Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

[SOLVED] Can i check PCB capacitance by LCR meter ?!

Status
Not open for further replies.

hm_fa_da

Full Member level 5
Joined
Sep 16, 2003
Messages
287
Helped
10
Reputation
20
Reaction score
4
Trophy points
1,298
Activity points
3,217
Hi all,

I'm making a board which the 32.768Khz oscillator in the MCU is too sensitive so i have to adjust the CL for Crystal exactly.
For that purpose, i need to know PCB and other components capacitance between two lines of crystal pins.
that part of circuit is:
Crystal.png
Can I use LCR meter to sense the total capacitance in board ?! how about check it when crystal is connected to 2pins of MCU ?
Can all LCR meters do it or need a specific function to do so ?

Thanks
 

Last edited:
Thanks, my first purpose of using LCR meter is to measure load capacitance of 32.768Khz crystal which is connected to pins No 3 & 4 of MCU.
The load capacitance is not only the caps (C13 & C14 in the PCB picture i sent), it's those caps + other capacitance which are capacitance between tracks and capacitance between pins of crystal and MCU after soldering.
I can measure PCB capacitance when crystal and MCU are not soldered. I want to know if i still can use LCR meter to measure capacitance after soldering those two components ? won't they make malfunction on C measurement ? (they are connected to tracks at same time).
 

I want to know if i still can use LCR meter to measure capacitance
after soldering those two components ?
Yes, you can.

Oscillator core circuit of MCU is a self-biased inverter.
To measure true load capacitance seeing from crystal, you have to set oscillator part of MCU active state.

See "LOAD CAPACITANCE" in https://www.yoketant.com.tw/yoketant/terms.htm#T9

If your LCR meter has DC-Cut characteristics, you can connect LCR meter directly.
 
Last edited:
In the shown layout, capacitance of each crystal pin against ground will be probably larger than differential capacitance, and fortunately can be better measured with most LCR meters. Measuring differential capacitance requires special considerations.

Generally I presume that the capacitance can be estimated with sufficient accuracy by PCB trace impedance calculation tools. The pin-to ground capacitance will be hardly larger than 1 (or maximal 2) pF and is unlikely to cause trouble, pin-to-pin much smaller. You should know however, if the oscillator circuit is expecting external load capacitors (like e.g. PIC18 secondary oscillator) or none at all like many RTC chips.

What means "too sensitive", by the way? I experienced some cases that a 32 kHz oscillator circuit didn't work due to too large load capacitors.
 
In the shown layout, capacitance of each crystal pin against ground will be probably larger than differential capacitance,
and fortunately can be better measured with most LCR meters.
I don't think so.
LCR meter measures Cdiff+Cg1//Cg2.
 
In the shown layout, capacitance of each crystal pin against ground will be probably larger than differential capacitance, and fortunately can be better measured with most LCR meters. Measuring differential capacitance requires special considerations.

Generally I presume that the capacitance can be estimated with sufficient accuracy by PCB trace impedance calculation tools. The pin-to ground capacitance will be hardly larger than 1 (or maximal 2) pF and is unlikely to cause trouble, pin-to-pin much smaller. You should know however, if the oscillator circuit is expecting external load capacitors (like e.g. PIC18 secondary oscillator) or none at all like many RTC chips.

What means "too sensitive", by the way? I experienced some cases that a 32 kHz oscillator circuit didn't work due to too large load capacitors.

Too sensitive means:
I have 3 different boards with 3 different STM32L MCUs, one with STM32L053, one with STM32L152 and one with STM32L433.
all three boards use external 32.768KHz crystal for oscillator circuit. and all need external caps for crystal.
the problem is 2 of my boards didn't start oscillating !
I searched net for a solution and found out many other people also have same problem with STM32 MCUs, and all refer to an application note AN2867 by ST for designing of PCB layout and load caps ... I saw it after PCB are made and i faced that problem, i can't change PCB layout at this time ! (Is my PCB layout that much bad ?!!)
it seems because of ultra low power consumption of STM32 series MCUs, they are too sensitive to capacitance or other parameters changes for starting and keep running of external 32.768Khz oscillator.
I never had such problem with AVR and AVR Xmega series MCUs which i used to work with.
in the AN2867 there is a formula for crystal load caps calculation:
an28.png

I want to measure Cs in the formula.
Even if oscillator starts to work now in lab, i worry for it's function later in other temperature or ... conditions on user side. so i want to calculate and choose the best value to minimize the probability of failure in future.
 

The relevance of keeping the specified load capacitance is primarily in minimizing the frequency error which is rather critical for RTC. The circuit will oscillate over a wider load capacitance range.

A possible problem comes up if the chip requires a quite different load capacitance than the nominal crystal load capacitance. In this case you would select a better suited crystal.

Cs in the application note is the sum of differential stray capacitance and series circuit of both single ended stray capacitances. I keep my point that it can't be well directly measured with a ground referred LCR meter. It can be calculated from three ground referred measurements. One single ended stray capacitance measurement aöllows already a good estimation of Cs.

I would expect Cs around 1 - 3 pF for the shown board (1.6 mm FR4).
 
I also highly doubt parasitic trace capacitance is your problem. Your traces are very short as-is, and should be negligible compared to the tolerable range of CL. I would instead suggest that the oscillator you chose is not a good match for the oscillator circuits used in those MCUs (wrong ESR or CL). I would contact STM and ask for recommended models. Or see what they use in reference designs.
 
I asked the question in ST forum too, and they replied refer to AN2867.
I had tested 3 different Crystals on my board, all same problem ! tried different Caps from 10pf to 33pf, same result ! (Crystal CL was 12.5pF).

I put my board in -10c temperature to check if oscillation can start or not, result was negative, oscillation didn't start at all.
but in STM32L4 Discovery board, it has no problem in -10c temperature, it starts to work normally.

I have STM32L4 and STM32L152 and STM8L discovery boards, i checked all their crystal circuits, almost none of them observed what is exactly mentioned in AN2867 (PCB part) !!!
There is no separate covering GND polygan around crystal as said in AN2867...

The last thing i faced to : I ran STM32L4 (STM32L476 MCU) discovery board to turn on LCD from 32.768 external crystal. when i touch crystal pins or even short circuit it by pans ! it still works, only frequency changes and LCD flickers !
But in my own PCB (STM32L443) when i touch only one crystal pin by pans, it stops working, MCU jumps to hardware error ... and wouldn't work again until power restart.

I removed STM32L152 discovery crystal and used it on my own PCB, still same problem with my board. of course stm32L152 discovery and stm32l4 discovery boards use different crystals.
the next i'm going to test stm32l4 discovery board crystal on my own pcb and my own crystals on discovery board.

Note: My board works on normal temp now, but seems too sensitive and stops working when crystal being touched or not starting oscillation if in -10c Temp.
 

Show me layers profile of your PCB.

Layers of metal
Thickness of dielectrics

Do you remove pattern underneath of crystal electrode ?

From https://www.edaboard.com/attachment.php?attachmentid=138558&d=1494082194,
Red is L1 plane.
Blue is L2 plane.

What value of thickness do you have between L1 and L2 ?

You should remove L2 plane underneath of crystal electrodes.

For example, assume five metals(L1~L5) PCB.

Crystal electrodes are L1 Plane.
Remove L2~L5 underneath of crystal electrodes.
 
Last edited:

The problem is found !
Thank you all for your helps, all problems were because of CSS (Clock security system) enabled in my MCU !
It stops OSC if a small failure occurs and won't start again until power restart (not even hardware reset). when running normally, if i touch crystal pins, CSS disables 32Khz OSC but CPU continues to run code. I've written a small code to turn on LCD clocked from 32Khz LSE, so when it stops, LCD also stops and i couldn't see if he cpu is working or not !
I used Keil debugger to monitor where the code is, if i've touched OSC before entering debugger, when debugger starts, it first resets the MCU but CSS bit is not cleared this way, so in new code running in debugger mode, i just saw that LSE doesn't start and CPU jumped to Error Handler().
in other hard environment (like -10c Temp) CSS just stops LSE after startup and the flag remains even if reset MCU. it prevents LSE start again.

- - - Updated - - -

Show me layers profile of your PCB.

Layers of metal
Thickness of dielectrics

Do you remove pattern underneath of crystal electrode ?

From https://www.edaboard.com/attachment.php?attachmentid=138558&d=1494082194,
Red is L1 plane.
Blue is L2 plane.

What value of thickness do you have between L1 and L2 ?

You should remove L2 plane underneath of crystal electrodes.

For example, assume five metals(L1~L5) PCB.

Crystal electrodes are L1 Plane.
Remove L2~L5 underneath of crystal electrodes.

I don't think it should be removed, the Blue plane is GND not other signals. in AN2867 is written to design PCB including GND plane under crystal ... :
cst.png
 

I don't think it should be removed,
the Blue plane is GND not other signals.
Wrong.

If recommended load capacitamce value is small and thickness between red and blue plane is thin, you have to remove blue plane underneath crystal.

This is very common sense regarding crystal, filter, etc.

Return to start of this thread,
was your origin of this thread "you doubts capacitaces seeing from crystal are too large compared to recommended load capacitor" ?
 

Wrong.

If recommended load capacitamce value is small and thickness between red and blue plane is thin, you have to remove blue plane underneath crystal.

This is very common sense regarding crystal, filter, etc.

Return to start of this thread,
was your origin of this thread "you doubts capacitaces seeing from crystal are too large compared to recommended load capacitor" ?

You are right, it makes capacitance but same time works as shield against noises. I just wanted to calculate the true load Caps for crystal. My PCB is FR4 1.6mm, so i think i still have room for adding caps to adjust the load Caps same with Crystal CL. I wanted to know how to measure the total capacitance between on crystal pins. of course if it is higher than crystal CL, i have to decrease caps or even remove caps or remove GND plane same as you said.

Thank you.
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top