Calibration Of Electronic Watch Timers ?

[Endeavor]

Dear All;

 

While working on adjusting the Beat Error of a Rolex, discussing beat-error issues with other forum members in the thread "Rolex 3135 Beat Error Adjustment" and trying to achieve an accurateness of less than 1ms, I posted myself the question how accurate is my electronic timer-machine?

I'm using the Watch-O-Scope, which is a PC based software program (written by forum member Svorkoetter). It analyzes the beat from a watch via a microphone.

To calibrate the program, a quartz-watch is used with a known daily rate.

However, from what I understand, most (if not all) electronic processors (PC or stand-alone) derive their time and calculations from the frequency of an oscillating crystal. Crystals come in many shape, forms and accurateness. There are many factors influencing the accuracy (therefore the accuracy of the Watch-timer) of a crystal frequency: power supply voltage, magnetic fields, electric fields, atmospheric influences, temperature and aging to name a few. The crystal frequency drifts over time, and needs, like any other precision instrument, periodically calibration. Some crystals drift less than others. OCXO (Oven controlled crystal oscillators) are fairly stable over time, but are more expensive and need time to warm up. I doubt that you will find them in low-end watch timers or in any PC.

Industry often uses the Rubidium oscillator as a calibration standard (10 MHz or 1Hz for example), but they are very expensive. Satellite time pieces for GPS are even more accurate.....but we are discussing the calibration of watch timers, not atomic clocks.

An accurate time reference is ample available; the atomic clock, but accurate frequency references not so.

Apparently a quartz-watch, with a known daily rate, may be the solution to calibrate the Watch timer, or determine the error of it.

It would be very nice if Svorkoetter (who designed the Watch-O-Scope) , or any other knowledgeable forum members could enlighten us how that works? Or in general, what accuracy to expect from a low- to medium-end timer machine (Stand-alone or PC), the expected drift and how to calibrate?

I for one, would be very interested to learn what "Ticks" inside the (black / grey) Timer-box and how to calibrate it.

Hope to hear.......

Roland.

Edited February 24, 2016 by Endeavor

[Rob]

Audio card calibration is an important topic for software spectrum analyzers too, check this page for lots of details about it: http://www.qsl.net/dl4yhf/speclab/frqcalib.htm

[JohnR725]

http://www.watchoscope.com/manual.html#calibrate

 

If you look on the website there is a description the how to calibrate your computer/software. The only catch is it also depends on the computer. I tried to calibration on two separate computers one was perfect the other I gave up on. On the other hand for most stuff the basic calibration is so close and the diagnostic abilities of the software is outstanding is not really an issue. Unless of course you're trying to time your Rolex to chronometer standard.

 

So if you're really trying to do something precise demanding the absolute best then a hardware-based timing machine is better. Plus they even make GPS calibration device to make sure your timing machine is running on time.

 

http://www.witschi.com/en/group-of-devices/measurement-of-electronic-watches/gps-empfaenger/gps-empfaenger.html

 

Unfortunately equally as expensive as the GPS calibrator is this device. It's a shame we couldn't rent this for a short period of time as it does more than just calibrating it would allow us to verify that everything is really working correctly and displaying everything correctly.

 

http://www.witschi.com/en/group-of-devices/measurement-of-mechanical-watches/microsignalgenerator/microsignalgenerator.html

 

 

 

 

[Endeavor]

Hi JohnR725 & Rob;

 

Thanks both again for your input ;)

By no means I'm trying to convert my Rolex to be a competitor with the atomic clock....... Since with Beat-Error, we are working with sub 1ms numbers, I'm just trying to get a feel for the accuracy of the equipment I'm working with and the calibration.

I will re-read how the calibration is done on the Watch-O-scope and compare my two computers. Similar to your experiences, one seems to work fine, the other erratic.

[Rob]

You can make your own GPS calibration unit to calibrate your PC sound card: http://protyposis.net/clockdrift/high-precision-audio-drift-measurements-with-gps/

[svorkoetter]

Just to clarify a few things ...

Watch-O-Scope, and any other purely PC-based watch timer (i.e. not something like MicroSet, which has its own calibrated hardware to do the timing), uses the crystal oscillator in your PC's sound card, not the one for the CPU itself. This oscillator is supposed to work at 44,100 cycles per second. (That's a simplification; it actually works at a much higher frequency, and is divided down depending on the sample rate of the audio being played, or the desired sample rate for recording, but 44,100 is one rate knwon to be supported by all sound cards, since it's the standard rate for CD audio).

 

Unlike the crystal in your quartz watch however (usually 32,768 cycles per second), the sound card rate doesn't need to be all that accurate. If it were off by 0.1%, it would result in a pitch error (when listening to music) of about 2 cents (there are 100 cents between two adjacent notes in a Western chromatic scale). There are very few people in the world who can perceive that small of a pitch error, and most musical instruments aren't even tuned that accurately. Many pop singers can't even come anywhere near that.

 

However, if a program like Watch-O-Scope is using that oscillator to test a watch (by counting the number of cycles from one tick to the next), a 0.1% error is a lot. How much? 86 seconds per day! That's why Watch-O-Scope has a rate correction setting.

 

With the rate correction setting initially left blank, testing a highly accurate watch (which for the purposes of adjusting a mechanical watch, almost any quartz watch would be) will show how much error is due to the sound card. If you test your 0s/d quartz watch and Watch-O-Scope tells you that it's gaining 30 seconds per day, then we know that the sound card oscillator is wrong by 0.035% (30 / 86400 x 100 = 0.035). By telling Watch-O-Scope what the reading is from your assumed-to-be-accurate quartz watch, it then knows what the error is in the sound card's oscillator. That 30 seconds a day correction tells Watch-O-Scope that your particular sound card oscillator is really oscillating at 44,084.7 cycles per second (if according to the sound card your accurate watch is fast, it means that really the sound card is slow). Now that Watch-O-Scope knows the actual frequency of your sound card's oscillator, it can be used to to measure the actual frequency of something else (like your ticking watch).

 

What makes this usable is that most sound cards are relatively stable. Part of this is just that on the scale of mechanical watch timing errors, crystals are pretty stable. Another factor is that the inside of the a computer, once warmed up, is generally at a pretty steady temperature, especially if you have temperature-controlled cooling fans.

 

Now let's look at the effect of sound card accuracy on beat error and amplitude.

 

The beat error is (half) the difference in spaces between ticks from one tick to the next. Another way of looking at it is if you consider every second tick to be correct, how far are the ones in between from where they are supposed to be. This is of course measured by the same sound card oscillator cycles as the watch's rate. However, reasonable errors in the oscillator don't really matter at all for this measurement!

 

Imagine that your sound card's oscillator was off by a whopping 1% (an error that would be detectable by most musicians). Even worse, imagine that it were fluctuating back and forth between -1% and +1% every few seconds (an error that would be audible to almost anyone; remember wow and flutter on tape players and turntables?). That would make the sound card useless for measuring daily rate, but it would have almost no effect on beat error and amplitude measurements.

 

Why? Because 1% is 1%. +/-1% of a day is +/-864 seconds, which would make daily rate measurements pointless. But +/-1% of a 1.5ms beat error is +/-0.015ms. That means a measured beat error of 1.5ms could be anywhere from 1.485ms to 1.515ms. But those would both still display as 1.5ms, since Watch-O-Scope (and other timegraphers I know of) only display one decimal place. So, sound card error doesn't really matter for measuring beat error.

 

Likewise, a 1% sound card error isn't going to affect amplitude measurement either. In fact, it won't affect it at all. The formula for calculating amplitude involves (among other things) dividing the lift time by the total time between ticks. Since both times will be wrong by the same percentage (unless the sound card oscillator is varying really fast), the error will cancel out.

 

To really know how accurate and stable your sound card is, a GPS one-pulse-per-second output fed into the sound card would be ideal. You could run Watch-O-Scope for hours, watching for fluctuations (or use the long-term test feature if you have the Pro version).

[Endeavor]

Very interesting and eductional information svorkoetter....thank you very much!

 

A big relieve that the sound-card crystal accuracy is hardly of any influence on the Beat Error...........however, I'm not out of the woods yet.

I just did a test on two computers;

- A HP Pavilion with a 1.58GHz AMD, running Windows XP

- Macbook, running Windows XP via Bootcamp.

 

Both were warmed up and both were calibrated, with a known quartz watch, as per Watch-O-Scope instructions.

I used first the HP and went through the positions DD, CD, DU, CU, each 5min+ duration. Changed the amplifier over to the Mac and did the same. To check repeatability, I went back to the HP and subsequently to the Mac again.

If we just concentrate of the measured Beat Error of my Rolex, I got the following readings (first measurement / second measurement);

 

HP;

DD: 1.9ms / 1.35ms

CD: 0.91ms / 1.85ms

DU: 1.13ms / 1.45ms

CU: 0.98ms / 0.98ms

 

Macbook;

DD: 0.75ms / 0.58ms

CD: 0.92ms /0.94ms

DU: 1.08ms / 0.8ms

CU: 0.98ms / 1.1ms

 

As you can see, the Macbook had far more consistent, and lower numbers than the HP. Question is, which one to believe? Obviously, I would like to believe the Macbook, but how to know? Is there a way to find out? Yes, a third computer....but with these two computers I ran out of options and a third one may confuse matters even more............

Hope you have some tricks on your sleeve.....

Edited February 24, 2016 by Endeavor

[bobm12]

I will never believe a Mac!!! and I'll take HP's with a grain of salt!  :)

 

Sorry, I couldn't resist!...and Sorry I can't shed any more light on the subject...learning here too!

 

Cheers,

 

Bob

[JohnR725]

Having compared various hardware-based and software-based timing machines it's really important that they all looked the same watch at the exact same time. There tends to be variation if you go from machine to machine. Physically moving the watch causes variations which means you need to let the watch settle down before timing. Then even if you do compare Several machines simultaneously they're averaging over different parts the waveform. Then the other amusing aspect is multiple machines which one is right?

[Endeavor]

JohnR725; I like to quote something interesting you said in the "Rolex 3135 Beat-error adjustment";

 

"So here's a thought on why beat at anything less than zero is bad. First what exactly is the hairspring for? We think about it as part of the oscillation system but it has another purpose which is it supplies the force for unlocking the escapement. So the strange way to think about this is by being out of beat it's almost like one side of the escapement is running at a different amplitude because it doesn't have as much energy to unlock the escapement. Everything we do with the escapement is bad for timekeeping if it's not adjusted correctly."

 

Facing the dilemma of which Timer-machine is correct, I'm wondering how many watches do actually run exactly in beat?

Next to that "out of beat" is bad for timekeeping, are there any other mechanical disadvantages? Meaning that if a watch runs with a BE of say 1ms (suppose we knew that accurately), but it keeps nicely time........is this "out of beat" for example bad for the movement ? Does it causes other issues like more wear and tear?

 

 

[JohnR725]

How about questions rather than answers. Timing machines are outstanding for diagnostics and telling us if the watches running on time at least at the instant it was on the timing machine. What do all the people out there do that don't have access to timing machines? Simplistic answer is without a timing machine you may not know you have a problem at all.

 

As far as I know being Out of beat doesn't wear the watch out. It's conceivable if being out of beat screws up timekeeping you can adjust the regulator to make up for the problem. But that isn't really the best solution to fixing the problem.

 

The other problem you may be having with the software-based timing machines is getting a good clean signal for the software to process. Just because you use the same amplifier doesn't mean they're all these devices will process the sound exactly the same. It'd be interesting to put the software into the Scope Mode The various computers and post the images For us to see.

[svorkoetter]

Putting the software into scope mode is a good idea. I mentioned (maybe in the other thread) that Watch-O-Scope works fine on my laptop. I forgot to add, "only if it's running on batteries". If the laptop is plugged into the wall, then there is so much noise on the audio input that it drowns out the signal. Apparently this is quite common with laptops, and affects any application using audio input.

[Endeavor]

That's a point I didn't consider. Both laptops were plugged in........ I will give it another try, time permitting.......Would be nice if I get good consistent readings!

Edited February 25, 2016 by Endeavor

[Endeavor]

Did partially another test. The Macbook, now on batteries, produced nearly the same readings as it did yesterday being plugged in. I did one long test with DD, in the others positions were of shorter duration.

 

Macbook Beat Error readings;

 

DD: 0.75ms

CD: 0.91ms

DU: 1.0ms

CU: 1.05ms

 

As for the HP-computer; the initial readings (on battery) were way off. I checked & manually adjusted the noise level, but then I noticed that as soon as the cooling fan kicked in, the noise/interference went off the charts. The HP has one of those fan which, as soon as you switch the computer on, goes berserk, only to discover 20 seconds later that there is nothing to cool. The steering of the fan is not subtle, it goes full on and then tapers down. To make a long story short, the HP turns out to be unsuitable for the Watch-O-Scope.

Leaves me with the Macbook, which so far has produced the most consistent readings, plugged in the wall or on battery.

From what I learned from svorkoetter about the crystal accuracy, being of hardly any influence on the Beat-error (unless the crystal is really poor), I can only assume that the numbers from the Macbook are reasonable accurate.

I included the "Macbook"-graph in the DD position. I have no idea whether is looks good enough of really bad?

 

[bobm12]

Which takes me to the conclusion that if we can build a pc without fans (it is possible) your reading might actually be more accurate since there won't be any fan noise...(?) I'm I correct?

 

If so, what would be the general requirements to run the software and have a very small box with an ssd drive running this? All solid state and fins to cool.

 

Cheers,

 

Bob

[svorkoetter]

You don't really need a PC without a fan. You just need to put the microphone far enough away from the PC when doing the tests. My PC has a fan, and my microphone is about 30cm away from the PC, and I haven't had any problems.

[Endeavor]

I don't think it's the audio of the fan. I think it's more an internal (magnetic or electrical) interference and probably also a voltage supply issue. The Macbook has a very subtle fan, even though it's older than the HP, and his no issues. The microphone is placed on foam (at least 40cm from the computer) and the computer was placed on a pillow to avoid any vibration transfer via the table.