Battery backed RTC not keeping time

[Bryan Smith]

Hey Guys,

My Minnowboard fails to keep time when power is removed from the board. I have a battery installed but not matter what happens it goes back to the date 1/1/2000 and time 00:00:00. I have the battery in the right way, but I even tried flipping it around to see if it made a difference. I even played around with putting a Jumper across J2 just to see if it had an effect but nothing works. I am running a RAID array from the Minnowboard and it's killing my I/O because the array has to keep resyncing due to the timestamp of the array being in the future.

I believe I have a Rev A1 board; is there a fix for this and has anyone else observed this?

Bryan 

[tak...@canalian.sakura.ne.jp]

My MinnowBoard behaves in the same way as your MinnowBoard.

RTC is forgetted.

I use date command and hwclock command on Debian7.

[sels...@gmail.com]

On 24/11/13 05:20, Bryan Smith wrote:
> I believe I have a Rev A1 board; is there a fix for this and has anyone else observed this?

I sort of noticed it, but thought I'd maybe done something wrong, or had forgot to copy the clock back to the rtc. Haven't had chance to investigate further yet.

[Bryan Smith]

Hey Guys,

The battery doesn't keep time within the distro or even in the UEFI. I also tested this from the UEFI shell with the date and time commands from the prompt. 

Hit escape at the count down before the board attempts to find the boot partition.

Shell> time

00:00:06 (UTC 00:00)

If you press the reset button and then interrupt the boot sequence again and run the time command you will see that the time is in fact being kept because power has not been removed from the board.

Disconnect the power adapter then run time again and you'll see that your clock has been reset back to 00:00:00.

If you install ntp it'll set the time for you on boot once the network comes up but that is wayyy too late in most cases. Even though that works the time fails to be retained when power is removed, so that is definitely not a fix.

I am running Angstrom, Debian, Arch, Ubuntu and Fedora on the Minnow and they all exhibit this behavior. If it doesn't keep time in UEFI it surely won't in the distro

Bryan

[Scott Garman]

Bryan and all: I've notified the folks at CircuitCo about this and
they're investigating the issue this week. I'll report what is learned
once I get further details.

Scott

--
Scott Garman
Embedded Linux Engineer - Yocto Project
Intel Open Source Technology Center

[Tord Andersson]

On Wednesday, December 4, 2013 12:39:35 AM UTC+1, Scott Garman wrote:

I'll report what is learned
once I get further details.

Any news on this subject? Tried to install RTC battery (BR1225) on rev A1 board with - terminal where + is printed (silkscreen mismatch as stated before).
Time was set with hwclock -w, read back succesfully with hwclock -r. Unfortunately, after removing the power, hwclock -r showed that the time was lost.

[Scott Garman]

Hi Tord,

This appears to have been a hardware issue on some batches of
manufactured boards, but not others. You can get a replacement board by
contacting CircuitCo at r...@minnowboard.org. Apologies for the
inconvenience.

Scott

[sels...@gmail.com]

On 09/01/14 16:08, Scott Garman wrote:
>
> This appears to have been a hardware issue on some batches of manufactured boards, but not others. You can get a replacement board by contacting CircuitCo at r...@minnowboard.org. Apologies for the inconvenience.

Scott,

Any ideas if it's fixable without an rma ? RMAing the board is rather dependant on where in the world you are and if it was bought through a distributor or whatever.

If it's something like just replacing D9 I'd opt to do that myself, if it's a problem with a bga or the raw pcb that's obviously less likely to be fixable.

[Scott Garman]

Hi Tord,

CircuitCo hasn't fully investigated the cause of this yet, and it's also
quite possible the RTC failed for different reasons on different boards.
So the most expedient way to resolve this will have to be through an RMA.

[sels...@gmail.com]

On 09/01/14 18:06, sels...@gmail.com wrote:

> If it's something like just replacing D9 I'd opt to do that myself, if it's a problem with a bga or the raw pcb that's obviously less likely to be fixable.

Update on this, D9 on my board has L4Z marking. Any information I can find suggests this is a BAT54, not the BAT54C that it's supposed to be which would be marked L43.

The BAT54 is a single diode between pins 1 & 3. Pin 2 being not connected. Observation with something as simple as a multimeter is that I have battery voltage at pin 2, but not at pin 1. Further investigation with a diode tester shows that it's not a BAT54S either, pin 2 really does appear to be completely isolated.

My board unsurprisingly loses it's date and time across a power cycle.

So I replaced D9 with a BAV70 which was the closest pin-compatible part I had available. It's not ideal as it has a higher voltage drop than the BAT54C.

My board now retains the date/time across a power cycle.

Conclusion: either someone ordered the wrong part, assuming all BAT54* parts were identical, or someone loaded the wrong reel onto the placement machine. I suppose it's also possible the diode manufacturer mis-labeled the reel, but either way it's likely to be a simple human error.

It's relatively simple to fix for those of us not worried about the warranty. That said, it's a manufacturing defect, so I'd be expecting CircuitCo to cover the return costs if I was doing an RMA.