hi, any idea about PWM duty extractor?

[william_rx]

i am doing a duty extractor of PWM signal for LED driver. The input PWM duty varies from 1~100% and input frequency varies also , from 200~10Khz. and i need to come up with an architecture to extract the duty of this input, and apply it to a new frequency,say 200~ 20KHz（external tunable). so, any idea of this architecture(expecially the duty extraction) is welcomed!


since i am doing on chip cmos design,so the size of circuit should not be bulky and analog circuit is prefered for their compact zise.

 

[ccurtis]

Interesting problem. I think it is going to be important to know the max rate the PWM varies and the max rate the frequency varies.

 

[Leftyretro]

i am doing a duty extractor of PWM signal for LED driver. The input PWM duty varies from 1~100% and input frequency varies also , from 200~10Khz. and i need to come up with an architecture to extract the duty of this input, and apply it to a new frequency,say 200~ 20KHz（external tunable). so, any idea of this architecture(expecially the duty extraction) is welcomed!


since i am doing on chip cmos design,so the size of circuit should not be bulky and analog circuit is prefered for their compact zise.

Just wire the PWM signal to a low pass filter ( A single series resistor and a capacitor to ground. The filtered DC voltage will represent the duty cycle of the PWM signal. Read this value input a A/D input and the rest will just be software scaling to then apply to your new PWM output signal.

Lefty

 

[william_rx]

Interesting problem. I think it is going to be important to know the max rate the PWM varies and the max rate the frequency varies.

indeed. the pwm input frequency is small, thus its duty varies according to frequency, but since human eye is not so sensitive to backlight change, thus i think 100hz (like monitor) is enough for the change of PWM duty. i just guess, the detail spec is not decided yet.

 

[william_rx]

Just wire the PWM signal to a low pass filter ( A single series resistor and a capacitor to ground. The filtered DC voltage will represent the duty cycle of the PWM signal. Read this value input a A/D input and the rest will just be software scaling to then apply to your new PWM output signal.

Lefty

Thanks for the idea. I did think of it as an option. but deeper thinking makes me frustrate. since the input frequency is small as 200hz, thus the R and C value is too big to implement on a silicon chip. even using switched capacitance filter is not so suitable, because of variation of input frequency varies a lot and a big C using miller effect of OPamp is also challenging, i need R=100K and C=100pF, with opamp Gain= 70 dB for a dominant pole of about 10 Hz, and C=100pf is definetely too large for on chip analog circuit.

My idea is to using analog integrator and analog divider. at each clock, integrate the "on" part and also integrate the whole one period, then divide them. but it has the problem of dynamic range (80db) since input frequency varies a lot. thus i am still thinking..

 

[colin55]

Put the signal into a PIC chip and you can get all the results you want.

 

[Sceadwian]

Why is an analog circuit preferred? A micro controller could detect the PWM duty cycle with extreme accuracy and output pretty much any equivalent frequency up to close to the RF range using a very small chip. I'm not even sure how you'd approach this from an analog basis but it would be significantly more complex, and much larger.

 

[william_rx]

Put the signal into a PIC chip and you can get all the results you want.

well, i have to design all by transistor level..because my project is CMOS analog design which is not so bulky.first i need to get an architecture.

 

[william_rx]

Why is an analog circuit preferred? A micro controller could detect the PWM duty cycle with extreme accuracy and output pretty much any equivalent frequency up to close to the RF range using a very small chip. I'm not even sure how you'd approach this from an analog basis but it would be significantly more complex, and much larger.

i have to do it on a chip design , an analog IC. so ,i cannot use any other chip, it is in transistor level. And first ,i have to come up with an efficient architecture. discrete component idea is also welcomed since they are basic similar circuit concept except the size of R L and C.

 

[colin55]

It can all be done with a single microcontroller chip and no external components.
Don't forget you are asking professional design engineers on this website and are getting a professional answer.
Your requirement is much more complex than you think.

 

[Ubergeek63]

so you need a 16 bit capture register clocked at 10MHz with an interrupt latency + service time of under 1uS to get a 1-99% PWM reading, and another to push out your "converted" PWM