While I'm waiting for my VoCore to really advance with the project, I'm thinking about ways to improve the accuracy of my sensor readings. I actually get enough information to determine if a machine is running and can somewhat tell in what state it is in but I'm depending one some very specific spikes that happen at the beginning and end of every run.

Right now the EmonLib is giving me basically three different current states with a bit of noise in each state.
Here a data plot of a complete run (right click->view to see it in full size).

The flat thing at the end is how its supposed to look when it idles but you guessed it, it doesn't always look like that. I get a spike to the mid level from time to time which really messes with how I want to write the software. I'm not able write complicated algorithms that can handle this random spiking.

If thats even possible, as a single spike sometimes means the start of a run and then minutes of just nothing until the machine goes into overdrive. So I will have get rid of this spiky noise that occurs in my circuit. I somewhat suspect that the  RF transmitterthat is pretty close to everything else might play a role in that.

I will try and see if the RPi directly connected to the sensor board has similar issues or if gets better. I might also try to shield the board somehow, the surrounding machines might emit quite a lot off electro-magnetic noise. Unfortunately thats a topic I'm not really familiar with yet so I have to read into quite a lot of stuff to understand and fix this issue.
I will also look into dedicated energy monitor boards and see what they did.

Regarding scematics, so far I don't have any. The circuit is pretty simple and I just did it on the go. The CT sensor circuit is just a copy of this circuit here:

http://openenergymonitor.org/emon/buildingblocks/how-to-build-an-arduino-energy-monitor-measuring-current-only

And yes I could also just use the full board from this (really great) project but thats just too much for this purpose, current is absolutely enough and I want to keep costs and complexity as low as possible.