Hello, PD. I'm new to the language, having found my way this direction due to a dissatisfaction regarding the limitations of various "modular synthesis" programs, and I must say that I'm having a great time! Anyway, thanks in advance for your help; I really hate bugging people on forums, but, being a niche language as it is, there isn't as much documentation available on Google as I would like.

Anyway, I'm building a synthesizer and, because I want the experience to be as educational as possible in the sense of getting a feel for the language, I'm trying to use only the most vanilla objects in PD. I came across a problem, though, when I found that I can't create ">~" objects, or any such objects to compare two signals rather than just two messages. Because of this, I'm rather stumped about how to make a variable duty-cycle pulse wave.

The tutorial at en.flossmanuals.net/PureData uses the expr~ object for this purpose, but it also claims that this is an object from pd-extended. Even if it isn't, the design philosophy seems to be different from most of the other vanilla objects (making use of dollar-sign variables), which makes me frown.

I can add signals to messages with +~, subtract them with -~. Is there a way to compare them without using expr~?

-David

expr~ and expr are included with the normal vanilla pd distributions. they are sort of 'vanilla externals'. the reason being that pd itself is released under the BBD license, and expr and expr~ are released under the GPL license.

hope that is correct anyway. it all confuses me a fair bit.

>>I'm rather stumped about how to make a variable duty-cycle pulse wave.

you could do it with a metro, delay, and vline~, but only up to 1000hz, cos metro has a hardcoded speed limit of 1000hz.

[loadbang]
|
| [r speed]
| |
[metro]
|
[t b b] [r dutycycle]
| \ |
[1 ( [delay]
| |
| [0 (
| /
[vline~]

so, to [r speed] you would send the inverse of your frequency, and to [r dutycycle] you would send the inverse of frequency multiplied by the duty cycle. duty cycle of 0.5 being a square wave.

Thank you, Mod! Your solution looks like just what I need, but are metro and delay accurate enough for the job, timing-wise? I had an idea for a solution that involved threshold~ and a counter that would alternate between sending a 1 and a 0 to the DAC~, but the timing wasn't accurate enough to produce musical tones, although it did sound like a square wave. The reason that I didn't try using metro and delay was because, after that, I assumed that objects that send messages rather than signals aren't suitable for the job, but I wouldn't be surprised if I misunderstood.

Anyway, I'll give your idea a try and see for myself.

the timing of [metro] and [delay] should be very precise, the problem usually arises when you convert messages into the signal domain. because pd uses a 64 sample block to do signal calculations, any values sent into the signal domain will only be sent once every 64 samples. [vline~] escapes that by sending its value between block boundaries.

Oh, wow. Thanks for that explanation -- it really is very enlightening. I had read about that 64-sample-block thing before, but I didn't make the connection. It's nice to know that the timing tools are precise; that really multiplies their uses. Thanks for your patience.

Mod, I just tried your solution and it worked the trick, but I'll have to find a way to get around that 1000hz limit. Now that I'm more knowledgeable, though, I'm sure I'll think of something.

-David

You could also look up a table that's filled half with 0's and half with 1's using [phasor~]. Just multiply the output of [phasor~] by however many values you want to look up, and add an offset. For example, you could have a table with 1024 values, the first 512 are 0's and the second 512 are 1's. Multiply [phasor~] by 512 and add a number between 0 and 512 to set the pwm. Then use the signal to [tabread~] the table.