oscillator IC desgin with CMOS process, internal Cap

[li-jerry]

Normally, if we use constant current to charge and discharge internal MOS Cap, the frequency range is quite wide because process and temperature. My question is if we need freqency in a narrow range, what shall I do? Thanks beforehand, my dear friend!

 

[andy2000a]

what kind frequency about your OSC ??
what is your narrow define ??

KHz -> use mos Cap , but becare mos_cap volt vs
Cap value

MHz -> ring cell

GHz -> LC

by the way , if you need accuracy timing /clock ,

1. use Trim digital counter , some battery chips
use osc send to counter , when wafer out use
CP(wafer test )measure osc frequency and use trim modify counter value get accuracy clock

2. use accuracy bandgap provide accuracy current
charge mos_cap , but MOS_cap have 2~10% error
by process , so you need Trim mos_cap arracy
(but if your mos_cap too small much more fringle
cap > Cgs )

 

[mikechiang]

you can using external resistor, then T=RC will be more accuracy

 

[Fom]

Use bangap for comparator to make it voltage independent.
Use another uncompensated bandgap with positive or negative temperature coefficient to cancell resistors (and others component) temperature coefficients to make it temperature independent. You can use the same bandgap with different taps for both these tasks.
And use adjustment during CP testing to make it process independent.

 

[wangjy]

What Form said is really true. I had found such kind of designs before. But temperature CP need AD converter to measure the temperature and then compensate according to measure temperature.

In some low cost microcontroller, the internal RCringosc freq is 8MHz(I dont know if it used pll)
and the precision is 1%. And there There is no AD
(For such low cost product, AD is impossible).

So can you imagine what kind of technology can be used to rech this precision.

 

[mss]

Dear all. In my process (0.6u 25HV option) I use nmos Cap. By simulation (Cadance) I have 5% accuracy. In SPICE distortion of Tox by corners arround 4%. In reality on wafer will I have this accuracy or not?
Sambody have experience in it?
P.S. I use HV NMOS CAP (Tox=690 +/-30A)

TNX beforehand.

 

[butterfish]

what kind frequency about your OSC ??
what is your narrow define ??

KHz -> use mos Cap , but becare mos_cap volt vs
Cap value

MHz -> ring cell

GHz -> LC

by the way , if you need accuracy timing /clock ,

1. use Trim digital counter , some battery chips
use osc send to counter , when wafer out use
CP(wafer test )measure osc frequency and use trim modify counter value get accuracy clock

2. use accuracy bandgap provide accuracy current
charge mos_cap , but MOS_cap have 2~10% error
by process , so you need Trim mos_cap arracy
(but if your mos_cap too small much more fringle
cap > Cgs )

agree with andy2000a!
if your frequency range<500 ppm, crystal oscillator is needed.