Ladder Multimode Filter is in the Shop 39€

This forum is for discussing Rack Extensions. Devs are all welcome to show off their goods.
User avatar
miscend
Posts: 1955
Joined: 09 Feb 2015

27 Nov 2015

Sounds good!

User avatar
selig
RE Developer
Posts: 11685
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

27 Nov 2015

eXode wrote:Just compare with u-he Diva or NI Monark for instance. Urs previously stated that it would likely not be possible to run Diva on a single voice on the RE platform. It's not as simple as using zdf designs (obviously). There are other aspects such as implementing oversampling (and downsampling) and also implementing nonlinear feedback. As far as I know, nonlinearities are one thing that consumes CPU.
Nonlinearities just mean some form of saturation etc. yes, it adds CPU because it's another process. It also often requires over sampling which adds CPU.


Sent from my iPhone using Tapatalk
Selig Audio, LLC

MDTerps2015
Posts: 416
Joined: 25 Jan 2015

28 Nov 2015

LMAO, you guys essentially talked me out of even trying this lol. Thanks a lot for the info and the on going convo but you have scared me half to death into even trialling this. Sounds good but it would probably just be a trophy for me.
150 paid RExtensions and still no Grammy

User avatar
eusti
Moderator
Posts: 2793
Joined: 15 Jan 2015

28 Nov 2015

MDTerps2015 wrote:LMAO, you guys essentially talked me out of even trying this lol. Thanks a lot for the info and the on going convo but you have scared me half to death into even trialling this. Sounds good but it would probably just be a trophy for me.
Just try it. It sounds really good. Didn't have any DSP issues with it, but that could be because I never tested it in a full song so far.

D.

User avatar
gak
Posts: 2840
Joined: 05 Feb 2015

29 Nov 2015

Alright, considering it's 19 bones I'll give it a whirl tonight. I'll letchu know about the CPU.

User avatar
gak
Posts: 2840
Joined: 05 Feb 2015

29 Nov 2015

Ok, it's "hungry" a bit (if you use more than one instance) but it's pretty cool. Hopefully, the sale is until Sunday as I've had a nip tonight and I'm not really sure what I'm getting out of it.

But CPU isn't DRASTIC like the mixfood stuff ;)

MDTerps2015
Posts: 416
Joined: 25 Jan 2015

29 Nov 2015

I just trialled this and it wasnt that big of a hit on my system. I got a four yr old Imac i5 at 2.7. I dont have a bunch of stuff or crap running in the background. I like im buying.
150 paid RExtensions and still no Grammy

User avatar
gak
Posts: 2840
Joined: 05 Feb 2015

29 Nov 2015

I think I'ma buy this at this great price and not get the synapse. Then I have to learn how to use filters creatively :oops:

nis
Posts: 7
Joined: 26 May 2015

30 Nov 2015

selig wrote:Interesting. Resistors don't add distortion in the digital world. You ether add it yourself or its not there. Not sure what their answer is getting at, because if there is saturation they have added it, at least as far as I understand that world.
That doesn't really make any sense?

Perhaps I was a bit unclear, but what I was trying to explain in the interview was just that we don't have a separate saturation unit (like a distortion effect), but that the saturation arises in the filter - like in the analogue world where saturation arises naturally when overdriving analogue components. (Including resisters).
selig wrote:IS this a different filtering "system", or is it that a different process was used to derive the algorithm (which might include analog anomalies)? If this is a "ladder" filter, then I would assume (though I'm certainly no expert) it MUST follow the circuit for a ladder filter, same as any other digital ladder filter. Inquiring minds would love to know - what exactly makes this different from the Ladder filter in Thor or Antidote? :)
We use the same model of the Moog Ladder filter as everybody else; four cascaded 1-pole low-pass filters with a feedback loop.

Antti Huovilainen shows in his seminal paper how to derive the differential equations:

http://www.mirlab.org/conference_papers ... /P_061.pdf

The equations on their own are actually not that complicated (high-school algebra). Where the fun begins, however, is when deciding on what numerical approach to take (https://en.wikipedia.org/wiki/Numerical ... _equations) when discretizing the system. In most cases, due to the feedback loop and the non-linearities, you end up with a set of mutually dependent trancedental equations (https://en.wikipedia.org/wiki/Transcendental_equation) which cannot be solved analytically.

From here you have as far as I know four options:

1. Resolve the feedback loops with a 1-sample delay. This is most likely what non-CPU-intensive models (like Thor) does. Vadim discusses in his book "The Art of VA filter design" (http://www.native-instruments.com/filea ... _1.1.1.pdf) the many drawbacks of doing so.

2. Disregard the non-linearities and resolve the equations analytically. Afterwards the non-linearities can be be reintroduced and you simply "pretend" that you have solved the equations. Vadim also discusses this approach in his book in Section 4.6. This is probably the approach that medium CPU-intensive plugins who brands themselves as ZDF (zero-delay feedback) filters like Synapsis AF-4 uses. Indeed we also use this approach in PMS-20. However, we are working on an update that uses approach 3.

3. Solve the equations explicitly by using a root-finding algorithm like Newton's Method or Broyden's method (https://en.wikipedia.org/wiki/Newton%27s_method). This is exactly what circuit simulators like SPICE (https://en.wikipedia.org/wiki/SPICE) does and obviously what Diva and probably also Monark does. This approach is very CPU-intensive but gives the best (most realistic) result soundwise. In Ladder we use a multidimensional newton solver that has been heavily optimised with SIMD (https://en.wikipedia.org/wiki/SIMD) to the equations that we are solving.

4. Use piecewise segmented functions to approximate the non-linearities and solve the equations analytically. This could also be the approach that Monark uses.

I hope this clarifies your question. Otherwise I would be happy to elaborate further.
JiggeryPokery wrote:The piece talks about the quality in the "high end"; it would be interesting to test and compare the results in 96 and 44.1. If there's no appreciable difference, then the "secret sauce" is ikely oversampling, which would largely explain the heavy CPU use, and all the guff about differential algebra can be safely ignored.
We use standard 4-X oversampling like many other plugins, e.g. Synapsis AF-4. Theoretically you could oversample your filter 100-X and whether you use approach 1. or 3. (as described above) probably wouldn't make a difference. Obviously if it was that easy then everybody would do it. However, if you stick with 4-X or 8-X oversampling I can assure that the "guff" about differential algebra becomes absolutely crucial.

And your idea that a test can be conducted using 44.1 KHz and 96 KHz samplerate from which it can be deduced whether oversampling is our "secret sauce" is simply put absurd. It doesn't require much understanding of DSP to imagine a scenario in which changing the samplerate from 44.1 KHz to 96 KHz wouldn't make a significant sonic difference, but where the underlying mathematical choices made in the algorithms would make a huge difference.
JiggeryPokery wrote:They also talk about avoiding one sample delay. So, they're using a zero delay feedback filter, like pretty much everyone else uses now.
As I discussed earlier all of the approaches 2., 3., and 4. are in the ZDF category. All ZDF means is that you've resolved the feedback loops somehow instead of using 1-sample delays.
jonheal wrote:Yikes! You guys make it all sound like a scam!
If your goal is to scam money from people then plugin-development would be a stupid path to take. Most plugin developers put their hearts into their plugins and work their asses off - because they love to make plugins. Including us. That's it. It is true that some companies use silly buzz-words to hype their plugins, but that doesn't make them evil scammers.
dhruan wrote:Nah, not a scam but how exactly it works is pondered here as the explanation provided by them doesn't seem to make sense.
I hope this post clarifies your questions. Otherwise I would be happy to discuss it further.

Best regards

Nis

Primal Audio
Last edited by nis on 30 Nov 2015, edited 2 times in total.

nis
Posts: 7
Joined: 26 May 2015

30 Nov 2015

dhruan wrote:Well fuck me... should have waited for the Black Friday sale, wasn't really thinking that they would discount it for less than half price so soon. Oh well...
Black Friday is not that big in Denmark so we didn't plan to do so and Propellerhead invited us to participate after we launched the plugin.

We are aware that this is a stupid situation for the people who bought the plugin in the week up to Black Friday, so feel free to email me at nis@primal-audio.com and we'll arrange something.

Sorry for your trouble.

Best regards

Nis

Primal Audio

User avatar
selig
RE Developer
Posts: 11685
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

30 Nov 2015

nis wrote:
selig wrote:Interesting. Resistors don't add distortion in the digital world. You ether add it yourself or its not there. Not sure what their answer is getting at, because if there is saturation they have added it, at least as far as I understand that world.
That doesn't really make any sense?

Perhaps I was a bit unclear, but what I was trying to explain in the interview was just that we don't have a separate saturation unit (like a distortion effect), but that the saturation arises in the filter - like in the analogue world where saturation arises naturally when overdriving analogue components. (Including resisters).
selig wrote:IS this a different filtering "system", or is it that a different process was used to derive the algorithm (which might include analog anomalies)? If this is a "ladder" filter, then I would assume (though I'm certainly no expert) it MUST follow the circuit for a ladder filter, same as any other digital ladder filter. Inquiring minds would love to know - what exactly makes this different from the Ladder filter in Thor or Antidote? :)
We use the same model of the Moog Ladder filter as everybody else; four cascaded 1-pole low-pass filters with a feedback loop.

Antti Huovilainen shows in his seminal paper how to derive the differential equations:

http://www.mirlab.org/conference_papers ... /P_061.pdf

The equations on their own are actually not that complicated (high-school algebra). Where the fun begins, however, is when deciding on what numerical approach to take (https://en.wikipedia.org/wiki/Numerical ... _equations) when discretizing the system. In most cases, due to the feedback loop and the non-linearities, you end up with a set of mutually dependent trancedental equations (https://en.wikipedia.org/wiki/Transcendental_equation) which cannot be solved analytically.

From here you have as far as I know four options:

1. Resolve the feedback loops with a 1-sample delay. This is most likely what non-CPU-intensive models (like Thor) does. Vadim discusses in his book "The Art of VA filter design" (http://www.native-instruments.com/filea ... _1.1.1.pdf) the many drawbacks of doing so.

2. Disregard the non-linearities and resolve the equations analytically. Afterwards the non-linearities can be be reintroduced and you simply "pretend" that you have solved the equations. Vadim also discusses this approach in his book in Section 4.6. This is probably the approach that medium CPU-intensive plugins who brands themselves as ZDF (zero-delay feedback) filters like Synapsis AF-4 uses. Indeed we also use this approach in PMS-20. However, we are working on an update that uses approach 3.

3. Solve the equations explicitly by using a root-finding algorithm like Newton's Method or Broyden's method (https://en.wikipedia.org/wiki/Newton%27s_method). This is exactly what circuit simulators like SPICE (https://en.wikipedia.org/wiki/SPICE) does and obviously what Diva and probably also Monark does. This approach is very CPU-intensive but gives the best (most realistic) result soundwise. In Ladder we use a multidimensional newton solver that has been heavily optimised with SIMD (https://en.wikipedia.org/wiki/SIMD) to the equations that we are solving.

4. Use piecewise segmented functions to approximate the non-linearities and solve the equations analytically. This could also be the approach that Monark uses.

I hope this clarifies your question. Otherwise I would be happy to elaborate further.
JiggeryPokery wrote:The piece talks about the quality in the "high end"; it would be interesting to test and compare the results in 96 and 44.1. If there's no appreciable difference, then the "secret sauce" is ikely oversampling, which would largely explain the heavy CPU use, and all the guff about differential algebra can be safely ignored.
We use standard 4-X oversampling like many other plugins, e.g. Synapsis AF-4. Theoretically you could oversample your filter 100-X and whether you use approach 1. or 3. (as described above) probably wouldn't make a difference. Obviously if it was that easy then everybody would do it. However, if you stick with 4-X or 8-X oversampling I can assure that the "guff" about differential algebra becomes absolutely crucial.

And your idea that a test can be conducted using 44.1 KHz and 96 KHz samplerate from which it can be deduced whether oversampling is our "secret sauce" is simply put absurd. It doesn't require much understanding of DSP to imagine a scenario in which changing the samplerate from 44.1 KHz to 96 KHz wouldn't make a significant sonic difference, but where the underlying mathematical choices made in the algorithms would make a huge difference.
JiggeryPokery wrote:They also talk about avoiding one sample delay. So, they're using a zero delay feedback filter, like pretty much everyone else uses now.
As I discussed earlier all of the approaches 2., 3., and 4. are in the ZDF category. All ZDF means is that you've resolved the feedback loops somehow instead of using 1-sample delays.
jonheal wrote:Yikes! You guys make it all sound like a scam!
If your goal is to scam money from people then plugin-development would be a stupid path to take. Most plugin developers put their hearts into their plugins and work their asses off - because they love to make plugins. Including us. That's it. It is true that some companies use silly buzz-words to hype their plugins, but that doesn't make them evil scammers.
dhruan wrote:Nah, not a scam but how exactly it works is pondered here as the explanation provided by them doesn't seem to make sense.
I hope this post clarifies your questions. Otherwise I would be happy to discuss it further.

Best regards

Nis

Primal Audio
Yes, now it's more clear what you're talking about, especially as concerns saturation - it's not THAT you added it or not, it's WHERE you added it (so it acts like the analog circuit acts). Thanks for taking the time to clarify here, we welcome your continued participation!
:)
Selig Audio, LLC

User avatar
gak
Posts: 2840
Joined: 05 Feb 2015

03 Dec 2015

Is there a dang manual for this thing or what? I jumped on it because the price was too good to be true (and it has cool filtering) but there HAS to be a way to understand how to adjust the HP/BAND/LP for specific types of filtering. Just playing with the knobs gets confusing after a while (and of course no presets for dopes like me! )

User avatar
eXode
Posts: 838
Joined: 11 Feb 2015

04 Dec 2015

gak wrote:Is there a dang manual for this thing or what? I jumped on it because the price was too good to be true (and it has cool filtering) but there HAS to be a way to understand how to adjust the HP/BAND/LP for specific types of filtering. Just playing with the knobs gets confusing after a while (and of course no presets for dopes like me! )
What is there to understand? You need one fader at 100% and the other at 0% to get that specific type of filtering. Setting both LP and HP to 100% gives you a NOTCH type response. Other than that it's just a question of mixing different "outputs".

User avatar
miscend
Posts: 1955
Joined: 09 Feb 2015

04 Dec 2015

When are you guys going to release another synth?

User avatar
gak
Posts: 2840
Joined: 05 Feb 2015

04 Dec 2015

eXode wrote:
gak wrote:Is there a dang manual for this thing or what? I jumped on it because the price was too good to be true (and it has cool filtering) but there HAS to be a way to understand how to adjust the HP/BAND/LP for specific types of filtering. Just playing with the knobs gets confusing after a while (and of course no presets for dopes like me! )
What is there to understand? You need one fader at 100% and the other at 0% to get that specific type of filtering. Setting both LP and HP to 100% gives you a NOTCH type response. Other than that it's just a question of mixing different "outputs".
Yes, but I was thinking there were known tricks or combos maybe.

Anyways, thanks, guess it's just pot luck.

User avatar
Data_Shrine
Posts: 517
Joined: 23 Jan 2015

04 Dec 2015

gak wrote:Is there a dang manual for this thing or what? I jumped on it because the price was too good to be true (and it has cool filtering) but there HAS to be a way to understand how to adjust the HP/BAND/LP for specific types of filtering. Just playing with the knobs gets confusing after a while (and of course no presets for dopes like me! )
I wrote to the dev about a manual for PMS-20, but got no answer (yet)
There is a (very short) manual for ladder filter : http://primal-audio.com/ladder/manual.html

User avatar
Puniho
Posts: 101
Joined: 15 Jan 2015

05 Dec 2015

Well, I love this filter and have been trying it out on everything! Here is a patch using just Thor and the Ladder filter:

https://www.dropbox.com/s/volyp3pfyd2zq ... r.cmb?dl=0

Post Reply
  • Information
  • Who is online

    Users browsing this forum: No registered users and 21 guests