Pulse width to voltage conversion. How?

[brett]

Hi.
I have a bunch of pulses of about 5 us width (at 10V) and 20us interval at 0 V.  ie about 40,000 pulses per second.
How do I convert the number and width of the pulses to a voltage?  (ie integrate or filter?)

Using LTspice, I've looked at non-inverting integrators, but can't seem to get anything to work (e.g. the LT1001 doesn't work at all.  The LT1223 is better, but bias currents etc cause havoc.) 
Maybe I should just build real ones?  Unfortunately, there seem to few applications of non-inverting integrators on the web.

Or maybe I want a filter instead of an integrator?
thanks

[earthtonesaudio]

Personally, I'd try a passive RC lowpass filter before anything else.

[R.G.]

Lowpass filter. The simplest one, as ET said, is a simple RC lowpass. If the lowpass response is very far from the base frequency, you get good results.

[Cliff Schecht]

With an RC filter, you have a trade-off between response time (large cap, very low frequency) and ripple. Play with your filter values until you find a good trade-off.

[brett]

Hi
Cliff's point is correct, if I use too large a cap, I get a little less than 10V with some ripple (up to 10V).  I suppose if I LTspice it, I can find a point where the output is 1 or 2V when the mark-space ratio is at its lowest, and see what response I get as the mark-space ratio increases.  I'll let you know how it goes.
thanks 

[Cliff Schecht]

Hi
Cliff's point is correct, if I use too large a cap, I get a little less than 10V with some ripple (up to 10V).  I suppose if I LTspice it, I can find a point where the output is 1 or 2V when the mark-space ratio is at its lowest, and see what response I get as the mark-space ratio increases.  I'll let you know how it goes.
thanks 

What you really want is a second order Butterworth or Chebyshev filter. You can easily achieve this with an op amp and a handful of caps and resistors. This will allow you to track changes quicker while still eliminating all ripple.