1
\$\begingroup\$

I have a GPSDO design I've been building and selling for a while now, and it's mostly successful, but sometimes the 3.3v buck converter to power the oscillator won't start up properly.

It isn't universal - most of them work, but it happens often enough that I think there must be some aspect of the datasheet that I am not heeding 100% that's causing the design to not be 100% stable.

When it doesn't work, the output is more like 1.7 volts instead of 3.3 volts. Momentarily shorting out the 10 µF cap on the input of the SC189Z is enough to get it to start up properly (but obviously that's a pretty heinous workaround).

The schematic is on the hackaday.io project page. The 2.2 µH inductors are Murata 1285AS-H-2R2MP2s (I couldn't handily obtain the part mentioned in the datasheet). The 10µF, 0.1µF and 0.01µF caps are ordinary X7R 0805 ceramic caps. The input power is a 6W rated wall-wart style 5VDC power supply. The OH300 requires 3W at startup reducing down to 1.1W steady-state (at 3.3V).

The relevant portion of the circuit starts with the buck converter, which is followed by an additional 2.2µH inductor and 10µF cap and 0.1µF cap as a ripple filter. From there, there's about a 1 inch long 32 mil wide trace over to the OCXO, where there's another 10µF filter cap, a 400Ω ferrite bead and a 0.01µF cap, then the power supply pin for the oscillator.

\$\endgroup\$
  • \$\begingroup\$ WHere are your startup Vin, Vout waveforms? It could be UVin or OVout \$\endgroup\$ – Sunnyskyguy EE75 Aug 23 '16 at 5:47
  • \$\begingroup\$ UVin and soft start look reasonably thought out, so it could be OVout. L3 and C29 form a 'lively' output filter, there is no dissipation, so perhaps it could overshoot? Scope the output and trigger on 3.3v+delta. If that is the case, you don't want to add series R, but a capacitively coupled shunt R, which will impact efficiency slightly, but you can control how much by the size of the C, should reduce overshoot. Otherwise, try driving EN after an input stabilisation period, cleaner than shorting Cin! Otherwise, post scope captures for good and bad startup. \$\endgroup\$ – Neil_UK Aug 23 '16 at 6:30
  • \$\begingroup\$ As you do not appear to be holding the GPS chipset in reset at start, it is possible that (depending on component tolerances) the buck is going into current limit at start up (not an uncommon issue); shorting the input cap starts the output into a (slightly) pre-biased load which may allow it to start up when you do that. \$\endgroup\$ – Peter Smith Aug 23 '16 at 12:11
  • \$\begingroup\$ The datasheet states that the minimum capacitance with a 2.2μH inductor is 10μF, so you are in fact at the minimums. As X7R devices exhibit not only DC bias effects but other things that reduce the effective capacitance, the loop may be unstable. See page 19 of the datasheet. (This is somewhat at odds of the corner frequency calculation on the same page). \$\endgroup\$ – Peter Smith Aug 23 '16 at 12:32
  • \$\begingroup\$ @PeterSmith The GPS receiver is separately powered by an LDO. The buck converter at issue powers only the OCXO and nothing else. The GPS module is starting normally. \$\endgroup\$ – nsayer Aug 23 '16 at 15:10

Your Answer

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

Browse other questions tagged or ask your own question.