Duty Cycle measurment instruments

[umesh49]

Hi,
I would appreciate if any one can help me to find some instrument to measure and log the duty cycle of a signal.
My requirements are;
1. PWM frequency of upto 50KHZ
2. I want to measure the duty of signal with a accuracy of 0.2% or better
3. I need to have a data logging capacity by instrument to log the data over time.
4. Duty cycle measurement shall be fast enough to capture the duty of every pulse.
5. My signal level is 0-5V.

Regards,
Umesh

 

[dick_freebird]

My TDS3054 oscilloscope, which is far from leading-edge these
days, has this among the waveform measurement functions.
But the 0.2% accuracy, I don't know. And do you know that
your PWM signal is actually 0.2% stable? Because mine are
often not, when you're talking about DC-DCs in real board
environments. Measuring to 0.1%, something that's not 1%
repeatable, is a fool's errand perhaps....

You want to capture every pulse, that's going to hurt. The
'scope can only single-sweep to some record-length and
you trade resolution for span in any instrument. Maybe you
really need edge-time capture and post processing, with a
hell of a record length (how many pulses, is "every"?) and
a timebase & trigger suited to the accuracy goal.

I think you want a lot of things vaguely, and would benefit
from thinking (or describing) more about what you're really
after and why. It's easy to want it all. Not so easy to find
and pay for.

 

[BrunoARG]

Maybe a MCU could -using its timer module- and with some external 1 bit A/D (or not since many cheap uC have comparator modules) measure the frequency and therefore the signal period, calculating when the rising or falling edges happen.

If the period is of 100us and the falling edge is detected at 20us since the rising one happened, then the duty cycle is 20%.

A clear disadvantage of this method is that it works only well at all with square waves, since for example, a triangular wave has no clear rising or falling edges and the 1 bit (higher or lower than) A/D conversion would mess it up.

 

[dick_freebird]

If you can assume edge times are trivial with respect to
high/low times, then a RC filter and a voltage logging tool
would seem to fit the bill (you would however want to
know VOH and VOL as well, if you can't say they are fully
railed, which on bipolar transistor outputs is certainly the
case). You might elect to buffer the switch-node through
a digital CMOS gate or very fast comparator and then
post-low-pass-filter. Your supply tolerance would enter
into the measurement error still, but VOH/VOL could be
assumed more close to zero.

I think your acuracy expectation is unreasonably high for
most applications and this will make a completely acceptable
solution harder to come by, certainly a cheap one. The
'scope based solution, you could just poll the 'scope over
serial-GPIB.