0
\$\begingroup\$

I am using KSZ8999 ethernet switch from Micrel and I have problem with 25 MHz external oscillator. There is constant 2,1 VDC on output pin which is VDD for the switch, so crystal is not oscillating. There should be no problem with crystal, because I tried to connect it to other design and it worked fine. I also tried using different values of load capacitors but with no result. There is also no short circuit between VDD and crystal pin. Can you please help me find out, what could cause this problem? Thank you for your answers.

enter image description here

R15 is now 510k, R16 is 0R, C41 and C42 is 15p. Crystal MPN is now GC2500081. Sorry for the weird schematic symbol, it was downloaded this way.

\$\endgroup\$
  • 1
    \$\begingroup\$ Instead of commenting on your own question, use Edit to add that remark to the question. There is constant 2,1 VDC on output pin which is VDD for the switch, so crystal is not oscillating What output pin, output pin of what? Why do you think that the 2.1 V DC is incorrect? How did you measure it? How do you conclude that the oscillator isn't working? Maybe the KSZ8999 chip is damaged? \$\endgroup\$ – Bimpelrekkie Mar 4 at 14:44
  • 2
    \$\begingroup\$ Crystal oscillators are highly sensitive to loading, and even just putting an oscilloscope probe on one can make it stop oscillating entirely. Are you sure that isn't the problem here? \$\endgroup\$ – Hearth Mar 4 at 14:51
  • 1
    \$\begingroup\$ To clarify: what I'm suggesting is that it's possible your crystal is oscillating, but it stops when you try to measure it due to the loading effect of the scope probe. \$\endgroup\$ – Hearth Mar 4 at 14:56
  • 1
    \$\begingroup\$ @glen_geek The reference design schematic shows a DC bias resistor of 500k across the crystal. Design guide at ww1.microchip.com/downloads/en/DeviceDoc/KS8999_V1.2_DP.zip \$\endgroup\$ – Peter Smith Mar 4 at 15:16
  • 1
    \$\begingroup\$ Also: I have two identical PCBs with this switch so I assume there is some systematical problem That's a tough statement unless you designed the PCBs with the Switch IC. Did you design these PCBs? What does the circuit look like? \$\endgroup\$ – Bimpelrekkie Mar 4 at 15:16
2
\$\begingroup\$

In the clip below from the reference design schematic in the demo board user guide, there is a 500K bias resistor across the crystal.

To fix the problem, you can probably solder an 0201 resistor across the package pins or a larger resistor across the crystal.

enter image description here

\$\endgroup\$
  • 1
    \$\begingroup\$ This is certainly a very common reason for crystals not oscillating. \$\endgroup\$ – Peter Smith Mar 4 at 16:29
  • \$\begingroup\$ Looks like a third party seller (Mouser) from that link you provided. From Microchip's website about their eval board it has different decoupling capacitor values with 18pF in lieu of 22pF. There was apparently a revision in 2007 that changed those capacitor values. \$\endgroup\$ – KingDuken Mar 4 at 16:35
  • \$\begingroup\$ I just added part of my schematic to the topic. I replaced 2M2 resistor with 510k, 22R with 0R and 27p capacitors with 15p (sadly I don't have 18p) but it's still not working. \$\endgroup\$ – MMA3 Mar 5 at 8:00
0
\$\begingroup\$

It's pretty hard for us to troubleshoot this with very little information.

Assuming you are correctly measuring the output pin with low enough loading, there are several possibilities.

  • The crystal could be defective or incompatible with the chip

  • The load capacitors could be grossly wrong value (eval board uses 18pF implying their crystal is designed for maybe 12-16pF load).

  • There could be some issue with leakage (unlikely, but possible) or PCB construction (open, short, bad via etc.) that is preventing the oscillator from working. Shorts under SMT crystals are a possibility. If the leakage was comparable to the bias resistor then it might not start. As crj11 says, you need the bias resistor (some ICs have it internal, this one does not).

  • The IC could be damaged

I would start by removing the crystal and load capacitors and replacing the crystal (and load caps) with a different one with compatible load capacitance of a similar level to the eval board, having a good look at the (cleaned) PCB under a microscope prior to re-assembly. Also make sure the solder and flux being used is appropriate. SMT capacitors are unmarked so checking the value with a meter prior to installation wouldn't hurt. Mistakes happen and an 18nF capacitor looks similar to an 18pF capacitor (usually the body is a bit darker in colour). Double check the bias resistor value as crj11 says, if it is too low (or too high) the oscillator will not work.

Pierce oscillator (from this document:

enter image description here

\$\endgroup\$
  • \$\begingroup\$ Since this crystal worked fine on other design, I conclude that it is not damaged. What do you mean "incompatible with the chip"? How could it be incompatible? I tried more values of load capacitors (15p, 27p, 33p) but nothing changed. I also checked connection between chip and crystal with multimeter and it is fine. I have two identical PCBs so I assume that both ICs aren't damaged. Just to mention, this is my design, I'm not repairing someone elses PCB. \$\endgroup\$ – MMA3 Mar 5 at 12:34
  • \$\begingroup\$ Because it worked before is not really a strong recommendation. It could have been marginal and there is some small change in chip parameters (eg. gm = transconductance of the on-chip amplifier) from batch to batch. This is a pet peeve of mine, there is really no guarantee that this type of oscillator will actually work, especially starting at temperature extremes. Suggest you try another brand of crystal, or even an HC-49 type, after checking other items. Also if it worked and then stopped, make sure you are not overdriving the crystal. \$\endgroup\$ – Spehro Pefhany Mar 5 at 12:55
  • \$\begingroup\$ P.S. It's a higher confidence if you've produced many batches, of reasonable size, and have done thorough testing of many (or all) units including starting at temperature extremes in an environmental chamber. \$\endgroup\$ – Spehro Pefhany Mar 5 at 12:59
0
\$\begingroup\$

Even that I didn't want to believe it, truth is that both switches are probably damaged. I measured only 50 and 130 ohms on these chips respectively between X1 and GND which is much lower resistance that it should be. I measured same resistance even when I disconnected pin from PCB, so the problem is inside IC. Anyway thank you all for your advices.

\$\endgroup\$
  • \$\begingroup\$ That’s unfortunate. I wonder if the ESD protection is less effective on the crystal connections. \$\endgroup\$ – Spehro Pefhany Mar 5 at 15:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.