am making actual use of a broken function now or only plan to use it in the future, if it's broken it needs to be repaired!

OK in this section here. In essence you're asking Ableton to improve in house SR conversion, that's fine, but as I stated, and the tests from the "mac only" site show, Ableton is meeting the standard for in DAW SR conversion. One can assume that they would be on par with the average results form the Barbabatch test, ( or beat them, the new audio engine etc. ) so then in that case the software is not broken.

I'm sorry, I was not making clear that this part of my post solely points to my bug-reports and not to this SRC thread. I fell into that because of the many replies accusing me of only wanting to make Ableton's life miserably.

My original post started like this (I changed the POS bit):

Don't take it personal, but I think the quality of Live's Resample Engine (SRC) is sub-par, at least when resampling very low sample-rate clips. I don't know if it's bugged or not, but either way it could probably be improved in a future release.

From there on everyone started jumping at me. One could wonder if I am really to blame for that mess.

MIDI timing issues, pops and clicks in 7, stability etc., these are all more important to me than an above average SR algorithm in Live.

I absolutely agree, and this is what I was talking about when I came up with the broken car jack thingy. I have a list of issues that are waiting for acknowledgement other than "we will be looking into this". (into what out of a dozend reports?)

[Tone Deft]

true.

my rants center around the notion that once you really dig into audio it's ugly, the touchstone is knowing how to interpret that ugliness and pick your battles accordingly.

[Tone Deft]

the quality of Live's Resample Engine (SRC) is sub-par, at least when resampling very low sample-rate clips. I don't know if it's bugged or not, but either way it could probably be improved in a future release.

IMO your test was highly flawed, too many variables.

in a scientific test you identify the variables you want to test, and pick tests that exercise each variable independently, one at a time.

[Timur]

request - revisit your test without the funky cell phone recording. if you've done that, repost, I'm not reading those long posts.

maybe put commentary at the bottom and BRIEF facts/observations at the top of a post. this shit isn't working.

Look at this post (reply to Robert), scroll down to the section where I quote your funky 17 KHz + 12000 dB statement. This is an all around "commentary at the bottom" with some "BRIEF facts/observations" on top of it for which I used your testfiles.

http://www.ableton.com/forum/viewtopic. ... 439#664439

[dcease]

EVERY DAW has this problem. define what's bad. show me how to SRC with Live to get a crap result.

correction... EVERY DAW has problems.

if you can't make a decent tune with live, maybe you should check yourself.

broken record. i'm done with this conversation. bye timur.

[Timur]

IMO your test was highly flawed, too many variables.

in a scientific test you identify the variables you want to test, and pick tests that exercise each variable independently, one at a time.

1. What variables do you keep talking about?

2. I repeated more than once that this was not scientific.

3. Again, what variables do you keep talking about? There was a source-file imported into Live and played back through my speakers through the very same audio-chain that I played the original through except for the playback software (Live is not able to playback at 8 KHz).

The full setup was:

a) The original file was played back through the original Nokia AMR for reference then comverted to WAV using "MIKSOFT Mobile AMR Converter".

b) The WAV was played back using Winamp's AudioBurst FX output plugin at original 8 KHz sample-rate through the Audiophile 24/96, for reference of how the original ought to sound.

c) The WAV was imported to Live and both the Default and Hi-Q SRC were used in comparison through both the Audiophiler 24/96 and the Creative X-Fi.

d) Several other SRCs were used for comparison both played back through my speakers (using both audio-interfaces) and rendered into resampled 48 KHz files for better comparison and then played back through my speakers (using both audio-interfaces).

Now I'm tired, the couch is waiting (damn I blew my ears today when listening to all those sweeping sounds and 17+ KHz testtones).

[Timur]

broken record. i'm done with this conversation. bye timur.

Bye dcease, I can't blame you.

[Timur]

Here's the shortform conclusion of doing the sine-sweep test with Tonedeft's files (as taken from my former post):

My half-educated guess is that most of the problems are based on aliasing. Especially the default SRC adds several additional sweeping harmonies of higher sweeping-rate to the original sweeping sound.

Both SRCs do that very audibly when processing the 8 KHz SR clip, especially when resampling to unnatural multipliers of the original SR (like 8 -> 44.1 KHz). When processing the 48 KHZ SR clip the Hi-Q SRC works audibly really good, but the default SRC only works transparent with 48 -> 96 KHz resampling.

So as long as you don't have to process low sample-rate content (8 KHz at least) Live's Hi-Q SRC will do the job quite adequate.

I still do worry about the high frequency energy being added which may need you to apply an additional low-pass/high-cut filter before entering the resampled signal into other effects. I would be glad if other could comment on that with their own experience.

Using an external SRC (are the any out there that offer different SRC methods/algorithms?) might become necessary if you strive for more transparent/neutral results. I hope everyone can live with this statement now.

[Tone Deft]

Look at this post (reply to Robert), scroll down to the section where I quote your funky 17 KHz + 12000 dB statement. This is an all around "commentary at the bottom" with some "BRIEF facts/observations" on top of it for which I used your testfiles.

http://www.ableton.com/forum/viewtopic. ... 439#664439

the 17kHz thing was directed at MW, he and I are a bit older than yourself, I wouldn't guess what your hearing is like.

whatever, that paragraph is more of you spouting bullshit, you do not know what you're talking about. bad assumptions, logorrhea. shit like this

My half-educated guess is that most of the problems are based on aliasing. Especially the default SRC adds several additional sweeping harmonies of higher sweeping-rate to the original sweeping sound.

pure idiocy.

wtf is your education on this, have you EVER taken one single class or seminar or convention or anything on audio?? what's your basis for all this bullshit? do you realize that this is a WORLDWIDE forum and there just might be people versed on this stuff reading your bullshit and seeing right through it?? I'm not even that well educated on it, I only have a bachelor's degree, I can't wait until a REAL DSP guy logs in and can REALLY rip you a new asshole.

Timur, you redefine n00b.

/goddamnit, now I'm being an asshole again, fuck it, this is audio, I'm not going to mince words about audio on this forum.

[Machinesworking]

true.

my rants center around the notion that once you really dig into audio it's ugly, the touchstone is knowing how to interpret that ugliness and pick your battles accordingly.

Right, again if you need the highest quality SR, then stepping outside Ableton will probably always be the way to go, independent companies like Izotope or Audio Ease spend more R&D on this stuff than a small DAW company like Ableton can afford to.
Also, in 99% of the cases for me and others, it's not a problem, therefore it's place on the list of demands for Live 8 is down near the bottom.

I would rather have elegant vertical resizing, more extensive zoom features, an event editor, more options for follow actions, warp tempo guessing improvements, a bug fix on the first note dropping out on third party instruments like Zebra, Automat, and Absynth, the ability to tear off the clip view like Reason can tear off the sequencer window, thus keeping to Abletons single window philosophy, yet giving use the ability to have an audio wave or piano roll on another screen full size!

ANY, of these things are more important to me, and honestly, probably one of the reasons that people get upset about this is that Ableton just spent a serious amount of time redesigning the audio engine as 64 bit in what amounts to pretty microscopic improvements in audio quality. No doubt sacrificing at least one of the above improvements for a small amount of people who were questioning Lives 'quality'.
Whoo Hoo! 64 bit summing! but still no decent vertical zooming, thanks audiophile types!

[Machinesworking]

fuck it, this is audio, I'm not going to mince words about audio on this forum.

THIS IS AUDIO!!

[Timur]

My half-educated guess is that most of the problems are based on aliasing. Especially the default SRC adds several additional sweeping harmonies of higher sweeping-rate to the original sweeping sound.

pure idiocy.
...
I can't wait until a REAL DSP guy logs in and can REALLY rip you a new asshole.

I don't understand what your problem is. First I did a "guess" then I thoroughly explained what can be heard when sending your testfiles through the different SRC routines. That invites anyone with true knowledge to comment on it. I do not know if the audible artifact can be defined as "aliasing" (I had a reason to guess that), but does it matter for the fact that they are present? Really, if you know anything about this then please let me and others share what you thing is the source of the artifacts. If not then maybe one of your DSP guys comes along and enlightens us.

As a user do I have to care why Live's SRC falls short? Nope, I only need to know its strengths and weaknesses. But I'd be interested for pure educational purposes nevertheless. Anyway, the artifact are there, and that is neither idiocy nor anything else you call it unless you can provide any evidence that it is.

(Do you notice how we are talking about my credibility again instead of the content? You keep ignoring the content and deny making any contribution. You are ruining the thread with this and keep anyone who's really interested in discussing the content from even looking at it. If that was your intention then you succeeded.)

[bgc]

...probably one of the reasons that people get upset about this is that Ableton just spent a serious amount of time redesigning the audio engine as 64 bit in what amounts to pretty microscopic improvements in audio quality. No doubt sacrificing at least one of the above improvements for a small amount of people who were questioning Lives 'quality'.
Whoo Hoo! 64 bit summing! but still no decent vertical zooming, thanks audiophile types!

And, as laid out in that swell video posted here and referenced on CDM recently, the 64-bit summing bus was implemented, potentially, for marketing purposes.

[Tone Deft]

MW - exactly how I feel about it. audio, serious business. there's my wife, there's audio and then there's everything else in my life that can fuck off.

Timur - I have not seen a test from you that's a scientifically sound, brief, to the point, obvious way to break Live's SRC. I am only interested in short posts.

I have guests coming for the weekend any moment now, I can't go on. at least one of the guests is a great guitar player, there will be jams, fuck this SRC ranting crap, MUSIC TIME!!!


fuck me, this is all over 8kHz SRC. as your sig says Timur, madness.