WatchUSeek Watch Forums banner
1 - 20 of 361 Posts

·
Registered
Joined
·
3,107 Posts
Discussion Starter · #1 ·
I believe there are at least 3 ezChronos owners around and since now I am looking into a 'HEQ firmware' for those models (and not necessarily just for my own watch) there might be a few questions involved:

- who has one (or more) ?

- have you tried programming a different firmware inside (for instance the supplied data-logger) ?

- do you have other temperature-related 'hardware' (similar to the mini-fridge that we have discussed in other threads) ?

- are you willing to do watch accuracy-measurements (or have you done so already - not necessary on the ezChronos but on any quartz watch - for instance with the video method) ?

There are still a number of problems to be solved but it seems that in the absence of a special (dedicated) hardware device to 'calibrate' each watch, the simplest approach would be to assume that the raw quartz from inside has a temperature response that can be well described with something like a*t^2 + b*t + c = 0 , but until we see otherwise we should assume that the a/b/c coefficients above are specific to each quartz and we should each try to measure those with decent accuracy - which IMHO would be feasible by having 3 timing measurements of around 2 weeks each at 3 'temperature points' - and since it is very unlikely that all of us will be able to keep the watch within 0.1 C for a full 2 weeks it is acceptable to keep it in a range of plus or minus 1-2C but with temperature-logging in effect for the interval (which will also help us in calibrating the results against the actual temperature that is seen by the watch). The 3 'target temperatures' should be something like 'fridge temperature', 'room temperature' and 'warm temperature'.

We should also discuss if it might be OK for this thread to become the 'home' for the 'ezChronos HEQ' or if we will just use some other approach.
 

·
Registered
Joined
·
2,473 Posts
Sounds good to me!

Let's see :
1. One, the 915Mhz version
2. Yes, I switched to the datalogger firmware thanks to your instructions (copied below)
3. No, just my fridge, office for room and wrist for worn ;-)
I had it in the fridge for 15 days but unfortunately didn't make a precise note of the time for accuracy testing (I was testing the Precisionist and 8F35), all I can see is that it is now running 20 seconds fast and looking at the Datalogger PC app it seems like I set it on 01/09, presumably at the correct time, so that's not too good. The problem with the datalogger mode is that you can't set the time directly on the watch. Will reset it now.
4. Yes

Other info

Logging capability : logging only temperature at the maximum 1/255s rate I got 11.25 days
start : 12.01.2011,13:29:22,0,17.6,0
end : 23.01.2011,20:55:22,0,3.5,0

How to flash to datalogger firmware :
Copied from another topic :
Catalin : To switch to the data logger you need to go to RFBSL mode (the one where you can program a new firmware over wireless) and then program

\Program Files\Texas Instruments\eZ430-Chronos\Recovery\Chronos Watch\Applications\eZ430_Chronos_ Datalogger_XXXMHz_1_5.txt

The datalogger is an alternate smaller firmware (4 top + 4 bottom modes, vs 7+8 on the 'normal' version) where you have to use a separate PC program to 'sync' (Chronos Data Logger vs. Chronos Control Center) but it can record temperatures - at first I was also logging the altitude (as it could be seen in the file above) but on the current segment I am only logging temperature and I am curious if I can go over one week without having to download and erase the data
 

·
Registered
Joined
·
3,107 Posts
Discussion Starter · #3 · (Edited)
...
3. No, just my fridge, office for room and wrist for worn ;-)
I had it in the fridge for 15 days but unfortunately didn't make a precise note of the time for accuracy testing (I was testing the Precisionist and 8F35), all I can see is that it is now running 20 seconds fast and looking at the Datalogger PC app it seems like I set it on 01/09, presumably at the correct time, so that not too good. The problem with the datalogger mode is that you can't set the time directly on the watch. Will reset it inow.
...
Yes, the code is very tight on the existing datalogger - but the bigger problem with the existing program is that when you set anything (from the PC program) you also change the internal 'temperature calibration' - I was very careful in my first two tests but in the 3rd (ended this Sunday) I have a 0.2C difference in the average :-d (which could also be from the fact that I had in tests two solar watches and I kept them and the ezChronos for a few hours on 2 days in a 'crown up, face to the sun' position instead of the normal 'face down on the warm source' position).

The other problems with the existing datalogger PC program and watch firmware are that reading time can only be done with 1 second accuracy and calibration values can not be read (or set with some decent repeatable accuracy).

Coding a slightly better one proved to be a small pain - the 'free CCS' (code-size limited) has a buggy library and for the moment can not be used reliably, and using msp430-gcc was also quite 'unfriendly' - since I had first to update my cygwin environment and then basically compile msp430-gcc from sources. Right now I am using that on one system and the 'full CCS' on another with a 30-days evaluation license (which can be extended with another 90 days). And all that time lost was before playing with the PC-part (where the existing programs are Tcl/Tk compiled stuff) :roll:

The current plan is to try to have a more usable pair of firmware + PC-program probably by late Saturday / early Sunday and after that at some point start the first of those 3 timing + temperature logging measurements on my ezChronos (I will also double-check on that with the video method, which seems to be usable on it). Of course that the software will also be posted so that other people will be able to start their own tests ;-)
 

·
Registered
Joined
·
15 Posts
Yes, the code is very tight on the existing datalogger - but the bigger problem with the existing program is that when you set anything (from the PC program) you also change the internal 'temperature calibration' - I was very careful in my first two tests but in the 3rd (ended this Sunday) I have a 0.2C difference in the average :-d (which could also be from the fact that I had in tests two solar watches and I kept them and the ezChronos for a few hours on 2 days in a 'crown up, face to the sun' position instead of the normal 'face down on the warm source' position).

The other problems with the existing datalogger PC program and watch firmware are that reading time can only be done with 1 second accuracy and calibration values can not be read (or set with some decent repeatable accuracy).

Coding a slightly better one proved to be a small pain - the 'free CCS' (code-size limited) has a buggy library and for the moment can not be used reliably, and using msp430-gcc was also quite 'unfriendly' - since I had first to update my cygwin environment and then basically compile msp430-gcc from sources. Right now I am using that on one system and the 'full CCS' on another with a 30-days evaluation license (which can be extended with another 90 days). And all that time lost was before playing with the PC-part (where the existing programs are Tcl/Tk compiled stuff) :roll:

The current plan is to try to have a more usable pair of firmware + PC-program probably by late Saturday / early Sunday and after that at some point start the first of those 3 timing + temperature logging measurements on my ezChronos (I will also double-check on that with the video method, which seems to be usable on it). Of course that the software will also be posted so that other people will be able to start their own tests ;-)
If youre interested in simply recording the temperature, there are cheap voltmeters with thermocouple input and data logging using a serial port on a PC. Im sure there are small stand alone temp logger that transfer its data via usb. On these themocouples, they can read 1-2 degrees off. IRC RTDs are more precise.
As far as taking data faster than every 1 second, I think watches has enough mass that its temp will not likely change faster than that, even if the environments temp changes wildly.

I cant wait to get my quartz tender built and start getting some baseline data on my Bulova,

Rob T
 

·
Registered
Joined
·
3,107 Posts
Discussion Starter · #5 ·
If youre interested in simply recording the temperature, there are cheap voltmeters with thermocouple input and data logging using a serial port on a PC. Im sure there are small stand alone temp logger that transfer its data via usb. On these themocouples, they can read 1-2 degrees off. IRC RTDs are more precise.
As far as taking data faster than every 1 second, I think watches has enough mass that its temp will not likely change faster than that, even if the environments temp changes wildly.

I cant wait to get my quartz tender built and start getting some baseline data on my Bulova,

Rob T
The major point is to have the temperature as read by the internal sensor from the watch - since that value will be used to implement the temperature correction :-d

Also the observation on 1-second limit was not related to the temperature readings - the problem is that the API to read the time from inside the watch only returns to 1 second accuracy - so video method is a better alternative than that :-d
 

·
Registered
Joined
·
15 Posts
The major point is to have the temperature as read by the internal sensor from the watch - since that value will be used to implement the temperature correction :-d

Also the observation on 1-second limit was not related to the temperature readings - the problem is that the API to read the time from inside the watch only returns to 1 second accuracy - so video method is a better alternative than that :-d
Sorry- My mistake. I didnt know youre T/C-ing your watch.
Is there a way to correct the pulses in the counters before its formatted for display? or later?

Regards,
Rob T
 

·
Registered
Joined
·
3,107 Posts
Discussion Starter · #8 ·
Sorry- My mistake. I didnt know youre T/C-ing your watch.
Is there a way to correct the pulses in the counters before its formatted for display? or later?

Regards,
Rob T
Surprisingly NONE of the 3 major firmware that I am currently aware of (the original TI full watch plus datalogger and the OpenChronos project) are using the internal hardware clock (RTC_A) but instead use the separate timer (TIMER_A) for 1-second interrupts and do ALL the timekeeping in software - which is slightly simpler to implement/change but is probably a little suboptimal in terms of power consumption - however I guess not by much since the software still needs to update the display every second, so the CPU is still waken-up from low-power mode every 1 second in 'normal mode'.
 

·
Registered
Joined
·
15 Posts
Surprisingly NONE of the 3 major firmware that I am currently aware of (the original TI full watch plus datalogger and the OpenChronos project) are using the internal hardware clock (RTC_A) but instead use the separate timer (TIMER_A) for 1-second interrupts and do ALL the timekeeping in software - which is slightly simpler to implement/change but is probably a little suboptimal in terms of power consumption - however I guess not by much since the software still needs to update the display every second, so the CPU is still waken-up from low-power mode every 1 second in 'normal mode'.
Given that the quartz crystal is probably of the regular 32k variety and with the +-15/month spread, have you figured out how well it would improve if you implement a stardard correction( based on a typical temp curve specs)? Do you know where the peak of the curve is on the temperature line? What is the 'ideal' average daily temp the manufacturer has chosen to use in calculating the error?
BTW- is the device's temp sensor designed to read environmental temp as opposed to the devices internal temp?
 

·
Registered
Joined
·
3,107 Posts
Discussion Starter · #11 ·
Given that the quartz crystal is probably of the regular 32k variety and with the +-15/month spread, have you figured out how well it would improve if you implement a stardard correction( based on a typical temp curve specs)? Do you know where the peak of the curve is on the temperature line? What is the 'ideal' average daily temp the manufacturer has chosen to use in calculating the error?
BTW- is the device's temp sensor designed to read environmental temp as opposed to the devices internal temp?
On the quartz - see my top post on how for the moment we will assume that the raw quartz is reasonably-well approximated by a quadratic equation, and we will try to determine a/b/c for that equation with 3 tests (consisting of about 2-weeks timings each at a temperature 'forced' toward 3 different points) - I am VERY curious on how similar (or dissimilar) the quartz crystals in the ezChronos samples found in this forum will be! :-!

On the temp - the ezChronos actually has inside two temperature sensors - one inside the MSP430 itself (in the ADC_12A part) and one inside the special pressure sensor - and only the second one is currently used. In theory a more special watch design could have tried to have that sensor oriented somehow towards the exterior - however in the ezChronos the sensor is well thermally isolated from the exterior by the plastic case and as such the reading from it is reasonably well correlated to the temperature of the watch, including the quartz - if the results will show that more temperature accuracy is needed it might be possible to also use the MSP430 temperature sensor and do some 'smart averaging' - but let's get there first and worry at that point :-d
 

·
Registered
Joined
·
3,107 Posts
Discussion Starter · #12 ·
...
The current plan is to try to have a more usable pair of firmware + PC-program probably by late Saturday / early Sunday and after that at some point start the first of those 3 timing + temperature logging measurements on my ezChronos (I will also double-check on that with the video method, which seems to be usable on it). Of course that the software will also be posted so that other people will be able to start their own tests ;-)
Small change of plans - the new version of the datalogger seems to be working but since the companion PC-program is a total nightmare I will work on it a few extra days for better testing and I will try to have things ready before the next weekend!
 

·
Registered
Joined
·
3,107 Posts
Discussion Starter · #15 · (Edited)
Do I need any new software to keep up to date with yours?
If you have already started timing tests (or plan on starting today) - it is still possible to use the existing datalogger plus the video method as long as you do not WRITE anything to the watch (in order to keep the same calibration for temperature) and you have a separate precise way to measure the existing calibration (meaning something like the watch is showing 21.3 but your calibrated device suggests 20.8 - so we know that all results are about +0.5 degrees C off).

If you have not - nothing else is needed now - towards the end of the week the new detalogger will be ready and you will need to program it; it will still work with the existing PC datalogger program but will also come with an extra PC program to get some extra info (see the question from webvan above).

And of course the PC should also have a way to sync with atomic internet time - ideally one that can be done 'on demand' and is telling some info on the errors - at least the AboutTime freeware (or ideally the NTP full packet).
 

·
Registered
Joined
·
3,107 Posts
Discussion Starter · #16 · (Edited)
No rush ;-) What did you change in your datalogger version?
Mainly I added in the 'timestamp info' the extra data on fractions of a second so that it might be possible to do the timing tests without the video method - but I hope that at least you and me will still do the video too - in order to have some extra info on the accuracy that we get ;-)

And also I added a way to extract from / set into the watch the information on temperature calibration - without that - each time you set anything (even if only you set the time or you change the datalogging interval) you also alter the temperature calibration!!!

And I have also done so far some other small things and tests - for instance I find very strange the info on the altimeter (which changes even if my house is fixed :-d - the reason for the change being the weather) and I changed that to display either that or the actual pressure (I used mmHg since I am quite familiar with that, with it could also be kPa if people are more familiar with something like that).
 

·
Registered
Joined
·
2,473 Posts
Great! Yes the altitude is pretty useless if the watch is kept in the same spot, haven't paid much attention to it.

The timestamp is a good addition, but how much logging time is it going to remove? I suppose there is no way to go around the 1/255s barrier? (256 bit coding).

As for temperature I have my watch sitting next to a Honeywell thermometer and after some effort they stay witin 0.1 dC. The problems I've had came after I'd been holding the watch in my hands so it warmed up. Had to put it down and let it settle before setting it again.

What's difficult when setting the time with the Datalogger PC app is that you need to hit sync to PC (after setting the PC time with Abouttime of course) and then set the watch as quickly as possible, hard to do that in less than a second, best I've done is 0.765 so far!

I just did a quick accuracy check after 4 days of testing :
Average temperature : 21.9dC
Drift equivalent in spy : 745 !

That's really bad, +2spd, how is yours performing?
 

·
Registered
Joined
·
3,107 Posts
Discussion Starter · #18 · (Edited)
Great! Yes the altitude is pretty useless if the watch is kept in the same spot, haven't paid much attention to it.

The timestamp is a good addition, but how much logging time is it going to remove? I suppose there is no way to go around the 1/255s barrier? (256 bit coding).
The extra time info is not logged (since the moment when the temperature reading was done is not relevant to millisecond accuracy) - it is only used in the direct communication with the PC program so as to calculate and set the time on the watch with enough accuracy! (as you also noted).

However I believe that interval could be increased with the right coding - I will look into it! (the main problem however is the PC application - for the moment it seems that I will not create something similar to that Tcl/Tk one).

As for temperature I have my watch sitting next to a Honeywell thermometer and after some effort they stay witin 0.1 dC. The problems I've had came after I'd been holding the watch in my hands so it warmed up. Had to put it down and let it settle before setting it again.

What's difficult when setting the time with the Datalogger PC app is that you need to hit sync to PC (after setting the PC time with Abouttime of course) and then set the watch as quickly as possible, hard to do that in less than a second, best I've done is 0.765 so far!
For me the temperature consistency is equally as important if not more - since that will be used to determine the coefficients for the quadratic equation of the quartz and any error there can be vastly amplified!

I just did a quick accuracy check after 4 days of testing :
Average temperature : 21.9dC
Drift equivalent in spy : 745 !

That's really bad, +2spd, how is yours performing?
Mine seems to be even worse - over +3 s/day (around 31-32 Celsius).

However that has to be expected - we now see the 'raw quartz' - not even corrected with the 10-second inhibition interval which is the standard in every single 1$ quartz ever since around 1980!

I will have to add on the bottom a mode with hour:minutes (since at some point with such raw errors it becomes tricky for the video method when only seeing the seconds and asking yourself if you are 10 seconds ahead or 70 seconds ahead :-d).
 

·
Registered
Joined
·
202 Posts
I have the 868 Mhz version but temperature and time precison is not my grail, well... not yet... ;-)
I was busy these last months so stopped to play with the ezChrono, my last work was to reduce the firmware size, removing useless stuffs like bluerobin, heart rate monitoring...
 
1 - 20 of 361 Posts
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top