The death of sofware modulars

dasz · Posted: Tue Sep 22, 2009 11:23 am Post subject:

whoops, looks like I got off topic.

to put it back on topic, I have used one of my G2's a lot less than the other. so I will continue to do so, to save the other one more (I have been using my newer one for the last 6 months and other than that maybe 4 months of use here and there).

So if it is a matter of time question (for example how many times can you save a patch in memory before memory goes), then I know one synth has done a hell of a lot more saves than the other.

I am curious how many people in addition to gomidas have had their boards go on them? Can we do a poll?
/Dasz

iPassenger · Posted: Tue Sep 22, 2009 11:29 am Post subject:

Dasz,

On the off topic front there is also that MFB Kraftwerg thingy, peaked my interest for a bit. Smile

dasz · Posted: Tue Sep 22, 2009 12:24 pm Post subject:

dorremifasol · Posted: Tue Sep 22, 2009 2:15 pm Post subject:

I don't own either of these, but to my eyes the Doepfer is better built (and nicer too!). MFB stuff seems great but built with compromises to lower the price. From MFB I'm in love with the MFB 522, absolutely great sounding analog drum machine! I must get one of those someday...

My brain is boiling lately with the idea of starting to build a real analog modular synthesizer to go with my G2... I only need a bit of space... oh and money Very Happy

Inventor · Posted: Tue Sep 22, 2009 3:15 pm Post subject:

I'm sort of ever so slowly building a software modular as you call it, but it's kind of different. It will have ChucK running on a netbook with two channels of audio out and two of audio in to an eChuck modular synthesizer that I'm designing (with lots of help). The boards under development for the synth are Karplus-Strong, Boolean Sequencing, and Song Sequencer.

Les

iPassenger · Posted: Wed Sep 23, 2009 2:14 am Post subject:

purusha · Joined: Mar 13, 2008 Posts: 131 Location: Ilkley

Here's someone who obviously isn't "mucking about" with his Doepfer kit. This belongs to engineer / producer - Ott (he's just posted this on another forum).

Posted Image, might have been reduced in size. Click Image to view fullscreen.

Posted Image, might have been reduced in size. Click Image to view fullscreen.

buzzr · Posted: Wed Sep 23, 2009 6:48 am Post subject:

yep,, analog is nice, but it takes up an entire room!

dasz · Posted: Wed Sep 23, 2009 10:06 am Post subject:

I'm quite infatuated with the Dark Energy.

The sound, routing, size, build quality, and MIDI implementation seem right.

The SEM also looks great. hmmmm
/Dasz

bar|none · Joined: Jul 06, 2009 Posts: 20 Location: WA

I have a 12U Eurorack modular and also a Pacarana Kyma system.

I however will say that one of the best software/hardware based machines for modular fatness is the Spectralis. Also a ridiculous bargain used.

There have been huge advancements from the pre 1.0 OS .97 and now we are at 1.04 OS. This machine really does not get the respect it deserves. Worth checking out. Sometimes I'll actually take a patch I made on the modular and build it again on the Specki to be able to call up later. The real analog filters and fixed filterbank + the sequencer is another reason to get this machine.

grimley · Posted: Wed Sep 23, 2009 4:14 pm Post subject:

Maybe the Future Retro XS:
http://www.future-retro.com/xsoverview.html

[edit] ok scratch that ... for some reason I thought that thing was cheaper.

3phase · Posted: Thu Sep 24, 2009 7:00 am Post subject: Re: The death of sofware modulars

ersatzplanet · Joined: Jul 24, 2005 Posts: 65 Location: Seattle, WA, USA

I just can't see why Clavia just doesn't spend some engineering time and convert the G2 Demo program into a full featured product and charge for it. I would buy it if it took advantage of my audio interface and did more than just 2 channels, took audio into itself from other programs or wave files, and used more of the full set of G2's modules. I have used the Demo to do Volta like control of my modular and it works great (but only 2 channels!). I would think that since the demo is a working product now that fleshing it out would not be a huge task, Why not turn all the engineering time spent on the Demo into a software only product with a decent profit margin? I guess Clavia is one of those companies that just moves on afrom older product/ideas and they just become dead forever.

dorremifasol · Posted: Tue Sep 29, 2009 11:00 pm Post subject:

I can think of at least 2 reasons:

- They still may put a new virtual modular generation on the market, so they are reserving the concept, usability, etc, for the future machine.

- The demo runs a emulation of the Motorola DSP, and that uses a lot of CPU power. Converting all the DSP code to native x86 code would be like doing it all from scratch.

Andreasfr · Joined: Apr 29, 2009 Posts: 3 Location: Copenhagen, Denmark

I was just thinking the exact same thing! I would pay for the G2 in standalone mode like ersatzplanet point out, even though I already have a G2 keyboard. I just soooo hope that Clavia's reason for holding back is the first one on dorremifasol's list and not the second.. That would be a real shame..

But also slightly OT; I've also been torn between the Dark Energy and the Kraftzwerg.
I've been trying to patch an intuitive noise box for live use together in the G2, but the 7 bit scale of midi data makes tweaking feel... well.. scaled and steppy. For instance, when coarse tuning an oscillator it always happens in semi tone-steps (beside the partial mode). I want it to be more smooth to feel more analog. Do you people know how to do this or is that why we are having GAS when looking at real analog modulars? Embarassed

purusha · Joined: Mar 13, 2008 Posts: 131 Location: Ilkley

BobTheDog · Posted: Mon Oct 12, 2009 5:55 am Post subject:

The DSP code will be motorola 56K not a high level language.

High level languages are not very good at generating code for DSPs which tend to have instruction sets and memory layouts quite different to a "normal" processor.

purusha · Joined: Mar 13, 2008 Posts: 131 Location: Ilkley

Not necessarily true I'm afraid, and I speak as someone who spent 5 years working on Texas Instruments DSPs (which are built on the Harvard architectural model).

Very Happy

I even wrote a custom real-time pre-emptive OS for TMS320 - in 'C' with a few hand-optimised assembler routines.

Most of the code for DSP systems these days is typically written in 'C', with assembler usually only used where the code needs hand-optimisation.

Many off-the-shelf DSP algorithm libraries are written in 'C'. Optimisation routines for 'C' are so good these days that it's often not worth delving into assembler (depending on scenario).

I'd hazard a guess that G2 is probably a hybrid - high level stuff written in 'C' with optimised assembler routines.

Like I implied in the previous post, it would entirely depend on the mix used in the G2 code-base.

Edit: a quick Google has turned this up:

http://www.allbusiness.com/technology/computer-software/267479-1.html

It looks like Motorola are using a Gnu 'C' cross-compiler for at least some of their systems.

BobTheDog · Posted: Mon Oct 12, 2009 7:13 am Post subject:

Mmm, I am not an expert but I haven't seen a c-compiler generate decent 56K, but I am willing to be proven wrong.

Also And I state yet again I am not an expert from what I have seen about DSP architecture a C compiler would not generate performance critical code, for instance on a 56K how would the compiler "know" how to map my C based FFT algorithm to use the special addressing modes available on the dsp?

I understand that C could well be used for non-performant parts of the system but for the G2 I cannot imagine there is much of that.

Cheers

Andy

BobTheDog · Posted: Mon Oct 12, 2009 7:14 am Post subject:

P.s. the gcc cross compiler for 56k is absolutely useless.

purusha · Joined: Mar 13, 2008 Posts: 131 Location: Ilkley

Antimon · Posted: Mon Oct 12, 2009 7:43 am Post subject:

I went to a job interview at Clavia once, and from what I remember they do have a fair amount of hand-coded low level stuff (assembler/machine code), to the level of moving individual bytes around (whatever that might mean). I got the impression (again, my memory may fool me) that C was mostly used for interface stuff.

If the code is that finely tuned on a low level, porting may be a considerable task (which I suspect it is - a soft synth G2 is such an unbelievable killer app that if you can put it out there with little effort I think it's criminal not to).

/Stefan

BobTheDog · Posted: Mon Oct 12, 2009 7:45 am Post subject:

As I said before I am not a 56K expert but:

The 56K is based on harvard but has two separate data memory spaces both separate from program space.

Management of data across the two data spaces in reference to addressing modes is important which is why c compilers are not good at generating code for it.

For example for efficient decimation-in-time FFT algorithm implementation the 56K is designed for efficient butterfly computation based on the real and imaginary parts being stored in the two separate memory data spaces

Also memory space is low another reason why compiled code is a No No.

Concerning the gcc compiler I have used it for 56k/ie32/ie64 and mips and it has always produced extremely inefficient code.

For an itanium project I did on a 512 processor machine the intel compiler generated code that performed nearly four times faster than the gcc compiler, a significant difference.

I surpose we will have to agree to disagree on this one Smile

purusha · Joined: Mar 13, 2008 Posts: 131 Location: Ilkley

purusha · Joined: Mar 13, 2008 Posts: 131 Location: Ilkley