problem using program change on VSTi in low latencies

[email protected] · Post by **[email protected]** » Sun Dec 12, 2004 8:37 am

In order to play VSTi as free instruments on a LIVE session i do 2 things:

1. Work under a low latency as 3ml
2. Assign program change from a clip to the VSTi
(When I like to play notes on that instrument I open another channel with "monitor ON" and send it to the instrument, but that does not effects the following problem)

The problem:

Every time a program change is applied I get an audio drop out on the master channel!
It appears that there is a jump of 50%-100% in CPU load for a split second.

i made few tests and it happens even with a minimum load (2 audio tracks with kick and bass line)
the problem still exist even with minimum load such as 3% CPU load!!!

i checked it with a wide range of VSTiS and the only ones the it doesn't happen with are Albino and Imposcar.
the rest : vanguard, minimoog, jangelist, 3z, reaktor amd many others will make audio drops every program change.

if I'd like to avoid the drops i have go as far as 48ml latency. In such situation it's impossible to play truly LIVE off an external keyboard

System details:

amd 64bit 3400+ processor + 1GB ram
RME multiface
Ableton 4.01 or 4.04

Help will be appretiated!!!

charlie p · Post by **charlie p** » Tue Dec 14, 2004 2:47 pm

what kind of drivers are you using..

make sure you use the ASIO drivers.

[email protected] · Post by **[email protected]** » Tue Dec 14, 2004 8:03 pm

Elementary

Using the RME drives as any other sound software

i must say that sending program change to external synth doesn't have any effect on the RME out put.
LIVE 4 manual doesn't mention program change in a VSTi context.
the only mentioning of the "program change" function is related to external synths only:

I'll be happy if even some one that haven't tried that before will just check it out on his machine..

Thanks

Machinesworking · Post by **Machinesworking** » Tue Dec 14, 2004 8:56 pm

Hmm? I'm using Reactor with program changes all the time, but first of all, it's for the sequencer in Reactor, and I have them queued in Session View to fire when a new session happens.
Secondly I got very used to 10ml latency, so I never set for less than that. The lower the latency the harder the CPU is pushed.
They say that standing a couple feet from a guitar amp produces that amount, and I came into keyboards that way, so I'm used to it I guess?

If you have specific VSTi that are giving you the result you want, then it's NOT Live that's at fault. Think about it, why would one VSTi be able to and another not? because one is coded better for changing presets at low latencies than the other.
So use the VSTi you have that work well changing presets in a live situation for that purpose.

[email protected] · Post by **[email protected]** » Tue Dec 14, 2004 9:49 pm

Machinesworking wrote: So use the VSTi you have that work well changing presets in a live situation for that purpose.

that a pretty big limitation and 10ms are to long to play with..any way only 48ms are enough to get the cpu steady.

[email protected] · Post by **[email protected]** » Sun Dec 19, 2004 9:31 am

any other with a similar experience?

tylast · Post by **tylast** » Fri Dec 24, 2004 11:08 pm

Yeah, I have the same issue with having clips change the patch in Reaktor. I get the CPU spike & audio drop out as well. I even tried adding a blank clip (with the program change) just before the scene that needed to have the program changed & I still the the spike & drop out. I thought it might just be Reaktor, but perhaps not.

It would really suck do a program change while I was playing live. Can't work around this either. I tried adding a second instance of Reaktor with the same synth. That raised my CPU usuage by 20%. No good.

tylast · Post by **tylast** » Sun Jan 23, 2005 9:26 pm

Apparently Reaktor is the cause here!

Frank Hoffmann · Post by **Frank Hoffmann** » Sun Jan 23, 2005 11:49 pm

When a virtual instrument receives a program change, it has to change and setup internal structures. This can cause performance peaks. Albino and ImpOscar seem to be pretty good optimized for program changes. Others are obviously not. What a VST does internally for a program change is out of our control.
With larger audio buffers on the sound card these performance peaks can be better compensated.

Frank

[email protected] · Post by **[email protected]** » Mon Jan 24, 2005 1:05 am

sad to hear!

i've noticed vanguard 1.1 does it better now.

for lower latency than 12ms i still get clicks but they are much less hard.

ReFx declaired better ajustment to ableton 4 in that last version, btu still the problem is there...

ewistrand · Post by **ewistrand** » Mon Jan 24, 2005 9:17 am

With Reaktor you aren't going to get past that in real time because of
the reinitialization that occurs,especially with complex ensembles.
You can get around the problem by building a snap switcher out of the snapshot module,with a morph control and A/B select button for the panel switches so you can automate it,but it still isn't a perfect solution.
ew

Hedroom · Post by **Hedroom** » Fri Jan 28, 2005 12:23 pm

Frank Hoffmann wrote:When a virtual instrument receives a program change, it has to change and setup internal structures. This can cause performance peaks. Albino and ImpOscar seem to be pretty good optimized for program changes. Others are obviously not. What a VST does internally for a program change is out of our control.
With larger audio buffers on the sound card these performance peaks can be better compensated.

Frank

I get it with impulse and simpler too!

Can badly-optimised plugins sitting in my plugin folder (even if not being used) affect performance of the native ableton instruments?

Frank Hoffmann · Post by **Frank Hoffmann** » Fri Jan 28, 2005 2:03 pm

Splashmas wrote:
Frank Hoffmann wrote:When a virtual instrument receives a program change, it has to change and setup internal structures. This can cause performance peaks. Albino and ImpOscar seem to be pretty good optimized for program changes. Others are obviously not. What a VST does internally for a program change is out of our control.
With larger audio buffers on the sound card these performance peaks can be better compensated.

Frank
I get it with impulse and simpler too!

Can badly-optimised plugins sitting in my plugin folder (even if not being used) affect performance of the native ableton instruments?

Of course not!

Simpler and Impulse are both sample based instruments and have to load the samples from hard disk into memory. All sample based instruments have the problem that they produce hugh performance peeks, when the preset is switched. This can't be avoided.

There might be other type of plug-ins that need to change much before a new sound can be create, like Reaktor for example.

Most VA-synths usually don't need so much performance.

Anyway, a preset change always creates a performance peek to some extend and therefore a chance for a drop-out. Best is to avoid preset changes in a song.

Frank

Hedroom · Post by **Hedroom** » Fri Jan 28, 2005 3:57 pm

Ok. Thanks for the quick reply.

[email protected] · Post by **[email protected]** » Fri Jan 28, 2005 5:30 pm

Frank Hoffmann wrote: Anyway, a preset change always creates a performance peek to some extend and therefore a chance for a drop-out. Best is to avoid preset changes in a song.

Frank

is there a chance for improvment on that aspect?
does ableton works with some VSTi developers on that isue?
if so - which ones?