In this article, I discuss my self-designed 10 MHz frequency source with OCXO.

An OCXO is a temperature stabilised crystal oscillator with high frequency stability. The OSC5A2B02 OCXO used provides a square-wave HCMOS output signal of 5 volts. The frequency can be accurately tuned via the Vref input, with this a 10 MHz output signal can be achieved with an accuracy and stability of better than 0.2 Hz. This corresponds to a stability < 20 PPB

After calibrating the OCXO circuit with an external frequency source, it is an ideal low cost built-in frequency standard for measuring devices, such as a frequency counter. The accuracy and stability of this circuit is more than adequate for tuning oscillators in HF, VHF and UHF transceiver's and that is exactly the purpose this circuit was designed for.

I had the circuit board fabricated at PCBWay. The Quality of the supplied PCBs is high, texts are easy to read, components fit well and all dimensions are correct. The PCB is excellent to solder and I am extremely satisfied with the final result.

• Output frequency 10 MHz
• Output voltage 5 volts 10 MHz HCMOS square wave
• Stability and accuracy after calibration < 20 PPB

• Power supply voltage 11 - 16 Volt DC
• Current consumption during warm-up first three minutes < 0.6 Ampere
• Current consumption steady state < 0.25 Ampere
• HF power supply decoupled with coils

• Warm-up time approximately 5 minutes
• High stability for measurements after 30 minutes.
• Stability and accuracy after calibration < 20 PPB

• PCB dimensions: length 9 cm, width 6 cm, height 2.8 cm

• Two PCB signal outputs
  1. Direct
  2. DC decoupled with 100 pF and 50 Ohm series resistor

• The decoupling components of output and power supply are easily bridged at your discretion via solder pads.
• All connections with standard male pin headers 0.1 Inch.

Are you interested in buying a ready-made, tuned one, or only the bare PCB? Let me know bart@pa7elf.nl

I designed this OCXO board as a built-in frequency source for my own frequency counter. This homebrew counter originally had a standard 10 MHz crystal as a reference source, but that proved not accurate enough for tuning my transmitting equipment. The frequency drifted due to temperature differences and was also not stable enough over time. I had equipped this counter in the past with a TCXO a Temperature Compensation Xtal Oscillator, but unfortunately it turned out to be quite a bit off the specified 10 MHz and that TCXO was not adjustable, so not calibratable.

With a cheap TCXO it is hit or miss and so I started looking for a reliable and more accurate frequency reference for my frequency counter and soon came across the OCXO the Oven Controlled Xtal Oscillator. The OCXO has little oven containing the oscillator crystal and the oscillator circuit. This small oven is kept at a constant temperature with a built-in control circuit. As a result, a change in ambient temperature has no noticeable effect and the OCXO is an order of magnitude more accurate than a TCXO

After delving into the different brands and types of OXCOs for sale, I opted for a used OSC5A2B02 from CTI This is a stable and professional OCXO that was used in base stations of telecom equipment such as GSM masts. The used ones can be bought for a fraction of the new value, giving you a good-quality OCXO for little money. A second advantage, the OCXO has already been in use for a while, so you no longer have to deal with the relatively high first-year turnover of a new unused copy. By choosing a used one, you will have a professional OCXO that is already well burnt in, for a low price.

• CTI OSC5A2B02 translated
• CTI OSC5A2B02 Chinese
• CTI OC25 series same specifications English
• CTI OSC5A2B02 dimensions

Output frequency measurements clearly show that the output frequency of the OSC5A2B02 OCXO drifts during oven heating. After and half hour, no more gradient can be seen and the output signal is stable and suitable as a frequency source for measurements. So let the OCXO warm up for at least and half hour before making a measurement. During the first three minutes, the current consumption is about 0.5 A (< 0.6 A according to the specifications). As soon as the furnace temperature approaches the working temperature, the current drops to about 0.23 A (< 0.25 A according to the specifications)

If the OCXO frequency source is built into a measuring device, it is convenient to connect the power supply of the OCXO in front of the DC on/off switch. As soon as you connect the measuring instrument to mains voltage, the OCXO starts to heat up you do not need to switch on the rest of the measuring instrument for this. If this is not possible because the measuring instrument's on/off switch is in the mains circuit, you can consider installing a separate 12 Volt 1 Amp power supply and connecting it in front of the mains switch.

For calibration, preferably GPSDO or a precisely set rubidium atomic clock is used. In addition, you need a phase meter or oscilloscope with X-Y input for measuring the phase difference between the OCXO and the calibration source. A frequency counter for coarse setting is also useful. It is useful to lacquer the setting potentiometer after calibration, for example with cheap nail polish. This prevents the potentiometer from being accidentally twisted.

The Calibration procedure consists of first coarse adjustment and then fine adjustment.In coarse adjustment, the OCXO is adjusted near 10 MHz so that phase meter is only slow or the oscilloscope shows a single slow-moving Lissajous figure. In fine-tuning, the Vref potentiometer is carefully turned to make phase meter or Lissajous figure helically stationary. The aim here is for an entire 360° cycle to be slower than 100 seconds. At a frequency of 10 MHz, this gives a calibration accuracy of 1 PPB equal to 10E-9 This is well within the specified long-term stability of the OSC5A2B02 OCXO. More accurate calibration makes no sense.

I like to think about good design. Below, using the diagram, I explain the design decisions I made and how I arrived at them.

• Diagram PA7ELF Frequency reference with 10 MHz oven crystal

All connections on the circuit board are made with standard male pinheabers with a spacing of 0.1 inch. This is ample for a 10 MHz signal in a shielded enclosure. The circuit can be easily connected with standard female wired pinheabers, or with separate female pinheabers. Coaxial and power cable can be attached here at your discretion.

The power supply connection is designed as a three-pin connection. The middle pin is plus the outer two are ground When using a three-pin connector, you avoid mix-ups with the output connector and there is little chance of misconnecting the power connector. The input of the power supply is decoupled with two coils, one in the plus and one in the minus. If the OCXO circuit is built into, for example, a frequency counter, this prevents an unwanted 10 MHz signal from passing over the power supply wires to other parts of the instrument These coils can easily be bridged via a solder bridge if required.

One of the mounting holes has a conductive ring with the option of connecting this ring to the PCB ground plane via a solder bridge. Before using this option, make sure you do not create any unwanted ground loops

The PCB has two outputs. The internal output is connected directly to the ground and output of the OCXO. The external output is DC decoupled. For this output, the ground and signal output is decoupled with a 100 pF capacitor. The signal output also has a 50 Ω series resistor. Both the capacitors and the 50 Ω resistor are easily bridged via a solder bridge. Before using this option, check that you are not creating unwanted ground loops.

To use the 10 MHz signal, in many cases one earth connection, via power supply or chassis connection, is sufficient. To use the ground of the power supply line as signal ground as well, bridge the ground coil. YMMV.

Several things can be deduced from the OSC5A2B02's datasheet.

• The power supply stability is ≤ 2 PPB for 5% voltage variation of the 5-volt supply.
• The tuning range for setting the frequency of the OCXO runs from 0 - 4 volts and has a range of - 2 to + 2 PPM.
• The modulation bandwidth of Vref is at least 10 kHz.

So there are stability gains to be made with a tightly stabilised supply voltage. A well-smoothed adjustable Vref is a must. This is why I chose a two-stage voltage stabilisation for both Vcc and Vref.

As a starting point, I chose a 12-volt supply voltage, this way there is enough control range. Later, I calculate the maximum and minimum supply voltage. I deliberately use analogue supply control and do not use switched-mode power supplies to eliminate the risk of unwanted signals. In the first step, the supply voltage of 12 volts is reduced to 9 volts with a 7809. In the second step, this voltage is reduced to 5 volts with a 7805 for the main power supply of the OCXO. The 7809 captures the largest voltage variations and should the output of the 7809 still vary slightly, the variations are captured by the 7805. All regulators are equipped with their own double-duty decoupling capacitors, to suppress trancients and prevent oscillation.

From the datasheet of the OSC5A2B02, it can be deduced that there is a reasonable spread for Vref and that it should be adjustable over a range of 0 to 4 volts. This was confirmed by measurements on the various specimens I bought. The value for setting 10 MHz is between 2 and 4 volts on my specimens. This makes it clear that a wide control range is needed for Vref. What is missing in the datasheet is a value for Iref the recorded current of the reference input. When measured, this turns out to be < 0.1 μA and hardly varies when changing Vref. This is an important value for making a good design.

For making Vref, I chose a precise 5-volt shunt regulator, LT1029 or LM336 Z5. These shunt regulators have high temperature stability, which is a must for the OXCO's reference voltage. A key advantage of these shunt regulators is their high temperature stability. This is a must for a temperature stable Vref. The stability is better than 1 mV over a temperature range of 0 to 100 °C. This translates to OCXO frequency stability of better than 1 PPB and therefore has no noticeable impact on the frequency stability of the OCXO.

The shunt regulator is fed with a series resistor from the 7809 and therefore already has a well-regulated input voltage. As a result, there will be hardly any variation in the current of the shunt regulator, which improves the stability of the output voltage.

For setting Vref, I chose a potentiometer circuit behind the shunt regulator. For the current through the potentiometer, I choose 1 mA. This reduces the load through Iref by more than 10,000 times, so the voltage distribution in the potentiometer, is hardly affected. I measured no variation in Iref, which can affect the voltage sharing in the potentiometer. As a result, the output voltage of the potentiometer circuit and hence Vref will have the same stability as the stunt regulator the potentiometer is connected to.

For the potentiometer, I chose a 25 stroke adjustment trimmer from the quality brand Bourns. This gives a stable and accurately adjustable reference voltage to set the OCXO exactly at 10 MHz during the calibration procedure. This proved to be a good move, as a 25 stroke trimmer proved to be no luxury during the calibration procedure. The adjustment steps to reverse the phase gradient hell to zero are still very small.

There are several reasons why I opted for a dual power supply arrangement in this design. The first is high stability of Vref and maximum smoothing of transcients that may be present on the supplied supply voltage. The second reason is to distribute the thermal dissipated power across two regulators. This keeps the temperature of both regulators low which is beneficial for the stability and lifetime of the circuit. I deliberately use an analogue supply control rather than switched power supplies to eliminate the risk of unwanted signals.

In my education, I learned to read datasheets well and did a lot of calculations on circuits and I still like to do that. The advantage of reading datasheets and doing calculations is predictable behaviour of the circuit I build. This means I spend less time experimenting and it is always satisfying to see after building that the result comes out close to my calculations.

For the shunt regulators used, the specifications are given at a shunt current of 1 mA, which is the current at which they give the best results. The minimum current at which these shunt regulators still work reliably is 0.7 mA

For the potentiometer circuit, a current of 1 mA is chosen the value of the adjustment trimmer then becomes: `(5V)/(1mA) = 5 KΩ` The series resistance of the shunt regulator is connected to the 9 Volt output of the 7809. The current through series resistor for the shunt controller and the potentiometer which is in parallel with the shunt controller is `1mA + 1mA = 2 mA` The voltage drop across the series resistor is `9 V - 5 V = 4 V`. The value of the series resistor is

`Rshunt = (9V - 5V)/(1mA + 1mA)= 2k`

`Rshuntmax = (9V - 5V)/(1mA + 0.7mA)= 2.35k`

If 2 KΩ is not available, preferably choose a slightly smaller value e.g. 1K96 of 1% to avoid the current through the shunt regulator becoming too low, causing it to malfunction. A 2K2 resistivity with a tolerance of 5% has a maximum value of 2.31 KΩ and is already on the edge of the maximum shunt resistance.

Because a reasonably high current flows through the 7809 and 7805 when they heat up, it is worth making a thermal load calculation for both regulators to determine whether they should be equipped with a cooling fin. First I make a calculation to determine what the maximum thermal resistance of a cooling fin should be.

Both regulators have their maximum heat development during the warm-up phase of the OCXO because that is when the highest current flows through the regulators.

Basis for limitations:

• The controller's maximum junction temperature Tj is 125 °C
• Heat resistance from junction to housing 0.5 C/W
• Ambient temperature Ta is 25 °C
• Initial supply voltage 12 volts
• Maximum heating current maximum 0.6 Ampere
• Stay site current maximum 0.25 Ampere

Maximum dissipated power 7805:

• During warm-up phase: `Pmax = (9V - 5V)*0.6A = 2.4W`
• During steady state: `Pmax = (9V - 5V)*0.25A = 1W`

Maximum dissipated power 7809:

• During warm-up phase: `Pmax = (12V - 9V)*0.6A = 1.8W`
• During steady state: `Pmax = (12V - 9V)*0.6A = 0.75W`

The highest heat development occurs at 2.4W, this is the basis for refraction of the maximum heat resistance for the cooling fin.

Maximum heat resistance of cooling fin: `((125C - 25C) / (2.4W)) - 0.5 "C/W" = 41 "C/W"`

I chose the UK14 SA 220 cooling fin from Fischer Elektronik. This is a compact U previous cooling fin that takes up little space and can be mounted on an upright voltage regulator with an M3 screw. The heat resistance of this cooling fin is 20 K/W which is equivalent to 20 C/W

• Datasheet Fischer Elektronik UK14 SA 220
• UK14 M3 @fischerelektronik.de
• Available in the Netherlands at Haje Electronics

Maximum temperature of the 7805 with the selected cooling fin and a 12-volt supply voltage.

• During warm-up phase: `Tj = (20 "C/W" + 0.5 "C/W") * 2.4W + 25C = 74.2C`
• During steady state state: `Tj = (20 "C/W" + 0.5 "C/W") * 1W + 25C = 45.5C`

Maximum temperature of the 7809 with the selected cooling fin and a supply voltage of 12 volts.

• During warm-up phase: `Tj = (20 "C/W" + 0.5 "C/W") * 1.8W + 25C = 61.9C`
• During steady state state: `Tj = (20 "C/W" + 0.5 "C/W") * 0.75W + 25C = 40.4C`

The are all relaxed values.

The minimum voltage drop, Vdrop that a 7809 needs to function properly is 2 volts. The minimum supply voltage of the circuit is: `Vmin = 9V + 2V = 11V`

The maximum supply voltage is the value at which the 7809 will not overheat with the selected cooling fin.

`Pmax = ((125C - 25C) / (20 "C/W" - 0.5 "C/W")) = 4,87 W`

The maximum voltage drop Vdmax across the 7809 during the warm-up phase is:

`Vdmax = 4.87W / 0.6A = 8.11V`

The maximum supply voltage:

`Vmax = Vdmax + 9V = 8.11V + 9V = 17.11V`

For some extra slack in the thermal housekeeping, I keep a Vmax of 16 volts for this design.

• `Vmin = 11 " Volts"`
• `Vmax = 16 " Volts"`

The maximum Junction temperature in steady state and 17.11 Volt supply voltage is lower, because less current flows through the 7809 in steady state.

• Steady state: power `Pmax = (17.11V - 9V)*0.25A = 2.03W`
• Steady state junction temperature: `Tj = (20 "C/W" + 0.5 "C/W") * 2.03W + 25C = 66.6C`

The values calculated above are 'worst case scenario'. The measured warm-up and steady state current are lower than the maximum values specified in the datasheet. The maximum measured warm-up current is 0.57 Ampere. The maximum measured steady state current is 0.23 Ampere. As a result, the actual temperatures will be lower than the calculated temperatures.

The 10 MHz OCXO frequency reference built into my already many, many times rebuilt frequency counter.

A second 10 MHz OCXO frequency reference was built into his Rotex RFC-250 by my brother PG0V. The Rotex operates with a 1 MHz reference, therefore 10 MHz of the OXCO is divided by ten with a 7490.

The 7490 consists of a two divider and a five divider. In this circuit, the output of the OCXO is connected to the input of the five-splitter and the output of the five-splitter is connected to the input of the two-splitter. The output of the tweeter produces a nice symmetrical TTL square wave of 1 MHz, which is connected to the oscillator input of the Rotex. The original crystal of 1 MHz has been removed Important detail, the smoothing capacitor of about 10 nF across the Vcc and Gnd connection of the 7490.