live 7 sample rate conversion poor? - comparison chart

muchachotron · Post by **muchachotron** » Fri Feb 06, 2009 11:24 pm

has anyone taken a look at this chart? compare the sweep graph between live 7 and, say, logic 8 on 10.5. logic looks clean and crisp. live 7 has grid lines all over the place. it actually looks like one of the worst ones in the comparison.

http://src.infinitewave.ca/

i am not a sound engineer so i do not fully grasp these charts. i would like some other opinions and views on this info. also, i am not bashing ableton live at all -- i am absolutely smitten with the software. i would just like to know what people think of these charts.

leedsquietman · Post by **leedsquietman** » Sat Feb 07, 2009 12:58 am

Never use DAWs for sample rate conversion is always one of the rules I stick to. I use Voxengo's R8Brain Pro, which is also recommended by mastering guru Bob Katz (among others) in his Mastering Audio : The Art and The Science (2nd edition).

Look at ALL of the tests. You will realize that those graphs are quite misleading in that even with the tiniest, in the vast majority of cases, inaudible errors, they look horrible on that 'sweep' grid. Look at the other tests with bar graphs and you'll see they are a much better representation and look a lot less shocking. The eye candy extremism is there to draw more traffic in and yet more people with more time on their hands than they know how to handle can draw it into a year long debate at KVR or whatever and keep the hits ticking over.

How much Sample rate converting are you planning on actually doing. If you are doing lots get a good tool designed as a sample rate converter, not a jack of all trades DAW or audio editor. I very rarely do any, as I find that recording at 24/44.1 is fine for my needs.

I have to say that Soundforge's sweep doesn't look that good on this graph, yet I have used it for SRC many times (in mastering other people's projects) and compared it against many of the others given I've used Protools HD, Adobe Audition, Vegas, Cubase/Nuendo. Logic, Wavelab and others and I didn't notice anything audibly different from Audition, whose graph looks clean. For me, if you can't hear the difference, it's not worth arguing about. And my hearing is pretty sensitive and I can definately tell the difference in many cases with dithering, which is another topic that could be discussed ad nauseum for years ....

Tone Deft · Post by **Tone Deft** » Sat Feb 07, 2009 1:07 am

what do the lines mean?
open Live, go to options-preferences-audio

turn on the tone generator, set it to 0dB.
that level of noise is the white line.

now turn the tone generator to -55dB
that level of noise is the bright orange line.

etc.

the white line is the input signal, it's there at 0dB, the other lines are artifacts, their shapes are due to harmonics and frequency folding and blah blah DSP geekshit.

ALL sample rate conversion is lossy and introduces noise. in interpreting these flaws you have to ask yourself at what level (how loud, in decibels) are these flaws acceptable. IMO -60dB is where artifacts start to not matter in my DAW (I expect much more from DAC and ADC conversion, for example.)

they chose a 44.1kHz to 96kHz conversion because 96k is not evenly divisible by 44.1kHz, the math has remainders and left over error bits.

hth

ethios4 · Post by **ethios4** » Sat Feb 07, 2009 1:18 am

I wonder if they used HiQ mode for testing Live ?

Tarekith · Post by **Tarekith** » Sat Feb 07, 2009 1:21 am

HQ mode has nothing to do with this test.

Tone Deft · Post by **Tone Deft** » Sat Feb 07, 2009 1:32 am

dunno what you mean.

my guess is that they made a PCM file at 96kHz with a rising tone. dragged it into Live and exported it at 44.1kHz.

ethios4 · Post by **ethios4** » Sat Feb 07, 2009 2:46 am

Manual p.109

If the High Quality switch is on, Live uses an advanced sample-rate conversion algorithm that provides better sound quality at the expense of a higher CPU load. Samples processed with the Hi-Q algorithm generate less distortion, particularly at high frequencies...

dazzer · Post by **dazzer** » Sat Feb 07, 2009 3:19 am

ethios4 wrote:Manual p.109
If the High Quality switch is on, Live uses an advanced sample-rate conversion algorithm that provides better sound quality at the expense of a higher CPU load. Samples processed with the Hi-Q algorithm generate less distortion, particularly at high frequencies...

pwned.

I never knew Live had this. I've just set it to Hi Q and suddenly my music sucks far less than it did before. Nice.

Tone Deft · Post by **Tone Deft** » Sat Feb 07, 2009 3:42 am

dazzer wrote:pwned.

+1

it would interesting to compare both modes in a similar test. I'm pretty sure those are Matlab plots (expensive math/DSP software), maybe we can find other software to try some tests. Bueller... ?

timothyallan · Post by **timothyallan** » Sat Feb 07, 2009 11:24 am

pg 506 dans la manual:

"Avoid using samples that are at different sample rates within the same project. If you want to work with such files, we recommend that you first convert them to the sample rate set for your audio interface in an offline application that is optimized for this task."

aaaand "Always render at 32-bit and at the sample rate set for your audio interface. If you need audio files at a different sample rate and/or bit depth, we recommend that you convert your rendered files in an ofine application that is optimized for these tasks, rather than in Live."

So they're basically saying Lives SRC is far from optimal... however, why should it be!?

ilia · Post by **ilia** » Sat Feb 07, 2009 12:54 pm

timothyallan wrote:So they're basically saying Lives SRC is far from optimal... however, why should it be!?

real-time constraints?

ilia · Post by **ilia** » Sat Feb 07, 2009 12:56 pm

Tone Deft wrote:
dazzer wrote:pwned.
+1

it would interesting to compare both modes in a similar test. I'm pretty sure those are Matlab plots (expensive math/DSP software), maybe we can find other software to try some tests. Bueller... ?

octave is free/open source.
http://www.gnu.org/software/octave/

leonard · Post by **leonard** » Sat Feb 07, 2009 1:17 pm

Tone Deft wrote: it would interesting to compare both modes in a similar test. I'm pretty sure those are Matlab plots (expensive math/DSP software), maybe we can find other software to try some tests. Bueller... ?

i knew i'd seen them before:
http://forum.ableton.com/viewtopic.php? ... e+audition

timothyallan · Post by **timothyallan** » Sat Feb 07, 2009 1:19 pm

ilia wrote:
timothyallan wrote:So they're basically saying Lives SRC is far from optimal... however, why should it be!?
real-time constraints?

Yep, that's what I think. Why waste CPU cycles on SRC when you could use it for something 'better'

dazzer · Post by **dazzer** » Sat Feb 07, 2009 1:47 pm

leonard wrote:
Tone Deft wrote: it would interesting to compare both modes in a similar test. I'm pretty sure those are Matlab plots (expensive math/DSP software), maybe we can find other software to try some tests. Bueller... ?
i knew i'd seen them before:
http://forum.ableton.com/viewtopic.php? ... e+audition

Cool link. Looks like it got discussed in depth a while back. Angstrom was right as well, dude in a thread about this on KVR keeps going on about "this crappy Live". el-oh-el.