Continue to Site

# how can i design a fix frequency oscillator

Status
Not open for further replies.

#### lylnk

##### Member level 1
i will design a full integrated oscillator. I want it output 60MHz square wave.

how can i get a accurate output?

for a RC osc, the poor +-20% resistor makes osc output frequency less accurate.

Is it possible to design a osc with untrimmed +-5% accuracy?

As far as I know, with current technologies, it is impossible to get an untrimmed accuracy of +-5%. Usually people use an offchip component to trim the oscillation frequency, or a crystal oscillator with a PLL.

Since the resistors and capacitors have a typical +/- 20% variation, the oscillator must be capable of adjusting its value in some kind of a feed-back loop.

What you could hypothesize is some kind of circuit that translates the frequency domain into a DC value of some sort that can be compared to a more accurate reference.

To my best knowledge, the best bandgap I every saw was +/- 6% over process, temperature, mismatch and 5-sigma, untrimmed. The circuit that will translate the frequency to the accurate reference will add additional mismatch.

Therefore, since there is no +/- 5% untrimmed reference to begin with and we have to leave some error for the theoretical "translation" circuit, we can conclude that it is impossible to generate a +/- 5% oscillator untrimmed on a typical IC process.

In practice, even if we could generate some kind of an accurate reference (i.e. external voltage), there is no accurate way of changing a frequency into a DC value. If you create a rectifier type circuit that has a filter at the front, you will get a DC value that is a function of the frequency at the point where the filter is rolling off. However, the filter varies all over with the C and R's, and the diode's Vt varies with temperature and process. I have never come across an untrimmed circuit that can accurately translate a frequency to a DC value.

Greg

hi
i think as my friends above said you should use some sort of feedback such as pll ,because they are made for this purpose ..... tuning and synch.

bset regards

try the digital program tuning

I use a nmos capacitor it's has accuray of +/- 10% in spec.
A compensation circuit emplies a nmos can cancel some corner variations of mos capacitor.

however, a poor resistor limits the frequency accuracy.

The spec requires +/- 10% untrimmed accuracy. I think it's possible over some wafer because the variation always smaller than process electronic design rule. But i can't make a guarantee over all wafer.

i will contact the author of circuit spec.

thanks for your help.

can i make a mixed resistor consist of several different type resistors to improve the accuracy of resistor?

The variation of different type resistor is unrelated. The tutal stdev can be smaller.

Are the variation of npolysab and ppolysab unrelated?
Are the variation of ndiffsab and pdiffsab unrelated?
Are the variation of polysab and diffsab unrelated?

I has little known to the foundry.

Added after 2 minutes:

Is the capacitor of metal finger accurate?

The metal finger capacitor use the side capacitor of two metal. In .18um, the space is small.

How accurate can the capacitor be?

Status
Not open for further replies.