晶体振荡器启振

wwm101

to:lianggu
thank you very much !

i'm sure it is the 100khz crystal .because i get it from a pspice lib.
  
r1  r2  m1 m2  consist as  a inverter amplify!


to:"Why is the power stimulus applied through PSUB? Strange!www"

i can't see . i don't connect  the psub to vdd!
i connect "b" pole of pmos to vdd to avoid the substrate effects.

i will make a trial use spetre tomorrow!
can you tell me how to confirm the min or min feedback res in theory?

lianggu

You are confusing me.

Based on your netlist, R1 is connected between PMOS's source and well, R2 is connected between NMOS's source and sub. Am I right?

Actually, there is no formula for exact Rmax and Rmin in any text book. But if you have this idea in your mind, you may get the Rmax and Rmin from simulation. As shown in my example, Rf must be higher than 50M. (You may fine tune Rf to get accurate Rmin value.) As what I said before, since input impedence of CMOS inverting amplifier is ideally infinity, Rmax might be quite large. Thus, you don't need to care about it. Good Luck!

wwm101

"R1 is connected between PMOS's source and well, R2 is connected between NMOS's source and sub. Am I right?"
yes ,but the pmos's well  connect to the Vdd,and nmos's sub connect the Gnd.

because rain ,today i don't go to lab to have a trial.
but i have a trial in hspice, i find when i cancel the initial value of the inductance of crystal or reduce initial value to a quite small value,the crystal  can't start up!
that's why,can you tell me ?
thank you very much!

lianggu

Hi, Buddy, you are pretty aggressive. I like it.

What is the purpose of those two 1MOhm resistors? Even if they are used for ESD, they are still not right.

For any kind of oscillator, it has two steady states, one is ZERO state, the other is OSCILLATION state. If no initialization when oscillator startup, it will be always in ZERO state. Thus, you need to give an initial condition to oscillator to push it leave for oscillation.

I did this kind of oscillator about 11 years ago. Although it is simple, but you can get some sense on analog design. More projects you did, More sense you would have. Sense is very important in analog/RF/microwave design.

wwm101

原帖由 lianggu 于 2007-6-17 11:40 发表
Hi, Buddy, you are pretty aggressive. I like it.

What is the purpose of those two 1MOhm resistors? Even if they are used for ESD, they are still not right.

For any kind of oscillator, it  ...

thank you for your help

one purpose for add 1mohm resistors,because conveniently  count the gm of inv amplify by hand.
because i can't get the Rds of pmos and nmos, and when the frequently of crystal oscillator is small there is a mix & min  value of gm.
( attachments is the paper introduce this theory)

[ 本帖最后由 wwm101 于 2007-6-17 12:41 编辑 ]

sk1105

原帖由 wwm101 于 2007-6-16 13:12 发表
to:lianggu
thank you very much !

i'm sure it is the 100khz crystal .because i get it from a pspice lib.
  
r1  r2  m1 m2  consist as  a inverter amplify!


to:"Why is the power  ...

one important thing is that you never use crystal model in your spice lib when you simulate crystal oscillaor.
you should have your own crystal model according to your crystal spec, such as Q? Rs? Cs? Cp? from your crystal datasheet.
another thing is that what is your crystal osc's spec, power? size? or just working properly?
it affects your circuit design a lot, because there are so many structures and which is yours.
when you get right crystal model and get your spec consideration done, which means you're ready to begin the circuit design.
FYI.

wwm101

to:sk1105
first thank you for your attention!
i use the Equivalent crystal circuit and the parameter form the xtal lib of pspice 9.02.
***********
* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
* 100khz frequency standard, DT cut, parallel resonant, Q=20000
*   calibration capacitance = 5pf
.subckt QZP100K 1 2
*
  lqz 1 11 lmod 588.873289
  .model lmod ind(tc2=3.5e-8)
  cs 11 12 4.30423941e-015
  rqz 12 2 18.5k
  cp 1 2 1.72169576e-012
.ends
******************

"another thing is that what is your crystal osc's spec, power? size? or just working properly?中国电子顶级开发网/o$j*yg Q5X'B(Ra
it affects your circuit design a lot, because there are so many structures and which is yours."

can you give me more information about how the crystal size/ power affect the design? because the crystal model in pspice and the equivalent crystal circuit don't include the crystal osc'c power &size.

wwm101

there is other problem:
we know in practical the oscillator start up by noise (main from source).So I simulate the noise by a vpwl source,but it can't start up .that is why?
have somebody know it?
for example:
* osc 100K
.PROT
.inc 'd:\hs\level3'
.UNPROT
M_M1         VOUT IN 0 0 n
+ L=16u  
+ W=0.8u         
+ M=10
M_M2         VOUT IN vdd vdd p
+ L=16u  
+ W=0.8u         
+ M=22
V_V1         vdd 0 3v
C_C1         IN 0  5p  
C_C2         0 VOUT  5p  

***crystal**
C_C4         IN VOUT  1.72169576e-012  
R_R3         IN N11450  18.5k
L_L1         N11450 N03952  588.873289 IC=1uA
C_C3         N03952 VOUT  4.30423941e-015  
****
R_Rf         IN VOUT  50meg  
.tran 1ns 20ms
.probe tran  v(in) v(vout)
.end
when I set the IC=1uA, the swing equal  to 3V(=Vdd). but when i set it to 0.1uA ,the swing equal to only 0.06V. finally i set it to 0.01uA ,the swing equal to 0.06v also. Later i cancel the intial value of inductance in stead of a vpwl source" vdd 0 pwl(0 0,10us 3Vdc,20u 0,30us 3vdc)",the swing equal to zero!

thanks!

[ 本帖最后由 wwm101 于 2007-6-17 19:09 编辑 ]

lianggu

I don't think the main noise comes from power supply, especially when battery is used which is pretty clean. The main noise should be thermal/shot noise and flicker noise.

Both ramp-up power supply and initial value of crystal are related to CAD algrithom. As what I know (maybe not right), Ramp-up power supply is used to generate proper operating point if necessary while initial value of crystal is used to set up oscillation. With too small initial value, the CAD algrithom can not get oscillation since it has an error to determine if iteration will keep going. (You may take some IC-CAD class to get more about it.)

60mV is just ripple, not real oscillation.

wwm101

gree with you "60mV is just ripple"
when i add vpwl source with initial value equal to 0.1u,it represent evidence.

thank you for your help very much!