IoTaWatt with commercial energy monitoring

Homebrew IoTaWatt
1

Hi,
I have just installed an IoTaWatt setup, and am generally very happy with what it’s doing! I’ve got a few more CT’s to purchase to cover the remaining circuits, but having a bit of a background in energy measurement, I was thinking that it might be interesting to do an upgrade.

The issue that the existing setup has is twofold:

  1. the processor is a touch under-powered, which means that we’re not sampling everything, and making assumptions.
  2. the ADC is a relatively low resolution (12 bit on a 100A CT will give a quantisation step size of ~25mA or 6W), which is ok given the CT’s that are used, but if you’re interested in the low power setup, then that’s a bit of an issue. I’d love to see accuracy in the 10’s of Watts, which means resolution to ~100mW approx.
  3. the CT’s only have a single calibration point (and as they are cheap, the low power measurement is relatively poor, as the Phase angle goes wonky when there’s not much current). It’d be a pain to calibrate much more so I’m generally not a fan of CT’s

I was thinking that a good upgrade may be to work with something like the ADE7816, which (whilst not cheap) allows for offloading all of the metrology functions, and will monitor every sample…
*Full disclosure: I work for ADI, so that’s where I’ve looked for a suitable product.

If we were to do this, then we could also move to a Rogowski coil, which should allow a much better linearity over the range.

The reason for making these changes are two-fold.

  1. I’m interested in understanding if I can make a Home battery (e.g. the Tesla power wall) pay, and if I can, what size should I select.
  2. I’m interested in running some NILM to try and disaggregate the power. For this getting second-by-second (or cycle-by-cycle) RMS readings, with phase-angle would be great.
    There are some nice algorithms in the NILM-TK that if I can get to understand better, may then be able to be made to run on the IotaWatt, as it will have much more bandwidth for the processing available, as it’s not sweating to do the metrology.

Is anyone interested in this? I’m highly likely to not get round to this for a few months, as I’m still finishing building my house, but if there’s interest more than myself, it’d be great, as I’m not much good with programming in node.

3 Likes
2

I would definitely be interested, I came across ADE7816 not that long ago and thought it would be great to build a multi circuit monitor around it.

3

It would be nice to have 4 those ADE7816 chips and 1+6 RJ45 connectors in box. In STP cable you could put 4 inputs interleaved to each chip. 1 connector for VTs, 6 for CTs. You could make (short) cable RJ45 - 4* 3.5mm for compatibility, or make custom cable like RJ45 - 4* SCT006. For VT make RJ45 - 4* barrel and 1to2 barrel adapters, so you could make single phase, split phase or 3 phase. With 3 phase there is 1 ‘extra’ input in each cable…

It would be quite stright forward to connect ships with spi (3+4 pins) but then you can’t use hsdc (high speed data capture) which works only when i2c is used. Using i2c is problematic because chip uses fixed address, so you need 4 i2c bus (4*2 pins) :frowning: And then also pins for hsdc (3 pins). Hsdc is activated via i2c so all chips can rest at same bus. So to make things work, you need to add i2c multiplexer (like TCA9544A) which makes program even more complex. Then only 2+3 pins would be enought. Maybe even add interrupt line :wink:

Anyway, would be great to get really small box with trulu accurate measurements :slight_smile: