I found an old wind up watch that I had not used in years. I got it 
serviced but it was running slow so I adjusted it. It went from running 
150 seconds a day slow to running 21 seconds fast. Not too bad for the 
first attempt, I thought.
Now that I know that the adjustment lever has the range, I will be 
scientific. So I recorded the tick on my tablet (1 GHz CPU) and analysed 
it. I also recorded a quartz watch as a reference. Damned if the wave 
form was > 0.6% out. That is nearly 9 minutes out in a day. It looks 
like I will have to revert to the old fashioned method of adjust, let it 
run for a day and repeat until I get fed up.
Could do better if it tried. :-(

-- 
Regards
     Stephen

Post a reply to this message

Stephen <mcavoys_at@aoldotcom> wrote:
> Now that I know that the adjustment lever has the range, I will be 
> scientific. So I recorded the tick on my tablet (1 GHz CPU) and analysed 
> it. I also recorded a quartz watch as a reference. Damned if the wave 
> form was > 0.6% out. That is nearly 9 minutes out in a day.

  It actually doesn't surprise me that CPU clocks aren't adjusted to
extreme accuracy because they don't need to. It's not like they have to
precisely sync with anything. This especially so because most modern CPUs
support throttling. (In fact, an error 0.6% sounds pretty small to me.)

-- 
                                                          - Warp

Post a reply to this message

On 29/05/2012 05:19 PM, Stephen wrote:
> It looks
> like I will have to revert to the old fashioned method of adjust, let it
> run for a day and repeat until I get fed up.
> Could do better if it tried. :-(

You could try using the atomic emission spectrum of caesium-133 to 
adjust the frequency of the quartz oscillator...

For reference, the frequency of photon generated by the transition 
between the two hyper-fine ground states of a caesium-133 atom is 
*defined as* 9,192,631,770 Hz. As in, that's the current SI definition 
of what one second *is*.

Post a reply to this message

On 29/05/2012 5:33 PM, Warp wrote:
>    It actually doesn't surprise me that CPU clocks aren't adjusted to
> extreme accuracy because they don't need to. It's not like they have to
> precisely sync with anything. This especially so because most modern CPUs
> support throttling. (In fact, an error 0.6% sounds pretty small to me.)

Well, you are right. I really wasn’t thinking. It was a thought I had 
last night before going to sleep and in the morning I went full steam 
ahead, at it. Maybe I just miss working for a living.
On some instruments I’ve worked on 2% was considered accurate enough on 
others you measured parts per million. When you expect to get a watch 
running to +/- 2 seconds a day it is galling to have to do it by iteration.

-- 
Regards
     Stephen

Post a reply to this message

On 29/05/2012 5:38 PM, Orchid Win7 v1 wrote:
>
> You could try using the atomic emission spectrum of caesium-133 to
> adjust the frequency of the quartz oscillator...
>

Sh*te! I forgot. I have one of those in the back shed, as well.

> For reference, the frequency of photon generated by the transition
> between the two hyper-fine ground states of a caesium-133 atom is
> *defined as* 9,192,631,770 Hz. As in, that's the current SI definition
> of what one second *is*.

*/- a minute a month will do me. Actually twice a day will do if I 
forget to wind it up. :-P

-- 
Regards
     Stephen

Post a reply to this message

>> You could try using the atomic emission spectrum of caesium-133 to
>> adjust the frequency of the quartz oscillator...
>
> Sh*te! I forgot. I have one of those in the back shed, as well.

0wned.

Post a reply to this message

On 29/05/2012 9:03 PM, Orchid Win7 v1 wrote:
>>> You could try using the atomic emission spectrum of caesium-133 to
>>> adjust the frequency of the quartz oscillator...
>>
>> Sh*te! I forgot. I have one of those in the back shed, as well.
>
> 0wned.

And never called me mother!

-- 
Regards
     Stephen

Post a reply to this message

> Now that I know that the adjustment lever has the range, I will be
> scientific. So I recorded the tick on my tablet (1 GHz CPU) and analysed
> it. I also recorded a quartz watch as a reference. Damned if the wave
> form was > 0.6% out.

There is no reason for a tablet (or any PC/phone) to have a clock 
generator with super accuracy, it's simply wasted expense when an 
accuracy of <1% will do for the majority of tasks.  That's why 
professional sound cards have an option to use the clock from an 
incoming signal (rather than the internal one) to keep everything in sync.

Try playing the same song (or your tick recording) on several different 
devices at the same time, you'll find they likely drift apart noticeably 
after only a minute or two.

You could try recording the speaking clock as a reference, that is 
probably more accurate than your tablet.

Post a reply to this message

On 30/05/2012 08:57 AM, scott wrote:

> Try playing the same song (or your tick recording) on several different
> devices at the same time, you'll find they likely drift apart noticeably
> after only a minute or two.

And here I was thinking that even cheap quartz oscillators are quite 
accurate...

Post a reply to this message

>> Try playing the same song (or your tick recording) on several different
>> devices at the same time, you'll find they likely drift apart noticeably
>> after only a minute or two.
>
> And here I was thinking that even cheap quartz oscillators are quite
> accurate...

Even cheap quartz oscillators are too expensive to just throw at a 
device like a tablet.  The audio clock is likely derived from elsewhere 
with whatever divide-down ratio was cheapest to implement (the frequency 
will likely not even be designed to be spot on 44100 if the numbers 
don't work out nicely).

Post a reply to this message

On 30/05/2012 8:57 AM, scott wrote:
>
> There is no reason for a tablet (or any PC/phone) to have a clock
> generator with super accuracy, it's simply wasted expense when an
> accuracy of <1% will do for the majority of tasks.  That's why
> professional sound cards have an option to use the clock from an
> incoming signal (rather than the internal one) to keep everything in sync.
>

It was a NTB (Not Too Bright) idea, a eureka moment where I slipped on 
the soap when I jumped out of the bath etc.
It’s not worth the trouble trying to make an escapement analyser when 
the software I have will only find the beats to a resolution of 1 
millisecond* and I don’t have a proper microphone. So the waveform I get 
is very noisy and automatically finding the beats is not very accurate.


> Try playing the same song (or your tick recording) on several different
> devices at the same time, you'll find they likely drift apart noticeably
> after only a minute or two.
>

I am sure that they would.

> You could try recording the speaking clock as a reference, that is
> probably more accurate than your tablet.

I was using the tablet as a recording device. It is the accuracy of that 
which is in doubt. Well it is not actually in doubt, it is not up to the 
requirements needed.
My quartz watch was for calibrating/adjusting the figures I got after 
analysing the wave form. But the time between ticks is not constant. The 
recording program and the sound card probably have too low a priority 
for accurate recording. Which is also an answer to Andrew saying that 
cheap quartz oscillators are quite accurate. That and the compensation 
for voltage and temperature won't be high spec.

* An accuracy of about +/- 7 minutes a day.


-- 
Regards
     Stephen

Post a reply to this message

On 30/05/2012 12:01 PM, scott wrote:
>>> Try playing the same song (or your tick recording) on several different
>>> devices at the same time, you'll find they likely drift apart noticeably
>>> after only a minute or two.
>>
>> And here I was thinking that even cheap quartz oscillators are quite
>> accurate...
>
> Even cheap quartz oscillators are too expensive to just throw at a
> device like a tablet. The audio clock is likely derived from elsewhere
> with whatever divide-down ratio was cheapest to implement (the frequency
> will likely not even be designed to be spot on 44100 if the numbers
> don't work out nicely).

A deviation in playback speed of +0.6% or -0.6% corresponds to a pitch 
deviation of about 10 cents. That's audible. I'd be surprised if they 
allow it to be quite that large.

(I would have expected the DAC to run at whatever speed is convenient, 
and then resample the signal in software... But what do I know?)

Post a reply to this message

> A deviation in playback speed of +0.6% or -0.6% corresponds to a pitch
> deviation of about 10 cents. That's audible.

If you play it back on the same hardware you used to record then there 
shouldn't be any pitch change :-)

But even if you played back something recorded elsewhere, do you really 
think a normal person would notice a 0.6% higher or lower pitch (without 
the correct reference pitch to compare with at the same time).  When 
they show films (shot at 24 fps) on PAL (25 fps) doesn't the video and 
audio come out 4% faster and higher pitch (or do they resample that)?

> (I would have expected the DAC to run at whatever speed is convenient,
> and then resample the signal in software... But what do I know?)

If the software knew what the real clock rate was with 100% accuracy 
there wouldn't be this discussion in the first place!  Anyway if you 
were to resample from (say) 43500 Hz to 44100 Hz you're going to get a 
lot of artefacts and use a lot of CPU (battery) - better to just play 
your 43500 Hz recording assuming it was at 44100 Hz.

You also have conflicting technical requirements in a device like a 
tablet.  For example the audio circuit needs the clock to be as steady 
and accurate as possible, whereas other parts of the circuits will 
definitely not want a steady clock frequency to avoid EMC/EMI issues. 
Guess which requirement will get forgotten when the EMC tests fail or 
the bean counters tighten the strings?

Post a reply to this message

On 5/29/2012 12:15, Stephen wrote:
> Sh*te! I forgot. I have one of those in the back shed, as well.

They're really not that expensive. I worked for someone who had three or 
four. You just have to keep them calibrated.

-- 
Darren New, San Diego CA, USA (PST)
   "Oh no! We're out of code juice!"
   "Don't panic. There's beans and filters
    in the cabinet."

Post a reply to this message

On 5/30/2012 0:57, scott wrote:
> Try playing the same song (or your tick recording) on several different
> devices at the same time, you'll find they likely drift apart noticeably
> after only a minute or two.

I was trying to play harmony with myself at one point, so I recorded the 
first track, then played it back on headphones while I recorded the second 
track. Within 8 or 10 measures, I was an entire beat off, when I merged them.

Very annoying. :-)
-- 
Darren New, San Diego CA, USA (PST)
   "Oh no! We're out of code juice!"
   "Don't panic. There's beans and filters
    in the cabinet."

Post a reply to this message

On 02/06/2012 9:19 PM, Darren New wrote:
> On 5/29/2012 12:15, Stephen wrote:
>> Sh*te! I forgot. I have one of those in the back shed, as well.
>
> They're really not that expensive. I worked for someone who had three or
> four. You just have to keep them calibrated.
>

More expensive than I would buy for one watch.
But not in Andrew's league of hundreds of millions of pounds. ;-)

-- 
Regards
     Stephen

Post a reply to this message

On 02/06/2012 9:21 PM, Darren New wrote:

>
> Very annoying. :-)

That's what I thought. Grrr!

-- 
Regards
     Stephen

Post a reply to this message

On 02/06/2012 09:21 PM, Darren New wrote:
> On 5/30/2012 0:57, scott wrote:
>> Try playing the same song (or your tick recording) on several different
>> devices at the same time, you'll find they likely drift apart noticeably
>> after only a minute or two.
>
> I was trying to play harmony with myself at one point, so I recorded the
> first track, then played it back on headphones while I recorded the
> second track. Within 8 or 10 measures, I was an entire beat off, when I
> merged them.
>
> Very annoying. :-)

Recorded with /what/?

I have a cheap digital 8-track at home, and I do /exactly this/ with it, 
and it works perfectly, every single time...

Post a reply to this message

On 6/3/2012 1:51, Orchid Win7 v1 wrote:
> Recorded with /what/?

This was a little pocket MP3 player thingie. It happened to be the only 
convenient play-back device I had handy at the moment.  I'm not surprised in 
hindsight it was off that much, but it was surprising enough that I did it 
two or three times before I realized it wasn't me messing things up.

-- 
Darren New, San Diego CA, USA (PST)
   "Oh no! We're out of code juice!"
   "Don't panic. There's beans and filters
    in the cabinet."

Post a reply to this message