Hey guys, I got another filter for you.

This is a linear-phase DC offset filter that can be used as an alternative to [hip~] when needed. It has a much faster response than [hip~], but at the expense of some delay. The amount of delay in samples can be provided as an argument. Smaller delays result in a faster response, but frequencies in the audible range may become attenuated. Larger delays improve the frequency response but slow the transient response.

So, the main advantages of this are the response time and linear-phase response. If you're dealing with a signal that has some wild DC fluctuations, [hip~] might not be fast enough. Increasing the cutoff for [hip~] will improve its response time, but can also introduce phase issues, so if phase is also important than [dcblock.mmb~] might be a better choice.

http://www.pdpatchrepo.info/hurleur/dcblock.mmb~.pd

Cool. I have a question. Do most sound cards have dc blocking on the input? Seems like in the past I had used a soundblaster for analog input like a multimeter but don't remember if I had to modify the soundcard. I'm asking because it would be cool to build something like a capacitive touchplate or a pressure sensitive resistor as an input device with all the logic controlled by pd. If most sound cards do have a dc blocking a pwm or pfm might work but this would introduce latency due to decoding. I guess it wouldn't be too hard to do the logic outside the computer but I thought it would be cool to make an input device with minimal components.

I don't really know. There was this guy I knew a few years ago who was trying to generate analog control voltages from his computer and discovered that many, but not all, audio interfaces use a DC offset filter at the output. And it's pretty easy to get rid of. If I recall, it's basically just a resistor that needed to be removed. He hacked an Mbox Pro to control a filter he built. It was pretty interesting, actually, because he found that most MOTU devices didn't have this built in, and as a result Max/MSP users were using MOTU devices to send CV signals. And then MOTU released Volta as a CV generator, despite the fact that you didn't need it if you already had a MOTU device.

But as far as the input goes, I have no idea. I'm not really an electrical engineering guy. I like 1's and 0's.

That's pretty cool about the MOTU CV out.

with a bit of soldering, you can get CV out from any DC offset filtered soundcard:

http://www.muffwiggler.com/forum/viewtopic.php?t=14328

Can you briefly explain when it is necessary to use the [hip~ 5] as shown in so many of Miller's examples? What is DC offset and does it occur in DSP or at A/DAC?

Many thanks, I've never known when to put one in, and where...

DC offset is when the oscillations of the signal are not centered around 0. The output of [phasor~], for example, is centered at .5, so it could be considered a sawtooth with a DC offset of .5. In the frequency domain, DC is the magnitude at 0 Hz. In filter design terms, [hip~] has a zero at 0 Hz. Wherever there is a zero in a filter, that frequency becomes completely filtered out, so [hip~] should completely kill DC. It works by using a weighted average of the current input and previous output of the filter. (The DC offset filter above uses the current input and many samples of the previous input.) Setting lower cutoff frequencies for [hip~] makes it take longer to filter DC, but has less of an effect on the other frequencies in the signal, so it's a matter of finding an appropriate trade-off.

As far as when to use it, it's probably not a bad idea to put one right before the [dac~]. But, of course, if there are no DC issues, you don't need it. If you know where the DC is occurring in the DSP chain, then you might want to put one there as you might get some unwanted artifacts or even more DC offset from further processing down the line. The reason why I made the one above, for example, is because I was working on a project that uses FM synthesis to generate band-limited waveforms. Part of the process uses an integrator, but because digital is not continuous time, the integration is approximate and causes DC drift. [hip~], in this case, was either too slow or caused unwanted phase issues, so the sawtooth ended up being curved.

Many thanks, I have a hunch that this is taking me a longer time to read than it did for you to write! I'll need to study filter design again to understand quite how this works and in the mean time throw a hip~ 5 at my dac~s.

Another patch for band limited waveforms? You are unstoppable.