IoTaWatt and emonTx Accuracy

Tags: #<Tag:0x00007f6e035ac8f0> #<Tag:0x00007f6e035ac828>

As resellers of the IotaWatt we would like to make clear our view of the accuracy of the IotaWatt.

The majority of users will likely get very good results from IotaWatt and we believe it is a great product.

Many of the points raised on the forum regarding limitations of the derived voltage reference for 3-phase monitoring, CT sensor, AC-AC adapter and other component tolerances are also true for the EmonTx and EmonPi. We are open about this.

E.g in our discussion on accuracy earlier the year and linked emontx learn page: OpenEnergyMonitor 9th of March 2018 meeting summary

Without calibration an IotaWatt using the same CT sensors and ACAC Voltage adapters that we supply through our shop are subject to the same errors resulting from these components as the EmonTx is.

The IotaWatt does have a higher resolution ADC and higher frequency sample rate, this is an improvement on the EmonTx, but its not possible for these factors to reduce the error associated with the sensors & components.

Error’s associated with sensor and component tolerance can be reduced with calibration. Both the IotaWatt and EmonTx can be calibrated to improve accuracy:

Other factors that can impact overall system accuracy include but are not limited to:

  • The amount of time spent sampling each CT channel per reading. The IotaWatt samples each CT channel for more time that the standard EmonTx discreet sampling firmware but less time that the EmonTx Continuous Sampling firmware. The amount of time spent sampling each channel on the Iotawatt is proportional to the number of CT’s and other activity such as WIFI.

  • Whether the unit calculates cumulative kWh at the ‘source’ (the unit itself) or at the destination, e.g: emoncms. Calculating cumulative kWh at the source removes errors that result from communication outages or any emoncms server downtime. The IotaWatt calculates cumulative kWh at the source, while the standard EmonTx approach to date uses the latter. We are working however to move to calculation at the source as part of the EmonLibCM firmware.

  • Use of derived voltage reference in 3 phase monitoring: Because the voltage of only one phase can be measured, the sketch must assume that the voltages of the other two phases are the same. This will, in most cases, will not be strictly true (see our Learn 3-phase page for explain why), therefore the powers calculated and recorded will be less accurate. Many users will likely find this approach sufficient, we use the same approach for the non-hardware modified emontx three-phase firmware. However, as @dBC highlights there can be situations where the error can be significant. Using per phase voltage measurement as supported by the IotaWatt ‘Direct three-phase’ can help mitigate this.

The IotaWatt and EmonTx are both designed for indicative energy monitoring, neither are designed to be revenue grade. We would like to improve on our understanding of accuracy over time as well as providing options for those keen to obtain higher accuracy measurements such as higher accuracy CT sensors and voltage measurement options.

Both the emonTx and IoTaWatt have merits for particular applications that go beyond a discussion of accuracy e.g

IoTaWatt advantages over emonTx:

  • Number of CT channels
  • Ease of setup
  • Support for multiple CT sensor models
  • User friendly calibration process
  • Over-the-air firmware updates
  • Support for multiple AC-AC voltage sample source

emonTx advantages over IoTaWatt:

  • Support for optical pulse sensor
  • Support for multiple temperature sensor
  • Battery operation option
  • Can be powered by single AC-AC adapter (additional 5V DC adapter not required)
  • Multiple emonTx can be used with a single emonBase which can also support emonTH remote temperature sensing
  • WiFi signal not required at consumer unit

Having used the emonTx at a number of sites now to review energy usage on the total site and individual circuits, it certainly delivers everything promised. It was never promised to have 0.5% revenue grade accuracy - equipment like this with cloud data logging and the features of emonCMS costs over $1000 and needs an electrician to install. I always do a verification check against a Fluke 175 (1% voltage, 2.5% current) and Fluke i200 (2.5% current) and the result is within specification. The strength of the emonTx is easy installation, battery operation, flexible, and the ability to analyse what is using the bulk of the power and when it is being used as well as how PV and the site load are interacting. Unbalance in three phase voltage has not been a problem.


If you have recorded values, or even just have a feel for the numbers, an indication of both the “usual” voltage imbalance and the worst case would be useful information. I appreciate it’s clearly dependent on the strength of the supply, and that’s wholly within the control of the owners of the distribution network. (And in the UK at least, you can never get them, even privately, to commit to other than limit values.)

Only just! They almost lost control of that here with widespread penetration of single phase PV systems; the LV side of the 11kV distribution transformers was starting to look like the wild west. They’ve taken back control though with new regulations. Inverters now must either output at a fixed PF of 0.9 lagging, or implement a volt-var response as per:

What I meant was, it’s not anything the individual user can control (even though, as we well know, en masse they clearly do).
It’s interesting that they call for a leading power factor to help push the voltage up, but they limit it to 0.95.

As an example, a job I was at today was 240.8 volts A-N, 238.8 B-N, 240.4 C-N, 415.9 A-B, 413.7 B-C and 415.6 C-A. The trusty Fluke read 239.8V when the emonTx said 241.8 V. This was a 3 phase supply to a motel in the middle of predominantly single phase urban. I’ll do a comparison between the three single phase revenue meters and the emonTx in a few weeks (the phases loads are very unbalanced). As a broad generalisation, I find the skinnier parts of the network with big phase unbalance have predominantly single phase loads - I certainly wouldn’t be surprised to see >20 volts variation between phases.

Here’s one Energex published a few years back. This is averaged out over a week. You can see that yellow phase is doing more than its fair share of the evening peak:

And here are some one-shot numbers I captured in a central Auckland commercial district (Parnell) a few years back:

1: 230.67V
2: 241.08V
3: 238.61V

1 Like

After 4 weeks of emonTx logging at this motel, results are in:
L1 revenue meter (0.5M) 4272 kWh, emonTx 4343 kWh, 1.6% error.
L2 1694 kWh vs 1717 kWh, 1.4% error
L3 1794 kWh vs 1821 kWh, 1.5% error.
Overall 1.6% error.
Highest loads seen L1 20.3 kW, L2 17.8 kW, L3 18.7 kW.
emonTx and 100A clip on CTs were straight from the box ie no change to factory calibration constants.
Only a one-off test, but I’m extremely happy with this level of accuracy