IMO we cannot (Yet) say that since some of us are proving otherwise. I'm not sure what you're suggesting about the harmonics…8cros wrote: Now we can say that the mixer changes sound. Even if it's masked.
We can rise to conspiracy theories.
It has a low level, but the effect on the harmonics greater than the fundamental tone.
You mean MClass EQ with idle settings? I may simply copy input to output when nothing is on.8cros wrote:In MClass equalizer for example no one rounding error.orthodox wrote:I would like to remind you that the machine epsilon is -138.5 dB for the Float32 type.8cros wrote:24-bit is not enough. Because in my case, after subtracting the signals I added +163.2 decibels...
So anything you may discover in that range is simply a digital noise that comes from rounding errors.
I do not like to argue. 32 bits floating point is 1540 dB dynamic range.orthodox wrote:You mean MClass EQ with idle settings? I may simply copy input to output when nothing is on.8cros wrote:In MClass equalizer for example no one rounding error.orthodox wrote:I would like to remind you that the machine epsilon is -138.5 dB for the Float32 type.8cros wrote:24-bit is not enough. Because in my case, after subtracting the signals I added +163.2 decibels...
So anything you may discover in that range is simply a digital noise that comes from rounding errors.
The 14:2 fader sets the gain factor not exactly equal to 1.000000. It is slightly different by a small margin. If you subtract its output from the original, you will get about -110dB. If you amplify it, you will get almost the same signal, but with only 4 correct mantissa bits instead of the original 23.
Null mixer test failed. There is noise but also harmonic. I really do not want to speculate.selig wrote:IMO we cannot (Yet) say that since some of us are proving otherwise. I'm not sure what you're suggesting about the harmonics…8cros wrote: Now we can say that the mixer changes sound. Even if it's masked.
We can rise to conspiracy theories.
It has a low level, but the effect on the harmonics greater than the fundamental tone.
Sent from my iPad using Tapatalk
It is not a case of dynamic range and 1540 dB does not apply here. The level of quantization errors is -138 dB. If you amplify the difference from the level of -110 dB to 0 dB, you are raising the level of rounding and truncation errors from -138 to -28 dB at the same time.8cros wrote:I do not like to argue. 32 bits floating point is 1540 dB dynamic range.
give an exampleorthodox wrote:It is not a case of dynamic range and 1540 dB does not apply here. The level of quantization errors is -138 dB. If you amplify the difference from the level of -110 dB to 0 dB, you are raising the level of rounding and truncation errors from -138 to -28 dB at the same time.8cros wrote:I do not like to argue. 32 bits floating point is 1540 dB dynamic range.
Example of what? I am expaining the actual behavior of the 14:2.8cros wrote:give an exampleorthodox wrote:It is not a case of dynamic range and 1540 dB does not apply here. The level of quantization errors is -138 dB. If you amplify the difference from the level of -110 dB to 0 dB, you are raising the level of rounding and truncation errors from -138 to -28 dB at the same time.8cros wrote:I do not like to argue. 32 bits floating point is 1540 dB dynamic range.
It looks like I knew where there was a mistake. I can not test the SSL in Reason's because SSL has a greater range than all other devices.orthodox wrote:Example of what? I am expaining the actual behavior of the 14:2.8cros wrote:give an exampleorthodox wrote:It is not a case of dynamic range and 1540 dB does not apply here. The level of quantization errors is -138 dB. If you amplify the difference from the level of -110 dB to 0 dB, you are raising the level of rounding and truncation errors from -138 to -28 dB at the same time.8cros wrote:I do not like to argue. 32 bits floating point is 1540 dB dynamic range.
OK, here is an example of what you call "1540 dB dynamic range". You take a normal (0dB) signal, lower it by 700 dB, raise it by 1400 dB, then lower it by 700 dB. It should stay the same.
And here is an example of quantization errors. You take signal A of 0dB, mix it with signal B of -140 dB, then subtract the original signal A from the mix. You won't get the signal B, it will be completely destroyed. In the case of SSL mixer, the signal B should be about -320 dB since the SSL is operating with Float64.
I did test as i was reading through with mention of comparing to original file . I used original file to test against bounces of all devices mentioned in this thread .WongoTheSane wrote:Well, yes, it's gone off the original topic, which was: is there a difference in the output signal between the SSL and the 14:2. Stranger said there was, and several people chimed in to note that his protocols were wrong. We provided several ways to test both hypotheses (#1: the output is different between a 14:2 and the SSL; #2: the 14:2 adds color). Both tests are really, really easy and literally take under a minute, but several other people have posted tests that try to answer both questions at once, while adding a third one: #3 Reason colors the sound.submonsterz wrote:errmmm so the thread has gone off the origanal thread topic ??
I am not trying to prove anything but the reality that people have bashed ie stranger in his findings when he was trying to help people and just saying I find he was correct and theres my proof and the it does not exist sayers are obviously not testing correctly to look below what reasons meters etc tell you.
I was just trying to help too in showing that theres is something there noise colour what ever it is it exists
That's where I (and I expect several others) object in the strongest terms:
1) you don't test three things at once. That's not how it's done.
2) you can't test whether X adds Y if you don't know what Y is (and are still unsure about what X is because it's been chained to an inordinate and superfluous amount of devices). Both your and 8cros' test show noise, not color (actually, 8cros' test shows phasing delay IMHO, neither noise nor color).
3) Occam: the explanation that requires the least assumptions is probably the right one. i.e. if you test only the pure output of a 14:2 and the pure output of the SSL and it DOES null perfectly bit for bit; and then you start piling gain methods and some "noise" starts to appear, the noise comes from your gain methods.
4) If you submit your tests to other, be clear on what you are testing and what you're trying to show. Noise is not color, and the tests for both are very different.
Right now, you've successfully convinced me that in some cases, some devices can add an infinitesimal amount of noise (but it's not something I would have doubted anyway); you haven't convinced me that it's color, nor that it is introduced by Reason, the SSL or the 14:2. Again, that was the point of the thread.
(to be clear: it's not a personal attack, I'm just defending the point of view of "the skeptics team". I have opened and studied the song you posted and agree it's interesting, but again: not Reason, not SLL, not 14:2, not color).
.Reason mixer doesn't have a channel count Limit. Just keep adding audio tracks and devices and be happy.Raveshaper wrote:The other has 64 channels, 8 sends, and can have things inserted into it.
If you add this much gain you might be seeing artefacts from rounding errors from the floating point representation of the samples inside your CPU and RAM. This is the digital noise floor.8cros wrote:24-bit is not enough. Because in my case, after subtracting the signals I added +163.2 decibels. But color 14: 2 and 6: 2 mixers in your example is easy to find. You need to add 94 dB to the master.
Yes very distinct differences . If people say they cannot see the differences from them plots eyes need testing ....Stranger. wrote:Wow-lol..
Ok/ further findings.
Case (for me) is not closed at all now,after building some new noise profiles....
Method used here>
All profiles were made over a 2-3min period from a playing static sample-- with no added gains,just roughly normal peak input levels.
1 x stereo channel feed direct to master section fx return 1+2 (to avoid any contaminate from mix channel) sent to IO output 1+2. >Profiled.
1 x stereo channel feed direct to 14:2 mixer >> direct out to IO 1+2. >Profiled.
1 x stereo channel feed direct to 6:2 mixer >> direct IO 1+2. >Profiled.
1 x stereo channel feed direct to mix channel >> direct IO 1+2. >Profiled.
1 x stereo channel feed direct to audio channel as a pass through (yes u can do this) >> direct IO 1+2. >Profiled.
Picture paints a gazzillion words. New conclusion to date- i am intrigued and surprised by results so far...Trying to make sure all my steps are totally systematic in approach+execution,and at this point,really do not have a clue what's actually going on here!.. lol- i never bothered looking at all this before,so,again,surprised by a variety of different tests all showing differences across the board.
Perhaps the ph team can take a closer peak into their own analasis and make some final clarifications.. *shrugz*
I made a mistake with the project. I was wrong. Mixer SSL limited -320dB and not 1540 as I thought.jam-s wrote:If you add this much gain you might be seeing artefacts from rounding errors from the floating point representation of the samples inside your CPU and RAM. This is the digital noise floor.8cros wrote:24-bit is not enough. Because in my case, after subtracting the signals I added +163.2 decibels. But color 14: 2 and 6: 2 mixers in your example is easy to find. You need to add 94 dB to the master.
Also make sure you don't accidentally have some audio input enabled, as those sometime also introduce some very low level noise (even when muted).
But we're not testing our eyes…submonsterz wrote:Yes very distinct differences . If people say they cannot see the differences from them plots eyes need testing ....Stranger. wrote:Wow-lol..
Ok/ further findings.
Case (for me) is not closed at all now,after building some new noise profiles....
Method used here>
All profiles were made over a 2-3min period from a playing static sample-- with no added gains,just roughly normal peak input levels.
1 x stereo channel feed direct to master section fx return 1+2 (to avoid any contaminate from mix channel) sent to IO output 1+2. >Profiled.
1 x stereo channel feed direct to 14:2 mixer >> direct out to IO 1+2. >Profiled.
1 x stereo channel feed direct to 6:2 mixer >> direct IO 1+2. >Profiled.
1 x stereo channel feed direct to mix channel >> direct IO 1+2. >Profiled.
1 x stereo channel feed direct to audio channel as a pass through (yes u can do this) >> direct IO 1+2. >Profiled.
Picture paints a gazzillion words. New conclusion to date- i am intrigued and surprised by results so far...Trying to make sure all my steps are totally systematic in approach+execution,and at this point,really do not have a clue what's actually going on here!.. lol- i never bothered looking at all this before,so,again,surprised by a variety of different tests all showing differences across the board.
Perhaps the ph team can take a closer peak into their own analasis and make some final clarifications.. *shrugz*
I hear you!Marco Raaphorst wrote:Are we testing how mad we are?
This is The Big Soap Thread. Put a safety warning on it. You might need help after reading this...
It's because of ancient aliens embedding our DNA with the lust for leaving no stone unturned in host sound qualityMarco Raaphorst wrote:Are we testing how mad we are?
This is The Big Soap Thread. Put a safety warning on it. You might need help after reading this...
That's because Soundcloud colors the sound.gak wrote:Besides, once it gets to soundcloud it will sound like shit anyways!
I always try to get my stuff to sound like shit BEFORE I upload it to soundcloud. That's how the real pros do it.gak wrote:Besides, once it gets to soundcloud it will sound like shit anyways!
"Reason is not measured by size or height, but by principle.” -Epictetus
@ -008' Sounds
Share and download patches!
Join ReasonTalk on Slack!