How do I filter a square wave to get an analogue voltage?

There is no linear way ot get a perfect average voltage from the waveform, but passing it through a low pass filter will reduce the strength of the higher frequency components of the square wave, without much changing the DC (average) component. Since the lowest frequency AC component is the fundamental frequency and low pass filters generally attenuate higher frequencies more than lower frequencies, the fundamental frequency is the hardest to acceptable eliminate with a simple low pass filter. But if you add a notch filter to the low pass filter, at the fundamental frequency, you can more easily get an acceptable average output. The combination of notch and low pass is often referred to as a Cauer or eliptical filter

In general, the higher the pulse frequency, the easier it is to filter out the AC in a given amount of delay time to produce an acceptable average voltage. That is, the delay between some change in the pulse duty cycle (percent of on time) and an acceptably stable average output voltage goes down as the pulse frequency rises, for a given filter complexity.

There are some links for low pass and notch filter designs.

http://en.wikipedia.org/wiki/Low-pass_filter

http://www.science.unitn.it/~bassi/Signal/TInotes/...

http://en.wikipedia.org/wiki/Elliptic_filter

All those filters are linear devices, because they have a proportional effect on any signal. But if you have the capability to do digital stuff (have a microprocessor of discrete logic) you can do lots of other solutions. For instance, you can combine a very simple low pass filter (that will have a lot of ripple in its average) with a sample and hold that is synchronized with the pulse, you can grab a sample of the average at the same place in each ripple cycle, so that the sample and hold output is much more representative of the true average with very little ripple, and also with minimal delay. This average will step to the new average when the duty cycle of the pulse changes.

http://www.analog.com/static/imported-files/tutori...

--

Regards,

John Popelish

I go along with most of the answers already given. However, voltmeters are supposed to read the RMS voltage of the ac waveform. The RMS is that value of dc voltage, which would give the same heating effect as an ac waveform. Thus, the only TRUE RMS measure is a meter involving a thermocouple. However, there is a relationship if sine waves are involved, which links the peak value to the RMS value. In your average voltmeter, a bridge rectifier inverts the negative half cycles so that a series of positive half cycles are produced. The average of these is NOT zero of course, and so gives an indication on the meter. On the assumption that sine waves are involved, the meter can then be calibrated to show the RMS value, even though the average of the rectified wave is what is actually measured.

Convert Square Wave To Dc

Fourier theorem sates that any periodic signal can be expressed as a summation of sinusoidal signals with different frequencies, so, if you put a band pass filter and center it at the frequency you want, you'll get a sinusoidal signal, of course it won't be pure, as your filter is not ideal.