LSE crystal for ST32MP1 - FC 135 38.768 kHz

[DMårt]

Hi!

I noticed this.

The FC-135 32.768 kHz crystall with CL = 12.5 pF, ESR max = 70 kOhm and C0 = 1.0 pF is suitable for MP1 products.

The gmcrit is 2.1631 uA/V.

When I do the math, i found out that gain margin is below 5.

C0 = 1.0*10^-12;
CL = 12.5*10^-12;
ESR = 70*10^3;
F = 32.768*10^3;
Cs = 7*10^-12; % Assume that stray capacitance is 7 pF
gm = 8.5*10^-6;

% Gain margins critical
gm_crit = 4*ESR*(2*pi*F)^2*(C0 + CL)^2
gain_margin = gm/gm_crit

% Capacitors
C = (2*CL - 2*Cs)*10^12

Output:

gm_crit = 2.1631e-06
gain_margin = 3.9295
C = 11.0000

Because the gm value of 8.5 can be found here

Why does ST recommend FC-135 32.768 kHz when the gm_crit_max is below the gm_crit value?

[STOne-32]

Dear @DMårt ,

Thank you for the valuable post, indeed the whole line is applicable to all 4 software drive levels from low to high .

 in case of 12,5pF CL crystal,using High drive is a MUST . So Gm using that mode becomes 13,5 and not 8,5 which is for Medium High level .

Hope it helps you .

STOne-32

[DMårt]

@STOne-32 

Thank you. Now I understand!

Do you think 10pF capacitors will work?

[STOne-32]

Hi @DMårt ,

Cs are in general between 2 to 3 pF with a good optimized PCB layout . CL1 and CL2 capacitors would be about 12,5pF - 3pF = 9,5pF * 2 so in range of 18pF in this case .

if Cs = 5,5pF ( worst case ) , CL1 and CL2 would be in range of 14pF . All of these calculations are based on A Crystal load of 12,5 pF to oscillate as close as possible to 32,768Khz frequency and have a low PPM drift vs Nominal frequency.

Hope it helps ,

STOne-32

[PatrickF]

HI @DMårt 

in fact the Gmcritmax in product datasheets already take the x5 margin into account.

Nevertheless, from my side, although 12.5pF will work with high-drive, I recommend to chose a crystal with CL around 6pF which allows to use default medium-high drive. Drawback of low CL could be higher pullability, which mean PCB load capacitors might need be adapted with real Cs if good LSE accuracy (<100ppm) is a must.

It seems the trend is to have smaller crystals with slightly lower CL value as mainstream (i.e. good availability and quantity pricing).

Regards.

[DMårt]

@PatrickF 

Thank you for your reply.

I think I will choose the Epson FC 135 LSE crystal due to the availability at the manufacturer JLCPCBs stock.

Q13FC13500004 | Seiko Epson | Crystals | JLCPCB

My goal is to produce an STM32MP board for a minium price. It's going to be open source as well.