Inexpensive Precision Frequency Standard by Neil Robin, WA7NBF Port Angeles, Washington Updated and revised: August 1, 2009

Byline:

Create a homebrew 10 MHz frequency standard for Frequency Counters under $70.00.  Accuracy better than 0.2 PPM over a years time!

Introduction:

One thing that’s always handy around the ham shack is a frequency counter to check accuracy of your transmitters.  The trouble is a highly stable oscillator which is usually out of reach for most amateurs is needed as the counters reference.  What follows is an approach that you can put together for under $70.

I bought a used low cost counter, Goldstar Model FT-2130 a few years ago on e-bay to measure the frequency of all my handheld and base station radios for HF, VHF and UHF.  I also build foxhunting systems which typically use old synthesized radios so I wanted to be sure they meet specifications before putting them to use.  The counter has a ÷10 prescaler so it can be used up to 1.3 GHz.  This is more than enough for all the amateur bands including 23 cm.

One of the first things you ask is how good is the time base for the counter?  You can't set a transmitter frequency accurately if you don't have confidence in the reference you're using.  Just as I thought, my built-in counter oscillator was low in frequency by 2.8 PPM (parts per million) after it warmed up for a twenty minutes.  It’s a type called TCXO (temperature compensated crystal oscillator).  They're relatively cheap and that’s why you find them in medium precision devices such as counters.  I had no idea what the drift characteristics would be over time, effects of line voltage or as temperature varied? It could be much worse than 2.8 PPM in routine situations.  In short, I didn't really know if this measured error was good or bad nor how it varied with environmental factors?

What influences accuracy in counter measurements:

Digital frequency counter accuracy is influenced by several factors.  But, you should know how much accuracy you need for a given measurement.  You can go crazy chasing something you really don't need.

For counters:

1. Accuracy of the crystal reference oscillator under the effects of temperature, aging, voltage, etc.
2. Length of measurement period sometimes called the Time Base (TB) setting.
3. LSD (Least Significant Digit) resolution.  Be careful of prescalers and ±1 count error.
4. Number of digits on display. Knowledgeable operators can work around this if you have too few.
5. Signal noise level. Poor S/N can reduce accuracy. Strive for at least 30 db S/N and limit the amplitude of the measured signal so that the counter hysteresis can reduce any remaining effects of noise.

Item #1 usually has the most influence in measurements when a long TB is used and input frequency is high.  When high accuracy is needed, use as long a TB as practical, 10 seconds being common but 1 second can be used in some circumstances.  Some counters have a 100 second time base but unless you have the reference oscillator to compliment it, you probably don't need it.  For very low frequencies, audio for example, a period measuring counter, using period averaging, gives superior results.

Item #3 and 4 comes into play when you display too few digits of resolution.  Most inexpensive counters have ±1 count error on the LSD.  This comes about because the reference oscillator and the input signal are usually asynchronous with undetermined phase relationship.  More expensive counters can reduce this error to 1/2 the amount.

In frequency counter measurements, the LSD resolution limits what you can achieve.  Each count of the LSD gives the error in PPM:

You want the PPM_{per count} value to be as small as possible to gain the best resolution.  Examining equation (1),  it’s clear that you want to avoid prescalers if you can and increase the Time Base to as long as possible.  Ten seconds is usually the practical limit.  At 147 MHz with a divide by 10 prescaler using 10 second TB yields 0.0068 PPM/count. This is significantly better than needed and it’s clear that the oscillator stability will be the limiting factor in overall accuracy.

It's rare today to find a counter with a ÷2 prescaler but if so; the TB selection will usually be increased by 2X to correct the display.

How much accuracy do I need?

When estimating the needs of radio transmitters, it’s usually the channel spacing that controls the oscillator stability requirements. Two meter radios have channel spacing in their synthesizer of 5 kHz.  You should not allow frequency instability beyond 5 KHz/2 or ±2.5khz, preferably less.  At 147 MHz this works out to:

2.8 PPM as measured above isn't bad for a low cost counter if all you need is 17 PPM.  Of course other bands will have different spacing requirements so this target would be band specific.  445 MHz with 5 KHz spacing requires stability of ±5.6 PPM.

I really wanted something better that I could trust over a long period of time.  One solution is most counters have a rear jack for an external reference signal.  This is the usual way to solve this problem, particularly in a lab environment.  Just plug in the "house frequency standard" and off you go.  One problem to overcome is the frequency needed? Most counters require 10 MHz as a standard reference plus adequate amplitude to drive their divider circuits.

Vectron oscillator

I purchased a high precision OCXO (Oven Controlled Crystal Oscillator) on e-bay that met all my specifications with flying colors but the signal output was 5 MHz.  They sell for about 35-$40 and are brand new.  Oven control means that the crystal and key oscillator parts are inside a very small oven in the main housing.  In this case, it’s kept at 80° C.

Of course, when you buy something without knowing its history or the integrity of specifications, you’re exposed to risks.  For something as important as a frequency reference, you'd like to be sure its stable.  The best way I've found is to recheck its performance a number of times over a year and note any serious deviation, particularly when they challenge published specifications. When things are sold "new" at auction such as e-bay, they may be deficient in some way and unusable for the original application or they may have just been excess.  Buyer beware, you may have a good deal or not.

I started checking its accuracy, at least once a month, and so far its better than my resolution capability of the error signal.  I'll go to once or twice a year once I build confidence that its making specifications or is stable within my needs.  This builds confidence that it’s really as stable as they claim.   Notice that ambient temperature extremes and aging have about equal effect; far better than needed for an oscillator in amateur service.  A copy of the complete specifications can be found here.

The Vectron oscillator was purchased from Alltronics which usually had several models but their source seems to have dried up recently. By maintaining a search on e-bay for OCXO's, you should hit pay dirt soon.  New, these oscillators sell retail for $250 and up.

The finished product and circuit theory:

The task is to interface the OCXO to the counters input standard frequency and amplitude requirements.  Since the counter expects 10 MHz and the oscillator produces 5 MHz, some type of frequency multiplier is needed to give a doubler function.  Most would consider a phase locked loop (PLL) which can be implemented with a single IC and a few components.  A problem with this approach is that if the 5 MHz oscillator fails the PLL will still run and generate an erroneous signal near 10 MHz that is uncalibrated.  Not a good idea for a frequency standard.  You can implement an "unlock" detector but this seems a roundabout way to solve the problem. Another concern was that many PLL's don't generate clean sinewaves which I wanted.

Years ago I use to build multiplier circuits for stable UHF signal references and wondered if I could still do it even though this was only 10 MHz? Figure one shows the approach I used.  A doubler, T1 and D2-D3 injects a signal (now 10 MHz) into the circuit, Q1, that would normally be a colpitts oscillator but purposely has limited feedback so as not to oscillate w/o the injected signal.  It might be best called a high "Q" bandpass filter.  If the 5 MHz source were to fail, the Q1 circuit would also fail and we'd have no output, a desirable outcome for a frequency reference.

Figure 1

The schematic excluding power supply can be found here.  T2 along with C2 and C4 form a 10 MHz resonant circuit.  Injection of the voltage doubler, a 10 MHz signal from T1 is via 1/2 turn winding into the torid core, T2.  The effective "Q" of this circuit can be increased by splitting C4 into two series parts with the common point connected to the emitter.  This creates a Colpitts oscillator and with enough feedback, can self-oscillate even when the Vectron oscillator is disabled.  I have also considered using a 10 MHz crystal as the resonant element but again be careful of feedback because of its very high "Q". Crystal filters are commonly built that could perform this function as well.  Q2 is an amplifier stage giving about 11 volts P-P (Peak to Peak) into an open circuit which is more than enough to drive my counter reference jack.

You might notice that I used three torid cores for the transformers of this circuit.  That’s not by necessity but only because I enjoy winding my own inductors and experimenting with values used.  T2 and T3 could just as easily be made with plain inductors that are reasonably low loss.  T2 could be implemented with a 5 uh inductor and a 1/2 turn wound over it for coupling. The circuit is tuned by adjusting C2 for maximum output signal.  It would also be a good test to check to be sure the output drops to zero when the Vectron oscillator is disconnected to be sure no self oscillations are present.

The entire circuit is run off of +12 VDC.  I designed it that way for several reasons:

1. The VECTRON unit requires 12 volts.
2. I wanted enough head room on the output signal to drive my counter, 11 volts open circuit.
3. In a emergency, I could use a 12 volt battery as a power source.  Although not regulated, it would be fine if I had to.
4. High efficiency +12 VDC ±5% switching power supplies are readily available at very low cost.  High efficiency is desired because this standard would be powered up 24/7 to maintain stability.
   

Calibration and verification:

OK, how do I calibrate a secondary frequency standard?  Their are two common ways for most amateurs.

1. Locate a "Frequency Standard" that has better performance then this unit, namely, stable to at least 1.0 x 10^-8 .  Be cautious that it's well maintained itself.  Usually you will find these devices in an electronics or communications labs.  You may have a friend that works at a facility that could provide this service.  One warning, to carry your standard into a lab requires that it be turned off which is something that is stressfull for frequency standards and can foil up performance for days.  You may find that the stress may not be worth it.
   
   
2. If you have an HF receiver, tuning to WWV or WWVH at 10 MHz is the most convenient way to check accuracy, .....to a limit.  Don't expect to do better than 1 x 10^-7 .  The stability of WWV is around 1 x 10^-13 but HF propagation degrades it down to the much lower figure.  Long term averaging can improve these figures but few amateurs have the equipment needed to make these corrections.
   
   As you approach a difference of a few hertz in beat frequency, you'll probably begin to notice erratic shifting of phase between the two signals.  This is due to constant changes in HF propagation and is what limits this method.  Many papers have been written on how to get the most out of WWV as a reference and can be found at: http://tf.nist.gov/timefreq/index.html

Number 2 is the only practical solution for most of us.   Their are other frequencies where WWV broadcasts references but 10 MHz is easiest to work with if your secondary is also 10 MHz.  You'll need suitable propagation at the time of calibration but its broadcast from more than one place in the world.

Figure 2 Using WWV as a reference.  Radio would be tuned to near 10 MHz

The idea is to beat the incoming WWV or WWVH signal with your new secondary standard and listen for the beat frequency which will be the error signal.  See figure 2.  Use the AM setting for demodulation rather than SSB since the carrier injection will add confusion.  The receiver is only needed to hear the weak signal from WWV.  You don't rely on the accuracy of any internal oscillators of the radio.  The secondary standard signal must be adjusted to a low level using attenuators and about the same level as WWV so as not to over power the receiver front end and block WWV .  

A few problems must be considered and addressed:

1. The "beat" will be very low in frequency as the two signals become nearly identical in frequency.  For 10 MHz references, an error of 1.0 PPM gives an error beat of 10 Hz.  Many communications receivers don't pass much less than 50-100 Hz in their audio circuits so hearing a tone is not likely at 10 Hz.  Most operators can hear a "flutter" of the beat frequency when it’s down below about 10 Hertz and begin to notice fluxuation on the "S" meter when AGC is set in a 'fast' mode.  In short, matching two signals to be less than 2 PPM (20 Hz beat frequency) with this method can be challenging but it certainly can be done.  An oscilloscope can also help judge when the two signals are getting close to each other in frequency.
   
   
2. WWV transmits various tones and signals during its one hour cycle and these can seriously interfere with the base carrier signal.  Their are dead periods of 15-18 minutes/hour where no tones are transmitted and that makes beat frequency measurements much easier.  Use these periods to judge the "beat" between the two signals.  For more information, consult: http://tf.nist.gov/stations/iform.html#stdt which has a table showing the transmission sequences.
   
   
3. Trying to get a match in frequency of better than 1 PPM requires tracking the "flutter" signal and adjusting the rate to well under 1 Hz.  I can reliably adjust the oscillator using this technique which gives a mismatch of only 0.1 PPM.  Ideally, you should have something that can see beat frequencies down to zero hertz and/or use an oscilloscope or some other near DC measuring device. What your doing here is making phase measurements. My standard runs with less than 1 Hz beat frequency which makes it less than ±0.1 PPM mismatch.  Listening to phase changes between two signals is easier than trying to hear say a 20 Hz tone when the receiver audio amplifier doesn't pass them well.  The good news; the oscillator specs limit doing any better than this and most amateurs would have little need for accuracy at this level anyway.
   
   
4. Using the "beat" frequency method doesn't tell you if your standard is higher or lower in frequency.  Of course, you can force a slight adjustment in the trimmer on the VECTRON oscillator and see if the error signal increases or decreases in frequency.  This is not recommended as you really don't want to tamper with the oscillator any more than neccessary.  Another way is to turn on SSB demodulation using USB (upper sideband) on the receiver and tune for a comfortable audio tone, first listen for WWV and then your standard.  By turning the frequency standard output, "on" and "off",  see if the beat frequency against WWV or your standard is the higher frequency. I've found that using an antenna switch to cut off the WWV signal works pretty good if the standard remains running at a lower level.   Human hearing is very sensitive to tone differences so you can probably tell the difference of 1 Hz with a 100 Hz tone.  This still doesn't give a quantitative value of the error but it helps in establishing direction of error. You could also measure the difference in frequency between WWV tone and the standard tone quickly to get the error frequency for quantitative recording.
   

If you have a friend that works at a communications lab or facility, they probably have a "house standard" that could be used to check your standard from time to time.  Be sure it’s really better and well maintained itself!  As these oscillators become more and more precise, they require careful handling which means no physical shock or wide changes in temperature.  They may not recover from temporary abuse and that includes your standard.  Never power down a standard oscillator if you can avoid it.  Some can take days to recover to full specifications.  For these reasons, I'd rather use WWV and leave my standard sitting on the shelf powered up at all times.

Summary:

Verifying performance with WWV is showing that the standard is quite stable.  So far, I've had less than 1 Hz beat frequency difference or "flutter" between the two signals.  I leave it run 24/7 so that it has no induced thermal shock over its life.  I use a Yaesu FT-757GX HF transceiver for the measurement.

The circuit is not an attempt for simplicity but to make it work with components I had available.  To build it again, I would probably try a 10 MHz crystal filter to build the high "Q" circuit.  Many variations are possible, please share any novel ideas you come up with.

Neil Robin
WA7NBF
neil@robin-wood.com