Hi,

in general I follow the path of "if it sounds good, it is good", but I wonder how...

Pds [lop~] and [hip~] affects the signal overall quality. Depending on what I feed through them, I kind of feel the the sound is affected negatively at times.

I can not hear any difference from summing the signal with [+~] or just attaching then to the same inlet~ of where they are going. What negative effect is supposed to happen when you not sum with [+~]?

I need a simple x-over to send 5 channels and 1 sub and did this thingie. Does not sound that great. I think.

The pre clip gain of course has to be adjusted for what is coming in. But I also like the sound of [clip~]. (And never mind the odd routing to dac~)

Thanks for input!

There's no difference between using [+~ ] and just inserting signals to one inlet. The signals are added in the code of each object anyway, so it is just the same.
I don't know what is happening outside the subpatch you posted, but adding five signals is very likely to create a sum that exceeds the digital audio range (-1 to 1), which will be clipped, even if you don't use [clip~ -1 1]. I'm guessing that the bad sound quality comes from this sum, even if you like clipping. I think it's best to add some limiter before that. zexy's [limter~] is a great object, but Pd's native [slop~] provides a compressor in its help patch that might work. But again, I'm just guessing here.

@alexandros said:

There's no difference between using [+~ ] and just inserting signals to one inlet. The signals are added in the code of each object anyway

Cool, thanks.

But what about creating 2,3,4 etc pole filters by series like i have done?

And pd documentation mention the quality of the vanilla filters, compared to "more expensive" (cpu). What is this about? Simply that they will not color the sound, or will they leave unvanted artifacts?

Sorry about the patch screenshot, it is a mess.

@alexandros said

...adding five signals is very likely to create a sum that exceeds the digital audio range (-1 to 1), which will be clipped, even if you don't use `[clip~ -1 1]

This surprise me! I though signals in pd basically had infinite headroom until it arrived at something that can not handle it. Or so it seems. And there is a huge difference in levels if you omit [clip~] even if it is set to -1 1. I use it ti protect ears, speakers, and sensitive ppl.

@cfry the quality mainly refers to how steep they are. The basic filters are just 1-zero 1-pole, simple recursive digital filters.
these do introduce phase differences in the frequencies but that usually isn't too audible. Other than that they are linear (meaning that the gain and phase difference at a specific frequency doesn't depend on input volume or phase) and time-invariant. so they won't have any other artifacts, unless you get roundoff error issues or something which also aren't audible usually. (however in very edge cases numerical issues can have an effect on the sound)
there was a case of this recently on the mailing lists but I can't find the full achive for some reason. here's part of the thread:
https://www.mail-archive.com/pd-list@lists.iem.at/msg23581.html
here is a description of the issue:

The test-patch is a 0.1 Hz oscillator sourcing a highpass filter
3.order with butterworth characteristic at 20 Hz.

Try this patch with pd-0.52-1-msw-i386 and amd64 aka 32-bit and 64-bit
and older pd versions.
The 32 bit version has 20 dB less noise than the 64 bit version (and
no oscillations).

Double precision filters of iemlib ("hp3_butt_dp~.pd") have less noise
than single precision filters.

IOhannes z. and Chrostof R. figured out, it depends on compiler
options -ffast-math and / or -fassociative-math.

but like I said it's unlikely that's an issue here..

@seb-harmonik.ar said:

@cfry the quality mainly refers to how steep they are. The basic filters are just 1-zero 1-pole, simple recursive digital filters.

Thanks, then I can relax on this.

but like I said it's unlikely that's an issue here..

Yes, sorry I posted such a messy patch.

But the general approach I use, vanilla filters in series to create a x-over for 5 speakers and a sub, is this a good idea? Or would you do it differently? Any idea on how steep the filters should be, and how much they should overlap? Just to learn a bit.

The actual speakers are basically scrap parts, so it will never be hifi, but that is another story.

@cfry I know that single hip~ and lop~ filters can act as crossover filters bc the highpass is made by subtracting the lowpass transfer function from 1. (and therefore if they're added together you'll get unity back)

I'm not too sure about more than 1 in series.. if you want to get into it their design is described in this chapter of miller's book. filters in series have their transfer functions multiplied. so if the transfer function of the lowpass is S and the transfer of the highpass is 1-S then 2 lowpasses wil have the transfer function S^2, and 2 highpasses will have the transfer function (1-S)^2. So adding filters like that and then adding them back together will give 1 - 2S instead of 1 so that makes me think it wouldn't be unity gain..

maybe someone who knows more on filter design will have a more conclusive answer