RF systems and precision measurements usually require a stable frequency reference to generate accurate RF carriers or to accurately measure the carriers generated.

I currently have a D-Star repeater running at home using converted NMT 450 base station equipment, the transmitter and receiver use TCXOs as a time base reference, each unit needs to be calibrated periodically to hit the narrow band carriers in D-Star.

I plan to replace the synthesizers in these units with a custom design, since they currently use modified (hacked more like) synth boards that aren’t likely to be reliable long term, I intend to build these units to accept 10 MHz inputs as synth references to remove the need for calibrations.

I have a Z3810AS GPS timing system, which includes two OCXO units controlled by GPS, the system has a 15 MHz output (redundant) and each unit has a 10 MHz test point (lower quality, more phase noise). We’ll pretend we can convert the 15 MHz to a more useful 10 MHz, in that case a device is needed to preferably use the 15 MHz derived signal as the time base.

If this disappears (power failure will cause the 15 MHz to be disabled until a lock is attained) then we may want to use one of the 10 MHz test points.

In the Z3810AS system the secondary unit which locks onto GPS first will be more reliable during startup, so we want to use the 10 MHz output from that unit as the secondary source until the 15 MHz is enabled once the system reaches a certain accuracy. We may alternately use a completely different source for the secondary, such as a rubidium oscillator or stable OCXO/TCXO.

In the case of complete failure of the Z3810AS or its power supply both signals will disappear, in this case we want a second redundancy that can provide a degraded but usable 10 MHz, for example a low cost TCXO.

The solution

To ensure reliable performance I decided to build a distribution amplifier that support redundant inputs with automatic switching to ensure these radios always receive the most reliable frequency source available. This amplifier has two inputs + an internal oscillator.

Each input + the internal signal is fed to a 74HC125 tri state buffer with separate enable signals, one of these is activated based on a priority encoder and the winning signal is fed to a 74HC14 driver IC that drives four outputs with a 50 ohm series terminated AC coupled CMOS output signal.

To detect loss of signal two 7474 latches are clocked from the internal clock, if the input clock is stuck in any state for more than two cycles this will be detected by the two latch chains, a NAND gate combines the signals to one output. This circuit is repeated for input 2, these two outputs then indicate the status of the clock inputs, and are fed to a priority encoder configured to choose input 1, then input 2, then internal clock if all else fails.

If an input is lost the circuit will switch inside of 2-3 cycles of 10 MHz, ensuring minimal disruption. If both external sources are lost it will switch to the internal reference until one of the other sources are reliable, ensuring some sort of 10 MHz source is always present.

Typical application

My particular application for this design is as follows:

Figur 1 - My example application

We assume the 15-10 MHz converter is present and working, in this case a failure of one of the RFTG units will cause the other to take over in providing 15 MHz, therefore we have redundancy into the 15-10 converter. If the RFTG system is reset then the 15 MHz is disabled until it reaches around 1 ppb or better and GPS lock is attained, in that time period there will be no signal from the 15-10 converter. The Ref-1 unit in the RFTG will reach a stable frequency faster (but will not be as short term accurate after full lock is attained), so we use that as the secondary input.

If the RFTG fails completely, for example a bad power supply, then the internal TCXO takes over and we may continue operation with reduced accuracy (~1 ppm or so, compared to around 1x10^-12 as is typical from a GPSDO).

This system therefore can handle three failures (hard reset of RFTG, failure of one RFTG unit or complete failure of both RFTGs) without failure of the output, the system as configured may therefore be called triple redundant by some definitions.

By adding a third source (external OCXO/LPRO) the system can truly be triple redundant, however a single failure point exists in the internal timebase.

Circuit descriptions

Input stage

Each external input is transformer coupled and 50 ohm terminated, the center tap on the secondary is used to set the common mode voltage to half the supply rail.

The balanced signal is fed to the inputs of an LT1713 high speed comparator, the Q output is used for the tri state buffer and the Qbar output is connected to the clock detection circuitry. This reduces the load on the Q output where signal fidelity matters.

Figur 2 - Input stage

The LT1713 is not ideal wrt. jitter but it is relatively available and simple, as the clock signal must also pass through standard 74 series logic it’s not a huge loss in fidelity.

Clock detection

The clock detection works by clocking two D latches from the reference clock and wiring the external clock to the Set port, in this case if the external clock is stuck high for two cycles of INTOSC the zero from the first latch will flow through the latch and the output will change state. The circuit is repeated with an inverter to check for stuck low situations and a NAND gate checks that both chains are set correctly.

Figur 3 - Clock detection circuit

Output driver

Each output port has a 74HC14 driver with 22 ohm in series and a basic edge shaving filter to reduce EMI. It is assumed that the high frequency output impedance of the 7414 is relatively high.

Figur 4 - Output driver

Test results

Stay tuned, still testing.

I intend to check this design for added phase noise using a spectrum analyzer and a pure signal source.

Schematic, PCB layout

The schematic, BOM and gerber files can be found here:

10daq\10daq-gerber.zip

10daq\BOM10daq.pdf

10daq\elsc10daq.pdf

Notes:

BOM error, transformer is Mini-Circuits, not Magnecraft

PCB layout: please note that the top layer of the PCB is a power plane, the ground plane is on the bottom layer. The mounting holes are connected to ground.

74 logic family: higher speed ICs are generally better wrt. jitter, for U2 and U1 use only CMOS types: HC, AC or better.

HC or better is recommended for all other ICs to reduce power consumption but is not required

Assembly notes:

None in particular, socketing the internal reference may be advantageous, but also increases the chance that it will fall out or develop a bad contact later. I recommend using a TCXO instead of a standard crystal oscillator to improve performance if you expect the external signals to disappear.

Here are 3D renders of the top and bottom layers:

Figur 5 - 3D render of top layer

Figur 6 - 3D render of bottom layer