cancel
Showing results for 
Search instead for 
Did you mean: 

32.768kHz clock doesn't start sometimes!

Posted on February 11, 2011 at 18:08

32.768kHz clock doesn't start sometimes!

#lse-rtc #lse-low-temperature
17 REPLIES 17
Nickname12657_O
Associate III
Posted on May 17, 2011 at 14:24

Ciao Fabio,

Could you please send /attach your Quartz Datasheet for sanity check it is compatible and matching our STM32F103Z RTC oscillator as we specify in our data-sheet and Application Note.

Regards,

STOne-32.

Posted on May 17, 2011 at 14:24

Ciao!

Here attached the datasheets: for the citizen cmr200t I bought the 6pF version (Farnell Italy part number 1652560 - see the citizen catalog page http://www.citizencrystal.com/khz.htm) and for the Epson MC-306 the 6pF version too (Digikey part number SE2417CT-ND).

I saw that, sometimes, putting my finger over LSE tracks on pcb helps oscillator starting. I think it's because of little capacitance injection...

Best regards

Fabio

Posted on May 17, 2011 at 14:24

Let me write another consideration: I tried another board, fully functioning, with the Epson Toyocomm MC-306 32.768kHz 6pF crystal. I found that it works fine even with no addictionally capacitor CL1 and CL2 or with CL1= CL2 = 12pF. So, STM32 is less susceptible to capacitor values (i.e. the oscillator starts with a wide range of external capacitors: the only thing changed is the frequency accuracy as application notes said).

May be my problem due to defective microcontroller? What tests can I do for verify this?

Best regards

Nickname12657_O
Associate III
Posted on May 17, 2011 at 14:24

Dear Fabio,

The most important parameters to allow the oscillator to start are the ''CL'' of the Quartz ( 6pF is the best)  and it ESR ( Motional Serial Resitance) : Using the Formula we provide in the application Note to have Gain Margin > 5.

The CL1, CL2 are just to tune the Frequency and to have an accurate Frequency, but have no effect indeed on the oscillation start-up.

Based on my experience, Epson MC-306  with CL=6pF  should be fine and is a good quartz, However it may happen that the received quartz is not the one with 6pF and in that case you will have an issue.  Avoid also to have dust on the PCB area near the Oscillator pads and going to the Quartz. any leakage on this part of the PCB may have an impact on the oscillation as well especially at low temperature. 

Cheers,

STOne-32.

Posted on May 17, 2011 at 14:24

Dear,

I tried different crystals (also different crystal models, so the problem probably isn't the crystal).

Then, I washed many many times the pcb near the crystal and microcontroller, but nothing happened: it seems that, in this board, the only way for starting this clock is to put a resistor (15MOhm) in parallel with crystal. I'll try with resistors bigger than 15MOhm and I'll use this board as demo or for r&d purposes...

Cheers

Fabio

fiim
Associate III
Posted on May 17, 2011 at 14:24

I use STM32L151-152, and 2 identical devboards(my).

On 1 devboard RTC starts good with 3 types of crystal and CL1=CL2=3..15 pF good!!

And on other devboard(the same!)the RTC does not

start at all(RCC_GetFlagStatus(RCC_FLAG_LSERDY) = RESET).

santi
Associate II
Posted on October 23, 2011 at 10:32

Dear,

I am having this exact problem on my very first two prototype boards with STM32F103RET6 rev.Z ( DEV_ID = 0x414, REV_ID = 0x1001). 

I have never found this with the development board that I have, the development board has rev.Y chip (REV_ID = 0x1003).

Is there something wrong with the rev Z chip?

Posted on October 23, 2011 at 17:26

Well without knowing exactly what crystal you're using it's hard to say. The 32 KHz oscillator is not mentioned in the errata. These circuits are hugely sensitive to capacitive loading, and resistance, as well as the physical characteristics of the PCB, traces, crystal, and soldering.

Without a clearer explanation of your parts, circuit, construction getting specific help may be hard. Try contacting your local ST FAE.
Tips, buy me a coffee, or three.. PayPal Venmo Up vote any posts that you find helpful, it shows what's working..
emalund
Associate III
Posted on October 24, 2011 at 15:14

layout, layout, layout

the traces to the xtal and the burden caps MUST be less than 1cm

the traces to the Vdd decoupling caps MUST be less than 1cm

EVERY Vdd pin must be decoupled

at least one of the Vdd pins must have a tantalum in addition to the ceramic

Erik