are the existing AA batteries alkaline? you could try switching to energizer lithium, they are a hair higher in voltage

and/or switch to a 3xAA holder - in any case 3.0V is not gonna be enough to keep the clock in good time on the other side of the usual diode(s)

maxtherabbit wrote on 2022-01-10, 23:07:

are the existing AA batteries alkaline? you could try switching to energizer lithium, they are a hair higher in voltage

and/or switch to a 3xAA holder - in any case 3.0V is not gonna be enough to keep the clock in good time on the other side of the usual diode(s)

Current ones are alkaline. I ordered some battery holders for 3xAA and 4xAA so will use one of those.

But which one to use? Because sometimes I see people also use 4xAA.

4 would probably work, but is in mild overvoltage territory when using lithium

If using alkaline it's fine, but alkaline are trash and can leak

maxtherabbit wrote on 2022-01-11, 00:13:

4 would probably work, but is in mild overvoltage territory when using lithium

If using alkaline it's fine, but alkaline are trash and can leak

Replaced the 2 alkaline with 3 lithium batteries. But still the clock runs too slowly. It's still getting behind a couple of minutes an hour or so.

Should I try 4 batteries or could the problem be the oscillator? Anyone knows how to "debug" such a problem? Don't have an oscilloscope btw.

Yoghoo wrote on 2022-01-13, 22:13:

maxtherabbit wrote on 2022-01-11, 00:13:

4 would probably work, but is in mild overvoltage territory when using lithium

If using alkaline it's fine, but alkaline are trash and can leak

Replaced the 2 alkaline with 3 lithium batteries. But still the clock runs too slowly. It's still getting behind a couple of minutes an hour or so.

Should I try 4 batteries or could the problem be the oscillator? Anyone knows how to "debug" such a problem? Don't have an oscilloscope btw.

3 lithiums is about 5 volts, which certainly should be enough

you can safely try 4 alkalines, that will get you up to 6V

maxtherabbit wrote on 2022-01-13, 22:49:

3 lithiums is about 5 volts, which certainly should be enough

you can safely try 4 alkalines, that will get you up to 6V

Now using 4 alkaline batteries but the problem is unfortunately not fixed. It seems like it's keeping the time quitte ok when the computer is idle or lightly used. When doing benchmarks etc the time gets behind with a couple of minutes each time.

No idea what the problem could be. Anyone has an idea where to look further?

Pretty sure that when the computer is on then the main PSU takes over from the battery for the RTC. So it shouldn't make a difference if you're doing benchmarks or not. Maybe noise on the power supply is causing problems for the RTC oscillator?

Have you got pictures of the board, with some close ups around the 32kHz crystal that will probably be on there somewhere (likely a small cylinder on its side)?

snufkin wrote on 2022-01-27, 19:09:

Pretty sure that when the computer is on then the main PSU takes over from the battery for the RTC. So it shouldn't make a difference if you're doing benchmarks or not. Maybe noise on the power supply is causing problems for the RTC oscillator?

Have you got pictures of the board, with some close ups around the 32kHz crystal that will probably be on there somewhere (likely a small cylinder on its side)?

See below for the mainboard. The 32.8k crystal is on the left from the 8 bit ISA slot. Is that enough information?

So between the crystal and (U1? maybe?) they should generate a 32kHz signal that goes to pin 72 of the 82C206. That chip stores the value for the RTC, along with other CMOS stuff. Pins 32 and 75 should be powered by the board when the PSU is on (probably read a bit less then 5V as there'll be a diode somewhere) and by the battery when the PSU is off. If the clock is significantly drifting then I guess that it'll either be the crystal, the support stuff for the crystal (there'll be some capacitors, maybe c6?) or the 206 itself.

2min/hour is a pretty big drift. Does it loose that overnight? Say if it's off for 8 hours, does it end up 15min out? If it keeps time overnight, then maybe the problem could be the power supply?

Incidentally, the board looks very similar to https://www.ultimateretro.net/en/motherboards/1251 , looking at the jumper numbering. Although JP25 (CMOS clear) is missing. There's no BIOS for that board on UltimateRetro, so if you can get a copy of yours then that'd be useful.

[edit: turns out there a quite a lot of similar boards, of which the Shuttle HOT-304 has manual and bios: https://www.ultimateretro.net/en/motherboards/4423 and the Octek Trout 386 has a manual: https://www.ultimateretro.net/en/motherboards/656#downloads ]

snufkin wrote on 2022-01-27, 22:36:

So between the crystal and (U1? maybe?) they should generate a 32kHz signal that goes to pin 72 of the 82C206. That chip stores […]
Show full quote

So between the crystal and (U1? maybe?) they should generate a 32kHz signal that goes to pin 72 of the 82C206. That chip stores the value for the RTC, along with other CMOS stuff. Pins 32 and 75 should be powered by the board when the PSU is on (probably read a bit less then 5V as there'll be a diode somewhere) and by the battery when the PSU is off. If the clock is significantly drifting then I guess that it'll either be the crystal, the support stuff for the crystal (there'll be some capacitors, maybe c6?) or the 206 itself.

2min/hour is a pretty big drift. Does it loose that overnight? Say if it's off for 8 hours, does it end up 15min out? If it keeps time overnight, then maybe the problem could be the power supply?

Incidentally, the board looks very similar to https://www.ultimateretro.net/en/motherboards/1251 , looking at the jumper numbering. Although JP25 (CMOS clear) is missing. There's no BIOS for that board on UltimateRetro, so if you can get a copy of yours then that'd be useful.

[edit: turns out there a quite a lot of similar boards, of which the Shuttle HOT-304 has manual and bios: https://www.ultimateretro.net/en/motherboards/4423 and the Octek Trout 386 has a manual: https://www.ultimateretro.net/en/motherboards/656#downloads ]

When not used the time is hardly drifting. I only see it drifting when using the pc. It can be a couple of minutes but at one time it was like 10 minutes when running a bunch of dosbench tests etc.

Is there a "how-to" somewhere to make a copy of the bios? And were should I put it?

Thanks for pointing out the similar boards. Funny that there are so many boards looking the same.

I've noticed some of the DOSBENCH tests can adversely affect the DOS clock

There should be 1 or 2x 1N4148 signal diodes in the circuit path between your battery header and the RTC chip. Check them

maxtherabbit wrote on 2022-01-28, 16:04:

I've noticed some of the DOSBENCH tests can adversely affect the DOS clock

SPEEDSYS can do that, it performs RTC Y2K test by setting the date to last second of 1999 and waiting for the clock to advance (and it will hang if the RTC isn't running properly). Possibly it doesn't properly set the seconds counter after restoring the original date/time?

Usually DOS only reads the RTC at boot and then keeps it's own time using timer interrupt. That timer can be wrong speed if the mobo clock generators were messed with (this would also affect any timed benchmark results, obviously). It shouldn't affect the RTC since the values will not be written back by DOS (so the two clocks are actually running in parallel) but any testing software that does RTC Y2K check or something similar could be using DOS-managed time to restore the RTC. And then it would look like it's going too fast or too slow but only during PC uptime.

First thing to check would be if the DOS-manged clock is actually running properly. Easiest way would be to start Norton Commander, note the time in upper right corner, and leave it running for a few hours, then come back and check time again. If it's not too fast or too slow then it must be a bug during RTC test I mentioned above.

The usual design for the 82C206 as per the datasheet calls for a 6V battery and 2 1N4148 diodes in series between the battery and the Vcc pin. I'm currently troubleshooting an issue on one of my 386 boards since using 2 x CR2032 coin cells "works" but the battery gets completely drained in about a week (using 4.5V or less is not enough to keep the RTC ticking).

I've noticed that the voltage drop from the diodes doesn't seem to be enough to keep Vcc under the operational range specified in the datasheet (4.75V to 5.25V). Two brand new button cells in series measure about 6.4V and I get around 5.7V in Vcc after the diode drop. Although the absolute max voltage to the 82C206 is 7V, it makes me wonder if straying too far from the operational 5V can cause the battery to be drained so fast, I'm currently considering testing with a 3rd diode in series.

I agree with the suggestion above, check the diodes and what voltage the Vcc pin of the 206 is getting when the main power is on and off.

Deunan wrote on 2022-01-28, 16:28:

SPEEDSYS can do that, it performs RTC Y2K test by setting the date to last second of 1999 and waiting for the clock to advance ( […]
Show full quote

maxtherabbit wrote on 2022-01-28, 16:04:

I've noticed some of the DOSBENCH tests can adversely affect the DOS clock

SPEEDSYS can do that, it performs RTC Y2K test by setting the date to last second of 1999 and waiting for the clock to advance (and it will hang if the RTC isn't running properly). Possibly it doesn't properly set the seconds counter after restoring the original date/time?

Usually DOS only reads the RTC at boot and then keeps it's own time using timer interrupt. That timer can be wrong speed if the mobo clock generators were messed with (this would also affect any timed benchmark results, obviously). It shouldn't affect the RTC since the values will not be written back by DOS (so the two clocks are actually running in parallel) but any testing software that does RTC Y2K check or something similar could be using DOS-managed time to restore the RTC. And then it would look like it's going too fast or too slow but only during PC uptime.

First thing to check would be if the DOS-manged clock is actually running properly. Easiest way would be to start Norton Commander, note the time in upper right corner, and leave it running for a few hours, then come back and check time again. If it's not too fast or too slow then it must be a bug during RTC test I mentioned above.

I didn't run SPEEDSYS as it's not working on DR-DOS 6.0. Did some other test today:

1) PC was off for 2 days as I am working on another retro PC. Did a ntp synchronize after turning it on and the difference was only 1 second.
2) Ran NC for four hours and running a ntp synchronize again. The difference was 7 seconds.
3) Did a whole bunch of benchmarks, music playing, games, Windows 3.0 and GeoWorks. After a ntp synchronize the difference was again 1 second.

So at the moment I can't reproduce the big gaps I saw earlier a couple of times (in the 10-20 minutes range). Will keep an eye out of course and will check the diodes when I will disassemble the PC again.