What is the situation with RTC Crystal on Cerb40? And how do I make it work?

So, Initially I got the recommended 32.768KHz 12.5 pF Crystal and soldered it onto my Cerb40. I soon found that it wasn’t working (I had written my own RTC code). After you turn on the crystal, it never reports ready… So, I started googling and found that the recommended crystal capacitance should be no more than 7 pF for an STM32F chip. So, I promptly ordered some 6 and some 7 pF crystals. I soldered on the 7pF crystal and although it takes 10 seconds to stabilize, it does report as ready and work! So, I got another cerb40 and the same crystal isn’t working for some reason. I tried a 6pF and a 7pF. Now, I went back and noticed this note on the Cerb40 Wiki page:

What!? I don’t remember that being there before…!

OK, so I think we can all agree that replacing tiny SMD capacitors has a low probability of success (at least for me) and should not be necessary. First, will there be a new revision that fixes this? (I assume yes) Is there a pdf of the schematic somewhere so I can see the circuit so I can try and figure out my own solution? Can someone explain the problem? Is there any crystal that might work without replacing these SMD capacitors?

Update: Just checked the Cerb40 wiki page changes and I see that the page was changed by @ Aron on August 9. Also, I see capacitors can be connected in parallel, so I am wondering if I can just solder another capacitor on top of the current ones… maybe that would be easier…

-Valkyrie-MT

The Cerb40 developer page has the design files, including a pdf schematic. I’d check the datasheet for capacitor calls, a it’s the only reliable source.

The last time your question about revving the design was asked, three was no clear answer.

Capacitors on top of each other will increase the value not decrease it so that will not work.

I just found it, clipping below. But, it is actually not on the Cerb40 developer page that I can see. I did find the zipped sch and brd files in the FEZ Cerberus page. No pdf that I can find either.

That’s good advice. I am looking up all of that right now… If I come up with a simpler fix, I will post it here.

Update: I have attached cerb40 schematic segment and a snippit from the relevent ST spec sheet. Let the head scratching begin… how to fix… hmmm.

Update2: Interesting… It mentions a resonator with integrated capacitors in the figure… It would be easier to short the 2 capacitors with a bead of solder…

When it says “integrated load capacitors” I believe it’s referring to a ceramic resonator (vs a crystal oscillator). They’re cheaper, but MUCH less accurate. Especially for the RTC, accuracy can be important.

Ok, I may be grasping at straws here, but it looks to me like the capacitors on the pcb are not necessary and the capacitors in the ST diagram are internal to the resonator/crystal package. So, I ran my code to the point that it polled for the LSERDY bit. As expected, the bit stayed at 0 - crystal not ready (this is with a 6pF crystal and no modifications to the Cerb40). Then, I tried shorting the 2 ends of C21 with a wire and that seemed to work, the LSERDY bit flipped to 1 after a few seconds. So I figured I’d just put some solder on top of C21 to short it, but the solder just would not stick to the top of the cap (Not surprising I suppose). As I am trying to get the solder to stick the whole capacitor comes off, so then I decide to just connect the two pads together. That did not work, crystal will not stabilize. Crap. So, then I figure if the assumption is that the caps are not necessary, I should disconnect the C21 pads. So, I pull out the solder sucker and get it off (picture attached). And wouldn’t ya know… it works every time. And it stabilizes within what seems to be less than 1 second!

I have since power cycled 4 times and the time is correct and retained with VBat every time. I have also let it run 20 minutes on just VBat and the time was accurate down to the second. Now I will test it overnight.

That seems plausible. So the load capacitance rating is inherent to the crystal I suppose…hmmm. It seems you can buy anywhere from 6pF to 12.5pF for the 32.768 kHz crystals.

Getting tired, going to bed. We’ll see if the time is still accurate tomorrow after running of VBat for a day.

You’re not using a ceramic resonator, you’re using a quartz crystal. You should need some sort of load capacitors, unless the trace+pin capacitance is enough by itself (the crystal was certainly much higher than it should have been, see my previous quote from the datasheet).

By the way, we tested it with no capacitors at all and it worked fine!

C21 I could probably get off, but now that the crystal is there, C22 is going to be a problem.

Anyway, it’s clearly a band-aid. The crystal should have a much smaller value, and two much smaller loading caps should be installed.

By the way, I did learn from reading the datasheet that normal startup time for 32 KHz crystals is in the 1-5 second range!

Please take a look at the attached image and compare it to what you received.

If the date code is 31th week of 2012 or higher then you got the correct capacitors on the board and you don;t need to do any hardware changes.

Holy cow you guys are blowing me away right now. First of all, do you people sleep?! Second…

So what you’re telling me is that removing C21 (do-able with my skills) is likely to be a legitimate fix and you guys have already corrected this in newer boards. That’s great news. So do they still say Rev 1.1? One thing that is strange is that the board I was working with last night, C21 was tan color and C22 was white (you can actually see it in the picture in post #5 of this thread, C21 was the same tan color as C1 before I removed it). That seems wrong to me if the color is an indicator of capacitance. According to the schematic, it looks like they should have both been 27 pF (18 pF on the updated one?).

So, the newer boards have a 18pF cap? So, what about the GHI recommended 12.5 pF crystal? Should that be 6 or 7 pF?

Thanks!

Whatever the answer is to Valkyrie-MT’s questions, can the Cerb40 dev wiki page get updated with specific instructions on how to update older boards? Thanks!

@ Joe- Is there a way to tell on the board itself if it is batch 3112A or later? I don’t keep mine in the original antistatic bag…

The hardware revision is the same.
You have to use a crystal with Cf=12.5 pF. you will have problems if you used 6pF crystal because it will require different C21 and C22

The new revision of the board, what capacitors does it use? Certainly, from the datasheet, we know that a 12.5 pF resonator is never appropriate, I don’t know if this applies to crystal or not.

@ ransomhall - Aron is updating the wiki page today.
Forr the date code, it is only on the label, so for you situation, I think you need to refresh your memory and try to remember when you bought it :slight_smile: if it before August 2012 (31st week) then you need to remove/replace the caps.

@ Joe - Fortunately my cognitive abilities still allow me to remember these kind of details :slight_smile: To double check I looked at my previous order dates on your site. All my Cerb40s are “older”. If I were to buy some Cerb40s today, would I get the latest batch, or are you still going through older stock?

@ ransomhall - [quote]If I were to buy some Cerb40s today, would I get the latest batch, or are you still going through older stock?[/quote]
Yes that is possible. But if you bought it before that date then it is for sure old.

@ Joe - what are the caps on the new version? Also, have the other Cerberus-family boards been updated as well?

@ godefroi - C21 and C22 are 18pF. and the crystal’s load capacitance must be 12.5pF