Berlin Modular (ancient thread)

Official support for: u-he.com
Post Reply New Topic
RELATED
PRODUCTS

Post

I only tried it on a not audio optimised linux/wine system, but three notes took about 75% of my phenom II quad core :)
It'll be better when I reboot into something with realtime optimisations (I hope)

It's a crazy synth for fx though. The fact that you can patch anything into anything and create crazy feedback looks makes this tremendous fun. It's what I like about the ms-20 too.

Post

Urs wrote: - Berlin Modular is for those who can not afford a full blown analogue modular synth but want some remotely similar fun. Or they just want to be able to save patches and play polyphonic.
and possibly live... :-o

Post

justin3am wrote:I've been trying to coax the ugliest sounds I can out of this thing. I think this comes close.

http://www.3amnoise.net/bipolarwaveshaping.mp3

some wonderful sounds there!

Post

mateo wrote:Will end users be able to create their own synths? Not that Bazille doesn't seem cool as is, but I'd be even more interested in a completely modular system.
Nope. Each system is a single, massive, highly optimized algorithm. Only this way can I achieve that quality at a maximum of 1 sample latency between two ends of a cable. A user configurable system would mean that I'd have to be a multitalented genius who can also write a compiler (think NI's Dr. Sync), or I'd have to make this up out of many, many jumps between various algorithms. Which I think would slow it down too much. However, we're not there, I can't say for 100% sure.

That said, thanks to SSE/AltiVec all 4 oscillators can be calculated at once (these use true sin() functions, not just wavetables like the "originals"). Same accounts for both filters (I could actually do 4 in the same time /mescatches head), the 4 multiples and the 4 lag generators. So it only makes sense to swap a whole section at once. And this is the basic difference that I think we'll have between the systems.

Post

But you are a multi-talented genius are you not?

Steve

Urs wrote:
mateo wrote:Will end users be able to create their own synths? Not that Bazille doesn't seem cool as is, but I'd be even more interested in a completely modular system.
Nope. Each system is a single, massive, highly optimized algorithm. Only this way can I achieve that quality at a maximum of 1 sample latency between two ends of a cable. A user configurable system would mean that I'd have to be a multitalented genius who can also write a compiler (think NI's Dr. Sync), or I'd have to make this up out of many, many jumps between various algorithms. Which I think would slow it down too much. However, we're not there, I can't say for 100% sure.

That said, thanks to SSE/AltiVec all 4 oscillators can be calculated at once (these use true sin() functions, not just wavetables like the "originals"). Same accounts for both filters (I could actually do 4 in the same time /mescatches head), the 4 multiples and the 4 lag generators. So it only makes sense to swap a whole section at once. And this is the basic difference that I think we'll have between the systems.

Post

You mean, you're not a multitalented genius ? :cry: :cry:

Post

Thanks :oops: - wasn't fishing for compliments though...

All I say is, there are some dudes out there with some incredible dev mojo. Magnus Lidstrom would be another one of those who can actually write such applications. I've looked into it, but it only leaves me scratching my head.

Post


Post

Urs wrote: That said, thanks to SSE/AltiVec all 4 oscillators can be calculated at once (these use true sin() functions, not just wavetables like the "originals"). Same accounts for both filters (I could actually do 4 in the same time /mescatches head), the 4 multiples and the 4 lag generators. So it only makes sense to swap a whole section at once. And this is the basic difference that I think we'll have between the systems.
Then how about the ability to add and remove whole sections? i.e. add another 4 osc section, if you need more? :)

Post

I wonder how over-the-top it could go though, because one instance of Bazille has to be processed by one CPU core, right? For monophonic patches, maybe.

Post

mateo wrote:
Urs wrote: That said, thanks to SSE/AltiVec all 4 oscillators can be calculated at once (these use true sin() functions, not just wavetables like the "originals"). Same accounts for both filters (I could actually do 4 in the same time /mescatches head), the 4 multiples and the 4 lag generators. So it only makes sense to swap a whole section at once. And this is the basic difference that I think we'll have between the systems.
Then how about the ability to add and remove whole sections? i.e. add another 4 osc section, if you need more? :)
Sheesh... then you also need to buy a new monitor and a new video card that can handle blitting all the cables...

Seriously though. At the moment the concept is like it is. I've been wanting to do this for years, and in all those years I havn't found a way to make it much more flexible.

It is not a replacement for Zebra in that way, it's really more of a fun device that somewhat follows the spirit of a Cwejman SE1 or a Roland System 100 (hence the colour I guess).

Post

Don't stress Urs, we at Urs-landia tend to get carried away.
You can only blame yourself for that :wink:

Post

spritex wrote:I wonder how over-the-top it could go though, because one instance of Bazille has to be processed by one CPU core, right? For monophonic patches, maybe.
It's a lot... the Win version in the beta team takes up 40% more cpu than the Mac version on teh same cpu because the compiler did not inline a single function. Now imagine a magnitude of not inlined functions...

As I said, it would have to be hand coded in assembly, and it probably would have to use self-modifying code. Which I havn't done yet. Maybe for version 3 then...

#---

On a side note, there is some potential for improvements left, e.g. currently all 20 inputs of the Multiples are evaluated many times for each sample. If I cut that down to blocks of, say, five inputs per iteration, depending on how many inputs are actually used, I guess we can alraedy save 10% for average patches. Complex patches would however use even more.

Trade-offs... trade-offs...

Post

Urs wrote:Seriously though. At the moment the concept is like it is. I've been wanting to do this for years, and in all those years I havn't found a way to make it much more flexible.
Please keep it that way. It will be even better for us if it's what you want it to be.

Post

It took me a bit of a while to figure some things out, partly because I'm not overly familiar with truly modular synths. But I spent pretty much the day playing with it (it's a good way to ensure new entrants for the next patch competition because all us lucky winners of competition #1 are gonna be a tad distracted ;-)). As per the email suggestion, I tried modulating the sequencer snapshot with the sequencer, some results are at http://blurk.net/Projects/Bazille-Rhythmic3c.mp3.

Post Reply

Return to “u-he”