The Hewlett‑Packard Z3815A is a GPS‑disciplined frequency reference originally designed for telecom‑grade timing applications, but today it has become a prized instrument among precision‑timing hobbyists, metrology labs, and the broader time‑nuts community. At its core, the Z3815A combines a high‑stability E1938A “hockey‑puck” oven‑controlled crystal oscillator (OCXO) with a robust GPS receiver subsystem, allowing it to deliver exceptionally stable 10 MHz and timing outputs.

Internally, the unit features a compact, layered architecture: a thin oscillator mezzanine board sits above the main motherboard, while the GPS receiver module interfaces through the rear panel’s antenna and I/O connectors. The design reflects HP’s late‑1990s/early‑2000s engineering style—serviceable, modular, and built for long‑term stability.

For collectors and precision‑timing practitioners, the Z3815A offers a compelling blend of historical engineering and practical performance. Its stable OCXO, disciplined by GPS, provides a frequency reference suitable for calibrating counters, stabilizing lab instruments, or anchoring a home time‑and‑frequency setup.

These units were relatively cheap and plentiful in 2017, and I took advantage of this and purchased three of them. They are genuine HP units, made in Korea in 1997 according the the HP tag. All three have the HP E1938A crystal oscillator. I did some initial testing, determined that they were all functional stored them for several years.

Recently I pulled them from storage, intending to bring them into service as sources for 10 MHz and 1 PPS house standards. I quickly realized that not one of them worked properly. They would power on and then fault out within a couple of minutes.

The Z3815A uses two Vicor DC to DC converters: One for the 5 V DC rail and one for the 15 V DC rail. The converters operate from the nominal input voltage which is 29-50 V DC for these units. The 15 V DC converter had failed in all of the units. I located the fault by simple voltage testing. Here are the converters in the Z3815A:

The converters have a built in heat transfer plate which is bolted to a heatsink:

The converter was removed by desoldering the through hole pins.

I then powered up the unit by applying 24 V DC to the power input (P10) and 15 V DC to the output pads of the removed converter. I connected a self-powered GPS antenna to the GPS Antenna input and no surprise, the unit powered up normally. Using GPSCon software, I could see the site survey running and the warmup/gps lock process moving along as expected. I put a fan on the chassis to cool the 5V converter and left the unit for several hours. Eventually it achieved lock and all the front panel indicators came to normal.

The 10 Mhz and 1 PPS outputs were functional.

Here are both DC to DC converters:

A bit of research showed that while the converters are still available on the surplus market, the price would exceed what I paid for the Z3815A. And on the surplus market, you never know what you may be getting. The converters add significant heat to the chassis: About 40 W on startup and around 25 W after warmup.

However both the 5 V and 15 V converters can be bypassed and the Z3815A will run properly. I powered the unit from my lab bench supplies and measured the following: 5 V DC @ 5A and 15 V DC @ 390 mA at startup, dropping to 5 V DC @ 3.2 A and 15 V DC @ 390 mA after warmup. Using the bench supplies, the Z3815A powered up and came to GPS lock as expected.

I removed both the DC to DC converters from all three GPS units. The next step will be to fabricate an interface box which has the appropriate inputs for GPS antenna and power (5 V DC and 15 V DC) and outputs. There are three buffered 10 MHz outputs and a 1 PPS output. A good quality power 5 V /15 V power supply (MeanWell) will power the units.