Crystals and Accuracy

[Screech]

I will be designing an elapsed signal pulse timer.
it will count from 0.00000 to 9.99999 seconds.
so, I will nead a 100,000 hz generator.

The idea is to use a 6.4 meg Hz crystal and cmos 4060 to divide by Q6(64) to give me the 100,000 Hz.
then counting with 6 4026 seven segment counter/led driver ic's.



I bought a 6.4000 Meg Hz. Crystal.
Connected it to a cmos 4060 chip, and It seems to be working fine.

Instead of the 10pF and the 39pF Trimer cap, I use two 22pF caps as thats what I had handy.

It seemed to run fine.
the out put of the chip was devided by Q(cant remember) to give an audio frequency that mached the tone from a downloaded pc tone generator.
Test cunducted by ear.

I also noticed that if I removed the caps, the crystal still seemed to function the same.
Is it better to not use the caps?

Question:
is the 39pF trimmer cap used to increase the /decreace the frequency?

I don't want to increase/decrease the frequency.
I want it to run at exactly 6.4MHz.

How do I know that the frequency is exactly 6.4MhZ and not 6.4035Mhz?

I don't have a mutimeter with Frequency function, and I think all multimeters have an error percentage anyway.


I know nothing about this field.
But do you think the crystal is outputing the exact 6.4000 MHz?
I guess it should , since it has 6.4000 printed on the crystal.
thanks

Here's a part print screen of a 4060 data sheet

 

[Someone Electro]

The capacitos are for the stability of the crystal.

Multimeters cant meshure that high freq. they mostly meshure up to about 100 kHz.Freqency counters do it or you can do that whith an osciloscope.

 

[Screech]

Oh, Ok electro,

But do you think the crystal is outputing the exact 6.4000 MHz?
I guess it should , since it has 6.4000 printed on the crystal.
thanks

 

[Nigel Goodwin]

I also noticed that if I removed the caps, the crystal still seemed to function the same.
Is it better to not use the caps?

The capacitors are an essential part of the oscillator circuit, it 'may' work without them due to stray capacitance, but it's important to leave them in place!.

Question:
is the 39pF trimmer cap used to increase the /decreace the frequency?

Yes, to trim the crystal to exactly the frequency you want.

I don't want to increase/decrease the frequency.
I want it to run at exactly 6.4MHz.

How do I know that the frequency is exactly 6.4MhZ and not 6.4035Mhz?

You need to accurately measure it, a standard technique is to compare it to a reference frequency, most countries have some reference radio transmitters, very accurately set to a specific frequency. You can 'beat' the harmonics against each other, and adjust for zero beat (rather like tuning a guitar! - in fact, thinking on, EXACTLY like tuning a guitar!).

Using a tone generated from a PC is likely to less accurate than the unadjusted crystal :lol:

I know nothing about this field.
But do you think the crystal is outputing the exact 6.4000 MHz?
I guess it should , since it has 6.4000 printed on the crystal.
thanks

Unlikely!, you need to adjust the trimmer so it's exactly correct - but do you feel lucky?.

 

[Screech]

Nigel says

you need to adjust the trimmer so it's exactly correct - but do you feel lucky?.

Yeah, I'll trim it and test it for a long time, to see if if goes out.

should I use the same capacitors as in the data sheet, even thou I'm useing a much higher crystal?

Thanks guys.

 

[eblc1388]

Even if you tune it to exactly 6.400MHz at a particular temperature, the frequency will change with different temperature.

You needs some means to keep the crystal oscillator at the same temperature if you want stable frequency output.

 

[audioguru]

Why not use the capacitor values (minus stray capacitance) that are spec'd by the crystal manufacturer?

 

[Nigel Goodwin]

Even if you tune it to exactly 6.400MHz at a particular temperature, the frequency will change with different temperature.

You needs some means to keep the crystal oscillator at the same temperature if you want stable frequency output.

Like everything, it depends on the accuracy you need - at one time "crystal ovens" were quite popular, these keep the crystal at a constant(ish) temperature to reduce frequency changes. They were generally quite crude devices, usually a big resistor switched by a mechanical thermostat.

However, reducing the crystals temperature range to 50-55 degrees C is much better than a range of -10 to +50 degrees C.

But it all depends what you need?.

 

[pebe]

Even if you tune it to exactly 6.400MHz at a particular temperature, the frequency will change with different temperature.

You needs some means to keep the crystal oscillator at the same temperature if you want stable frequency output.

You can purchase crystals with a very low temperature coefficient. Probably a bigger problem at that frequency is trying to tune them accurately with the capacitors.

A possible way to ensure accuracy might be to take the one-second output from a digital watch and use it to synchronise your timer somehow.

 

[eblc1388]

He would be better off using a TCXO crystal oscillator module which is quite affortable and usually fulfil the frequency stability requirement of hobby purpose.

But I don't know whether they have TCXO of 6.4000MHz but they definitely have 1.0000MHz or 10.0000MHz.

Edit: you beat me, pebe.

 

[lord loh.]

I once made an Xtal oscillator and was viewing the waveform on the CRO. The waveform actually registered a noticable change when I touched the crystal package. My guess was it was body heat. My friend thought it was mechanical pressure. I think he is right. The point I am trying to make is that it might not be very accurate.

 

[chesapeake]

valtrex - **broken link removed**

[EDITED by mechie - adverts for pills are NOT WELCOME :evil: ]

 

[Dean Huster]

A TCXO is really not enough more stable than an uncompensated crystal oscillator to justify the significantly higher price. As I recall, from an article written by a Hewlett-Packard engineer on the various types of frequency standards, a TXCO is only about 4 times more stable than an uncompensated xtal oscillator. An ovenized oscillator is much better, even more so if it uses a sophisticated proportional temperature control to eliminate the sharp on/off cycles of a mechanical thermostat.

I would guess that for your application, a trimmed but uncompensated crystal oscillator will be sufficient.

You can take the 100KHz output, connect it to a wire and turn a shortwave receiver to WWV at 5, 10, 15 or 20MHz, bringing the oscillator close enough to the receiver to set up a beat note. Then adjust the oscillator to zero beat the tone.

Dean

 

[panic mode]

this might be helpfull

**broken link removed**

**broken link removed**

 

[mechie]

Crystal Accuracy

Just how accurate do you want ?
A typical crystal will give something like 100ppm stability ?
That's 0.0001 % ?

 

[Screech]

Re: Crystal Accuracy

Just how accurate do you want ?
A typical crystal will give something like 100ppm stability ?
That's 0.0001 % ?

Mechie, It's running 0.02% too fast.
I'm getting an error of 18 seconds per 24 hours and thats using that 39pf trimmer on that data sheet.

Looks like I nead to use more pfs then the 39pf specified.
Is it ok to us more?

I also did not use a voltage regulator.Does it matter with crystals?
i used 4.5Volts (3* 1.5 volts AA batteries)

What does 100ppm mean?
Thanks

 

[Nigel Goodwin]

Re: Crystal Accuracy

Just how accurate do you want ?
A typical crystal will give something like 100ppm stability ?
That's 0.0001 % ?

Mechie, It's running 0.02% too fast.
I'm getting an error of 18 seconds per 24 hours and thats using that 39pf trimmer on that data sheet.

Have you zero beat the oscillator against a reference source?, or how have you set it up?.

Looks like I nead to use more pfs then the 39pf specified.
Is it ok to us more?

I also did not use a voltage regulator.Does it matter with crystals?
i used 4.5Volts (3* 1.5 volts AA batteries)

I would expect the frequency to change as the batteries go flat, digital watches do!.

What does 100ppm mean?

ppm is "parts per million".

 

[Screech]

Nigel asks

Have you zero beat the oscillator against a reference source?, or how have you set it up?.

I bought some more caps to tune it (27 pf on both sides of the crystal).

I hooked up the seconds pulse of it to the parallel port.
Wrote a program in QBasic to tell me the time when the port receives a low/high input signal.
Qbasic's timer command has a resolution of 0.01 second.
I'm now getting an error or 0.01 second every 3000 seconds.
or an error of 0.28 seconds / 24 hours when compared to my pc clock. I think my pc clock is accurate to within a second per day when updated daily from the internet clocks

 

[Nigel Goodwin]

Nigel asks

Have you zero beat the oscillator against a reference source?, or how have you set it up?.

I bought some more caps to tune it (27 pf on both sides of the crystal).

The crystal is tuned by altering the DIFFERENCE on the two sides of the crystal, so by adding to BOTH sides you're not really changing anything 8)

I hooked up the seconds pulse of it to the parallel port.
Wrote a program in QBasic to tell me the time when the port receives a low/high input signal.
Qbasic's timer command has a resolution of 0.01 second.
I'm now getting an error or 0.01 second every 3000 seconds.
or an error of 0.28 seconds / 24 hours when compared to my pc clock. I think my pc clock is accurate to within a second per day when updated daily from the internet clocks

Sorry, but that's a REALLY poor way of trying to adjust it, the accuracy of the QBASIC program is going to be pretty poor, nowhere near as accurate as your untrimmed crystal!.

As your resolution is 0.01 seconds, by definition the accuracy can be no better than + or - 0.01 seconds, and the way the QBASIC timer command works will make it worse than that!.

 

[Screech]

Nigel says

The crystal is tuned by altering the DIFFERENCE on the two sides of the crystal, so by adding to BOTH sides you're not really changing anything

Thanks for that info.


Nigel says

As your resolution is 0.01 seconds, by definition the accuracy can be no better than + or - 0.01 seconds, and the way the QBASIC timer command works will make it worse than that!.

I ran it continuously for a few hours and got an average of 0.01 seconds error per 3000 seconds, but i see what you are getting at.

Next time I will not run Qbasic.
will just let it run over night (12 hours) and see visually how many seconds its out.

I also have a seven segment thing showing seconds.

eg.
when it's 10:00:00 pm on my computers clock, I'll turn on the crystals power .
After 12 hours, I'll see if the seconds are still syncronizing with the pc clocks second hand.

Oh, I'll use a voltage reg too (i forgot to use it today).

Thanks