No Noise Oscillator Problem

I am having a problem getting my oscillator for my CS5528 (24-bit ADC) to start. It's 38.768khz, has no resistors or caps in the oscillator circuit, and is the clock which drives the onboard controller for the CS5528. The problem is as follows:

1. The oscillator will start in this circuit on a breadboard.
2. A 4-layer (1 gnd, 1 +5V, 2 signal layers) board was designed and components soldered. Lead traces were made as short as possible (they're equal though).
3. My testing is done underground (I work in a basement) with very little noise. One computer with a pretty hefty case, no cell phone reception, no wireless phones. HOwever, an oscope reading of my skin shows that my body is picking up some of the 60hz from the power.
4. I had this problem the onboard PIC18F4550 oscillator, but if I simply touch the pins of that oscillator it will start right up. This problem seems to have gone away.
5. I changed the CS5528 on the PCB to a new one, and it does the same thing.
6. Per the data sheet, there are no capacitors on the pins of this oscillator.

I did some reading on oscillator circuits and apparently they need a little noise to get them started. If this is true, is it realistic that I have so little noise on my PCB that the oscillator doesn't start? If so, how do I resolve this problem? This PCB will be in a shielded metal box when it's done, so there shouldn't be many sources of noise. I have read you can put a carbon film resistor to add noise, but I'm not sure where it goes. Between ground and the crystal, in series, not sure.

I'm going crazy with this problem! I'm going to output a 38.768 PWM signal from my PIC to drive the chip just to make sure it's the oscillator.

Thanks for advice,
Matt

Added after 2 hours 4 minutes:

I hooked up another PIC to this CS5528 XIN pin, with a 50khz square wave, and my circuit works fine! I'm happy it's definently the oscillator, but what I don't know is how to get an oscillator to reliably start... Advice?

Matt

The "noise" required to start an oscillator is internal noise, not external. Your "quiet" environment isn't the problem.

The CS5528 data sheet is quite poor and gives no information about connecting a crystal that I can see. It is quite common for 32kHz clocks to need quite a bit of capacitance - from 20pF to 50pF from both sides to ground. Give that a try. Otherwise, some resistance across the crystal 100k to 10M ohms may be worth trying.

Keith.

Thanks Keith, I'll give that try. The crystal is in a PCB, so is it okay to give the crystal a square wave (that may be a different frequency then the resonance frequency of the crystal) when I need to excite the XIN pin, or should I remove the crystal?

In a breadboard, the crystal did not need a capacitor. Does the breadboard have enough stray capacitance that it accounts for this?

Thanks,
Matt

Your report suggests that the problem is rather due to additional stray capacitance of the 4-layer PCB. It may be a problem
of the particular crystal exemplar. Some have higher series impedance than others.

does your adc has two diff supply?
for digital and analog portion separately?

if so , a diode between VA+ and VD+ can be connected and
check the result.
place diode anode at VD+ and cathode at VA+.

please find whether the problem disappears.

srizbf
11thmay2010

In a breadboard, the crystal did not need a capacitor. Does the breadboard have enough stray capacitance that it accounts for this?

Click to expand...

Yes, breadboards have a LOT of capacitance. That's probably why it worked in the breadboard.

Try adding more capacitance to your circuit. It may require a some trial and error.

dksoba said:

Thanks Keith, I'll give that try. The crystal is in a PCB, so is it okay to give the crystal a square wave (that may be a different frequency then the resonance frequency of the crystal) when I need to excite the XIN pin, or should I remove the crystal?

In a breadboard, the crystal did not need a capacitor. Does the breadboard have enough stray capacitance that it accounts for this?

Thanks,
Matt

Click to expand...

If you are going to externally drive the Xin pin you should remove the crystal.

It is difficult to say which has the most stray capacitance - a breadboard or PCB. The layers are very close together on a 4 layer PCB, but areas are also smaller and the track lengths will be shorter, so inductances lower. I notice that Crystal do not show any capacitance in their diagram but I doubt they have 30pF per pin. I think experimentation is required. You may find that extra capacitance to just one pin works best.

One other thought, is the crystal package the same in both the breadboard and PCB and is it soldered in the same way? Some surface mount packages have extra pins and it can make a difference if you ground the spare pins or not.

Keith.

I had the same problem with the CS5528 and 32kHz crystal -which brought me here- but finally i got it working. So for future reference i'd like to add my solution:

-Add two capacitors of 33pF on both ends of the crystal to ground.
-A resistor of 10MOhm between XIN (pin12) and XOUT(pin 13).
-Connect XIN directly to one end of the crystal
-The other end of the crystal goes via a 300kOhm resistor to XOUT.

Tested with several brand crystals and worked all the time!

Bas.