23 Apr
2005
23 Apr
'05
1:02 a.m.
In a nutshell, my issue is trying to slow down recorded audio and
maintain decent sound quality.
Here's the extended explanation:
A number of years ago I recorded some of my live performances on a
stereo 7-inch reel-to-reel deck. Just recently I borrowed a friend's
Fostex Model 80, which is an 8-track deck. I am able to pull off both
channels of my music, but the Fostex plays it *way* too fast and does
not have any speed adjustment that I can see.
What I have done so far is use Ardour with JACK running at 48KHz to pull
off the music. Then, I exported it to a WAV and used AlsaPlayer to
playback the audio at approx 25% normal speed. My problem is that the
quality is quite poor. Sure, the original recordings are less than
stellar to begin with, but my results are less than expected. Note that
I am using an M-Audio Delta 44 to record through, so soundcard quality
should not be an issue.
I'm looking for some suggestions as to how to go about doing this the
proper way.
1. Is my approach reasonable? Record with Ardour and then slow it down
with something else?
2. AlsaPlayer plays back my audio, but how do I permanently change the
speed and save it?
3. Would a higher sample rate in JACK help the quality?
4. Anyone have a Fostex Model 80 manual that tells how to slow down the
speed to begin with?
Alan
23 Apr
23 Apr
3:37 a.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to slow them down
Alan Taylor wrote:
What I have done so far is use Ardour with JACK
running at 48KHz to pull
off the music. Then, I exported it to a WAV and used AlsaPlayer to
playback the audio at approx 25% normal speed. My problem is that the
quality is quite poor.
The most obvious reason I can see for this is that the effective sample
rate is reduced to 12kHz via this method.
If you can increase your soundcard sample frequency, this should yield a
dramatic improvement (96kHz sample rate becomes effective 24kHz).
Beyond that, I suspect you need to find a way to slow that tape :)
HTH
Rob
6:36 p.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to slow them down
On Sat, 2005-04-23 at 08:44 +0100, Rob Fell wrote:
The most obvious reason I can see for this is that the
effective sample
rate is reduced to 12kHz via this method.
If you can increase your soundcard sample frequency, this should yield a
dramatic improvement (96kHz sample rate becomes effective 24kHz).
That makes sense, but I seem to be having trouble getting JACK to run at
anything above 48KHz. When I start it at 88.2KHz or 96KHz, I
immediately get messages such as "XRUN callback (xx skipped)", where xx
is a number between 50 and 300, depending on which sample rate and
frames/period I choose. I get at least 50 skipped callbacks per second
without doing anything but starting jackd.
If I am not mistaken, the Delta 44 should be able to handle at least up
to 96KHz. I believe my kernel is set up properly since I can record 6
or 7 tracks at 48KHz and 512 frames/period in RT mode for an extended
period of time without any xruns at all.
Any suggestions?
Alan
7:29 p.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to slow them down
Alan Taylor <alan-linuxaudiouser(a)lowb.org> writes:
On Sat, 2005-04-23 at 08:44 +0100, Rob Fell wrote:
When you double the sample rate you generally need to double the
buffer size. So, -r96000 -p1024 would be equivalent to -r48000 -p512.
Your latency is the same, either way.
--
joq
The most obvious reason I can see for this is
that the effective sample
rate is reduced to 12kHz via this method.
If you can increase your soundcard sample frequency, this should yield a
dramatic improvement (96kHz sample rate becomes effective 24kHz).
That makes sense, but I seem to be having trouble getting JACK to run at
anything above 48KHz. When I start it at 88.2KHz or 96KHz, I
immediately get messages such as "XRUN callback (xx skipped)", where xx
is a number between 50 and 300, depending on which sample rate and
frames/period I choose. I get at least 50 skipped callbacks per second
without doing anything but starting jackd.
If I am not mistaken, the Delta 44 should be able to handle at least up
to 96KHz. I believe my kernel is set up properly since I can record 6
or 7 tracks at 48KHz and 512 frames/period in RT mode for an extended
period of time without any xruns at all.
Any suggestions?
8:02 p.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to slow them down
On Sat, 2005-04-23 at 18:32 -0500, Jack O'Quin wrote:
When you double the sample rate you generally need to
double the
buffer size. So, -r96000 -p1024 would be equivalent to -r48000 -p512.
Your latency is the same, either way.
No matter what frames/period I choose, my results are the same... lots
of XRUN callbacks skipped. I've tried 512, 1024, and 2048. It won't
let me set 4096 at all.
Alan
5:08 a.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to slow them down
Hi Alan!
What I have done so far is use Ardour with JACK
running at 48KHz to pull
off the music. Then, I exported it to a WAV and used AlsaPlayer to
playback the audio at approx 25% normal speed.
I think the optimal solution concerning quality was to record at 192 kHz
and process the resulting file at 48 kHz...
Ciao,
HippiE
24 Apr
24 Apr
1:17 p.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to slow them down
Alan Taylor <alan-linuxaudiouser(a)lowb.org>rg>:
In a nutshell, my issue is trying to slow down
recorded audio and
maintain decent sound quality.
1. Is my approach reasonable? Record with Ardour and then slow it
down with something else?
Alsaplayer sounds crap in varispeed. I'd use Ardour for slowing things
down as it uses libsamplerate for that purpose. In the Mixer strip,
just above the pre-fader plugins, it says "1" which is for playback
speed of that track. [Shift-] Click that to change it. You can then
export to wav or whatever.
3. Would a higher sample rate in JACK help the
quality?
Yes. But I take it from other messages in this thread that you have
trouble running jack with higher rates. No clue what's in the way.
Wolfgang
3:15 p.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to slow them down
On Sun, 2005-04-24 at 19:16 +0200, Wolfgang Woehl wrote:
Alsaplayer sounds crap in varispeed. I'd use
Ardour for slowing things
down as it uses libsamplerate for that purpose. In the Mixer strip,
You are correct. While the results are still not ideal (due to the
original sample rate of 48KHz), the overall quality surpasses what I got
out of alsaplayer. Thanks.
Yes. But I take it from other messages in this thread
that you have
trouble running jack with higher rates. No clue what's in the way.
Further inspection of my IRQs shows that my Delta 44 shares with my
ethernet controller and my USB controller. My next step is to move my
card around and try to get a better IRQ.
Thanks again for your help.
Alan
4:52 p.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to slow them down
Further inspection of my IRQs shows that my Delta 44
shares with my
ethernet controller and my USB controller. My next step is to move
my card around and try to get a better IRQ.
irq 10 would be a good choice. Usually it's the first available high
priority interrupt.
Wolfgang
6:22 p.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to slow them down
On Sun, 2005-04-24 at 22:51 +0200, Wolfgang Woehl wrote:
irq 10 would be a good choice. Usually it's the
first available high
priority interrupt.
I tried all 4 empty slots in my PC and was only able to get IRQ5 or
IRQ11. IRQ5 is where I was before... sharing with the video card, the
gigabit ethernet controller, and the USB v2 controller.
Now I'm in IRQ11, sharing only with the USB v1.1 controller. Instead of
300 XRUNs per second at 96KHz, I get only one. Also, instead of 50
XRUNs per second at 88.2KHz, I get only one every 30 seconds.
So, there's marked improvement, but still not good enough to pull off a
whole reel of music. I don't know why IRQ10 seems to be left unused
altogether. I figured my PC would have assigned an empty IRQ before
making my soundcard share an already used IRQ. Time for more research.
Alan
10:16 p.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to
IOf this is a one-time deal I would disable UBS in the bios and rip the
tracks....
but that's just me....
On Sun, 2005-04-24 at 22:51 +0200, Wolfgang Woehl
wrote:
--
When asked the definition of "pi":
The Mathematician:
Pi is the number expressing the relationship between the
circumference of a circle and its diameter.
The Physicist:
Pi is 3.1415927, plus or minus 0.000000005.
The Engineer:
Pi is about 3.
irq 10 would be a good choice. Usually it's
the first available high
priority interrupt.
I tried all 4 empty slots in my PC and was only able to get IRQ5 or
IRQ11. IRQ5 is where I was before... sharing with the video card, the
gigabit ethernet controller, and the USB v2 controller.
Now I'm in IRQ11, sharing only with the USB v1.1 controller. Instead of
300 XRUNs per second at 96KHz, I get only one. Also, instead of 50
XRUNs per second at 88.2KHz, I get only one every 30 seconds.
So, there's marked improvement, but still not good enough to pull off a
whole reel of music. I don't know why IRQ10 seems to be left unused
altogether. I figured my PC would have assigned an empty IRQ before
making my soundcard share an already used IRQ. Time for more research.
Alan
10:47 p.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to
On Sun, 2005-04-24 at 22:12 -0400, timg(a)expressmart.com wrote:
IOf this is a one-time deal I would disable UBS in the
bios and rip the
tracks....
but that's just me....
Yeah, you make a good point. This is a one time deal. I've got about
10 reels to pull off and after I'm done, then it's all over.
Unfortunately my mouse and keyboard operate off of that particular USB
controller. Maybe I'll try to temporarily change to PS/2 devices just
to get this done. Thanks for the idea.
Alan
25 Apr
25 Apr
11:05 a.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to
On Sun, 2005-04-24 at 22:12 -0400,
timg(a)expressmart.com wrote:
That's what I would do....
all things being equal... blah the simplesst soloution ... blah smoething
like that....
;)
IOf this is a one-time deal I would disable UBS
in the bios and rip the
tracks....
but that's just me....
Yeah, you make a good point. This is a one time deal. I've got about
10 reels to pull off and after I'm done, then it's all over.
Unfortunately my mouse and keyboard operate off of that particular USB
controller. Maybe I'll try to temporarily change to PS/2 devices just
to get this done. Thanks for the idea.
Alan
3:11 a.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to slow them down
Alan Taylor <alan-linuxaudiouser(a)lowb.org>rg>:
On Sun, 2005-04-24 at 22:51 +0200, Wolfgang Woehl
wrote:
Alan, if you have APIC on your board you could shuffle things around in
the BIOS. (I'm not sure, I think on APIC boards the irq 10 rule does
not apply.)
Wolfgang
irq 10 would be a good choice. Usually it's
the first available
high priority interrupt.
I tried all 4 empty slots in my PC and was only able to get IRQ5 or
IRQ11. IRQ5 is where I was before... sharing with the video card,
the gigabit ethernet controller, and the USB v2 controller.
Now I'm in IRQ11, sharing only with the USB v1.1 controller. Instead
of 300 XRUNs per second at 96KHz, I get only one. Also, instead of
50 XRUNs per second at 88.2KHz, I get only one every 30 seconds.
So, there's marked improvement, but still not good enough to pull off
a whole reel of music. I don't know why IRQ10 seems to be left
unused altogether. I figured my PC would have assigned an empty IRQ
before making my soundcard share an already used IRQ. Time for more
research. 
4:02 p.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to slow them down
El dom, 24-04-2005 a las 18:22 -0400, Alan Taylor escribió:
Now I'm in IRQ11, sharing only with the USB v1.1
controller. Instead of
300 XRUNs per second at 96KHz, I get only one. Also, instead of 50
XRUNs per second at 88.2KHz, I get only one every 30 seconds.
Try disabling that USB controller using BIOS, if not used, you should
get at least a little performance boost.
So, there's marked improvement, but still not good
enough to pull off a
whole reel of music. I don't know why IRQ10 seems to be left unused
altogether. I figured my PC would have assigned an empty IRQ before
making my soundcard share an already used IRQ. Time for more research.
I don't know which is your mobo. IRQ assignments on VIA chipsets are not
much predictable.
Cordially, Ismael
5:39 p.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to slow them down
On Sun, 24 Apr 2005, Alan Taylor wrote:
On Sun, 2005-04-24 at 19:16 +0200, Wolfgang Woehl
wrote:
Actually, for the real-time varispeed, ardour does *not* use
libsamplerate, but a handcrafted linear interpolation from Steve
Harris. This is also unsuitable for high quality archival work.
If you can use ardour or alsaplayer to estimate the exact amount
of resampling you need, you can then use:
sndfile-resample -by <amount_ratio> infile.wav outfile.wav
to do a high quality resampling. You'll want to use the inverse
of the varispeed you find to get the correct resampling. The
resulting file will have a weird sample rate, but if played back
at the original sample rate of your capture will sound correct.
You can replace the samplerate field in the output wav file with
the attached python script. Run it as follows (example assumes
a 48000 original capture):
python wavsrmod.py 48000 outfile.wav
As mentioned by others, the higher capture rate you can achieve
the better quality you'll get, but depending on the source, 48k
might be good enough. Hope this helps.
jlc
Alsaplayer sounds crap in varispeed. I'd use
Ardour for slowing things
down as it uses libsamplerate for that purpose. In the Mixer strip,
You are correct. While the results are still not ideal (due to the
original sample rate of 48KHz), the overall quality surpasses what I got
out of alsaplayer. Thanks.
26 Apr
26 Apr
4:24 a.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to slow them down
Jesse Chappell <jesse(a)essej.net>et>:
On Sun, 24 Apr 2005, Alan Taylor wrote:
Jesse, thanks.
Wolfgang
On Sun, 2005-04-24 at 19:16 +0200, Wolfgang Woehl
wrote:
Actually, for the real-time varispeed, ardour does *not* use
libsamplerate, but a handcrafted linear interpolation from Steve
Harris. This is also unsuitable for high quality archival work.
If you can use ardour or alsaplayer to estimate the exact amount
of resampling you need, you can then use:
sndfile-resample -by <amount_ratio> infile.wav outfile.wav Alsaplayer sounds crap in varispeed. I'd use
Ardour for slowing
things down as it uses libsamplerate for that purpose. In the
Mixer strip,
You are correct. While the results are still not ideal (due to the
original sample rate of 48KHz), the overall quality surpasses what
I got out of alsaplayer. Thanks.
30 Apr
30 Apr
6:03 p.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to slow them down - SOLVED
On Mon, 2005-04-25 at 17:38 -0400, Jesse Chappell wrote:
If you can use ardour or alsaplayer to estimate the
exact amount
of resampling you need, you can then use:
sndfile-resample -by <amount_ratio> infile.wav outfile.wav
to do a high quality resampling. You'll want to use the inverse
of the varispeed you find to get the correct resampling. The
resulting file will have a weird sample rate, but if played back
at the original sample rate of your capture will sound correct.
You can replace the samplerate field in the output wav file with
the attached python script. Run it as follows (example assumes
a 48000 original capture):
python wavsrmod.py 48000 outfile.wav
The above method resulted in a file that still sounded way too fast.
I'm not sure what I did wrong. Knowing that I wanted to slow down my
recording to 25% of it's orginal speed, I tried using your suggestion of
the inverse "-by .750", I tried the normal representation of 25%
"-by .250", and I tried going directly to 25% speed with "-to 12000".
All resulted in a wav file that was too fast, and got even faster after
running your python script.
After messing around quite a bit, I tried doing the steps in reverse
order. I ran your python script directly on my Ardour-exported wav file
"python wavsrmod.py 12000 reel1.wav". That resulted in a 12KHz file
that was playable by mplayer at the correct sounding speed (but XMMS
doesn't like to play 12KHz files it seems). Then, I ran
sndfile-resample to change it back to 48KHz. The end result is the best
quality conversion thus far.
Thanks for all your help!
Alan
8:51 p.m.
New subject: [linux-audio-user] Pulling off old 2-track reels and trying to slow them down - SOLVED
On Sat, 30 Apr 2005, Alan Taylor wrote:
The above method resulted in a file that still sounded
way too fast.
I'm not sure what I did wrong. Knowing that I wanted to slow down my
recording to 25% of it's orginal speed, I tried using your suggestion of
the inverse "-by .750", I tried the normal representation of 25%
"-by .250", and I tried going directly to 25% speed with "-to
12000".
All resulted in a wav file that was too fast, and got even faster after
running your python script.
Actually, since you wanted the new speed to be 0.25 of the
original, the inverse of 0.25 = (1 / 0.25) = 4.0 ... so using -by
4.0 then using the python script to set the sample rate to 48000
would have also worked.
But your method is effectively the same thing (in reverse) so
the end result file is the same.
Glad I could help, but it sucks that the tape was that much
faster. Re-recording at 96k would probably really help to
maintain fidelity.
jlc
6969
days inactive
6976
days old
linux-audio-user@lists.linuxaudio.org
18 comments
8 participants
participants (8)
-
Alan Taylor -
HippiE -
Ismael Valladolid Torres -
Jack O'Quin -
Jesse Chappell -
Rob Fell -
timg＠expressmart.com -
Wolfgang Woehl