When I play a project, Sonar is not playing random midi notes. It's not the same notes every pass. In one project it happens with Rapture pro,TTS, and AD2. All midi tracks have this problem.
I opened another project and had the same problem.
I was using [2017.05] 23.5.0 build 32, when it first happened, so I updated to [2017.06] 23.6.0 build 24.
Still have the problem. I made sure the drivers are up to date.
Is there anything I could have done to cause this?

Preferences / MIDI / Playback and Recording / Playback : Prepare using: Xms.
Try different values, like >= 500ms and <=100ms.

azslow3
Try different values, like >= 500ms and <=100ms.

I am sure this should be between 500 and 1000. Some do need more than 1000 ms but I would suggest starting with 500 and increment the value until SONAR stops dropping notes.

scook

azslow3
Try different values, like >= 500ms and <=100ms.

I am sure this should be between 500 and 1000. Some do need more than 1000 ms but I would suggest starting with 500 and increment the value until SONAR stops dropping notes.

Till someone can prove my theory wrong, no VST synth in the world "need" more then 50ms to play notes correctly:
1) VST receive MIDI events exactly for one buffer/block size:

VST SDK 2.4
Events are always related to the current audio block. For each process cycle, processEvents() is called once before a processReplacing() call (if new events are available).

2) 44.1kHz / 2048 samples means ~46ms buffer
3) all VST instruments are ready to be used live. Everything over ~30ms is not "playable" live. I mean nothing can be "prepared" in such situation.
4) VST event/audio processing does not distinguish between live and recorded.
 
So for several years I try to understand why black magic with huge numbers works. If my theory is right, that is .... just a workaround for a bug In this case, 50, 100, 300, 1000 are "magic" numbers, which under particular conditions avoid that bug. I mean that is nothing with physical meaning (in ms).
 
PS. Note that my theory is for VST only. DX (f.e. TTS-1) has different logic.

Far from being magic numbers they are a practical answer to the issue encountered by this user. What is the alternative from a user perspective? Whether the fundamental reason for the setting is a bug or a choice to buffer this data to allow other processing that cannot be as buffered as easily without adversely affecting performance, I cannot say.

Thanks azslow3 and scoot.
It was set at 250ms, and I changed to 500ms.
That flute solo sounds right for the first time in a week.
If I only knew how to mark this thread solved.
 
God bless ya,
Dave G

Edit the initial post and modify the subject line.

Thanks scoot 
You learn new things every day, Today I learned 2
DG