Wow, I hadn't even thought of CPU load. I'd say that makes your approach the method of choice for lower grade CPUs/DSPs, definitely for me.<div><br><br><div class="gmail_quote">On Mon, Oct 17, 2011 at 7:38 PM, Edward Diehl <span dir="ltr"><<a href="mailto:diehl@umich.edu">diehl@umich.edu</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;"><br>
I've also been wondering about the best workflow for midi work. I<br>
want to record solo piano using Pianoteq and possibly other sounds<br>
using Linuxsampler or Yoshimi. That is, I'll use the same midi track<br>
to drive 1 or more sound generation programs. I find that the CPU<br>
requirements of doing midi seem to be quite high so I've decided that<br>
the best approach is to do a 2 step process:<br>
<br>
1) Sequence my midi track using just A3 and Pianoteq and no effects.<br>
do all the midi takes, editing. Use low latency since audio is<br>
monitored.<br>
<br>
2) "Bounce" midi to audio tracks (with Pianoteq, Linuxsampler, etc)<br>
and then mix/master those. Might even bounce to Mixbus since I own<br>
it already. Use high latency to minimize DSP demands.<br>
<br>
Initially I tried working with a midi track + audio buses like this:<br>
<br>
a3 midi track -> Pianoteq -> a3 bus -> a3 master<br>
a3 midi track -> Linuxsampler -> a3 bus -> a3 master<br>
<br>
I could add fx to either the bus or the master. The above set up<br>
requires routing midi and audio externally of Ardour using Jack,<br>
though I don't think that is a big issue. I'll automate the setup<br>
using a scripts start all the programs I'm using and use aj-snapshot<br>
for the Jack connections. One issue is that when recording midi you<br>
need to use low latency (I'm running at 8.7 ms) so you can monitor<br>
properly. The low latency makes the DSP problem worse.<br>
<br>
Below I have some results about the Ardour3 CPU requirements. The CPU<br>
usage I gauge from the Ardour DSP number. I have found that if the<br>
DSP reaches 100% bad things happen, like lots of xruns.<br>
<br>
Here are some measurements of Ardour DSP using Ardour3 Alpha10<br>
<br>
Setup is:<br>
a3 midi track -> Pianoteq -> a3 bus -> a3 master<br>
<br>
DSP requirements<br>
IR EQ LS LS+EQ<br>
Just playing 45% 54 67 82 92<br>
Play + record 57% 65 75 92 99<br>
Playback 50% 60 72 88 99<br>
<br>
IR = Add IR plugin on master track<br>
EQ = Add LinuxDSP MKIIGraph EQ on Pianoteq bus<br>
LS = Add LinuxSampler bus. Route midi also to LinuxSampler playing<br>
SalamanderGrand and make new A3 bus for it.<br>
LS+EQ Add LinuxDSP MKIIGraph EQ on Linuxsampler bus<br>
<br>
DSP varies a bit for above measurements. Seems to go up if you play<br>
faster. You can see that I run up against overloading the CPU when I<br>
drive 2 audio buses from a midi track with a few fx. Hence, I've<br>
decided that I need to do midi recording and audio mastering in the 2<br>
step process I mentioned above.<br>
<br>
My laptop is a Dell 1420 with Intel Core2Duo T7100 1.8 GHz.<br>
<br>
_______________________________________________<br>
Ardour-Users mailing list<br>
<a href="mailto:ardour-users@lists.ardour.org">ardour-users@lists.ardour.org</a><br>
<a href="http://lists.ardour.org/listinfo.cgi/ardour-users-ardour.org" target="_blank">http://lists.ardour.org/listinfo.cgi/ardour-users-ardour.org</a><br>
</blockquote></div><br></div>