| Author |
Message |
1:11

Joined: Oct 15, 2010 Posts: 468 Location: The Desert
Audio files: 109
|
Posted: Sat Mar 12, 2011 3:48 pm Post subject:
Quick Question Subject description: Signal Processing and MicroChips |
 |
|
i've been learning lately that different microchips can alter the character of the sounds being processed through them.
i'm having trouble understanding why.
If the signal is just 1's and 0's, why is there still color being added?
Is it purely an effect of the A/D-D/A converter?
and does this still effect processing done entirely "in the box"?
Also, does this mean if i mixdown on one computer, another computer mixing the same signal will sound different?
Thanks ahead of time.
-1:11= |
|
|
Back to top
|
|
 |
DrJustice

Joined: Sep 13, 2004 Posts: 2112 Location: Morokulien
Audio files: 4
|
Posted: Sat Mar 12, 2011 5:11 pm Post subject:
Re: Quick Question Subject description: Signal Processing and MicroChips |
 |
|
The question may be quick, but a complete explanation is not so quick - this is a much debated issue on the net, with a lot of confusion surrounding it.
| 1:11 wrote: | i've been learning lately that different microchips can alter the character of the sounds being processed through them.
i'm having trouble understanding why.
If the signal is just 1's and 0's, why is there still color being added?
|
It's not necessarily the difference between the (processor) chips themselves that produces different sound, but rather the way the chips do the processing. Even on two quite different processors, say an 8 bit one from manufacturer A and a 32 bit one from manufacturer B, you can run software that produces the exact same result down to the last bit.
However, different chips, in combination with the software they run, can alter the character of the sound mainly in two ways:
- They can run different algorithms. E.g. an oscillator or filter of essentially one specific type can be programmed in an almost infinite number of ways, taking different mathematical approaches. This may result in audible differences, from the minute to the obvious. Using different algorithms is the way do differentiate your 'product', and manufacturers will usually keep their algorithms secret to ensure that they have their unique sounds.
- They can run the same algorithms but use different numeric precision (aka "word" lengths, typically 8, 16, 24, 32 or 64 bits) and/or number formats (floating point or fixed point). In this case there may not even be an audible difference if the number of bits used to represent a number is sufficient (say 24 vs 32 bits). More bits are obviously better if one wants high fidelity. Note that an 8 bit processor can do e.g. 32 bit maths exactly like a 32 bit processor, but it will need to do a lot more work, combining four 8 bit words to make up a 32 bit word. For computational efficiency, one usually choose a processor with a native word size and format that matches the application - e.g. an 8 bit AVR CPU is fine for a simple lo-fi synth, while a CPU with at least 32 bit floating point numbers is usually desirable for a high quality synth or audio workstation.
| Quote: | Is it purely an effect of the A/D-D/A converter?
|
Well... It can be. But with modern A/D-D/A converters you'll be hard pressed to hear a difference if they're fed the same data and use the same sample rate. To hear a difference between two good 24 bit converters, which is pretty much the standard today, you'll need to listen extremely carefully - most people will never hear a difference. The analogue electronics that make up the buffer/amplifier stages before an A/D or after a D/A and the power supply is likely to colour the sound more than the converters themselves. Still, this colouring will be minute unless bad components or circuit design is used. Note that some people will claim that there are huge differences, but that's mostly experts with "golden ears".
| Quote: | and does this still effect processing done entirely "in the box"?
Also, does this mean if i mixdown on one computer, another computer mixing the same signal will sound different?
|
A PC or Mac with a CPU made by AMD will sound the same as one with a CPU made by Intel if the two run the same software and use the same A/D-D/A converters.
If they run different software, all bets are off - But even so, two different pieces of digital audio workstation (DAW) software used to play back and mix down the exact same tracks through the same (or similar) type of D/A converter should sound the same unless there's something fishy with the software (there have been fishy DAW software around). Straight up playback and mixing is uncomplicated; just read the files (tracks) and literally add them together, sample by sample. Sound processing such as EQ and reverb is complicated, and different plugins may sound vastly different.
I'm very tired so I'll stop typing now... There's a lot more to say, but these are the basic premises unless I missed something. Hopefully others will correct the above if needed, and add further points if we're lucky
DJ
-- |
|
|
Back to top
|
|
 |
1:11

Joined: Oct 15, 2010 Posts: 468 Location: The Desert
Audio files: 109
|
Posted: Sat Mar 12, 2011 8:21 pm Post subject:
|
 |
|
That was a fantastic answer!
Thank You for taking the time to thoroughly explain the basics, i think i'm finally starting to understand it.
..that helped out quite a bit!!
Rad, it's always so cool when you learn new things...
edit:
Would fishy DAW's include those that use extensive "beat matching", time warping ect. as their main feature? (even when all warp settings are off)
Also, are you still able to push a floating point processor to a point where colorization occurs, like you can with a fixed point?
Again thank you, this is a great help. |
|
|
Back to top
|
|
 |
DrJustice

Joined: Sep 13, 2004 Posts: 2112 Location: Morokulien
Audio files: 4
|
Posted: Sun Mar 13, 2011 3:49 am Post subject:
|
 |
|
I'm glad you found it useful
| 1:11 wrote: | Would fishy DAW's include those that use extensive "beat matching", time warping ect. as their main feature? (even when all warp settings are off)
|
No, most probably not. When doing 1:1 playback (not 1:11 ), there's no reason why a correctly programmed DAW should affect the sound to any degree worth worrying about (but see below). One example of a fishy DAW mixer is the old Cubase VST, which lost a lot of precision during the mixing, certainly enough to colour the sound. In fact it was left with only 12 of the most significant bits intact after mixing 16 tracks with 16 bit audio (which is baffling).
| Quote: | Also, are you still able to push a floating point processor to a point where colorization occurs, like you can with a fixed point?
|
In theory, yes. If you mix a very high number of channels there will be slight rounding errors. However, I don't think it will be a problem for must users. E.g. Cubase SX, which was non-fishy and just used pure addition when mixing, has an error of only a fraction of a bit (leaving almost all 16 bits intact) after mixing 16 tracks of 16 bit audio.
(I once did an analysis of DAW summing errors, using specially written software, to arrive at the results above. Unfortunately I think the resulting article disappeared along with the old SOS forum...)
When it comes to fixed point, you can actually get better results than with floating point. It depends on the word lengths being used, and fixed point DSPs also have a trick or two up their sleeves. In practice most DAWs use the native CPU's floating point format (usually with audio stored as 32 bit floats and the math operations performed at 64 bits). There are exceptions, such as Pro Tools TDM which uses fixed point DSPs to good effect, i.e. better than 32 bit floating point.
DJ
-- |
|
|
Back to top
|
|
 |
nobody

Joined: Mar 09, 2008 Posts: 1687 Location: Not here
|
Posted: Sun Mar 13, 2011 7:49 am Post subject:
|
 |
|
| 1:11 wrote: |
Also, are you still able to push a floating point processor to a point where colorization occurs, like you can with a fixed point?
|
Along these lines, you can deliberately colour your sound with oversampling or something even more basic like pushing a 1-bit flip-flop at a very high rate.
A little AM with sidebands comes to mind when thinking about intentional colourisation (assuming you how know AM works and this is what you want).
Sorry if I'm OT.  |
|
|
Back to top
|
|
 |
DrJustice

Joined: Sep 13, 2004 Posts: 2112 Location: Morokulien
Audio files: 4
|
Posted: Sun Mar 13, 2011 8:19 am Post subject:
|
 |
|
| audiodef wrote: | | Along these lines, you can deliberately colour your sound with oversampling or something even more basic like pushing a 1-bit flip-flop at a very high rate. |
Oversampling can theoretically colour the sound a little bit since it involves filtering which is non ideal. However, I don't think it's generally really audible to mere humans. Of course, oversampling is often used to run an algorithm or AD/DA converter at a higher rate to avoid colouring the sound or getting aliasing and so on.
I'm not sure what you mean by "pushing a 1-bit flip-flop at a very high rate".
And no, not OT at all - firstly because it just isn't and secondly because this is schmooze
DJ
-- |
|
|
Back to top
|
|
 |
1:11

Joined: Oct 15, 2010 Posts: 468 Location: The Desert
Audio files: 109
|
|
|
Back to top
|
|
 |
DrJustice

Joined: Sep 13, 2004 Posts: 2112 Location: Morokulien
Audio files: 4
|
Posted: Sun Mar 13, 2011 11:54 am Post subject:
|
 |
|
^ That's a pretty high rate of flip-flops right there
DJ
-- |
|
|
Back to top
|
|
 |
Antimon

Joined: Jan 18, 2005 Posts: 4145 Location: Sweden
Audio files: 373
G2 patch files: 100
|
Posted: Sun Mar 13, 2011 12:15 pm Post subject:
|
 |
|
A reflection: when someone stops by a desk at work and starts by saying "Quick question", the following conversation will usually not be the briefest during that day.
I wish those conversations were as interesting as this thread though!  _________________ Antimon's Window
@soundcloud @Flattr home - you can't explain music |
|
|
Back to top
|
|
 |
Acoustic Interloper

Joined: Jul 07, 2007 Posts: 2082 Location: Berks County, PA
Audio files: 90
|
Posted: Sun Mar 13, 2011 8:25 pm Post subject:
|
 |
|
| Antimon wrote: | A reflection: when someone stops by a desk at work and starts by saying "Quick question", the following conversation will usually not be the briefest during that day.  |
Quick question: Why do you think that is? _________________ When the stream is deep
my wild little dog frolics,
when shallow, she drinks. |
|
|
Back to top
|
|
 |
Antimon

Joined: Jan 18, 2005 Posts: 4145 Location: Sweden
Audio files: 373
G2 patch files: 100
|
Posted: Mon Mar 14, 2011 5:07 am Post subject:
|
 |
|
| Acoustic Interloper wrote: | | Antimon wrote: | A reflection: when someone stops by a desk at work and starts by saying "Quick question", the following conversation will usually not be the briefest during that day.  |
Quick question: Why do you think that is? |
Sociology probably. It's all sociology's fault.
Let me stress again - this is a great thread! Please keep posting these quick questions. I think that the discussion that starts after quick questions at work are usually beneficial too.
Maybe "quick question" is often short for "here's something that I think is poorly defined, help me clarify". The misconception here is that definitions are easily dealt with, when in fact sometimes the major part of work that you spend on a problem has to do with defining what the problem is.
/Stefan _________________ Antimon's Window
@soundcloud @Flattr home - you can't explain music |
|
|
Back to top
|
|
 |
1:11

Joined: Oct 15, 2010 Posts: 468 Location: The Desert
Audio files: 109
|
Posted: Mon Mar 14, 2011 10:59 am Post subject:
|
 |
|
LOL!!
About 10 minutes after posting, i was thinking about that....not so quick...like:
Quick question- What is the meaning of Life? hehe
It IS a cool thread...i've read the Doctors response about 10 times already, lots of good info to know!!
i did have some more "quick questions" in the edit on my response above.
edit:duh, missed the Dr.s Response...sheesh.
The problem i'm having is that WAV's mixed from another DAW/Comp and are import/exported to a different system seem to sound different while in my main DAW (compared to playing wavs straight).
i'm using Live, and even when all sound processesing tools are off, there seems to be coloration. And when i export to somewhere else, it seems different.
Audio, hope you weren't offended by my post, was just being a jokester. Care to expand on the intentional colorization? |
|
|
Back to top
|
|
 |
1:11

Joined: Oct 15, 2010 Posts: 468 Location: The Desert
Audio files: 109
|
Posted: Mon Mar 14, 2011 11:59 am Post subject:
|
 |
|
Ok...so i need to look into all the settings then.
i DO use a large handful of tracks, but the percieved color happens when soloing as well.
On proprietary algorithms, namely iTunes/Apple,
Are they possibly using an alg that makes the signal sound good on their products?
Like, if i am using iTunes to convert into MP3, is their processing set up to make it sound the best on Apple products?
Going through my settings with a fine tooth comb.... |
|
|
Back to top
|
|
 |
Antimon

Joined: Jan 18, 2005 Posts: 4145 Location: Sweden
Audio files: 373
G2 patch files: 100
|
Posted: Mon Mar 14, 2011 12:19 pm Post subject:
|
 |
|
Live does have a bunch of fancy manipulative stuff that always seems to come on by default. I've found, especially when importing external wav files from Nanostudio, that I need to go through every sample and turn off "Warp" to get it to sound like I want it. Curiously, I have similar problems on the Octatrack (reported a bug on it), seems like a general problem with these time-shift-centric devices... _________________ Antimon's Window
@soundcloud @Flattr home - you can't explain music |
|
|
Back to top
|
|
 |
jksuperstar

Joined: Aug 20, 2004 Posts: 2503 Location: Denver
Audio files: 1
G2 patch files: 18
|
Posted: Mon Mar 14, 2011 12:29 pm Post subject:
|
 |
|
I doubt Apple is doing that, since I don't think the iPod sounds all that good anyway. You're ear buds would probably have a larger impact than any tweaking that they can do.
One thing to note about DSPs vs. general purpose (GP) computing (intel, AMD, etc). The GP processors typically adhere to IEEE-754, a standard that says for example "when you do a floating point add of two numbers, convert them into fixed point *this* way, then add them, then normalize to something close to 1.000 x 10^x as possible to make the final answer available as another floating point number". All that converting back & forth...for each operation, can add up in the long run.
DSPs however are typically hand-coded to optimize out all those conversions, as the DSPs allow many more operations like: "For adding 16 floating point numbers, convert them all to fixed, then add them all, then normalize". A GP computer would more probably add two at a time, with conversions to fixed/float/normalizing happening for each group of two. |
|
|
Back to top
|
|
 |
Acoustic Interloper

Joined: Jul 07, 2007 Posts: 2082 Location: Berks County, PA
Audio files: 90
|
Posted: Mon Mar 14, 2011 3:32 pm Post subject:
|
 |
|
| Antimon wrote: | | Live does have a bunch of fancy manipulative stuff that always seems to come on by default. I've found, especially when importing external wav files from Nanostudio, that I need to go through every sample and turn off "Warp" to get it to sound like I want it. Curiously, I have similar problems on the Octatrack (reported a bug on it), seems like a general problem with these time-shift-centric devices... |
Live seems to do some compression on its mixing, too. It could be auditory masking -- I have never taken a look with a scope to decide -- but in listening to recorded output from Live, either it is compressing some low-end signals and then expanding them when more dominant signals leave the mix, or my ears/head are doing it via masking. _________________ When the stream is deep
my wild little dog frolics,
when shallow, she drinks. |
|
|
Back to top
|
|
 |
DrJustice

Joined: Sep 13, 2004 Posts: 2112 Location: Morokulien
Audio files: 4
|
Posted: Mon Mar 14, 2011 4:01 pm Post subject:
|
 |
|
| Acoustic Interloper wrote: |
Live seems to do some compression on its mixing, too. |
I know some synths do that to avoid clipping when several voices or parts are playing. It could makes sense for Live too, when it's used for live performance, but generally, In the case of DAW's, I think they should be upfront about it and maybe include an option to switch of such things. IMO a DAW should not really have a sound, but be fully transparent unless you specifically want your audio processed.
DJ
-- |
|
|
Back to top
|
|
 |
Antimon

Joined: Jan 18, 2005 Posts: 4145 Location: Sweden
Audio files: 373
G2 patch files: 100
|
Posted: Tue Mar 15, 2011 4:31 am Post subject:
|
 |
|
Now that you mention it, I often see that the master meter in Live goes up in the red without any obvious audible effect, which doesn't correspond with my idea of digital clipping. If I put a lot of boost on the signal it will clip though.
I'm pretty sure that this isn't because Live leaves a lot of headroom, because the volume is what I'd expect it to be in these cases, close to max. So some kind of compression most likely is taking place.
/Stefan _________________ Antimon's Window
@soundcloud @Flattr home - you can't explain music |
|
|
Back to top
|
|
 |
Acoustic Interloper

Joined: Jul 07, 2007 Posts: 2082 Location: Berks County, PA
Audio files: 90
|
Posted: Tue Mar 15, 2011 5:05 am Post subject:
|
 |
|
It works to good effect with my duets with algorithmic improvisation such as finger picking guitar or banjo atop scrabble-to-midi. When I lean in, my algorithmic partner falls back somewhat, and when I lighten up on my picking, the generated voices get louder, seeming actually to take the lead. There is seeming interplay between my playing and the machine's. After noticing this on a webcast recording one time, I started using compressors intensionally to accent the effect. So, now I have no idea whether it was Live or my perception, but it's a useful approach. _________________ When the stream is deep
my wild little dog frolics,
when shallow, she drinks. |
|
|
Back to top
|
|
 |
1:11

Joined: Oct 15, 2010 Posts: 468 Location: The Desert
Audio files: 109
|
Posted: Tue Mar 15, 2011 10:15 am Post subject:
|
 |
|
It's looking like you guys are picking up the same type of coloration i'm experiencing.
Just as Interloper mentioned, haven't used a scope or other analysis tools on the audio yet, but i'm thinkin it'd be a good idea. Cuz i might have to switch DAW's for creating...
It does kinda sound like masking or a touch of compression.
The meters in Live have always been a mystery to me. Cuz you're right Antimon, you don't hear clipping when it's tapping in the red.
But i'm thinking when you have it in the green, it's lower volume than other DAW's.
When you listen to any of my tracks, they all have a lower volume than other peoples tracks on the radio...yet when mixing down, the meters slightly bump the reds.
Feels like might be approaching the cause of the perceived coloration, RAD!!
[/code] |
|
|
Back to top
|
|
 |
Acoustic Interloper

Joined: Jul 07, 2007 Posts: 2082 Location: Berks County, PA
Audio files: 90
|
Posted: Tue Mar 15, 2011 4:03 pm Post subject:
|
 |
|
I use Live as an instrument. I use Live effects fairly heavily, so to me its "colorization" is like any other instrument, e.g., a good guitar. It's not just that I don't "mind" it, I "use" it. It's a very powerful instrument.
When I want a clean recording, I ask my son to do it with ProTools LE. I usually don't bother much, because limited time being what it is, this is all pretty ephemeral stuff for me. I let my streaming program capture a WAV or MP3 when I am contributing to a webcast. I am usually satisfied with that. The clinkers in my finger picking are a bigger source of aggravation for me. _________________ When the stream is deep
my wild little dog frolics,
when shallow, she drinks. |
|
|
Back to top
|
|
 |
1:11

Joined: Oct 15, 2010 Posts: 468 Location: The Desert
Audio files: 109
|
Posted: Fri Mar 18, 2011 1:17 pm Post subject:
|
 |
|
YES, Live IS an Instrument.
i couldn't do what i do with out it!!
i suppose that'd be a better way of looking at it.
Think i'm gonna use something else for compiling tracks, but i'll probably always use Live for performances and glitching stuff out! |
|
|
Back to top
|
|
 |
1:11

Joined: Oct 15, 2010 Posts: 468 Location: The Desert
Audio files: 109
|
Posted: Fri Mar 18, 2011 2:41 pm Post subject:
|
 |
|
Another Question for the Experts!!
It was said that a large number of tracks can possibly effect performance processing...SO....
What about using a large number of individual samples on each track?
i use anywhere from 10-20 individual samples on each track, on about 10-16 tracks. So sometimes sample #'s are triple digit.
Last night i noticed when processing video along with audio, dips in processing quality. Especially with a large # of video clips....so maybe there's a limit to just audio too......
Is this the reason you get an out of the box DSP? |
|
|
Back to top
|
|
 |
DrJustice

Joined: Sep 13, 2004 Posts: 2112 Location: Morokulien
Audio files: 4
|
Posted: Sat Mar 19, 2011 9:46 am Post subject:
|
 |
|
| 1:11 wrote: | | What about using a large number of individual samples on each track? |
If you're using a polyphonic or multi timbral instrument on a track, then its currently sounding voices or parts must be mixed down to be presented to the track as a single signal. Normally this would not be a problem for the audio quality, unless the total number of simultaneously playing voices for all tracks is very large (perhaps hundreds). That is, as long as only pure mixing is taking place.
| Quote: | | Last night i noticed when processing video along with audio, dips in processing quality. Especially with a large # of video clips....so maybe there's a limit to just audio too...... |
Is this using Live? How is the drop in quality manifesting itself? Usually, if the number of voices and tracks is too large for the processor, you get drop outs as the audio can't be processed in time - but exceeding the capacity does not normally result in colouration as such. Note that many very tiny drop outs can sound a bit like crackling distortion.
| Quote: | | Is this the reason you get an out of the box DSP? |
Well, yes and no... Previously mentioned systems like Pro Tools, and also UAD and the like, use extra DSPs both to ensure quality and capacity. But unless you get drop outs, your system should have the capacity for the work it is doing. If written to do so, software can potentially take action resulting in loss of quality as the capacity of the system is nearing exhaustion. I don't know if/which software does that - maybe other members have some info on it.
DJ
-- Last edited by DrJustice on Sat Mar 19, 2011 11:28 am; edited 1 time in total |
|
|
Back to top
|
|
 |
jksuperstar

Joined: Aug 20, 2004 Posts: 2503 Location: Denver
Audio files: 1
G2 patch files: 18
|
Posted: Sat Mar 19, 2011 11:08 am Post subject:
|
 |
|
| I only recommend this because I replaced my studio with one recently: The Sonic|Core SCOPE XITE. It's a DSP box, full of 18 DSPs, has a really great collection of synths and effects, as well as what I would call bast-in-class mixing. It can handle 64 ASIO channels, so I send each track from Ableton to the XITE for mixing in real time, and bypass ableton's mixer altogether. That's the best of both worlds to me. I'd agree Ableton seems to do things so it never sounds "bad", as it is a live instrument, and so otherwise you can get yourself into trouble...on stage. But I've never quantized it. |
|
|
Back to top
|
|
 |
|