My company is using the standard 6MHz oscillator circuit design described in Fig 6 of the TI datasheet for the TUSB2046B hub controller. Perhaps 20% of our new boards are coming back with a dead oscillator within months of use. I have been able to repair them temporarily by replacing the IC and/or the 27pF capacitors, only to find out the next day that the problem has come back. Either the design is marginal, or there is a PCB layout issue, or something else I haven't yet identified. Looking for a permanent solution!
Here is what I've tried, but each case to no avail:
- Substituted a 5K potentiometer in place of the 1.5K start-up resistor. I adjusted it down to 200 ohms, and adjusted it up – after turning power off each time. No value even got close to starting the oscillator.
- Replaced the ECS crystal ECS-60-20-5PX-TR with an equivalent FOX crystal.
- Changed the 27pF caps to 22pF caps (0402 package SMT)
- Changed the 27pF caps to 30pF
- Applied both a heat gun and a cold spray can to see if temperature had an effect.
- Added a 1000pF cap across the 0.1uF bypass cap on the power pins of the hub controller IC.
- Added a 10uF cap across the 4.7uF tantalum filter caps on the hub controller IC.
- Removed the 4.7uF tantalum filter cap across the IC, leaving 4uF of capacitance due to other IC's on the 3.3V power supply.
Several times the oscillator worked for a while after I replaced the 27pF caps with an equivalent NPO 0402 and 0603 package. In two of those cases, I washed the surrounding PCB area with alcohol, and found the oscillator died again. After drying out a while, it again worked. But after a few hours, it died again. Similarly, replacing the IC itself appeared to fix the oscillator problem, but only temporarily.
JMueckl
