They updated libpd to the latest last week. More info to follow.....

;-~   Gemnotes-0.1alpha: Live music notation for Pure Data http://sharktracks.co.uk/

From: Rich E reakinator@gmail.com To: Ed Kelly morph_2016@yahoo.co.uk Cc: PD List pd-list@iem.at; pddev pd-dev@iem.at Sent: Tuesday, 20 March 2012, 16:37 Subject: Re: [PD-dev] Filter design for iPhone

What do you mean by break, crashes?  If yes, what version of libpd are you using?  Up until about a month ago, floating point exceptions (such as NAN) would crash libpd, whereas pd-vanilla ignores them.  Peter B. recently added a commit that changed this:

https://github.com/libpd/libpd/commit/27c848a56d3d27e62231f8b0d9ef17c7c56cd9...

Anywho, you'll probably want to take care of those NAN's anyway. :)  Not sure why pd vanilla ignores SIGFPE.

cheers, Rich

On Tue, Mar 20, 2012 at 4:57 PM, Ed Kelly morph_2016@yahoo.co.uk wrote:

Hi,

I'm trying to design or find a resonant lowpass filter for use with libPd for Openframeworks running on an iPhone 4.

1. I've tried the rjlib filter attached (filtertest_list.tar.gz) and it works fine in Pd, but instantly breaks when I run it on the iPhone - all other playing audio cuts out, and it is impossible to reset. I suspect that the buffers fb1 fb2 ff1 ff2 ff3 are hitting NAN values and staying there, but I haven't tested this.

2. I've had a go at implementing the filter from http://www.musicdsp.org/archive.php?classid=3#25 as a Pd external (mvcf~.tar.gz) and it is very efficient CPU-wise. However, higher values of f or q cause the internal filter buffers to hit NAN and stay there (tested in Pd - I haven't tried this on the iPhone yet).

I would ideally like to find out why the second attempt breaks at high q and/or f, as what I want is a 24dB/Octave moog-ish filter. I'm sure those of you with higher math skills will intuitively know why, but I'm just learning calculus now (my early education was a bit chaotic). I'm also limited by the fact that this is for a commercial project (although I'll be happy to share the code :) so I can't just take a GPL external and recompiled libpd with it.

Ed

Gemnotes-0.1alpha: Live music notation for Pure Data http://sharktracks.co.uk/ _______________________________________________ Pd-dev mailing list Pd-dev@iem.at http://lists.puredata.info/listinfo/pd-dev

mvcf~ - Well, it breaks when the Q is high and freq is moderately high, or at high freq and Q, but it seems to be that the higher you get with either frequency or Q it has a greater probability of breaking (all  dac~ outputs at 1 (or possibly 1e+07), no more audio can be played by the patch.   What I mean is that with the noise~ object as the input, the filter breaks after about 1-3 seconds when Q=0.70-0.8, f=0.7-0.8, and there are variations on this depending on the settings. Something is profoundly un-robust about this filter, but it's cpu load is very low. I want to have a stable version for not much more CPU.

I will look at the RJDJ filters mentioned also. I would like not to have to worry about filters when programming libpd audio engines. It would be nice to get a really efficient IIR with a musical response that uses low CPU into the core - I am not an expert at using poles and zeroes however (but I'm working on my higher math often these days, with the Khan academy :) Some of this is subjective, perhaps the "musical" nature of filters, but certain types of filter are assumed by many (musicians, DJs etc) in this part of the world to be normally available. The 24dB lowpass resonant filter a la Moog is one of them, and a fundamental analogue processor as well. I understand the analogue better than the discrete, so I'm anxious to know what limit is reached in the coefficients of the filter that causes the undefined result (NaN).

But I'll try the other rjlib filters first! Ed

Gemnotes-0.1alpha: Live music notation for Pure Data http://sharktracks.co.uk/

From: Joe White joe@rjdj.me To: Ed Kelly morph_2016@yahoo.co.uk Cc: PD List pd-list@iem.at; pddev pd-dev@iem.at Sent: Tuesday, 20 March 2012, 17:50 Subject: Re: [PD-dev] Filter design for iPhone

Hi Ed,

I just tested your filtertest_list patch in RjDj and it worked exactly as expected. As Rich suggested maybe you're using an outdated version of libpd?

If Frank's around I'm sure he would know for certain but if I remember rightly the [u_lowpass] / [u_beequad] combination is not really designed for dynamically changing parameters. Maybe take a look at [e_reslop], [e_lop2] and [e_lop4] for some other filter implementations.

Regarding your second attempt, at what frequency does it 'break' and what's your sampling rate?

Cheers, Joe

On 20 March 2012 05:57, Ed Kelly morph_2016@yahoo.co.uk wrote:

Hi,

I'm trying to design or find a resonant lowpass filter for use with libPd for Openframeworks running on an iPhone 4.

1. I've tried the rjlib filter attached (filtertest_list.tar.gz) and it works fine in Pd, but instantly breaks when I run it on the iPhone - all other playing audio cuts out, and it is impossible to reset. I suspect that the buffers fb1 fb2 ff1 ff2 ff3 are hitting NAN values and staying there, but I haven't tested this.

2. I've had a go at implementing the filter from http://www.musicdsp.org/archive.php?classid=3#25 as a Pd external (mvcf~.tar.gz) and it is very efficient CPU-wise. However, higher values of f or q cause the internal filter buffers to hit NAN and stay there (tested in Pd - I haven't tried this on the iPhone yet).

I would ideally like to find out why the second attempt breaks at high q and/or f, as what I want is a 24dB/Octave moog-ish filter. I'm sure those of you with higher math skills will intuitively know why, but I'm just learning calculus now (my early education was a bit chaotic). I'm also limited by the fact that this is for a commercial project (although I'll be happy to share the code :) so I can't just take a GPL external and recompiled libpd with it.

Ed

Gemnotes-0.1alpha: Live music notation for Pure Data http://sharktracks.co.uk/ _______________________________________________ Pd-dev mailing list Pd-dev@iem.at http://lists.puredata.info/listinfo/pd-dev

...

I'm anxious to know what limit is reached in the coefficients of the filter that causes the undefined result (NaN).

I haven't seen the code, but I just want to make you notice that adding together -Infinity and +Infinity results in a NaN ; so does subtracting two infinities of the same sign.

So, the NaN might happen when two expressions that are supposed to partially cancel each other, happen to both overflow, in different directions.

There are various possible causes for NaN, but with formulas that only involve +, - and *, the possibilities are a lot more limited.

Hmmm. I think that's why it puzzles me so. Can you see a / anywhere here?

while (n--) {     i1=(*in++);     fc1 = (*fc++);     res1 = (*res++);     q = 1.0f - fc1;     p = fc1 + 0.8f * fc1 * q;     fcoeff = p + p - 1.0f;     q = res1 * (1.0f + 0.5f * q * (1.0f - q + 5.6f * q * q));     i1 -= q * xb4;     t1 = xb1;     xb1 = (i1 + xb0) * p - xb1 * fcoeff;     t2 = xb2;     xb2 = (xb1 + t1) * p - xb2 * fcoeff;     t1 = xb3;     xb3 = (xb2 + t2) * p - xb3 * fcoeff;     xb4 = (xb3 + t1) * p - xb4 * fcoeff;     xb4 = xb4 - xb4 * xb4 * xb4 * 0.166667f;     xb0 = i1;     *out++ = xb4;  }

No reciprocals, no divisions, just +,-,*. I cannot help thinking I must be reaching a "limit" in the calculus sense...

Ed

| Mathieu BOUCHARD ----- téléphone : +1.514.383.3801 ----- Montréal, QC

My mistake

Ah. As often happens, as soon as I have pressed "send" and posted the question, the answer pops out. xb4 = xb4 - xb4 * xb4 * xb4 * 0.166667f;

is replaced by...

xb4 = xb4 - xb4 * xb4 * xb4 * 0.01f;

...and it works.

and then breaks again

Gemnotes-0.1alpha: Live music notation for Pure Data http://sharktracks.co.uk/

From: Ed Kelly morph_2016@yahoo.co.uk To: Mathieu Bouchard matju@artengine.ca Cc: PD List pd-list@iem.at; Joe White joe@rjdj.me; pddev pd-dev@iem.at Sent: Saturday, 24 March 2012, 8:47 Subject: Re: [PD-dev] [PD] Filter design for iPhone

...

I'm anxious to know what limit is reached in the coefficients of the filter that causes the undefined result (NaN).

I haven't seen the code, but I just want to make you notice that adding together -Infinity and +Infinity results in a NaN ; so does subtracting two infinities of the same sign.

So, the NaN might happen when two expressions that are supposed to partially cancel each other, happen to both overflow, in different directions.

There are various possible causes for NaN, but with formulas that only involve +, - and *, the possibilities are a lot more limited.

Hmmm. I think that's why it puzzles me so. Can you see a / anywhere here?

while (n--) {     i1=(*in++);     fc1 = (*fc++);     res1 = (*res++);     q = 1.0f - fc1;     p = fc1 + 0.8f * fc1 * q;     fcoeff = p + p - 1.0f;     q = res1 * (1.0f + 0.5f * q * (1.0f - q + 5.6f * q * q));     i1 -= q * xb4;     t1 = xb1;     xb1 = (i1 + xb0) * p - xb1 * fcoeff;     t2 = xb2;     xb2 = (xb1 + t1) * p - xb2 * fcoeff;     t1 = xb3;     xb3 = (xb2 + t2) * p - xb3 * fcoeff;     xb4 = (xb3 + t1) * p - xb4 * fcoeff;     xb4 = xb4 - xb4 * xb4 * xb4 * 0.166667f;     xb0 = i1;     *out++ = xb4;  }

No reciprocals, no divisions, just +,-,*. I cannot help thinking I must be reaching a "limit" in the calculus sense...

Ed

| Mathieu BOUCHARD ----- téléphone : +1.514.383.3801 ----- Montréal, QC

Pd-dev mailing list Pd-dev@iem.at http://lists.puredata.info/listinfo/pd-dev

I think I've got a handle on this now. As you said, xb4 spirals out of control, so there are a couple of clamping factors that reduce xb4, hopefully below a threshold where it is stable.

It is of course, unstable if the frequency is <0 or >1 (where 1 is the Nyquist frequency) but there is a "safe" mode

I'm going to give it a try with this filter and see if any of my colleagues can break it :)

Ed

Gemnotes-0.1alpha: Live music notation for Pure Data http://sharktracks.co.uk/

From: Peter Brinkmann peter.brinkmann@googlemail.com To: Ed Kelly morph_2016@yahoo.co.uk Cc: PD List pd-list@iem.at Sent: Saturday, 24 March 2012, 15:08 Subject: Re: [PD] [PD-dev] Filter design for iPhone

On Sat, Mar 24, 2012 at 5:20 AM, Ed Kelly morph_2016@yahoo.co.uk wrote:

My mistake

...

Ah. As often happens, as soon as I have pressed "send" and posted the question, the answer pops out. xb4 = xb4 - xb4 * xb4 * xb4 * 0.166667f;

is replaced by...

xb4 = xb4 - xb4 * xb4 * xb4 * 0.01f;

...and it works.

and then breaks again

Inf - Inf is NaN.  If the value of xb4 spirals out of control, that may happen here.  That would also explain why it takes longer to fail when you reduce the last factor.

Gemnotes-0.1alpha: Live music notation for Pure Data http://sharktracks.co.uk/ ________________________________ From: Ed Kelly morph_2016@yahoo.co.uk To: Mathieu Bouchard matju@artengine.ca Cc: PD List pd-list@iem.at; Joe White joe@rjdj.me; pddev pd-dev@iem.at Sent: Saturday, 24 March 2012, 8:47 Subject: Re: [PD-dev] [PD] Filter design for iPhone

...

...

I'm anxious to know what limit is reached in the coefficients of the filter that causes the undefined result (NaN).

I haven't seen the code, but I just want to make you notice that adding together -Infinity and +Infinity results in a NaN ; so does subtracting two infinities of the same sign.

So, the NaN might happen when two expressions that are supposed to partially cancel each other, happen to both overflow, in different directions.

There are various possible causes for NaN, but with formulas that only involve +, - and *, the possibilities are a lot more limited.

Hmmm. I think that's why it puzzles me so. Can you see a / anywhere here?

while (n--) {     i1=(*in++);     fc1 = (*fc++);     res1 = (*res++);     q = 1.0f - fc1;     p = fc1 + 0.8f * fc1 * q;     fcoeff = p + p - 1.0f;     q = res1 * (1.0f + 0.5f * q * (1.0f - q + 5.6f * q * q));     i1 -= q * xb4;     t1 = xb1;     xb1 = (i1 + xb0) * p - xb1 * fcoeff;     t2 = xb2;     xb2 = (xb1 + t1) * p - xb2 * fcoeff;     t1 = xb3;     xb3 = (xb2 + t2) * p - xb3 * fcoeff;     xb4 = (xb3 + t1) * p - xb4 * fcoeff;     xb4 = xb4 - xb4 * xb4 * xb4 * 0.166667f;     xb0 = i1;     *out++ = xb4;  }

No reciprocals, no divisions, just +,-,*. I cannot help thinking I must be reaching a "limit" in the calculus sense...

Ed

---

| Mathieu BOUCHARD ----- téléphone : +1.514.383.3801 ----- Montréal, QC

---

Pd-dev mailing list Pd-dev@iem.at http://lists.puredata.info/listinfo/pd-dev

---

Pd-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list

...

[...] I will look at the RJDJ filters mentioned also. I would like not to have to worry about filters when programming libpd audio engines.

Absolutely.  If a patch works with Pd Vanilla, then it ought to work with libpd.  Would you mind filing a bug report at GitHub, ideally with a minimal example that reproduces the problem? Thanks,     Peter

Sure, although this is ofxPd from Dan Wilcox, so I'm not sure what the difference is between this and the iOS or Android versions.

Best, Ed

Hi Peter,

Oh dear! I'm not allowed to release this to anybody! I did eventually come up with a way to get round this - a new filter external that is now part of the app. If and when I get an iPhone I'll try to recreate the problem with a less commercially sensitive patch. Best, Ed

Gemnotes-0.1alpha: Live music notation for Pure Data http://sharktracks.co.uk/

From: Peter Brinkmann peter.brinkmann@googlemail.com To: Ed Kelly morph_2016@yahoo.co.uk Cc: Joe White joe@rjdj.me; PD List pd-list@iem.at; pddev pd-dev@iem.at Sent: Friday, 23 March 2012, 0:04 Subject: Re: [PD-dev] Filter design for iPhone

Hi Ed, The attached files are just the Pd patch and abstractions.  We'll probably need the entire app to get a sense of what's going wrong here. Cheers,       Peter

On Thu, Mar 22, 2012 at 11:52 AM, Ed Kelly morph_2016@yahoo.co.uk wrote:

...

...

[...] I will look at the RJDJ filters mentioned also. I would like not to have to worry about filters when programming libpd audio engines.

...

Absolutely.  If a patch works with Pd Vanilla, then it ought to work with libpd.  Would you mind filing a bug report at GitHub, ideally with a minimal example that reproduces the problem? Thanks,     Peter

Sure, although this is ofxPd from Dan Wilcox, so I'm not sure what the difference is between this and the iOS or Android versions.

Best, Ed