What do I use to measure the ppm of a crystal.

[trevwhite]

Hi all

I often read crystals have a tolerance based on ppm. What equipment would I use to actually measure a crystal tolerance.

Thanks

Trev

[barry14]

The tolerance is not based on ppm but is given in terms of ppm (parts per million).  For example, a 10 MHz crystal with a tolerance of 5 ppm can vary in frequency by 50 Hz (1 ppm of 10 MHz is 10 Hz).  To check if the crystal is in tolerance you will have to insert it in the proper circuit to get it to oscillate and then measure the output with a frequency counter. The frequency counter should have a tolerance at least 3 times better than the crystal (preferably 5 times or more better).  Thus to measure a 5 ppm crystal, the counter should have a tolerance of 1 ppm or better.

[mos6502]

A cheaper way would be to use the crystal to run a microcontroller. Whip up a little clock program, set it using an accurate clock source like a GPS and let it run for a week (or a month, or a year). Then check how much it drifted vs. the GPS.

[trevwhite]

Thanks for the replies. My main requirement at the moment is to check how accurate an rtc is. I will have a seconds pulse that I can measure. I dont really have hours and hours to wait and compare.

I looked at frequency counters but they all seem to be rf based. I just need something that tells me how accurate a seconds pulse is. I would like to be able to measure other frequencies but I do not have fixed limit right now.

Any recommendations for a frequency counter?

[nctnico]

Another factor of a crystal is that it can be detuned based on the amount of capacitive loading and for which kind of operationg the crystal was intended. Depending on the frequency range almost any generic frequency counter will do (even the ancient HP ones like the HP5334) but you'll need a good reference clock source and the frequency counter will need to have an input for an external clock.

A network analyser is probably the best tool because it can show both parallel en series resonant frequencies but you will still need a stable reference clock.
See: http://www.electronics-tutorials.ws/oscillator/crystal.html

This is the result of a crystal connected to network analyser which has crystal/resonator analysis:

[mikeselectricstuff]

Most frequency counters have a period mode, which will measure low frequencies like your 1Hz output to about 0.1ppm
Bear in mind a 'vanilla' frequency counter will have a crystal timebase, likely to be accurate to maybe 10ppm.
A TCXO based one will be down to a few ppm, and an ovened osc maybe below 1ppm.
Or get one of the various cheap rubidium reference oscs and use that as a reference ( any decent counter will have an external reference input)
For a single-task test type application, a cheaper option would me a micro clocked from one of the accurate sources above would do the job pretty easily.

 

[Kalvin]

If you want to go to DIY and microcontroller (MCU), you would need a stable reference clock, say 10MHz, derived either from a TCXO or GPSDO. Typically a MCU has timer/counter unit which can be used as an event timer: The timer will increment by one for each MCU clock cycle ie. at 10MHz. An external signal edge (rising, falling, or both) can trigger an event, and the current timer value will be latched, and the MCU can read the timer value. Use the 1Hz clock signal as the event signal. For example, if the MCU is running at 10MHz the timer will accumulate 10 000 000 pulses in one second. That means that you can obtain 0.1 ppm resolution. Use a faster MCU and use the microcontroller's internal PLL, or longer pulse accumulation time, if you need a higher resolution. Keep in mind that the PLL will add some jitter, though.

See this pdf-document and reciprocal counting / time stamping for the principle: http://www.npl.co.uk/upload/pdf/20060209_t-f_johansson_1.pdf

[trevwhite]

Wow this is great stuff. Thanks. So I see some cheapo ocxo on eBay but I don't trust them really. I need to know I can trust the thing. Any recommendations as a microntroller based measurement is actually quite practical for me.

[nctnico]

When using a microcontroller approach you have to be carefull not to make +1/-1 mistakes IOW make absolutely sure you program the timer correctly.

[zlymex]

A better frequency source(say 0.1ppm) is required when check 1pps(1 pulse per second) source with ppm accuracy.
A good frequency source of 0.1ppm is often 10MHz and comes from a GPSDO(very trustworthy), a rubidium standard or a very good OCXO.
A frequency counter is a comparison tool that may or may not have a good frequency standard in it.
In short, you need a comparison tool(such as a frequency counter or DIY) and a good time base.

[trevwhite]

Another factor of a crystal is that it can be detuned based on the amount of capacitive loading and for which kind of operationg the crystal was intended. Depending on the frequency range almost any generic frequency counter will do (even the ancient HP ones like the HP5334) but you'll need a good reference clock source and the frequency counter will need to have an input for an external clock.

A network analyser is probably the best tool because it can show both parallel en series resonant frequencies but you will still need a stable reference clock.
See: http://www.electronics-tutorials.ws/oscillator/crystal.html

This is the result of a crystal connected to network analyser which has crystal/resonator analysis:

Thanks for this analysis, that is very nice. What instrument did you use to take those measurements?

[nctnico]

I used an Anritsu MS4630B (10Hz to 300MHz) network analyser for this measurement (which was just for taking it out for a spin and see what it can do).

[retrolefty]

I see no discussion on probe capacitance/induction effect on the "accuracy" of measuring any one specific crystal. If the objective is to be within a certain frequency in an existing circuit then it's most likely any frequency counter probe will change the circuit conditions enough to shift the measurement from what it is when not subjected to probing.

 Measuring indirectly in the digital domain using a micro is one method but would only be valid for that micro's crystal oscillator circuit. An old trick is to make one of the 'padding caps' a high Q trimmer cap and just adjust for desired 'exact frequency'.

 Eliminating this 'problem' for a micro-controller application can be to set the micro fuses for external clock signal input and use a TCXO (or better OCXO) external oscillator module and be done with the 'problem'. 

Remember : measurement changes behavior