two questions need your help

I have two questions here.
in on-chip OSC circuit, R (internal resistor for reference current generation, Iref=Vref/R) and C will all vary with process deviation, and the deviation for R or C can reach +/-30% or larger, so the OSC frequency can change a lot relative to the typical value when process deviation occurs. in order to calibrate the process deviation due to R and C, frequency calibration is always used, such as using programmable current source to charge or discharge the capacitor. now ,i want to know:

1, how to calibrate the process deviation simultaneously caused by R and C? and will the programmable current source be ok? or any other methods should be used? i need your help, some papers or advice will be appreciated.

2, there is a calibration formula used to calibrate the clock frequency when using the chip, that is, we collcet many samples of tested clock frequency (to an excel table), and obtain a formula to get a register value, send this value to the chip to change the programmable charging current, so the frequency can be calibrated to the desired value. now, i want to know, how to get the calibration formula? what is the procedure of getting the formula? pls help me, thanks .

thanks all.

The answers to you question will depend upon the chip you are calibrating. What is the chip part number?

If you use a programmable current source which loads (and perhaps unloads) a capacitor, then the osc. frequency is proportional to the current. That means by selecting the correct current, you can calibrate your osc. to the required frequency. Such calibration can compensate for all process variations (slow, fast) which affect Vth, R & C values, and -- via Vth -- comparator switching levels.

If f_org is the original frequency at the default current I_org (prior to calibration), and f_req is the required frequency, then the new current I_req to be programmed/trimmed is I_req = I_org * (f_req / f_org) .

Remaining VT variations depend on the respective "quality" (i.e. VT-independence) of the programmed/trimmed current.

but as i know, the calibration formula is a complicated formula, such as a*x^2+b*x+c=y (register value), and assuming the register is 6-bit controlled (the current is 32u,16u,8u,4u,2u,1u separately), a,b,c is the calibration coeficient, x is the measured frequency (such as 1.13Mhz for a desired 1MHz value), and y is used to calibrate the frequency. what is the procedure to get the calibration formula 'a*x^2+b*x+c=y' ? pls explain it more clearly, thanks all.

I guess you mix up a 2^nd order tempco algorithm with a binary trimming scheme?

yes, but how to get the calibration formula 'a*x^2+b*x+c=y' ?

lhlbluesky said:

how to get the calibration formula 'a*x^2+b*x+c=y' ?

Click to expand...

What for?
How should this parabolic equation help you in any way to calibrate/trim the necessary current for the required oscillator frequency with a binary trimming scheme?