| Author |
Message |
Guest
|
 Posted:
Tue Nov 08, 2005 7:06 pm Post subject:
dac clock rates |
|
|
Hi there,
I'm currently researching CD/MP3 mixing devices which allow the user
(typically a DJ) to alter the speed/pitch of the track being played to
allow them to beat match the song with another.
Does anyone know of any audio DACs which allow the output clock rate
(typically 44.1KHz) to be changed in real-time to allow the speed of
the track being output to be altered? I know this will not affect the
pitch as this is more of a DSP task.
I'm trying to design a device which will allow the clock rate to be
changed in real-time so a DAC which supports this would be excellent!
Any help/info/advice appreciated!
Thanks, Alex
|
|
| Back to top |
|
 |
Richard Crowley
Guest
|
| Posted:
Tue Nov 08, 2005 7:28 pm Post subject:
Re: dac clock rates |
|
|
<dancedynamix@hotmail.com> wrote ...
| Quote: | Hi there,
I'm currently researching CD/MP3 mixing devices which
allow the user (typically a DJ) to alter the speed/pitch of the
track being played to allow them to beat match the song
with another.
Does anyone know of any audio DACs which allow the
output clock rate (typically 44.1KHz) to be changed in
real-time to allow the speed of the track being output to
be altered?
|
Which ones *don't* allow you to do that?
Changing the output clock rate is likely not the usual way
of doing that since it presents other problems.
| Quote: | I know this will not affect the pitch as this is more of a
DSP task.
|
Of course it will affect the pitch. That is why you need
the DSP computing.
| Quote: | I'm trying to design a device which will allow the clock
rate to be changed in real-time so a DAC which supports
this would be excellent!
|
Good luck. Might be instructive to research why people
with more resources than you declined to pursue that method. |
|
| Back to top |
|
|
mr_alex
Guest
|
| Posted:
Tue Nov 08, 2005 7:37 pm Post subject:
Re: dac clock rates |
|
|
Ok, thanks for the reply. Can you give me more insight into the
problems that can occur with just changing the rate at which it is
being played back?
The range would only go from roughly 40KHz upto 50KHz to get a +/- 10%
speed range (which is common for most turntables and CD mixers). I
didn't think this would affect the pitch as it's only the gap between
the samples which is being changed? Pitch shifting is done via DSP and
formant shifting etc...
| Quote: | Good luck. Might be instructive to research why people
with more resources than you declined to pursue that method.
|
I am trying to conduct this research but I find nothing on the WWW that
is why I have started to use the newsgroups. Do you have links to this
information?
Cheers, Alex |
|
| Back to top |
|
|
Guest
|
| Posted:
Tue Nov 08, 2005 7:43 pm Post subject:
Re: dac clock rates |
|
|
dancedynamix@hotmail.com wrote:
| Quote: | Hi there,
I'm currently researching CD/MP3 mixing devices which allow the user
(typically a DJ) to alter the speed/pitch of the track being played to
allow them to beat match the song with another.
Does anyone know of any audio DACs which allow the output clock rate
(typically 44.1KHz) to be changed in real-time to allow the speed of
the track being output to be altered? I know this will not affect the
pitch as this is more of a DSP task.
|
Do you want to bet it won't?
The pitch will most certainly be affected, directly in proportion to
the
change in the clock frequency.
| Quote: | I'm trying to design a device which will allow the clock rate to be
changed in real-time so a DAC which supports this would be excellent!
|
And that WILL change the pitch
Varispeed is, in fact, a very tough DSP task that DOES NOT
involve using a variable clock frequency. You need a seriously
sophisticated algorithm to do this without seriously screwing
up the audio in other respects.
No one who does high-quality, constant-pitch varispeed
does it with a variable DAC clock. No one. Why? Because it
can't work. |
|
| Back to top |
|
|
mr_alex
Guest
|
| Posted:
Tue Nov 08, 2005 7:54 pm Post subject:
Re: dac clock rates |
|
|
I used to experiment with my sound card where I could change the clock
from 44.1KHz to 48KHz and all that would happen was that when I played
a 44.1KHz wav it would sound faster (i.e beats quicker) but the pitch
(i.e. key) would remain the same. So i reckon the pitch wont be
affected only the speed - this will suit DJ's as the tracks will remain
in the key they were orginally recorded.
| Quote: | You need a seriously
sophisticated algorithm to do this without seriously screwing
up the audio in other respects.
|
Can you be more specific as to how the audio will be affected?
Cheers, Alex |
|
| Back to top |
|
|
Guest
|
| Posted:
Tue Nov 08, 2005 8:05 pm Post subject:
Re: dac clock rates |
|
|
mr_alex wrote:
| Quote: | Ok, thanks for the reply. Can you give me more insight into the
problems that can occur with just changing the rate at which it is
being played back?
The range would only go from roughly 40KHz upto 50KHz to get a +/- 10%
speed range (which is common for most turntables and CD mixers). I
didn't think this would affect the pitch as it's only the gap between
the samples which is being changed?
|
It most certainly will. Consider the following:
Let's start with a tone at middle a at 440 Hz. Each cycle of that tone
completes in 1/440 or about 2.27 mSec.
Now, let's assume that we've recorded that tone on a CD at a sample
rate of 44.1 kHz. That means that each cycle of this 440 Hz tone takes
up almost exactly 100 samples (100.227... to be precise).
That means that everytime your DAC puts out 100 samples, it puts
out a complete 440 cycle. Fine.
Now, let's do what you want to do: change "only the gap between
the samples." Let's increase the sample rate by 10% to 48.5 kHz.
That means the samples, instead of being 22.676 uSec apart,
are now only 20.61 uSec apart, about 10% less.
Now, remember that our 440 tone required 100 samples for
each cycle. At the faster sample rate, you STILL have only 100
samples, but now, instead of taking 2.27 mSec per cycle, it takes
2.06 mSec. And a tone whose cycle is 2.06 mSec long has a
frequency of 484.6 Hz.
484.6 Hz is not 440 Hz, I hope you agree. It's 10% higher. Why?
Because you increased your clock rate by 10%.
Time is time, and you only have one time scale: the time scale
that you're fiddling with to change the DAC clock is the SAME
time scale that affects pitch and the SAME time scale that effects
tempo. IF all your doing is changing that toime scale, you will
simultaneously affect the DAC clock, the pitch AND the tempo.
| Quote: | Pitch shifting is done via DSP and formant shifting etc...
|
Yes, it is. And since change the DAC clock also changes the pitch,
you need pretty good DSP to do pitch-contant tempo changes, and
it DOES NOT affect the DAC clock speed.
| Quote: | I am trying to conduct this research but I find nothing on the WWW that
is why I have started to use the newsgroups.
|
Probably because all the current work on high-quality time, tempo
and pitch shifting (they're very closely related) is leading-edge
proprietary development in cmmercial products.
If you're willing to negotiate a licensing agreement, I can put you
in touch with some of the true experts in this field. But be prepared
to fork over a sufficient chunck of cash to pay for quite a few
man-years
of fairly esoteric DSP. |
|
| Back to top |
|
|
Richard Crowley
Guest
|
| Posted:
Tue Nov 08, 2005 8:19 pm Post subject:
Re: dac clock rates |
|
|
"mr_alex" wrote ...
| Quote: | Ok, thanks for the reply. Can you give me more insight into the
problems that can occur with just changing the rate at which it is
being played back?
The range would only go from roughly 40KHz upto 50KHz to get a +/- 10%
speed range (which is common for most turntables and CD mixers). I
didn't think this would affect the pitch as it's only the gap between
the samples which is being changed? Pitch shifting is done via DSP and
formant shifting etc...
|
You have some significant learning to accomplish.
Changing the clock rate always changes the pitch.
That is why you need DSP to counteract the effect.
| Quote: | Good luck. Might be instructive to research why people
with more resources than you declined to pursue that method.
I am trying to conduct this research but I find nothing on the WWW
that
is why I have started to use the newsgroups. Do you have links to this
information?
|
There are several chip vendors selling DSP chips who
likely have tutorials, sample code, etc. available. You
are investigating a rather sophisticated area of technology,
the information isn't going to fall into your lap.
You could pursue a post-graduate degree in mathematics
and take it on as a dissertation project. |
|
| Back to top |
|
|
Richard Crowley
Guest
|
| Posted:
Tue Nov 08, 2005 8:29 pm Post subject:
Re: dac clock rates |
|
|
"mr_alex" wrote ...
| Quote: | I used to experiment with my sound card where I could change the clock
from 44.1KHz to 48KHz and all that would happen was that when I played
a 44.1KHz wav it would sound faster (i.e beats quicker) but the pitch
(i.e. key) would remain the same. So i reckon the pitch wont be
affected only the speed - this will suit DJ's as the tracks will
remain
in the key they were orginally recorded.
|
Your perception of key/pitch is faulty if you think it was not
affected by changing the clock rate. Or else your sound card
was automatically doing the DSP behind the scenes. Note
that SoundBlaster cards are notorious for pulling tricks like
this. |
|
| Back to top |
|
|
mr_alex
Guest
|
| Posted:
Tue Nov 08, 2005 9:28 pm Post subject:
Re: dac clock rates |
|
|
Thanks for the responses - my understanding has greatly increased from
all your posts!
I've heard of a method of dropping samples (decimating) to decrease the
sample rate and then inserting samples with the average value between
the two to increase the sample rate.
Would this be an effective method? How much DSP processing power would
this require to be done in realtime?
Alex. |
|
| Back to top |
|
|
Guest
|
| Posted:
Tue Nov 08, 2005 10:15 pm Post subject:
Re: dac clock rates |
|
|
mr_alex wrote:
| Quote: | Thanks for the responses - my understanding has greatly increased from
all your posts!
I've heard of a method of dropping samples (decimating) to decrease the
sample rate and then inserting samples with the average value between
the two to increase the sample rate.
|
You're confusing several concepts here. "Decimation" is a means of
reducing the sampling rate. It's not just dropping samples., which by
itself will cause alisaing distortion to occur. Proper decimation
usually
involves first applying an anti-aias filter to the stream, the cutoff
of that
filter must be less than half the final sampling rate. Only after
filtering
can you then decimate. Sometimes the process first incolves over-
sampling before filtering before decimation. But all this requires DSP
to get it to work.
| Quote: | Would this be an effective method?
|
No, you're simply changing the sampling rate. Doing it this way will
keep the pitch contant, but it will also keep the tempo the same as
well.
If your filter and decimate by two, you'll have have the samples, but
you'll be playing them half as often, the result being exactly the same
pitch and tempo that you started with. If you then go back to the
original
smaple rate, you'll end up with something that's twice the tempo AND
twice the pitch.
One way that it CAN be done is to analyze the music over a time window
that's long enough so that pitch information is averaged out, but
narrow
enough that tempm information remains. The analysis determines
whether there are repetitive waveforms in that interval. If there are,
then
some repetitions of the WAVEFORM are splice out (to increase tempo)
or repeated (to slow tempo). The window then moves a little bit to the
next interval and does it again.
| Quote: | How much DSP processing power would this require to be done in
realtime?
|
That's not the really hard question, as it can be done on a 400 MHz-
class Pentium. The REALLY hard question is how much DSP expertise
does it take, and the answer to that is so much, that I think there are
probably no more than a dozen people on the planet that have been
able to implement almost totally sonically transparent real-time vari-
speed that's dynamically changeable.
One example of a product where it is actually done in real time can
be found on 25-seven.com. It is, in essence, a real-time varispeed
algorithm with a pretty sophisticated control program on top of it.
At variations of +-15%, it's very difficult to detect the change
beyond
the tempo change. It's targetted towards a different market than what
you're talking about, but the underlying pronciple is the same. The
horsepower comes as much from the quality of the algorithm and thus
the expertise of the implementors as it does from the hardware on
which it runs.
The bottom line being that simply buying enough DSP horsepower
and reading a couple of web links is not going to swell the ranks of
the experts in high-quality continuous real-time varispeed. You have,
with all due respoct, and ENORMOUS education in front of you, measured,
quite literally, in many years. |
|
| Back to top |
|
|
Codifus
Guest
|
| Posted:
Tue Nov 08, 2005 11:22 pm Post subject:
Re: dac clock rates |
|
|
mr_alex wrote:
| Quote: | Thanks for the responses - my understanding has greatly increased from
all your posts!
I've heard of a method of dropping samples (decimating) to decrease the
sample rate and then inserting samples with the average value between
the two to increase the sample rate.
Would this be an effective method? How much DSP processing power would
this require to be done in realtime?
Alex.
Isn't this topic kind of like the re-invention of the wheel? Haven't |
existing CD DJ tunrtables, the most famous being the Pioneer CDJ-1000
and the Technics SL DZ 1200, been doing this for some time now?
CD |
|
| Back to top |
|
|
mr_alex
Guest
|
| Posted:
Tue Nov 08, 2005 11:34 pm Post subject:
Re: dac clock rates |
|
|
It is, I just am trying to understand how it is done and what the
easiest way of implementing it is. |
|
| Back to top |
|
|
Guest
|
| Posted:
Wed Nov 09, 2005 1:42 am Post subject:
Re: dac clock rates |
|
|
mr_alex wrote:
| Quote: | It is, I just am trying to understand how it is done and what the
easiest way of implementing it is.
|
To do it right is far from easy. |
|
| Back to top |
|
|
mr_alex
Guest
|
| Posted:
Wed Nov 09, 2005 2:08 am Post subject:
Re: dac clock rates |
|
|
So, with the method of altering the clock speed of the DAC what are the
actual problems with the audio? I've heard of time-stretching problems
when the difference is as much as 100% but when the range is only
+/-10% how can this affect the audio?
You spoke of aliasing problems when decimation occurs and I understand
this (it must sound horrible!) but what are the type of errors that
occur with altering just the clock rate?
Thanks, Alex |
|
| Back to top |
|
|
Guest
|
| Posted:
Wed Nov 09, 2005 3:02 am Post subject:
Re: dac clock rates |
|
|
mr_alex wrote:
| Quote: | So, with the method of altering the clock speed of the DAC what are the
actual problems with the audio?
|
The pmajor problem is that it just does not work. Changing the DAC
clock
speed WILL change BOTH the tempo AND the pitch. If you want to change
one and not the other, chnaging the DAC clock speed WILL NOT WORK.
Despite your experience to the contrary, IT WON'T WORK.
| Quote: | I've heard of time-stretching problems
when the difference is as much as 100% but when the range is only
+/-10% how can this affect the audio?
|
Well, if all you are doing is changing the DAC clock speed, it will
affect
the audio by changing the pitch as well as the tempo.
YOU CANNOT SEPARATE ONE FROM THE OTHER BY CHANGING THE
CLOCK SPEED. Do you understand that?
| Quote: | You spoke of aliasing problems when decimation occurs and I understand
this (it must sound horrible!) but what are the type of errors that
occur with altering just the clock rate?
|
Sigh, let's try this again:
If you change the DAC clock speed, YOUR WILL CHANGE THE TEMPO
AND THE PITCH AT THE SAME TIME BY EXACTLY THE SAME AMOUNT.
If you increase the DAC clock speed by 10%, the tempo will increase
by 10% AND the pitch will increase by 10%. If you double the DAC clock
speed, the tempo will double AND the pitch will double.
Please LOSE the idea that you can change tempo and not pitch by
changing the DAC clock rate. YOU CAN'T.
Okay? |
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in