24 Sep
2005
24 Sep
'05
11:23 a.m.
I would like to build an application to assist in setting the "Beat" of
a pendulum type clock. This is a tool used by a clock repair person to
set a clock to keep accurate time. It is basically a low frequency
oscilloscope with some specialized functionality.
A microphone is placed inside the clock case and attached to the mic
input of the sound card. I want to have a display of the resulting tick
and tock waveforms. There are two primary functions that are needed to
accurately set a clock's beat.
One is to be able to compare the tick waveform to the tock waveform. A
clock that is off balance and hence off beat will have a difference in
the sound of the tick and the tock. Tick being the pendulum swinging in
one direction vs the tock being the opposite direction swing.
The other function is to measure the time between the tick and the next
tick or full cycle of the pendulum swing. For instance, a lot of
pendulum clocks are set to have a pendulum cycle time of exactly one
second. Measuring this to say, hundredths of a second allows for
accurate adjustment of the clock.
I am looking for pointers to libraries and sample code that might help
me get started with this project. Once completed, it would be released
as an Open-Source project for use by any clock repairer that wants it.
I have not seen any open-source projects of this type and the commercial
units that perform this function cost several hundred dollars.
24 Sep
24 Sep
1:31 p.m.
On Sat, Sep 24, 2005 at 11:23:34AM -0400, CapnLinux wrote:
I would like to build an application to assist in
setting the "Beat" of
a pendulum type clock. This is a tool used by a clock repair person to
set a clock to keep accurate time. It is basically a low frequency
oscilloscope with some specialized functionality.
...
Interesting project. I see one problem: the sample rate on most audio
hardware is much less accurate than what would be required for a clock.
That means that either you must use the computer's real time clock
(not extremely precise either), or some external timing reference,
e.g. a 1 Hz pulse from a GPS receiver, or a reference frequency.
In either case you would need to relate the audio clock the other
one. Since there is significant jitter on the only observation of
the audio clock you have (audio card interrupts as seen from user
space), the only solution is use a delay locked loop to filter out
the jitter and find the exact relation. I write this sort of thing
professionally, and if you need some help with this please feel free
to contact me directly.
--
FA
4:31 p.m.
On Sat, 2005-09-24 at 19:22 +0200, fons adriaensen wrote:
(not extremely precise either), or some external
timing reference,
e.g. a 1 Hz pulse from a GPS receiver, or a reference frequency.
One way I've used is to feed more accurate reference clock to the word
clock input of some pro sound card. Some reference clocks can be
programmed to output some specific frequency.
--
Jussi Laako <jussi.laako(a)pp.inet.fi>
5:20 p.m.
On Sat, Sep 24, 2005 at 11:31:39PM +0300, Jussi Laako wrote:
On Sat, 2005-09-24 at 19:22 +0200, fons adriaensen
wrote:
For a 'pro' card that would do the trick, provided you *have* an accurate
clock generator. But pendulum clocks being big, heavy and difficult to move,
one would prefer to have this sort of app on a laptop. If you could find some
frequency reference that can be input as an audio signal on a second channel,
that would be ideal - no jitter to deal with. There are some cheap portable
clocks that sync to a broadcast time service. Maybe it's possible to extract
a usable signal from these.
--
FA
(not extremely precise either), or some external
timing reference,
e.g. a 1 Hz pulse from a GPS receiver, or a reference frequency.
One way I've used is to feed more accurate reference clock to the word
clock input of some pro sound card. Some reference clocks can be
programmed to output some specific frequency.
6:19 p.m.
This is exactly the type of discussion that I was hoping to stimulate.
The accuracy issue is paramount to the success of this project. A
reference source is something that I had been considering. Wondering if
a standard source could be used to calibrate the program initially so
that resultant measurements could be done without the reference being
present.
Not sure I understand the references to a 'pro' card vs what might be
available in a standard MB based sound card or laptop as portability may
be a desired feature for use outside the repair shop. Can one of you
clarify a little? I am new to some of this terminology.
Thanks - TheCapn.
On Sat, 2005-09-24 at 23:24 +0200, fons adriaensen wrote:
On Sat, Sep 24, 2005 at 11:31:39PM +0300, Jussi Laako
wrote:
On Sat, 2005-09-24 at 19:22 +0200, fons
adriaensen wrote:
For a 'pro' card that would do the trick, provided you *have* an accurate
clock generator. But pendulum clocks being big, heavy and difficult to move,
one would prefer to have this sort of app on a laptop. If you could find some
frequency reference that can be input as an audio signal on a second channel,
that would be ideal - no jitter to deal with. There are some cheap portable
clocks that sync to a broadcast time service. Maybe it's possible to extract
a usable signal from these.
(not extremely precise either), or some external
timing reference,
e.g. a 1 Hz pulse from a GPS receiver, or a reference frequency.
One way I've used is to feed more accurate reference clock to the word
clock input of some pro sound card. Some reference clocks can be
programmed to output some specific frequency.
7:08 p.m.
On Sat, Sep 24, 2005 at 06:19:06PM -0400, CapnLinux wrote:
This is exactly the type of discussion that I was
hoping to stimulate.
The accuracy issue is paramount to the success of this project. A
reference source is something that I had been considering. Wondering if
a standard source could be used to calibrate the program initially so
that resultant measurements could be done without the reference being
present.
That will probably not be enough. Not only is the sample frequency of most
audio cards not very accurate, it's not very stable w.r.t. temperature
variations either.
Not sure I understand the references to a
'pro' card vs what might be
available in a standard MB based sound card or laptop as portability may
be a desired feature for use outside the repair shop. Can one of you
clarify a little? I am new to some of this terminology.
'Consumer' audio cards and those integrated on a motherboard take their
clock either from a cheap xtal oscillator or (horror) from e.g. a PCI
clock. 'Pro' cards usuallly have the option of using an external clock.
If you could find an accurate, stable (and portable) reference source
that could be input *as an audio signal*, that would make things rela-
tively easy, much easier anyway than syncing to anything that is input
as 'data'.
A very accurate reference frequency can be obtained form e.g. the BBC
radio transmitter on 198 kHz, and this can be received over a large
part of Europe (no idea where you are...). Use a simple PLL chip to
receive it, divide the oscillator frequency by 32 and input as audio.
Most regions have similar reference transmitters available.
Another option is to use the mains frequency as a reference. This
is maintained to very high long term accurary (as many clocks depend
on it), but its short term precision could be bad.
--
FA
25 Sep
25 Sep
10:14 a.m.
El sáb, 24-09-2005 a las 14:22, fons adriaensen escribió:
On Sat, Sep 24, 2005 at 11:23:34AM -0400, CapnLinux
wrote:
This is a good question to CapnLinux: what kind of precision do you
need?
Fons: why do you think the sample rate is not accurate enough?
Please, don't interpret I'm trying to refuse those statements, I just
want to now the "whys" so as to explain it to some friends and ask them
for help (as you can read in the other mail)
Cheers, Damian.-
I was married by a judge. I should have asked for a jury. -- Groucho
Marx
I would like to build an application to assist in
setting the "Beat" of
a pendulum type clock. This is a tool used by a clock repair person to
set a clock to keep accurate time. It is basically a low frequency
oscilloscope with some specialized functionality.
...
Interesting project. I see one problem: the sample rate on most audio
hardware is much less accurate than what would be required for a clock.
10:14 a.m.
El sáb, 24-09-2005 a las 20:12, fons adriaensen escribió:
[snip]
A very accurate reference frequency can be obtained form e.g. the BBC
radio transmitter on 198 kHz, and this can be received over a large
part of Europe (no idea where you are...). Use a simple PLL chip to
receive it, divide the oscillator frequency by 32 and input as audio.
Most regions have similar reference transmitters available.
Another option is to use the mains frequency as a reference. This
is maintained to very high long term accurary (as many clocks depend
on it), but its short term precision could be bad.
Very interesting project. What about using some external signal
generator? As you need only one signal, I think there must be something
accurate enough for doing this (I mean, not a multi-frecuency signal
generator). Perhaps a professional tunner? I've heard that some of them
have compensation for temperature, for example.
Anyway, I'm gonna ask some friends about this.
Please, keep us up-to-date with this project.
Cheers, Damian.-
"Any sufficiently advanced technology is indistinguishable from magic"
Arthur C. Clarke
4:22 p.m.
On Sun, Sep 25, 2005 at 11:14:31AM -0300, Lisandro Damián Nicanor Pérez Meyer wrote:
Very interesting project. What about using some
external signal
generator? As you need only one signal, I think there must be something
accurate enough for doing this (I mean, not a multi-frecuency signal
generator). Perhaps a professional tunner? I've heard that some of them
have compensation for temperature, for example.
If by 'tunner' you mean the sort of thing used by radio builders to tune
their stuff, that will not do. These things need frequent recalibration.
But for a few tens of dollars you could get an potted 1MHz xtal oscillator
that you would need to recalibrate maybe once a year just to be sure.
Divide the frequency by some easy factor, say 256, to get into the audio
range and you're done. A temperature compensated version would be a bit
more expensive. I'll put the question to the hardware engineers in my
design team tomorrom, and probably I can come up with some concrete
examples.
--
FA
4:22 p.m.
On Sun, Sep 25, 2005 at 11:14:47AM -0300, Lisandro Damián Nicanor Pérez Meyer wrote:
This is a good question to CapnLinux: what kind of
precision do you
need?
Fons: why do you think the sample rate is not accurate enough?
An audio card that is 0.1 percent off would be quite good for most non
professional uses. This error is about 1/60 of a semitone in musical
terms, and a 3 minute song would be longer or shorter by no more than
a fraction of a second. But most consumer audio cards will show larger
errors.
A clock with that same precision would be off by almost one and a half
minute per day. To set a clock you'd want a relative error of say 1e-6.
That's a factor 1000 better.
--
FA
4:52 p.m.
El dom, 25-09-2005 a las 17:26, fons adriaensen escribió:
On Sun, Sep 25, 2005 at 11:14:47AM -0300, Lisandro
Damián Nicanor Pérez Meyer wrote:
Thanks!!!
No hay preguntas tontas, solo tontos que no preguntan. personaje, en un
mail del LugFi.
This is a good question to CapnLinux: what kind
of precision do you
need?
Fons: why do you think the sample rate is not accurate enough?
An audio card that is 0.1 percent off would be quite good for most non
professional uses. This error is about 1/60 of a semitone in musical
terms, and a 3 minute song would be longer or shorter by no more than
a fraction of a second. But most consumer audio cards will show larger
errors.
A clock with that same precision would be off by almost one and a half
minute per day. To set a clock you'd want a relative error of say 1e-6.
That's a factor 1000 better.
26 Sep
26 Sep
11:26 a.m.
New subject: [OT] Re: [linux-audio-dev] Setting the Beat of a Clock
El dom, 25-09-2005 a las 17:19, fons adriaensen escribió:
On Sun, Sep 25, 2005 at 11:14:31AM -0300, Lisandro
Damián Nicanor Pérez Meyer wrote:
Well, a teacher of mine also suggested the GPS with square-wave output
too.
Cheers, Damian.-
Gabardinas "Windows 95". Se cuelgan solas.
Very interesting project. What about using some
external signal
generator? As you need only one signal, I think there must be something
accurate enough for doing this (I mean, not a multi-frecuency signal
generator). Perhaps a professional tunner? I've heard that some of them
have compensation for temperature, for example.
If by 'tunner' you mean the sort of thing used by radio builders to tune
their stuff, that will not do. These things need frequent recalibration.
But for a few tens of dollars you could get an potted 1MHz xtal oscillator
that you would need to recalibrate maybe once a year just to be sure.
Divide the frequency by some easy factor, say 256, to get into the audio
range and you're done. A temperature compensated version would be a bit
more expensive. I'll put the question to the hardware engineers in my
design team tomorrom, and probably I can come up with some concrete
examples.
6792
days inactive
6794
days old
linux-audio-dev@lists.linuxaudio.org
11 comments
4 participants
participants (4)
-
CapnLinux -
fons adriaensen -
Jussi Laako -
Lisandro Damián Nicanor Pérez Meyer