How would you configure and use [bob~] to demonstrate what the ladder filter fuss is all about? For some reason (inexperience? poor taste?) I keep returning to [vcf~].

Edit: after an hour or two of monkeys at the typewriter, I'd say that the 2 main differences from [vcf~] are the roll-off slope ([bob~] never completely mutes signal even when the cutoff freq is 8 hz) and the resonance sounds brighter and grittier (bob at resonance 3 sounds like vcf at q 22 to me). There's got to be more to it than that, right?

little note: your filter does not resonate at the cutoff frequency.
top: how it should be, bottom: how it is

now, solve this

@solipp If you have a solution why don't you share it? What is going on inside the second [pd ladder] sub patch I wonder...

Experimenting a bit, it becomes obvious that resonance screws with the filter gain coefficient

Using these crazy ass formulas for filter gain and and feedback I get correct frequency response at resonance close to 1:

Source: https://www.researchgate.net/publication/241280072_POWERWAVE_-A_HIGH_PERFORMANCE_SINGLE_CHIP_INTERPOLATING_WAVETABLE_SYNTHESIZER

It looks like this when I implement it:

I know it aint pretty... It also beefs up the CPU use quite a bit, which I sort of dealt with by by keeping it in a sub patch with normal block size and no oversampling. Another downside is the soft clipping has got to go- it screws up the precision.

I wonder if there is a simpler way to approximate this stuff...

Anyway, have fun finding all my design flaws 4plop~.pd

I wonder if there is a simpler way to approximate this stuff...

Turns out there is. A bit of polynomial magic and I got rid of the [exp~] module

Also ditched the resonance/feedback formula as I didn't see it making a greater difference in frequency response than it did in CPU usage

Updated version with precise response at resonance around 4 and cut off < 11025Hz - still no soft clipping...
4plop~.pd

Hi boca can you provide the whole patch?

@tungee said:

Hi boca can you provide the whole patch?

I already did in my last post

@bocanegra said:

Updated version with precise response at resonance around 4 and cut off < 11025Hz - still no soft clipping...
4plop~.pd

I dont understand, there is only one input . Shouldnt be 3 inputs there (signal input, cutofffreq and rsocnance)?

That is the coeff sub patch you are looking at. This is what I get when I open the attached .pd file from here:

Ah ok now its clear, there was a subpatch ok. If i use this in my automatonism module, it fires up when dragging the slider some Nasty highpeaks. So i installed the anaclip again in it.
Also: the patch of yesterday was stable until 8000hz at 5 resonance. now i can go upto 11khz and reso 10 . Thats nice! GREAT JOB BOCANEGRA :DD

@bocanegra http://www.willpirkle.com/Downloads/AN-4VirtualAnalogFilters.2.0.pdf is a good read. I translated the TPT laddder filter to pd in the [pp.ladder~] object in audiolab (you can download it with deken). Here is a version with an outlet for each pole and the "analog clipper" from above : TPT-ladder.pd
Another solution is [bob~] of course : https://github.com/pure-data/pure-data/blob/master/extra/bob~/bob~.c

@solipp The first link returns an empty pdf...

But thanks a lot anyway. http://www.willpirkle.com/Downloads/ has a lot of useful stuff. Never knew about TPT filters. Also, I have added your audiolab abstractions to my library long time ago and learn a lot from going through the patches seeing how stuff is done- but the content is massive so I haven't sifted through even 10% yet

A question: in 4plop above, there's 1 - coefficient done separately four times. Couldn't this be optimized by doing [*~ -1] --> [+ 1] only once and feeding that result into the 4 rpoles? Bc it looks to me like they all get the same signal input.

hjh

@ddw_music said:

A question: in 4plop above, there's 1 - coefficient done separately four times. Couldn't this be optimized by doing [*~ -1] --> [+ 1] only once and feeding that result into the 4 rpoles? Bc it looks to me like they all get the same signal input.

True! Nice catch

Fixed version 4plop~.pd

with some other minor improvements as well: Resonance/cutoff compensation and anaclip are back in the mix

Here are all available Pure Data Filters collected as automatonism modules in action:

In your Boca Ladder abstraction i mixed the normal tanh with the analog clipper,which in my opinion gives the best results. If in exp~ the quotient is ==5 , only analog clipping is used via switch. if both mixed are together the sounds gets some superb irregular resonances. the bite parameter in the module amplifies the resonance:

i also spent a oversampling switcher for 2/4 *oversampling, where with 4 OS the snapping of modulation sources sounds greater
imho.

Here is the test-file
Battle of the filters Vol.zip

Im late to the party. But for all it's sins Bob~ retains very clear transients, something that real analogue filters do very well too. Horses for Courses. I abso love the filter implementations above and they sound great in context, but this week trying them as a sub for bob~ in a fast transient environment, they smudged every thing thrown at them. Sure Bob~ sounds thinner but I found it more musically useful in fast percussive situations. Also has better pronounced resonance. It depends on the musical context. I love the Matrix 12 filter implentation, mind blowingly good. Now me thinks, with the wonderful wizardry displayed in the above filters has anyone attempted somethin in the class of the Pultec EPQ-1. By listening to the examples above, Im sure it could be done

@Element-S I can't help you with the Pultec EQP-1 implementation, but since you're talking about being late to the party it might be worth pointing out that at least some of [bob~]'s weaknesses, mentioned in this thread, may have been the result of a bug, that was fixed around three years ago. If I remember correctly, the output wasn't taken after the fourth stage of the filter but before the first (= input + feedback). So no surprise, that this was considered a bad implementation of the Moog ladder filter.

Edit: Had to look it up ... It was actually a bit different from what I remembered: The incorrect output was the first instead of the fourth state variable. So the output was taken AFTER the first stage.

@Element-S said:

I love the Matrix 12 filter implentation, mind blowingly good.

I have been meaning to update that for awhile now, greatly improved the resonance and distortion characteristics but need to put in some time to cut some of those improvements so cpu usage is sane, there is a balance in there somewhere. Will see about finishing it up tonight/tomorrow and will upload it in the polemix thread.

somethin in the class of the Pultec EPQ-1

I kind of like that idea, the fixed frequency bands of the EQP1/MEQ5/old passive EQs and mixers has always appealed to me, I like the workflow. I might play with this, maybe have something up next weekend. Never dawned on me to implement a passive mixer with all of their quirks and limitations in pd, would be a nice break from the DAW and its endless tweaking, getting the cross talk between channels right might be a trick and is a must to make its simple switched left/center/right panning work. Sounds like a fun project, 1930s style studio in pd.