My quest for a solid 10Mhz signal finally took it’s penultimate step recently, when I purchased a Trimble Thunderbolt GPSDO off eBay.
The unit arrived packed well, and in good shape. It even came with a power connector, which I had not expected. After rewiring it properly (it seemed “off”), I hooked up my Rigol DP832 for the ±12V supplies, and +5V supply, and powered the unit up. Everything looked okay except the +5V rail was drawing about 3x what it should. Oh, and the 10Mhz output being “out” by +26Hz (I wondered if this was because the unit was still powering up, so it didn’t bother disciplining the oscillator yet).
Hooking up a serial connection got no output, and probing the serial connection with the scope showed very low level signals (less than 1Vpp). Opening the unit, I found that the RS232 transceiver (a 232IBE, which is now called a ICL232, or any other number of MAX232 derivatives) was HOT. Everything else looked good for the unit, so my assumption is the chip was damaged. I clipped the leads and replaced the chip with a MAX202 made by TI. It’s pin compatible, but has much better tolerance for ESD and Latch-up (which seemed to be what was going on with the failed chip).
Sadly, the MAX202 I had was not the wide version, so I had to use some 30 gauge wire to fully connect the chip. Powering the unit back up, I got serial data! But, after a full survey, the Tboltmon.exe software still showed the Oscillator out by 26Hz (or as it said, 2660ppb (parts per billion)), and the DAC registered as at a rail (in this case, 4.999990V). Checking the output of the 5V reference, it showed -0.6V. Huh. The part is an LT1021DCS8-5. I didn’t have one of these, but ordered a replacement after lifting the pin, and feeding that pad with the +5V rail through a 470ohms resistor (to limit the current), it still didn’t work. The 5v reference output was still 0v, and the 5v being fed in was being dragged down to 3v on the pad side. So whatever the reference was feeding was clearly dragging things down. Tracing out the signal, pin 6 (output) of the 5V reference (U15) goes to the R83 pair (both 10k), then to pin 3 (the non-inverting input of amp A) of LT1014 (U17). Pin 1 (output from amp A) goes to R99, R79, then C232, and back to Pin 12 (non-inverting input for amp C) on U17. Then back out on pin 10, and to the 74AC174 (U14) on the other side of the board. From reading here, this seems to be all to provide a very stable power supply for U14 because it is providing the very monotonic DAC, and 10Mhz signal.
From testing everything, it appears that the 5V ref, and quad amp are both dead on my unit. But U14 looks good, and seems to be being powered via one of its protection diodes on a “high” pin. I ordered parts and waiting for those to arrive.
Replacing the parts was easy, just clipping all the pins and then “wiping” off the pins with a hot iron, then soldering down the replacement parts. The 5V reference was there (5.00000V on my Keithley 196, which is amazing), and the 10Mhz signal was actually disciplining! Interestingly, though, I couldn’t see any of the satellites the unit was tracking, and anytime I went to check configurations, all I got was ?’s. Looking at the RS232 driver, I could see TTL on it’s output, and those went directly to the Trimble CPU. So, there ends the repair, as that part is “unobtanium”. Sad. So, the unit works great, it just takes 30-45 minutes to boot since it doesn’t store any configuration information. =/ Sadness. I’ll keep it, or maybe send it to a youtube channel I watch to try to further repair. =)
Special thanks to everyone on the Time-Nuts mail list, as without them to at least be a foil for my emails, I probably wouldn’t have gotten very far in the repair. =)