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.

What cause a crystal paused?

Status
Not open for further replies.

eepty

Full Member level 2
Joined
Oct 21, 2005
Messages
143
Helped
2
Reputation
4
Reaction score
2
Trophy points
1,298
Activity points
2,611
I have a ARM M0+ MCU connected with a 32.768KHz crystal. The crystal is connected to a load capacitor on each of its lead. The load capacitors are connected in parallel so one of its lead connected to the crystal and another lead connected to the reference ground.

I have made a number of board and they run normally. However recently I make some more boards and I found that the device seems paused for several seconds after power on. The observation is after I power on the device, it run normally for about 20seconds, and then it paused for about 30seconds, and then it runs normally again.

To try to investigate the problem, I modify the software to generate a pwm signal at one of the MCU pin, the pwm signal's clock source is the external 32.768KHz crystal. Having checked with a oscilloscope, I found the pwm signal start at the very begining when I power on the device, and then it paused for a while, and then it returns normal again. (Note: I am checking the pwm signal output from a MCU lead instead of chekcing the crystal lead directly, because I found that the crystal lead signal stopped when I touch the probe to it).

Then I changed the crystal to a new one, the pwm signal does not pause anymore.

Then I solder back the original crystal to the board, it runs normal again!8-O (That is it does not pause anymore after power on)

The original crystal was new and was bought from a trusted supplier

I would like to ask what can cause this problem?:thinker:
Thanks a lot!
 

When I was playing with a 32 kHz crystal, I got it working in a simple circuit. It required a lot of adjusting of component values. Evidently crystals can only operate within a narrow environment.

Later I must have overdriven it, because its frequency is not always stable, and it no longer starts to oscillate so easily. I believe my mistake was that I applied supply voltage through a low resistance. It was just a few V but that was apparently enough to alter the crystal's performance.

I don't know that this is the same as your problem but it tells how we need to be mindful of a crystal's fragility.
 

Hi,

The load capacitors are connected in parallel so one of its lead connected to the crystal and another lead connected to the reference ground.
For the oscillator the configuration is like the capacitors are connected "serially".
So if you find a "load capacitance" in the crystal's datasheet you need to use "twice" this value for each capacitor with your configuration.

But i don't think this is the problem here.

Klaus
 

Hi,


For the oscillator the configuration is like the capacitors are connected "serially".
So if you find a "load capacitance" in the crystal's datasheet you need to use "twice" this value for each capacitor with your configuration.

But i don't think this is the problem here.

Klaus

Yes, I have doubled the value. I have also added the stray capacitance of the PCB and the pin capacitance of the MCU pin to calculate the load capacitor.

- - - Updated - - -

Hi,

Try to use a series resistor on positive side of the crystal.

- - - Updated - - -

Hi,

Try to use a series resistor on positive side of the crystal.

But the crystal does not have a positive side.

- - - Updated - - -

When I was playing with a 32 kHz crystal, I got it working in a simple circuit. It required a lot of adjusting of component values. Evidently crystals can only operate within a narrow environment.

Later I must have overdriven it, because its frequency is not always stable, and it no longer starts to oscillate so easily. I believe my mistake was that I applied supply voltage through a low resistance. It was just a few V but that was apparently enough to alter the crystal's performance.

I don't know that this is the same as your problem but it tells how we need to be mindful of a crystal's fragility.

The two leads of the crystal is connected directly to the MCU pins.

- - - Updated - - -

One weird observation: When I pull down the hardware reset pin of the MCU to reset the MCU, a longer pull-down time (or reset time) will result a longer pause time of the crystal.

For example, if I just pull down the reset pin for about 1 secound, the MCU reset and the crystal did not pause. If I pull down the reset pin for 5 sec, after reset, the crystal pause for 10sec. Then if I pull down the reset pin for 15sec, the crystal pause for 20sec after reset.
 

Could also be a dry joint that makes intermittent contact or heat-related disconnections.
To see if you are getting some strange interaction with the firmware, try creating a program that turns on the 32kHz oscillator and then effectively does nothing - perhaps using the oscillator to flash a LED. IF that stops then the problem is likely to be with the oscillator/crystal; if not the there is some other interaction with the firmware.
Susan
 

I recently had an annoying intermittent problem with a crystal oscillator.

What I finally discovered was the metal can around the crystal (underneath) was occasionally shorting to the two pads on the circuit board. Fitting a couple of small insulating washers between board and crystal solved the problem.

It's obvious in hind sight, but was pretty frustrating until I found it.

Shorting both pins to the can obviously completely stopped oscillation.
But shorting just one side to the can may (?) produce some symptoms such as small jumps in frequency or amplitude, or erratic starting.
 

At one stage, I had a lot of quality problems with these small cylindrical 32.768 kHz crystals for a project.

The test below helped to identify possible suspect crystals. I applied a 1 Vp-p pulse train (at 1/2 the crystal frequency) as shown through a 1 pF capacitor over the crystal and monitored the response with a x100 scope probe. Using a normal x10 scope probe may load the crystal too much reducing the Q.

Good crystals showed a slow amplitude build up and a long decay (up to a second or two) when the pulse was terminated. This is due to the high Q of the crystal. Suspect crystals may show a low flat amplitude, not ringing between pulses and instantaneous envelope decay when terminating the pulse source.
 

Attachments

  • crystal_test1.png
    crystal_test1.png
    190.9 KB · Views: 150
  • crystal_test2.png
    crystal_test2.png
    11.8 KB · Views: 183
Last edited:
I am sure that a lot of these very low cost Chinese crystals are those that failed some kind of quality check.

I have noticed that the frequencies are usually either side of what it should be, and if you test a lot of them there are very few that are correct, but cluster either side of correct.
There may also be some problems with low Q with some of these crystals.

If you are buying "bargain" Chinese from e-bay, like twenty crystals for a dollar, or really unbelievably cheap, you may not always be getting top quality.
Just something to think about.
 

Thanks all, according to the response, the crystal for the microcontroller really has caused problem to many people.

Now I am thinking that it may be I am using a wrong load capacitor. Reviewing the microcontroller's latest datasheet, it has a remark that if a crystal which is recommeded a 12.5pF load capacitance, I do not need to connect any external capacitor to it. From the datasheet, the microcontroller internal capacitance is 3.05pF or 3.29pF. Why it suggests that I do not need to connect any external capacitor? (My crystal is 12.5pF load capacitance)

MCU datasheet:


In fact I have tried to remove the external capacitor. The problem seems improved, the signal exist for a longer time, but still it then stopped totally.

Then I test the external capacitor value by trial ane error, if I use a 8pF external capacitor, the problem gone...

I have studied some documents about how to calculate the external capcitor before. It teach me that the value of the required external capacitor is 2 x (Cxtal - Cmcu - Cpcb).

Cxtal : Crytal stated load capacitance in the datasheet (datasheet say 12.5pF)
Cmcu: The internal capacitance of the MCU pin (MCU datasheet is about 3.15pF)
Cpcb: The stray capacitor of PCB (can roughly assume 0.5pF)

So my calculated value is 18pF. Is that correct?
 

Could also be a dry joint that makes intermittent contact or heat-related disconnections.
To see if you are getting some strange interaction with the firmware, try creating a program that turns on the 32kHz oscillator and then effectively does nothing - perhaps using the oscillator to flash a LED. IF that stops then the problem is likely to be with the oscillator/crystal; if not the there is some other interaction with the firmware.
Susan

Tested with a very simple program, the result is the same so I think the problem is something else.
 

Hi,

So my calculated value is 18pF. Is that correct?
Yes. Correct.

****
With a some_MHz Xtal osciallator i found out that a 100R from inverter output to Xtal improves start up time.

I don´t know if this helps with the low frequency XTAL.
It seems the circuit has low Q and needs something like a kick start.

You could play around with some resistors.

Klaus
 

Now I am thinking that it may be I am using a wrong load capacitor. Reviewing the microcontroller's latest datasheet, it has a remark that if a crystal which is recommeded a 12.5pF load capacitance, I do not need to connect any external capacitor to it. From the datasheet, the microcontroller internal capacitance is 3.05pF or 3.29pF. Why it suggests that I do not need to connect any external capacitor? (My crystal is 12.5pF load capacitance)

There are several points:

1. None of the crystals run free: they are loaded by the microprocessor. This is a regenerative load that helps the crystal to vibrate (phase shift).
2. The crystal inside the can is pretty small and it holds too little charge and we need external capacitors to increase the charge (at the cost of the voltage; the Q will be lower) else the crystal will not run at the natural frequency. Because the crystal is being driven, it will show forced vibrations.
3. The crystal stops when the phase shift becomes insufficient. Correct capacitive load is therefore the most important factor.
 

One weird observation: When I pull down the hardware reset pin of the MCU to reset the MCU, a longer pull-down time (or reset time) will result a longer pause time of the crystal.

For example, if I just pull down the reset pin for about 1 secound, the MCU reset and the crystal did not pause. If I pull down the reset pin for 5 sec, after reset, the crystal pause for 10sec. Then if I pull down the reset pin for 15sec, the crystal pause for 20sec after reset.

This would not make sense, unless there is a long-time-dependent characteristic inside the MCU. Possibly an RC network of some kind (very high R, very low C). I don't see how a 32kHz crystal can generate several seconds of time-dependent behavior. Therefore it must be inside the MCU. It might be the result of accident, who knows. Anyway as a result oscillations are unreliable, or borderline.

I agree with post #12 which suggests resistors, as a simple and easy thing to try. Try hooking up a high-ohm resistor in various positions at the crystal: (a) across its terminals, (b) from a terminal to ground, or (c) from a terminal to supply +. Etc.

The simple circuit I played with had a 1M or 1.5M resistor across the crystal. I tried omitting it, but it was necessary. It's a high value, yet it allowed oscillations to start and sustain.
 

I presume that the datasheet specifies the oscillator circuit clearly and that no external bias resistor is required according to the datasheet. The case that a bias resistor is required though should be considered as silicon and/or documentation bug.

If the oscillator is not running due to crystal quality, we can expect a too high ESR. It can't be fixed by an external circuit. Us a known good crystal.
 

The pdf from post #15 has this typical schematic for an MCU. It resembles the circuit I tried (although I used a plain inverter gate, not an MCU).

2695249800_1459364657.png


Normally I see MCU schematics without the resistors near the crystal. Therefore I suppose an MCU contains them internally. Their values may have a lot to do with success of the crystal oscillating. Maybe they are easy to alter through accidental exposure to excess voltage or current?

- - - Updated - - -

Therefore once the crystal, or the MCU, becomes suspect, it could be worth trying to add various resistors at different points, in hopes of making oscillations more stable, and easier to start up?
 

I presume that the datasheet specifies the oscillator circuit clearly and that no external bias resistor is required according to the datasheet. The case that a bias resistor is required though should be considered as silicon and/or documentation bug.

If the oscillator is not running due to crystal quality, we can expect a too high ESR. It can't be fixed by an external circuit. Us a known good crystal.

The crystal I used is from Abracon, ABS25-32.768k-2. I attached the datasheet.
How can I check if it is the crystal problem?
 

Attachments

  • Abracon - abs25.pdf
    1 MB · Views: 103

Hi,

did you test Rs and Rf like in post#17?

Try 1M as Rf.
What is the result?

and then additonally 100 as Rs.
What is the result?

Klaus
 

As mentioned in #8, a 10X probe (I presume that is 10M) loads the oscillator sufficiently so that it stops. Looking at #17, I was wondering whether Rs should be 1M and Rf should be around 100 (or so). A small value of Rf will reduce the oscillation amplitude to a small value (perhaps a few mV or so) and a large value of Rs will reduce the feedback and overloading.

Or, as usual, I am thinking the wrong way??
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top