How many effects has the integrated DSP?

What impulse lengths are useful to you?

Which increases the CPU hunger - the length or the complexity of the IR file? (or is it both?)

thanks for even thinking at all about a reverb unit.
know what reverb is MY fave? the sp404 stock one!
no one asked? I KNOW!

1 Like

I didnā€™t ask but Iā€™m glad to know :smiley:

And seconded, the thought of a reverb, let alone a convolution one, makes me very happy. Howbouta SP404 impulse response once this gets sorted out, @jonsimon ? :slight_smile:

1 Like

What do you mean by complexity of the IR file? Assuming that the IR file is just a WAV file then the contents do not have an effect on the CPU usage, just the length, i.e. number of samples. I suppose more sparse representations are possible such as a collection of delayed IR, or, parametric IR but that is probably outside the scope of this discussion.

I just had a look thorough my modest collection of IRs and the longest is 11 seconds at about 3.2Mb.

Iā€™ve not particularly felt these to be lacking in any way, so something around that kind of length would be great if possible.

i would be very happy with a stereo convolution unit with sample length of about 2 seconds! maybe sample rate insensitive if possible. but less is also good as long as it is stereo.

1 Like

Extremely short impulse lengths are of interest to me as well as longer ones.

Kind of had reverb written off on the basis of it eating up too many precious cpu cycles but this could be amazing if do-able, even if only for sound design stuff, rather than mix verb

1 Like

Sorry Brian, it was probably a dumb question :smiley:

Any size would be great! Maybe have the length determine the CPU load?
But cap it at 20sec?
As long as you can set the size and decay can be massive for ambient applications then Iā€™ll be over the moon!

And while we are talking about IRs, another unit like this tonebox could be a lovely addition to the amazing list of units.

1 Like

Yep, kind of what I was hinting at above. Convolution is can be used for more than just reverb. :slight_smile: Used in speaker/cabinet modeling, filtering, synthesis. Having a generalized convolution unit (or maybe specific types of convolution units) under voltage control could get pretty wild!

2 Likes

Holy moly how good would this be! I would absolutely love a selection of convolution units, or at least one unit that can be used as a reverb/colouration box. Maybe with an option to disable or turn the decay and size to 0 for it to become just a signal colouring tool.
Too many ideas right now hahahaha

1 Like

Yes, convolution is used to great effect in my Axe FX Ultra II (guitar modeller) - imagine reamping inside the ER-301. Now the ER-301 can go to 11 :slight_smile:

1 Like

The AxeFX has been on my gear lust list forever, along with the Kemper Profiling amp which is convolution based. Iā€™ve never quite been able to get myself to pull the trigger on either though Iā€™ve gotten close. So far, the Roland VG-99 still seems to satisfy my urge to take my guitar into outer space when I get that urge. :slight_smile:

How do you like yours?

Yes, having convolution inside of the modular would be mind melting, I think! Especially the way Brian builds things- very smart and very flexible!

The Axe FX II uses IRā€™s in the Cab block. Iā€™ve used many modelers (and amps) over the years and the Axe is the best of them IMO. I recently had a friend over to me (Ireland) from New Zealand, heā€™s a session guitarist / bassist. He is such a fantastic guitarist, plays effortlessly and beautifully. It was his first time using the Axe and he was blown away. Iā€™m afraid Iā€™ve just cost him a few quid!

Itā€™s a deep machine, but hey, here we are in the OD forum, like thatā€™s going to scare you :slight_smile:

2 Likes

Re: IR convolution

So I tried implementing the UPOLS (uniform partitioned overlap-save) algorithm for convolution.

(Reproduced from p.110 of Partitioned convolution algorithms for real-time auralization, Frank Wefers)

Unfortunately, the CPU load is probably too heavy for serious workā€¦

Test conditions:

  • 48kHz
  • Mono

Results for UPOLS:

 IR length | CPU load
----------------------
 200ms     | 10%
 400ms     | 24%
 800ms     | 54%
 1000ms    | 67%

:crying_cat_face: Sorry guys. There is, however, one more algorithm (GUPOLS) to try but itā€™s a bit more complicated to suss out and the improvement is probably just 70% of the CPU load of UPOLS.

1 Like

Hi guys, I heard that a new CPU will be available as an upgrade, is this true ? If it is, how much of an upgrade on processing power will it be ?

Nothing has been decided yet. That is 2018 agenda.

Okay, will keep my ears open ā€¦ thanks