T-Ladder Filter conversion

[aliasant]

Here is something I made that sounds interesting but could use some more work to be efficient enough.

Its a Ladder filter. The original is from Reaktor and it uses about 87cycles right now.
It is hopped by 64. This cold be in green but I wanted it in stream and I think that when it is all asm'd
and hopped it wont be that slow any way.
Right now it is divided into several code blocks + some primitives.

Good luck with it.

T-Ladder from Reaktor.rar

(104.27 KiB) Downloaded 255 times

[infuzion]

Neat, thank you!

BTW, it is best not to use any hops at all until the near the final version; you cannot find the difference between an original module and an faster version. I just wasted 20 minutes finding that out.

Also, you do not need to .RAR OSMs; SM automatically ZIPs them very well.
cheers!

[aliasant]

Neat, thank you!

BTW, it is best not to use any hops at all until the near the final version; you cannot find the difference between an original module and an faster version. I just wasted 20 minutes finding that out.

Also, you do not need to .RAR OSMs; SM automatically ZIPs them very well.
cheers!

oh. ok. Ill think about that in the future

[MichaelBenjamin]

i must say as dr infuzion already mentioned you gain nothing by the (surely well-meant) packing to rar, since the osm is already compressed, and on the downside we must take the extra step of unpacking to the disk before opening, instead of instant opening. but this is only minor compared to your effort of putting this filter together, so thank you dr zoidberg:)

[infuzion]

...as dr infuzion already mentioned....but this is only minor compared to your effort of putting this filter together

Dr.? LOL
Yes, thanks for putting it together. I'm only saving you time by suggesting no need for the RAR

As you can see, there is 99.99% no difference in the Code merging I've done so far, the only blip is in the beginning from the hopped sqrps I forgot.

[aliasant]

i must say as dr infuzion already mentioned you gain nothing by the (surely well-meant) packing to rar, since the osm is already compressed, and on the downside we must take the extra step of unpacking to the disk before opening, instead of instant opening. but this is only minor compared to your effort of putting this filter together, so thank you dr zoidberg:)

OK. I got to clearify this. I dont pack it with rar to save space. I pack it because rar uses error correction.
zip files sometimes gets messed up but if the file is "rared" the error correction prevents this. Works well for me.

[aliasant]

...as dr infuzion already mentioned....but this is only minor compared to your effort of putting this filter together

Dr.? LOL
Yes, thanks for putting it together. I'm only saving you time by suggesting no need for the RAR

As you can see, there is 99.99% no difference in the Code merging I've done so far, the only blip is in the beginning from the hopped sqrps I forgot.

Well Dr. Infuzion

That was nice. When the TIRs are done the ladder part can be put together

Then those 2 parts can be put together and finally it all can be ASM'd =)

What do you guys think of the sound?

[MichaelBenjamin]

the sound very nice, similar to the reaktor lp2 - creamy but with a bit more juice than the lp2. i have a look later, no asm and code for me for some days.

[infuzion]

Here is what I have so far, but I need to move on to other projects, so I'll let you guys have an attempt to improve what I have so far. The "merging" version is 100% the same, but not "TIR Code Lower HF" nor "TIR Code Aliasing?". The "Lower HF" might be better, since the peaks near 16kHz are not as pronounced, though the original does seem to have a slightly steeper slope.

Also, the first 20% of the Freq knob is dead, so I think the Freq Algorithm 2 green maths need to be altered to use the full [0..1] modulation range IMHO.

Have fun!

[aliasant]

Ill have a look tomorrow.

I think the freq algo is a bit too exact in the lower end. Can easily be adjusted with the exponent.

Update: I set the Exponent in the Freq Algo to 3 and that feels better.

The "T-Ladder TIR Code Aliasing?" has a problem. Havnt found what yet but the freqs are wrong.
If I want say 200Hz it actually gives me about 400Hz And if I set it at 400 it gives me about 800 so its doubled I guess.
Wonder why?
If you remember doing something that could have caused that let us know.


Update 2: haha. (feeling like a nutty professor) If you change the double Pi (6.2831898) to Pi in the green math the freq is back to normal with an interesting side effect The aliasing above nyquist ( I think) is gone and it looks more like a traditional filter. But.... It is not as accurate above 1kHz

Check the difference between the "T-Ladder TIR Code Lower HF" and the "T-Ladder TIR Code NOT Aliasing?"

[infuzion]

Thanks for the replies.
I goofed with the Stream-primitives math conversion into Code with the TIR calculations. I think it should be:

Code: Select all

streamin TIRfeed;streamin w;
float TIRa0;
streamout TIRb1;
streamout TIRc0;
streamout TIRd1;
float denorm=0.00000001;
TIRa0 = TIRfeed+TIRb1;
TIRb1 = (TIRa0+denorm)-denorm;
TIRc0 = (w*TIRa0)+TIRd1;
TIRd1 = (TIRc0+denorm)-denorm;

I think I did not understand fully the "Delay by 1 sample". However, it is still the cause for the filter being off from the original. So, I'll still need help there, unless my Code version is more desirable.

Also, I think the Frequency Analysis Graph II.1 should switch off that Tick100 when no notes are being played. When ever I load that graph, my system CPU% on my laptop jumps up to 90%. Just think of all the electricity you'd save!
cheers.

[aliasant]

Ok. I changed to the new code but the aliasing version still has double the freq. I dont understand why.

I did make a new analyzer. It has selectable update rate. from 1 / sec to 100/ sec
Will use that from now. I dont have any problems like you mentioned and the latest beta6 feels much faster.

[infuzion]

...the "Delay by 1 sample". However, it is still the cause for the filter being off from the original. So, I'll still need help there, unless my Code version is more desirable.

We still need help/advice here; anyone please?

[aliasant]

OK. I made some pictures from Reaktor that describes how the parts inside this thing works.

I made this into a 1 sample delay because that was what I thought would be correct at the time.

[rl]

right, z^-1 or 1/z is just another notation for a one sample delay (z-domain transfer function)