I just got my IoTaWatt hooked up, and I’ve started learning how to use it. This thing is super awesome.
I’ve noticed that when my HVAC starts up, the inrush current seems to cause the readings drop below zero. Is this normal?
One thing that’s possibly relevant - I have a split system. I only have one CT on the air-handler (doubled) and one on the condenser (also doubled).
I broke it down and took the calculated Outputs out. It looks like it went negative only on my calculated outputs, because the Condenser inrush current wasn’t reflected on the main inputs. So I guess my real question is what would cause the condenser inrush current to not be captured by the main CTs?
With calculated Mains (A+B)
With original main A and main B.
IotaWatt samples the CT inputs independently. Mine typically samples at 30 to 40 times/second, which means each channel sampled about twice a second. If the inrush current/power is short-lived, it will show up on one circuit, but could be gone by the next circuit’s sample time.
To reinforce what @frogmore is saying, try zooming out to a period greater than one-hour. The intervals will go from 5 second averages to 10 seconds or more as you zoom further out. At some point, probably early on, those downspike differences should disappear. IoTaWatt isn’t an oscilloscope. At best it’s data resolution is about 350-400ms. It’s just not physically possible to store all of that data for any length of time, so it strikes a balance by saving the time weighted average every 5 seconds.
That makes sense. Thanks for the explanation!
Purely curiosity, is that a “too much data to fit reasonable lengths of time on a SD Card”, or “too much data to transfer to the SD Card in realtime”? Or something else altogether?
Both I think. We were talking going to finer granularity. A years worth of 5 second data is just shy of 2 GB. Going to 1 second would be 10GB. Fat32 max file size is about 4GB.
With respect to writing it out, there isn’t a whole lot of extra processor time left and it gets harder to do more without impacting sampling.
Not to mention when people also want to fit uploading it elsewhere as well, right? And the vast majority of us aren’t monitoring anything sensitive enough to need that granularity 
I’m impressed with the year-long size, I’d have thought it would be larger. Nice work 
A lot of this is dated and a lot of it is timeless. Mine is a hardcover first edition, bought new about 50 years ago. The Art of Computer Programming. Volume 3. Sorting and Searching. Second Edition : Donald E. Knuth : Free Download, Borrow, and Streaming : Internet Archive
In a past life I wrote a mainframe sort utility and file reorganization software. The data organization and searching is much more advanced on modern servers, but IOT devices have pretty much the same kinds of memory and storage constraints that mainframes had in the 70’s and 80’s.
