ませ１りすか

Accurate frequency calibration is particularly important for digital modes such as FT8, and even more so when operating on 50 MHz and above.

WSJT-X includes a very useful tool for this purpose: FreqCal.

A common way to think about FreqCal is:

That works, but I think FreqCal is even more useful if we use it one step earlier.

My preferred sequence is:

1. Let the radio stabilize first

Do not begin precision adjustment immediately after power-on.

The reference oscillator will normally drift as the radio warms up. In practical amateur use, 30 to 60 minutes may often be sufficient, but I do not consider a fixed number of minutes to be the real criterion.

A better approach is to use FreqCal itself: watch a stable reference signal and confirm that the measured offset has stopped moving consistently in one direction.

2. Use FreqCal to adjust the radio itself first

If your transceiver has an internal reference-frequency adjustment, I recommend correcting the radio itself before relying on software compensation.

On the IC-7300, for example, this is the REF Adjust setting. Other transceivers may use a different name.

Switch WSJT-X to FreqCal mode and receive a reliable reference signal.

During FreqCal operation, WSJT-X tunes the radio so that the reference carrier should appear at approximately 1500 Hz in the audio passband.

For the initial coarse adjustment, I prefer to use the highest reference frequency that can be received reliably at that time.

Adjust the radio's reference until the measured carrier is as close as practical to the expected 1500 Hz position.

This is only the first step. Do not assume that a single frequency proves the radio is perfectly calibrated.

3. Check several frequencies

Once the rough REF adjustment is close, let FreqCal cycle through several reference frequencies.

With a modern synthesized transceiver, the remaining frequency error should approximately follow a straight-line relationship with frequency.

This multi-frequency check is important.

A radio may appear almost perfect at one frequency while still showing a residual frequency-dependent error across the band.

If necessary, make a final small adjustment to the radio's hardware reference, then leave it alone.

4. Start the final FreqCal measurement

Now the radio's hardware reference is fixed.

Enable Measure and let FreqCal run through the chosen calibration frequencies several times.

In current WSJT-X versions, existing Slope and Intercept correction values are automatically disabled while Measure is active.

Therefore, you do not need to manually set Slope and Intercept to zero before collecting new measurement data.

The measurement results are stored in:

5. Use suitable reference signals

Use only signals that are reliably receivable at your location.

It is not necessary to force every nominal calibration frequency into the measurement. Propagation conditions change, and some frequencies may be weak or unusable at a particular time.

The WSJT-X User Guide also notes that major-city AM broadcast stations can serve well as calibration signals at the low-frequency end of the spectrum.

I agree — they are very useful.

But there is an important practical point.

If you have several suitable AM stations, I think it is perfectly reasonable to include all of them initially.

Then collect several FreqCal cycles and look at the results.

Example from my station in Tokyo. The AM frequencies shown here are local examples only.

6. Before solving, inspect DF

This is the additional step I now recommend.

Before simply accepting all collected measurements, look at the reported DF values in frequency order.

After the radio's hardware reference has already been reasonably adjusted, the remaining frequency error should normally follow an approximately linear trend with frequency.

For example, if the residual error is increasing with frequency, a higher-frequency reference would generally be expected to continue along that same trend.

If one station at a higher frequency suddenly shows a distinctly smaller deviation than the lower-frequency stations, or otherwise reverses the overall trend, that point deserves closer examination.

7. Recalculate with suspicious AM points removed

A practical procedure is:

8. What do Slope, Intercept and StdDev mean?

Slope

Slope is the frequency-dependent part of the error, expressed in ppm.

A smaller absolute Slope means that less residual error grows with frequency.

Intercept

Intercept is the fitted error at 0 MHz, expressed in Hz.

It is not an actual measurement at 0 MHz. It is the intercept obtained by extending the fitted straight line.

StdDev

WSJT-X defines StdDev here as the root mean square deviation of the averaged measurements from the fitted straight line.

In practical terms:

9. What happened in my Tokyo measurement

At my location, two AM frequencies appeared inconsistent with the overall trend:

• 594 kHz
• 1422 kHz

I therefore compared four calibration data sets.

Removing both AM frequencies produced a much tighter fit to the remaining calibration data.

The final solution I adopted was:

10. Apply and verify

Once you are satisfied with the data selection and the calculated fit, use Apply to set the resulting Slope and Intercept in WSJT-X.

Then clear Measure and run FreqCal again.

With Measure off, the calibration correction is active, so this gives a useful visual check of the final result.

If you use JTDX with the same radio and the same CAT frequency reference, the same calibration parameters can also be used there.

FreqCal is much more interesting when used this way.

It is not only a frequency-correction function. It is also a useful tool for understanding the complete frequency behavior of the radio and the reference signals being used to calibrate it.