WP34s Differential Time and shift

[MarkHaysHarris777]

My two Timex watches (Ironman Shock) are handy because their time(s) shift is opposite -- one is half second fast per day, the other half second slow per day. The difference between them tells me when they were synchronized, and the coordinated UTC, for my zone, is then half way between. A third timepiece (like the WP34s with crystal) can then be set accurately +/- second from the other two as absolute reference based on their differential times (or alternately, the WP34s once determined may be used as one of the differential time markers.

I am noticing that the time shift for my WP34s is almost precisely 1 sec fast per day; half minute fast per month (on average). A program can be written which based on the date and 34s time, along with the difference time in a second time piece (input data) could then reset the 34s for correct present time eliminating the shift mostly.

I am wondering if I can narrow the shift a bit on the 34s short of replacing the crystal... how the two caps affect the timing... and whether it would be easy to have the firmware (configurable) be able to adjust for the shift ??

TIA

cheers
marcus

[Katie Wasserman]

(02-09-2015 11:43 PM)MarkHaysHarris777 Wrote:  I am wondering if I can narrow the shift a bit on the 34s short of replacing the crystal... how the two caps affect the timing... and whether it would be easy to have the firmware (configurable) be able to adjust for the shift ??

TIA

cheers
marcus

I used a trimmer cap for that. Here's a thread showing what it looks like. Not pretty but effective.

[matthiaspaul]

(02-09-2015 11:43 PM)MarkHaysHarris777 Wrote:  I am noticing that the time shift for my WP34s is almost precisely 1 sec fast per day; half minute fast per month (on average). A program can be written which based on the date and 34s time, along with the difference time in a second time piece (input data) could then reset the 34s for correct present time eliminating the shift mostly.

In fact, some decades ago I wrote a small DOS utility named CORRTIME to silently correct the RTC time of a PC on every bootup and further adjust its correction factor whenever the clock was set manually. So, yes, it's definitely doable... ;-)

However, from the electrical design perspective the correct solution is what Katie suggests, "pulling" the crystal with a trimmer cap. (Of course, there are other solutions as well, but they consume more energy and are therefore not well-suited for a calculator powered by coin cells.)

Quote:and whether it would be easy to have the firmware (configurable) be able to adjust for the shift

Hm, a function to receive a time packet from an attached emulator running on a PC (and having an NTP, GPS, DCF77, RDS etc. synchronized time) might be a nice feature. Either by setting a special "synch time" flag and working silently in the background, or by issuing a specific command.

Greetings,

Matthias

[Dave Frederickson]

The HP 75C/D and 71B both employ software mechanisms to calibrate their RTC's.

(02-10-2015 03:27 AM)Bit Wrote:  To be even more useful, this could be included in the firmware and the correction could be automatically performed (if necessary) whenever the date or time is queried.

Just like the 71B and 75.

[BarryMead]

(02-09-2015 11:43 PM)MarkHaysHarris777 Wrote:  A program can be written which based on the date and 34s time, along with the difference time in a second time piece (input data) could then reset the 34s for correct present time eliminating the shift mostly.

If you reserved a register for "Last Correction Time" (Say R99) then whenever you run your program
it could use the difference between R99 and current time and compute the correction factor, then it
could update the current time and save the current time into R99 for the next calibration.

This way your calculator's time gets corrected whenever you run the update program, but consumes
only the normal idle/off power in the mean time.

[Bit]

(02-10-2015 01:58 AM)BarryMead Wrote:  

(02-09-2015 11:43 PM)MarkHaysHarris777 Wrote:  A program can be written which based on the date and 34s time, along with the difference time in a second time piece (input data) could then reset the 34s for correct present time eliminating the shift mostly.

If you reserved a register for "Last Correction Time" (Say R99) then whenever you run your program
it could use the difference between R99 and current time and compute the correction factor, then it
could update the current time and save the current time into R99 for the next calibration.

This way your calculator's time gets corrected whenever you run the update program, but consumes
only the normal idle/off power in the mean time.

To be even more useful, this could be included in the firmware and the correction could be automatically performed (if necessary) whenever the date or time is queried.

[MarkHaysHarris777]

(02-09-2015 11:51 PM)Katie Wasserman Wrote:  I used a trimmer cap for that. Here's a thread showing what it looks like. Not pretty but effective.

hi Katie, nice to meet you, and thanks! I was hoping for a pic like that... 'preciate it! Well, its like my dad used to tell me when we would fly together (SuperCub) back in the 70's (after bouncing to a landing in a cross wind, followed by a wing-tip ground loop)... "son, any landing you walk away from is a good landing!" Amen.
So, any circuit that gets the job done (effective) is beautiful! I noticed the stuff you did with the printer... did the 'pull' up of the crystal affect the ability to IR print in any way? yes, I was hoping the timing would be at least as precise as a Timex watch, but on the other hand (and this is important) the WP34s is the only calc I have that has timing at ALL!


Cheers
marcus

[matthiaspaul]

(02-10-2015 01:44 AM)Dave Frederickson Wrote:  The HP 75C/D and 71B both employ software mechanisms to calibrate their RTC's.

Interesting! I just learned that a similar software for the HP 49G/HP 49g+ was written by Heiko Arnemann some years back. Besides software clock adjustment, it also adds support for daylight saving time (DST) and timezone management:

http://users.belgacom.net/ea/Heiko/HP49/index.htm CLCKADJST 3.1 (2007-08-20)

An older version is also available from HPCALC:

http://www.hpcalc.org/details.php?id=4977 CLCKADJST 3.0 (2006)

Greetings,

Matthias

EDIT:

Found another solution for the HP 48S/SX:

http://www.hpcalc.org/details.php?id=2698 CLK by Hewlett-Packard (1990)

[matthiaspaul]

(02-10-2015 01:58 AM)BarryMead Wrote:  If you reserved a register for "Last Correction Time" (Say R99) then whenever you run your program it could use the difference between R99 and current time and compute the correction factor, then it could update the current time and save the current time into R99 for the next calibration.

Yes, and once the semi-constant calibration factor for a particular unit has been established after a few weeks, it should be stored in flash, thereby freeing up the register again.

I recall that my old CORRTIME utility could optionally maintain two correction factors, depending on if the PC was switched off and on. This was down to my observeration that the clock was off by different factors depending on if the machine was powered or the RTC running on its backup battery, probably due to the slight difference in voltage (constantly high versus variable and lower). However, this required to run the tool before switching off the PC as well.

So, for most accurate results, we might take the battery voltage and the temperature into account as well. Also, a temperature sensor might be a nice addition to the WP 34S in general.

As far as I see, the WP 34S' hardware has enough free I/Os left for use. Since the firmware already includes code for the analog/digital conversion of the battery voltage, an analog temperature sensor like the Texas Instruments TMP20 (http://www.ti.com/product/TMP20/technicaldocuments) appears to be more desirable than using one with digital read-out (via I2C or a similar protocol). The code overhead to read out the sensor should be minimal, and the electrical interface is simple as well: One pin for power (via a port output to reduce power consumption), another for GND, and a third pin for the analog signal. Fully operational the device drains 4 µA, and truely zero ampere if powered off. Basically, anyone who could install the crystal would be able to add the temperature sensor as well.

Greetings,

Matthias

[Steve Simpkin]

For many years I had a simple program that would automatically launch every night at 2:00am on my HP-48SX and nudge the system time by the average amount it was off per day. I refined the correction value over many months and the result was a very accurate time clock.

[matthiaspaul]

(02-11-2015 09:07 PM)matthiaspaul Wrote:  Since the firmware already includes code for the analog/digital conversion of the battery voltage

I meanwhile had a look at the sources and have to correct myself. The code to detect the battery voltage does not use one of the four free ADC inputs, but the internal brown-out detection logic. One of the advantages of this approach is that no external voltage divider is needed, so less energy is wasted (the reference voltage for the ADCs is clamped to 1.8V).

However, this means that there is no particular advantage of using an analog temperature sensor over other types of sensors. The eight spare I/O pins include the signals necessary for SPI. An I2C interface would be available in parallel to the ON key (and would therefore be difficult to use).

Greetings,

Matthias

[Katie Wasserman]

(02-10-2015 06:35 AM)MarkHaysHarris777 Wrote:  

(02-09-2015 11:51 PM)Katie Wasserman Wrote:  I used a trimmer cap for that. Here's a thread showing what it looks like. Not pretty but effective.

hi Katie, nice to meet you, and thanks! I was hoping for a pic like that... 'preciate it! Well, its like my dad used to tell me when we would fly together (SuperCub) back in the 70's (after bouncing to a landing in a cross wind, followed by a wing-tip ground loop)... "son, any landing you walk away from is a good landing!" Amen.
So, any circuit that gets the job done (effective) is beautiful! I noticed the stuff you did with the printer... did the 'pull' up of the crystal affect the ability to IR print in any way? yes, I was hoping the timing would be at least as precise as a Timex watch, but on the other hand (and this is important) the WP34s is the only calc I have that has timing at ALL!


Cheers
marcus

Nice to (sort of) meet you too Marcus! The IR is sensitive to timing and modulation frequency but not to anywhere near the extent it's possible to pull a crystal. In fact I can get the HP IR printer to work using an RC clocked microcontroller with frequency variations up to 2%. The verbage in the IR printer specs about the tight timing accuracy needed is just to maximize the range. It'll work with all sorts timing inaccuracies.