[NEKNIM]

This thing is really cool. I was going to download one of the iPhone apps but I'm glad I stumbled across this. One of the things I did was pull up youtube videos of people using physical timegraphers. I was able to get some readings that seemed to match but it was difficult as theres usually background noise or people talking. Also, if the person recording changes the position of their camera mic it affects the measurements. On one video I used you can see clearly in the graph that something is off on one side of the pallet fork, as one line is nice and the other is a mess of dots. What was cool to see was when this program picked up the tiks a very similar pattern emerged. You guys can give this method a try (if you don't own a physical timegrapher yourself) but its not going to give the most consistent results in terms of comparison.

I tested my watches, which are mostly cheap sub $200 nothings and one tag aquaracer (ETA 2824-2), with my built in Mac microphone and a set of Apple EarPods. Both of which gave pretty much exactly the same reading. One thing I noticed when I tested my Tag, which has been losing about 5 minutes a month, was that if I placed the watch on a flat surface I got a reading of around +2s/d in most positions. However, if I put the watch on my wrist and taped the microphone to the watch I got a reading of about -9s/d. This would be the number I would expect as that equates to almost -5min/month. I feel like this could be due to temperature but it seemed to confirm my observations over the last few months.

The reason I wanted a timegrapher program was purely out of my own curiosity to see what my watches would register. I only actually learned what a timegrapher was today when I googled "how to test watch accuracy?" so take my mini review from a complete newbie's perspective. Anyway, thanks for the program!

 

[evilphoton]

I'm seeing some inconsistency between daily error in what tg is reporting vs. the daily error that the watch is actually doing.
i set my watch on the mic of my mac in all 6 positions and it reports an average of about -20s/d. actually wearing it though, it runs almost +20s/d fast - yes, 40s difference.
I thought maybe tg it was having difficulty hearing the through the case, so i took the back off, but it reports the same.
I really don't think it's the software based on the feedback here and my GP is 5 position adjusted (never touched) and tg shows +1 to +2s/d and this is just about what the watch actually runs. but I'm at a loss determining the difference with this watch.

Has anyone else noticed something like this? I assume this is some sort of user error but i haven't been able to determine what. it seems to just be this watch. it's an 007 with a 7s26c.

 

[Paleotime]

You could have an issue with the balance arcs becoming short in the transitions between positions. If your watch wrist is pretty active then the short/fast arcs in these transitions could be an issue - even if you settle into a good rate once a position is established for a little bit.

 

[evilphoton]

thank you! I can see that and it explains why when i first set the watch on the mic, the balance arc seems low (say 250 degrees) and the daily error is fairly low. then after it sits for that minute (it actually stabilizes in probably 15 seconds or less), the arc has increased to say 290 degrees and the error has increased. the change is similar each time i set the watch down in a different position. i can see wearing the watch during the day and constantly moving that the balance would run in a fairly constant state of transition, as you say.

I've slowed the watch quite a bit to make it fit with how i wear it, using tg as a reference that i didn't move the spazzy little lever too far... since if i can see the lever move, it's way too much.

You could have an issue with the balance arcs becoming short in the transitions between positions. If your watch wrist is pretty active then the short/fast arcs in these transitions could be an issue - even if you settle into a good rate once a position is established for a little bit.

 

[1afc]

Thanks contrate_wheel. Great effort to write such good software and then supply it as open source.

Ran first time on Win 7. This just makes life so much easier. Time with watch with something like watch check or a piece of paper, open it up and make a change that should be very close to what you need.

Thanks everyone else for providing feedback for other items like mikes and amplifiers (which I'm trying to stay away from).

 

[beastomaniac]

It seems like that the latest version doesn't recognize greater amplitudes than 300 degrees. Does anybody else have same kind of results?

 

[1afc]

I really don't think it's the software based on the feedback here and my GP is 5 position adjusted (never touched) and tg shows +1 to +2s/d and this is just about what the watch actually runs. but I'm at a loss determining the difference with this watch.

So as we understand, you say you have a GP that tg measures +1 to +2s/d and when you wear the watch you get the same reading. So you can get the software to perform on one watch.
What are the timing errors on the GP and the 007? I don't know what is acceptable but I have just tweaked a few of mine to be 0.1ms.

With the 007 I would start by measuring the watch in one position, say face up, and then put the watch on the shelf, face up, for a day and see what it does.
If you do this in a number of positions and find most are ok then there appears to be one setting which is going very fast.
If you can find that position then you probably need to go to a watchmaker and tell him the story.

 

[revance]

Very cool, I look forward to playing with this.

 

[contrate_wheel]

Hi friends,

we finally have a calibration procedure!

The new version of tg (0.3.1) is available in the usual places. A summary
of the install instructions for several operating systems is here
https://github.com/vacaboja/tg

Before I go into the details, let me thank pbnelson for his install
instructions: he actually stumbled upon a bug in the Debian distribution
that interferes with one of the packages needed by tg
(https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=718221). Fortunately,
the install sequence suggested by pbnelson works around the bug, so I have
included it in my instructions on github. Finally, I would like to thank
all other contributors to this thread for many hints and suggestions, the
past ones and hopefully the future ones!

As any computer program that emulates a timing machine, tg necessarily
takes its time reference from the clock of the audio card of the computer.
There is anecdotal evidence that audio cards have usually relatively
stable clocks. However, unfortunately, these clocks are often affected by
a constant deviation from true time, sometimes of many seconds per day. To
correct for it, one must measure the deviation by comparison to a more
accurate time source, and then inform the program of its value. With tg,
this can be done either manually or through an automatic calibration
procedure.

The manual setting is straightforward, just input the rate of the sound
card's clock, in seconds per day, into the "cal" field.

The automatic calibration feature works by comparing any analog quartz
watch (producing one beat per second) to the sound card. Operationally,
you put a quartz watch on the microphone and then you click on the
calibrate button.

Tg will try to pick up the watch's beat.

When it does, you will be notified.

Then you will have to wait for a calibration period of about 15 minutes,
and, at the end, two outcomes are possible: luck

and no luck.

If the calibration procedure succeeds, the calibration constant will be
updated automatically. To leave calibration mode, just press the calibrate
button again.

Now to some technical details. Most of you will already know, but it
might be surprising to some, to learn that quartz watches usually do not
beat at a constant rate. In general, the crystal oscillator controlling a
quartz watch is set to beat at a considerably fast rate, say +10 s/day.
The circuitry in the watch is then programmed to periodically apply a
correction that precisely compensates the rate of the crystal. This
mechanism, called "inhibition compensation", is cheaper than implementing
an equally accurate oscillator directly in hardware. For those interested,
it's easy to find more information on the internet, for instance one might
want to take a look at this:
http://tf.nist.gov/general/pdf/2276.pdf

As an example of inhibition compensation, take a look at the trace of a
common quartz watch generated by tg's calibration routine.

You can see very clearly that the watch is running fast most of the time,
and every minute the compensation mechanism brings it back to the correct
average rate. The screenshot has been taken after the termination of tg's
calibration procedure. The blue lines are roughly parallel to the zig-zag
trace of the watch, this indicates that tg has computed its average
correctly, therefore the resulting calibration is accurate (referenced to
that particular watch, of course). One might ask why, when the inhibition
mechanism operates, the trace jumps back quickly, but not as abruptly as
it should. This happens because each dot on the graph is actually a
short-term average (this is done to minimize the influence of audio
noise), so it takes a few seconds after the compensation mechanism has
fired for the average to converge to the new value.

There is one last caveat that I have to point out: high end quartz watches
might use very complicated compensation mechanisms that operate over long
periods of time. See, for instance, this very informative post by forum
member dwjquest
https://www.watchuseek.com/f9/eta-2...rature-correction-method-unveiled-300533.html
and this excellent discussion of several high accuracy watches
More about affordable High-End Quartz
As a consequence, somewhat contrary to intuition, it is not recommended to
use a high end watch for calibration, unless, of course, you feel
comfortable evaluating the trace and understanding what is going on. Also
radio-controlled watches are not a good choice, because, self-adjusting
every day or so, they prevent one from getting a feedback of how actually
accurate they are. The best reference to calibrate tg is a cheap,
commonplace, quartz watch that you know by experience is within plus or
minus a few seconds per month.

Please keep in mind that the rate of a sound card's clock is not really a
figure of merit for consumer-grade products, because sound cards are not
designed for accurate timekeeping. So, a sound card that has a rate of,
say, +4 s/day, is not, on this account, inferior to one that has +0.4. It
is just the luck of the draw. Stability, here, is the important thing, and
there is no a priori relationship between rate and stability, so +4 card
might or might not be more stable than the +0.4 one. Temperature is
probably the biggest factor impacting the stability of the rate. As I
said, there is anecdotal evidence that this should not be a concern, see
for instance this (section 2.3)
http://protyposis.net/download/608

Barring bugs, I believe that the core features of tg are now almost
complete. I will consider adding some built-in screenshot or report
generation mechanism: there has already been some work done in this
direction by a github user named wahlstedt. Also an audio input selector
is likely to be added (currently, tg can only use the default input). If
you find that any other essential feature is still missing please let me
know!

 

[pbnelson]

Looks great! I pulled the latest source, built it on my Linux Mint (i.e. Ubuntu/Debian) system, and it started right up.

I calibrated tg using a Swiss Legend watch (a middle-of-the-road quartz). After 30 minutes it had settled down to a -4.7s/d calibration factor.

I then compared tg to my TG1000 using the same simultaneous monitoring technique I used before. The tg program is now much more accurate, reporting only 1 s/d slower than the TG1000. I manually lowered the calibration from -4.7s/d to -4.0s/d, and then tg consistently reported the same error rate as the TG1000.

Unfortunately the beat error was still off by a half second. The TG1000 consistently reports 0.2ms, and tg consistently reports 0.7ms.

 

[Snakepottery]

Looking forward to tryout no the new version, however, not being a OS X specialist (just a user), not sure how I update the software. Running the install command tells me it's already loaded.
Any help anyone?
Thanks

Snake


Sent from my iPad using Tapatalk

 

[1afc]

For you Vostok fans, Meranom informs me that the lift angle of the auto is 42 degrees and any amplitude over 200 is fine.
My Dirskie is running at 255 degrees .

I didn't specifically ask about the manual winds but expect they will be exactly the same.

 

[1afc]

For you Vostok fans, Meranom informs me that the lift angle of the auto is 42 degrees and any amplitude over 200 is fine.
My Dirskie is running at 255 degrees .

I didn't specifically ask about the manual winds but expect they will be exactly the same.

 

[Klaus Tickalot]

A really great new feature !!!!

Quick feedback:
I did the calibration using a 10 year old Tissot V8 with ETA G10.211 (battery one year old). Maybe 10 or 15 minutes later the calibration was set to +5,1s/d
The calibration-graph did not look so "pretty" as shown in the example, but it seems to have worked properly and corresponds even better to real-life. Using the previous versions I ended up in adjusting some of my watches a bit slower or at a maximum to +/-0s/d to have a better timing. Now I know that it should have been 5,1s/d

MANY THANKS AGAIN |>

p.s. How about adding a beatrate of 3600bph e.g. for selecting a quarzmovement with a "pretty graph" in the menu? (no begging, just an idea)

 

[dmnc]

Looking forward to tryout no the new version, however, not being a OS X specialist (just a user), not sure how I update the software. Running the install command tells me it's already loaded.
Any help anyone?
Thanks

Snake

Sent from my iPad using Tapatalk

This is an idiosyncrasy of homebrew with a head only formula.

I'm away for a few days but will check the formula all works ok when I get back. However, assuming no dependencies changed instead of upgrade run 'brew uninstall tg' followed by a fresh 'brew install tg --HEAD'.

Sent from my iPhone using Tapatalk

 

[NEKNIM]

For those wondering how to install the update on a mac, here is what I did that seems to work.

run, in terminal, [brew uninstall dmnc/horology/tg --HEAD]
Then run, [brew install dmnc/horology/tg --HEAD]

This uninstalled the old and reinstalled the new.

 

[Snakepottery]

Perfect, thanks gents! I did think that would be the solution but wasn't 100% sure of the syntax. Pretty obvious really

Thanks
Snakey

Sent from my iPad using Tapatalk

 

[Guido Velasquez]

Excelente trabajo. Muchas gracias. Hemos estado disfrutando casi desde su inicio el programa. Felicitaciones por la característica de Calibración. Muy fácil del configurar. Me gustaría solicitar que integraran el modo Clock. Para relojes de péndulo. Se trataría solamente de medir "rate" y "beat error", sin hacer medición de amplitud. Utilizando solo el tercer sonido de la secuencia del escape.

Excellent work. Thank you very much. We have been enjoying almost since its inception the program. Congratulations on Calibration feature. Very Easy to configure. I would request that integrate the Clock mode. For pendulum clocks. It would be only measured "rate" and "beat error", without amplitude measurement. Using only the third sound of the escape sequence.

Watch-O-Scope here compared to tg calibrated.

 

[Coug76]

For you Vostok fans, Meranom informs me that the lift angle of the auto is 42 degrees and any amplitude over 200 is fine.
My Dirskie is running at 255 degrees .

I didn't specifically ask about the manual winds but expect they will be exactly the same.

Vostok 22mm calibers (2209, 2214, 2234 etc) run 46° lift.

Vostok 24mm calibers (2409, 2414, 2415, 2416b, 2426, 2432 etc) run 42° lift. There is no difference in lift angle between manual wind and auto (I've transplanted balances from manual to auto wound movements quite successfully). I can't say if the 42° lift angle holds true for older 24mm movements such as the 2403.

Vostok 28mm calibers (2809) run 44° lift.

Hastily spouted for your befuddlement

 

[Guido Velasquez]

Inconsistency in data beat error.

Dear Contrate_Wheel. I have had many hours of doing tests with the program you do not a have done. I still admired the performance.

However I have found on several occasions some inconsistency in the data delivery on the beat error.

It is possible that the algorithm is choosing the third sound of the entire event 3, (1: Unlock 2: Impulse and 3: Drop or blockage) However, the sound of falling or blockage that is the stroke of tooth wheel exhaust against locking face of the pallet, is not very stable, ie not as stable as coming from the true clock regulator that is the balance wheel.

Here also I have two screenshot made almost at the same instant. The first with tg and another with other software. I tested with two other software and the results are consistent with each other but not with tg.
View attachment 8596810 View attachment 8596834
Therefore asymmetry that is displaying the program is not the steering wheel but relationship escape wheel / pallets.

If the graph is derived from the time between an unlocking and the next, this screenshot is not corresponding.
View attachment 8596778
The amplitude is not recorded. But if it is registered the beat error. If the amplitude is read from the event unlocking fall, How you are determined by the software here?

It may be that I'm doing something wrong. If it were so I need to make the necessary corrections.

Guido