RTC Running slow on Apollo 1260

[rabidgerry]

Developments with this issue so far.

After much discussion and thought I ended up replacing the clock chip on the Apollo. So far things seem good. I replaced the OKI chip for a Epson chip and literally swapped and did nothing else.

The card is holding time it seems.

First test it held time over an hour, then another half an hour, then over night. However over night it was perhaps 8 - seconds behind. Then I dunno what happened but it ended up 10 minutes behind whilst the machine was still on. I have noticed this before though after playing games the clock can be behind all of a sudden.

Anyways I reset it this time and then set the time and date. All day today it has remained a nice 5 seconds faster than my phone.

Is the issue completely resolved? I do not know. Lets give some more time and see. One thing is certain that it is not doing what it was before.

The scalos clock module still remained frozen which was bizarre so I just deleted it, although it was nice to see the clock in the title bar.

More updates throughout the day with this.

Humungous thanks to Daedalus and Paul1981 for all their help. If it wasn't for Daedalus I would not have tried the Epson chip in place of the OKI and if it wasn't for Paul1981 I wouldn't even know where the clock was on the card and what was what! Also thanks for the general mentoring too.

Fingers crossed on the outcome.

[daxb]

Notice that programs/games exists that stop the clock as long as they run. They are rare but just in case.

[rabidgerry]

Originally Posted by daxb

Notice that programs/games exists that stop the clock as long as they run. They are rare but just in case.

I find this happens with Micropose Formula 1.

So you are saying this is rare but can happen with some games/programs?

[daxb]

Yes, the coder don't know or care for it. In the old floppy only days it wasn't necessary because there was no exit to WB or DOS and usually a reset was done to load something else. As far as I know the program/game use/stop the a clock timer. After exit you get the time before the program/game was started. A reset or with "SetClock <LOAD|RESET>" should fix this.

Btw. if your Amiga is connected to internet/LAN or plan to do. You can/should set the clock via a time sync server. Maybe your router provides this service already. So you don't need to adjust the clock manually.

[hooverphonique]

The RTC (Oki, Ricoh, etc) is loaded into the CIA chip TOD timer (time of day) on boot - the TOD doesn't progress correctly if certain interrupts are disabled or TOD registers are meddled with, which is why some hardware-banging software will cause the time of day to get skewed..

As dabx said, you can do a setclock load after exiting the software in question to reload the TOD. If it is the actual RTC that runs slow, then that doesn't fix it of cause.

[nogginthenog]

If you have ethernet you can just set the time on each boot. That's what I do in my battery-less 4000.

[rabidgerry]

Na the whole thing just seems f**ked now. It's lagging behind again having held it for at least 24 hrs with tests. Attempted to lift pins 16 and 17 to bypass the old crystal that was still on there and well that seemed like to big a job after fitting the dam thing in the first place and removing the old one. Not sure what else is left to try but it's definitely screwed. Might be easier to remove the old crystal in case of interference from it possibly? I dunno. The thought of using solder wick again to try and de-solder that whole chip is giving me nightmares. I don't know enough to diagnose the causes or other possibilities. Could be time to give up.

Ethernet, well I don't have that setup and wouldn't be pursuing that direction I don't think, although I get it is a solution to not having an RTC.

[nogginthenog]

Note: Not familiar with this board.

Traditionally a clock circuit (like in a digital watch) would have a crystal with the frequency of 32.768 khz. Not seeing anything like this on the pictures of this board.

What frequency is the crystal connected to pins 16/17? Could it simply be that the 50Mhz (or whatever you have) is not accurate?

[rabidgerry]

Well I have circled what I believe to be the real time clock crystal. This is not my own personal card but it's the same as mine.



However the chip that I swapped out has it's own crystal so this old one should not matter.

No idea about the frequency of the old crystal or how to test that out.

[Exodous]

I can't see what model clock chip you swapped for, but did you just swap the chip out or make some other modifications?

If an external oscillator is present, some chips will still use that, even if they have an internal oscillator - to bypass it you usually have to tie the XT and _XT pins low (ground)

To do this, you would need to remove the oscillator and associated capacitors and then tie the chip side of capacitors to ground, or bend up or snip the pins on the chip so they don't go into the circuit board and solder them directly to a ground pin.

[rabidgerry]

Sorry I should have put that in ages ago.

The chip I swapped out for the OKI is an Epson RTC72421B.

I have been getting help from Daedalus regarding this issue of the chip swap and I know he has done the same thing as me with successful results and no removal of existing parts such as crystal.

Albeit I think he done the swap on different pieces of hardware, so not an Apollo 1260 like me.

Perhaps the new chip is trying to use the old crystal??

see here for a little more info

http://eab.abime.net/showthread.php?t=85395&page=9

[nogginthenog]

Definitely a 32.768 khz clock crystal. These are very accurate so I don't think that is the problem.

[rabidgerry]

Well the dead ends keep comin'

[Exodous]

This chip, with an internal osciallator is accurate to +/-20ppm (pulses per million) which works out at 1.728 seconds accuracy per day, so at most it should gain or lose that much time every day.

On boot up the Amiga should run "setclock load" to get the time from the chip and into the on-board clock circuit. The CIA chips then keep it running whilst the machine is powered on.

As noted by other posts above, if software plays with the hardware, then it is possible for it to become inaccurate during that session, but on the next boot it should be back to the correct time.

The time should only be written back to the clock chip when ussing a "setclock save" command.



Looking at the application notes at http://www.west-l.com/uploads/tdpdf/ih680254.pdf , the bottom of page 2 says pins 16 and 17 should not be connected externally as they are connected internally to VDD (+5v nominally) - I'm pretty sure these are where the oscillator circuit would be connected on the original chip.

I don't know what impact it would have on the circuit with them connected, as logic says if they're not supposed to be connected to anything, they should just be dummy within the chip packaging and not connect internally, but it may be worth trying to disconnect them and seeing if that makes any difference.

[rabidgerry]

Originally Posted by Exodous

This chip, with an internal osciallator is accurate to +/-20ppm (pulses per million) which works out at 1.728 seconds accuracy per day, so at most it should gain or lose that much time every day.

On boot up the Amiga should run "setclock load" to get the time from the chip and into the on-board clock circuit. The CIA chips then keep it running whilst the machine is powered on.

As noted by other posts above, if software plays with the hardware, then it is possible for it to become inaccurate during that session, but on the next boot it should be back to the correct time.

The time should only be written back to the clock chip when ussing a "setclock save" command.



Looking at the application notes at http://www.west-l.com/uploads/tdpdf/ih680254.pdf , the bottom of page 2 says pins 16 and 17 should not be connected externally as they are connected internally to VDD (+5v nominally) - I'm pretty sure these are where the oscillator circuit would be connected on the original chip.

I don't know what impact it would have on the circuit with them connected, as logic says if they're not supposed to be connected to anything, they should just be dummy within the chip packaging and not connect internally, but it may be worth trying to disconnect them and seeing if that makes any difference.

Ok well I already knew about the software making the clock slow with certain titles but on a working RTC I knew that a soft reset corrected the time again. Only one game I know of or have experienced recently was doing this but ultimately this isn't an issue. That was more just a passing comment I guess.

What I seem to be experiencing now is the clock is going very slow again.

So for reference I was able to get it to hold date and time for just over 24hrs before it slipping behind say 30 seconds all of a sudden.

Other behaviour I noticed was soft reset would then set it behind even further.

Right now what I am experiencing is it's falling behind after cold boots and soft resets. Sometimes after soft resets it will not move on during the few seconds it takes for it to boot again, it will literally start again from where you last saw it before the soft reset. Other times I will set the clock and have the Amiga switched on say for an hour. Then I soft reset to test it out and I immediately will see the clock fall behind another 30 seconds or something similar. If I reset again it will fall behind a little bit more before doing the behaviour I noted above where it picks up from where it left off and doesn't seem to continue with time through the boot.

I know last weeks tests were using a workbench disk to boot from to try and eliminate software but I would argue that with other RTC on a different accelerator it was behaving fine.

I will see if I can disconnect those pins properly then. My first attempt didn't work out. I may have to remove the entire chip again.


****I checked the Amiga time there having reset the chip last night at some stage and the bloody thing is in time and has held time over night which was a time of 4-3.5 seconds ahead of my phone - the mystery deepens****

[Exodous]

I've just been looking and I've not got a "setclock load" in my startup sequence. Having just looked back, it also wasn't in a backup from my 3.0 ROM A1200 back in the mid 90's when I had a 4MB RAM board with RTC.

I wonder if 3.0+ ROMs get the clock time on startup automatically and doesn't do it on a soft reset - what happens if you open a shell/cli and run "setclock load" does the clock change?

[crazyc]

Seems like quite a palaver for a RTC. Is it possible to just run one off your motherboard clockport and ignore the one on the apollo?
I dont know if this would cause any issues etc so hope someone more knowledgeable will comment.

[rabidgerry]

Originally Posted by Exodous

I've just been looking and I've not got a "setclock load" in my startup sequence. Having just looked back, it also wasn't in a backup from my 3.0 ROM A1200 back in the mid 90's when I had a 4MB RAM board with RTC.

I wonder if 3.0+ ROMs get the clock time on startup automatically and doesn't do it on a soft reset - what happens if you open a shell/cli and run "setclock load" does the clock change?

As far as I know setclock load just displays the time in the shell window right?

Also I have been setting my time using the "date" command

so I have been doing this

date 21:00:00

then enter

other times I have been setting the time by using Amiga Test Kit tools where you reset the chip and then can set the time.

I will try this command now.

Just for the record the last time I checked the RTC was still keeping time.

Originally Posted by crazyc

Seems like quite a palaver for a RTC. Is it possible to just run one off your motherboard clockport and ignore the one on the apollo?
I dont know if this would cause any issues etc so hope someone more knowledgeable will comment.

You're right But I'm now so frustrated I kind of want to get to the bottom of the issue.

I tried that and I think what happened was it held onto the time correct but was wrong in scalos. I also got rid of the scalos clock program thinking it was perhaps corrupt but I don't think it was actually at all. I just think what ever this issue I have was freezing that display clock in the title bar. I am currently getting my time now from Dclock which loads on boot. It basically is just an automatic clock display that you stick the icon for it in wb-startup drawer and it loads the ACTUAL clock time for you without having to go in and look for it. It was my substitute for the scalos clock plug in.

[rabidgerry]

ok typing "setclock load" didn't do anything.

What is supposed to happen?

[Exodous]

"setclock load" should take the time from the chip and put it into the OS, with no output unless there is an error. You then follow it with "date" to display the date.

Just setting the date using "date" only sets it for that session, "setclock save" writes it to the RTC module.

Amiga Test Kit will be effectively doing the same when it sets the clock.