In this video, Dave Jones tunes a frequency counter’s internal frequency standard by comparing it to a rubidium frequency standard on a scope. He triggers on the latter, freezing it, causing the former to move back and forth with respect to it, and adjusts a variable capacitor in the counter until the counter’s wave slows down; ideally it would freeze, too, but the variable cap is too sensitive for Dave to tune it to the exact frequency of the rubidium.

Why does the counter’s wave move at all instead of freezing and just displaying a slightly (even imperceptibly) longer or shorter wavelength than the rubidium’s? Higher frequency, shorter wavelength, lower frequency, longer wavelength. What would you need to do to display two frozen traces of slightly different frequencies?

* Zero-beating is the process of tuning one frequency to another by adding them together, noting the resulting frequency (the “beat” frequency), and tuning the frequency to be adjusted such that the beat disappears.

 

What would you need to do to display two frozen traces of slightly different frequencies?

You would need to be able to trigger the scope separately on each channel and waveform.
Don't know if any oscilloscopes can do that.
But the difference would be so small that it would be imperceptible when the two frequencies were close.

That's why you do the zero beat approach, as it can detect down to near zero difference in frequency.

 

Why does the counter’s wave move at all instead of freezing and just displaying a slightly (even imperceptibly) longer or shorter wavelength than the rubidium’s?

The oscilloscope is locked to the Rubidium standard, hence it shows a continuous change of phase of the other clock, which differs in phase for each sweep of the display, hence a "motion" to the left or right

What would you need to do to display two frozen traces of slightly different frequencies?

A Single sweep stored in memory (instead of continuous sweep) will display the 2 frozen traces.

 

What would you need to do to display two frozen traces of slightly different frequencies?

In an analogue 'scope, where the two channels are generated by displaying each channel on alternate timebase sweeps, generally one of the options for trigger is alternate trigger. This means sweep one displays and is triggered by channel 1 and the next sweep displays and is triggered by channel 2. This way both waveforms will be stationary.

I don't know whether any digital 'scopes have this feature.

 

In an analogue 'scope, where the two channels are generated by displaying each channel on alternate timebase sweeps, generally one of the options for trigger is alternate trigger. This means sweep one displays and is triggered by channel 1 and the next sweep displays and is triggered by channel 2. This way both waveforms will be stationary.

I don't know whether any digital 'scopes have this feature.

Wow, I didn't know any scope had that feature. That's really cool!

 

The oscilloscope is locked to the Rubidium standard, hence it shows a continuous change of phase of the other clock, which differs in phase for each sweep of the display, hence a "motion" to the left or right

A Single sweep stored in memory (instead of continuous sweep) will display the 2 frozen traces.

Aah, that’s it. Thanks.

 

Wow, I didn't know any scope had that feature. That's really cool!

For the old TEK2465 select VERT Trigger Source, and ALT field trigger.

 

The drift is the desired effect! It enables you to accurately compare and adjust a source against a standard. If you have sufficient control you can get your source accurate to about 1 part in 10^8.

 

The oscilloscope is locked to the Rubidium standard, hence it shows a continuous change of phase of the other clock, which differs in phase for each sweep of the display, hence a "motion" to the left or right



A Single sweep stored in memory (instead of continuous sweep) will display the 2 frozen traces.

I prefer the X-Y display. I put my counter 10 MHz output from the rear panel into one axis of the oscilloscope. I put the rubidium standard 10 MHz signal into the other axis. Now I watch the Lissajous drift around and adjust the counter time base to stop the drift.

I discovered an interesting fact. That there is no such thing as zero beat. You can get as close as you like but never zero. I had my two frequencies so close that it took over an hour for the pattern to drift one revolution. How close is that? Well, one rotation per second corresponds to one part in 10 million, or 100 parts per billion. Since there are 3600 seconds in an hour, having the pattern drift around fully in an hour means the two frequencies are apart by about 0.03 parts per billion or around 30 parts per trillion. Three parts in 10 to the eleventh power, if I do my mental arithmetic properly. So if I have a 1 GHz signal I can expect it to be within 0.03 Hz. Close enough for most purposes, I think.

Stating it another way, and I don't want to do the mental arithmetic, it would be maybe a millisecond error in one year. You do the calculations by knowing how many seconds in a year and dividing by three parts to the eleventh power.