1-20MHz crystal oscillator

[eekenneth]

20 mhz crystal oscillator

hi All

i am going to design a low power (<10uA) 1-20Mhz crystal oscillator
I want to know some starting points how to design and what kind
of things I have to take care?

Would you please tell me more on circuit level and how to simulate?
it's good if there's reference design and simulation example

many thanks

 

[papyaki]

crystal oscillator ac analysis

Hi

You will find some information here :

h**p://users.telenet.be/educypedia/electronics/analogosciltypes.htm

* = t

 

[E-design]

why used 20mhz crystal oscillator

If you can find a copy of the book "CRYSTAL OSCILLATOR CIRCUITS"
by Robert Matthys, ISBN 0-89464-552-8 (The book might be out of print)

This gives all the details on design with theory, practical examples and verification.

(Thought I saw it once uploaded here... long time ago)

E

 

[eekenneth]

oscillator 20mhz

Do u know if anyone can upload this book again ?

 

[Najfi]

20mhz crystal oscillator

there is very good application notes by Atmel and motorola on this topic
here some more references
https://foxonline.com/oscbooks.htm

 

[buta]

crystal oscillator circuits, matthys, upload

Is the osc frequency adjustable from 1 to 20MHz or fixed?

 

[sunjimmy]

davidwong crystal

Hi,

Is it possible to design a 1-20MHz oscillator with such low current (<10uA) ?
Assume the pad capacitance is 10pF, 0.1V voltage swing @20MHz, it will take about 20uA average current.
I don't know much about this topic, maybe someone can share his experience with us : )

 

[eekenneth]

20mhz crystal current

The design should suit for 1-20MHz crystals and it should be adjustable whenever user
plugs in different crystals into the pads.
Anyone has the experience in doing crystal oscillator design and tell me the procedures

 

[buta]

20mhz crystal size

This book "Design of Crystal and Other Harmonic Oscillators" by Benjamin Parzen (1983, ISBN 0-471-08819-6) introduces procedures to design a transistor based crystal oscillator.

I FEEL there are two problems:
1. 10uA
2. variations of crystal's parameters.

I FEEL (but not sure) you can try to build a osc with CMOS invertors and replace those serial / parallel capacitors with variable one. CMOS gates provide high open loop gain and it MAY overcome the variation of crystal's parameters.

 

[ysz]

20mhz crystal oscillator circuit

If you can find a copy of the book "CRYSTAL OSCILLATOR CIRCUITS"
by Robert Matthys, ISBN 0-89464-552-8 (The book might be out of print)

This gives all the details on design with theory, practical examples and verification.

(Thought I saw it once uploaded here... long time ago)

E

do you have the book yet? could you upload it again? tk u

 

[Fom]

20 mhz crystal oscillator datasheet

8 years ago I designed a 32 kHz crystal oscillator for 1.2u CMOS process. My task was to reduce current consumption as much as possible. I've got 1.5uA after optimization not only oscillator invertor but second and third invertor that used for increasing driving capability of oscillator. The main problem the signal on input of invertor is pure sine wave, so feedtrough current is significant. If you try reduce this feedtrough current by transistor size the gain will be not enough for oscillation.

 

[ysz]

ac analysis in crystal oscillator

8 years ago I designed a 32 kHz crystal oscillator for 1.2u CMOS process. My task was to reduce current consumption as much as possible. I've got 1.5uA after optimization not only oscillator invertor but second and third invertor that used for increasing driving capability of oscillator. The main problem the signal on input of invertor is pure sine wave, so feedtrough current is significant. If you try reduce this feedtrough current by transistor size the gain will be not enough for oscillation.

the problem is how to sim the oscillator circuit, with hspice, the circuit is difficult to work.

 

[Fom]

20mhz crystal accuracy

I don't know about how Hspice will simulate that. I used Spectre. First you should run AC analysis to select proper size of an inverter. Use a linear frequency scale with very narrow range close to the expected oscillation frequency.
Then you verify oscillator by transient analysis with "conservative" option. But be patient, you have to wait quite long time until you get visible oscillation. Use Saving checkpoint option to stop and restart from simulation from checkpoint. The total start up time of oscillator (and simulation time as well) should be about Q*T Q - quartz quality factor, T - oscillation period.
Good luck.

 

[ysz]

20mhz crystal circuits

I don't know about how Hspice will simulate that. I used Spectre. First you should run AC analysis to select proper size of an inverter. Use a linear frequency scale with very narrow range close to the expected oscillation frequency.
Then you verify oscillator by transient analysis with "conservative" option. But be patient, you have to wait quite long time until you get visible oscillation. Use Saving checkpoint option to stop and restart from simulation from checkpoint. The total start up time of oscillator (and simulation time as well) should be about Q*T Q - quartz quality factor, T - oscillation period.
Good luck.

Fom, thank you, I got the correct result by add the accuracy.

 

[rambus_ddr]

20mhz crystal

This book about crystal osc almost about bipolar technology, where have CMOS ?

This book "Design of Crystal and Other Harmonic Oscillators" by Benjamin Parzen (1983, ISBN 0-471-08819-6) introduces procedures to design a transistor based crystal oscillator.

I FEEL there are two problems:
1. 10uA
2. variations of crystal's parameters.

I FEEL (but not sure) you can try to build a osc with CMOS invertors and replace those serial / parallel capacitors with variable one. CMOS gates provide high open loop gain and it MAY overcome the variation of crystal's parameters.

 

[France]

cd4060 - variable oscillator

I don't know about how Hspice will simulate that. I used Spectre. First you should run AC analysis to select proper size of an inverter. Use a linear frequency scale with very narrow range close to the expected oscillation frequency.
Then you verify oscillator by transient analysis with "conservative" option. But be patient, you have to wait quite long time until you get visible oscillation. Use Saving checkpoint option to stop and restart from simulation from checkpoint. The total start up time of oscillator (and simulation time as well) should be about Q*T Q - quartz quality factor, T - oscillation period.
Good luck.

Hi Fom
I kenw first to run AC analysis. It was decision the gain(db).
So, How does build the config by transient analysis.

 

[Fom]

crystal osc +voltage swing

Your AC analysis should include all parts of oscillator (inverter, quartz, resistor and capacitors). After AC, when you find proper inverter size (or gain), before run transient analysis, you should set conservative simulation option (basically limit integration step 10 times less than oscillation period and reduce voltage, current tolerance). Then switch ON VDD power supply (VPULSE) and wait. When you get visible oscillation you can stop simulation and save this state as starting point for next simulation. In Spectre it is called Save Checkpoint.
After that you can release I-V tolerance to speed up you simulation, change power supply from VPULSE to VDC and proceed simulation from saved Checkpoint.
It was my simulation procedure.

 

[France]

quartz oscillator 20mhz

Fom, thank you.
I kenw the transient analysis.
you said "set conservative simulation option" to indicate convergence option, is right ?
About frequency of the oscillator by transient analysis.
If my frequency is 27MHz.
It's period is 37ns or difference value after transient. analysis.

 

[nj_jack]

matthys crystal oscillator circuits

I want to know whether osc frequency is variable or not. if it is fixed, everything is easy to do. if not, you can adopt VCO, which is a good choice.

 

[buffollo]

this link also is usefull...
h**p://www.eetasia.com/ARTICLES/2001APR/2001APR23_AMD_AN1.PDF#search='oscillator%20pad%20crystal'