I got a GPSDO time reference for my home lab a few months ago and while I've been happy with the inexpensive model, I've wanted a few available outputs to go off to various instruments. There are commercial amplifiers available and many video distribution amps can be repurposed for reference frequency distribution, the overall design seemed simple and there was some extra space in my small GPSDO's case, so I decided I'd design one to fit in the case and be as unobtrusive as possible. This page documents the design and gives you all the information you need to make one of your own, including a guide to finding the signals you need to tap off of when installing it. Of course, I cannot be held liable for building it up, modifying your GPSDO, or damaging anything in the process of it, but for people like me, it's a fun, low risk project with a useful result.

The design:

The point of the project was to fit into the case my BG7TBL 2016-05-31 was already in, and since the extruded aluminum case had a matching board groove on the top as well as the bottom, I had a bit of clearance over the heatsink on the primary regulator and some open space on the front panel to drill holes for more BNCs. I could easily fit 4 BNCs, a 5th may be possible, but it would be quite a squeeze without moving the antenna cable. While my unit is configured for square wave output and I haven't come across an instrument that needs a sine reference specifically, I wanted the flexibility to have both options as outputs. My initial attempt at laying out a board was similar to what I ended up with, but included an opamp for the sine wave output and a header with a jumper for selecting the output.

This design worked alright, but it had some disadvantages. Because it was laid out as a simple follower amp without feedback resistors, the input needed to be connected to keep the opamp from drawing tons of current and the signal was small - it would work fine if you added a 2-3x gain, but after an output resistor to limit current, the resulting output signal wasn't very high in amplitude. Fixable problems, but as it turned out, not worth the effort. A decent, low noise opamp was going to cost a few dollars, not an expensive part, but much more than the other parts on the board (except the connectors), and the same functionality could be achieved with a square wave and a filter, so I set out to design one. It turned out that two capacitors and an inductor in standard values did a great job of smoothing the signal out, and while it's a bit lumpy with a high impedance input, it's a great looking sine wave on a 50 ohm load.

I also did away with the selectors for sine and square, since the board was going to be inside a box where things weren't accessible, and I was able to remove some other headers for alternative signal inputs. Because of the way the GPSDO is designed, the OCXO produces a sine wave which is squared up and fed into the CPLD for timing measurement, then out to a driver and the BNC, but the sine wave output variants of the GPSDO do a similar filtering technique for the output, so the input signals were there for either option and the output was going to be a filtered square wave even in the original design.

The second version was the final one, with fewer parts and expense and essentially the same performance. It even includes some mounting holes for structural support if used in a different housing than the BG7TBL GPSDO case. I opted for local regulation to be sure it would not load down the main regulator on the GPSDO board since it ran warm and already had a heatsink, but the finished unit only needs about 40mA in operation.

Building the distribution amplifier:

This project is fully documented, the PCB layout files and schematic can be found here and were made in CircuitMaker, which is a free PCB design tool (not without its faults, but powerful for free software). I've also included the schematic here for reference.

The bill of materials is small and consists of:
The PCB (10cm x 4cm double sided)
4x BNC connectors, I used a Molex 0731385033 but the footprint seems fairly standard - get decent quality ones so they last!
2x 74LVC14 14 pin SOIC hex Schmitt triggers or comparable TTL level inverter
1x AZ1117 series 5V regulator in a SOT-223 package
2x Radial electrolytic filter capacitors with a 5mm lead spacing and less than 20mm height, I used 330uF 50V 105C rated Nichicon caps, must be rated to above input voltage
2x 4.7uF 0805 ceramic capacitors
2x .1uF 0603 ceramic capacitors (for bypassing)
2x 1nF 0603 ceramic capacitors (for bypassing)
4x 75 ohm 0603 resistors
1x 1 Megaohm 0603 resistors
2x 2 pin 2.54mm male headers, I used one straight and one right angle

Those parts will build you a board with four square wave outputs, then for each one you want with a sine wave output instead, add:
1x 1uH 0603 inductor
1x 680pF 0603 capacitor (NP0 as it is for a filter)
1x 470pF 0603 capacitor (NP0 as it is for a filter)
1x 1nF 0603 capacitor for AC coupling the output

Assembly of the board is fairly straight forward for most parts, and the pads are left large for hand soldering (large enough for 0805s to fit on the 0603 pads). The only complicated part is the output filtering: each channel can be configured for square or sine wave output, for a square wave, populate R1, 2, 3, or 4 (whichever corresponds to the channel), then put a solder bridge in place of L1, 2, 3, or 4 (matching the number of the channel again), and bridge C9, 10, 11, or 12 that corresponds to the channel you are configuring. For sine wave output, populate all of those passives on the channel you want - the two capacitors which are part of the filter, the inductor, and the capacitor on the output to AC couple it.

Installation:

This section assumes you are installing the board in a BG7TBL 2016-05-31 model GPSDO, the one it was designed for. Several other dated BG7TBL GPSDOs will be nearly identical, but very early ones using branded GPSDO modules may not physically accommodate the board. The amplifier could realistically work in almost any chassis, but I obviously can't provide installation instructions for all of those. When your board is built up, you need to locate and tap off the input signals. A list of parts needed are as follows:
1x 2 pin 2.54mm male straight header
1x length of shielded single conductor wire
1x short two pin to two pin female header cable
1x female two pin housing and crimp pins
Electronic safe silicone or other glue to give a little strain relief to your mod

The first is easy, solder a 2 pin straight header to the terminals marked DC In next to the DC input jack - this is where your board will be powered from. Then you need to tap off a signal source, and I've highlighted a few good sources for square waves in light blue on this image of the bottom of my board. You'll want to use a single conductor shielded cable to run the signal up to the amplifier, and luckily they can be found very cheaply in small quantities - I got a one meter length from Digikey for just over a dollar. If you have any doubt about which pin the source is, measure it with your scope - it's not a very square looking wave on mine, but it certainly is not a sine wave. Near the signal source, find a grounding point for the shield of your cable, some good choices are marked in purple on the image. Terminate one end of the cable in the two pin female housing, this will connect to your amplifier board, then cut to length and attach to the points you've located for your square wave source and ground shield - I went through the holes on the board to solder on the underside and then used some electronic safe silicone adhesive (does not smell like vinegar when curing) to stick it in place and provide a little mechanical strength to the mod. As you can see, since I did this mod for the first design versions, I had two tap off points - the orange boxes are around potential sine wave sources.

The yellow box is unrelated and just shows where the sine wave and square wave variants of the GPSDO differ, the filtering section for the 10MHz output. After the electrical part of the mod is done, you just need to drill the holes for the BNCs in the front PCB plate. They are tough to measure, but you can get exact BNC spacing from the PCB design files, then just remember that the board will be installed on the PCB ridge in the chassis directly opposite of the one for the main board, so it will be installed upside down. I ended up measuring a bit and eyeballing a bit, and it worked out fine enough. Then you thread in the BNCs through the newly cut holes, screw their nuts down, and connect up the power and signal connectors. Then you should be able to just slide the thing into the case, and when the screws are back in the front plate, you are finished!

Performance

Performance seems perfectly usable and generally decent. There is a little ringing on the square wave output and both outputs should be terminated with 50 ohm impedances (1M impedance will work, but the waveforms look bad), but they are consistent. With a 75 ohm limiting resistor, as listed in the schematic, you get a 2Vpp output on the square wave and just over a 1Vpp output with the sine wave configuration. The board used about 38mA without any outputs connected and just over 39mA with two connected, so power consumption is quite low. The image below is after a couple minutes of running with infinite persistence on. The yellow trace is the built in square wave output, the pink sine wave is the sine configuration output from the distribution amp, and the blue square wave is the square output from the distribution amp.

The whole mod costs around $12 in parts, with close to $10 of that in BNC connectors, the board was another $10 but only because I had to order several to get it made. Given that it effectively takes up no space and gives you four more outputs, this simple and cheap mod seemed to be a great value for me! If you try it as well, I hope you see similar results.

December 14, 2016