EmonTX3 vs EmonTX4 accuracy (also vs SDM120/SDM230 or inline meter)

So my original EmonPi in 2015 is still going strong, and over the years I have added EmonTX3 in the garage and about 15 months ago a EmonTX4 with the improved voltage reference to the main house CU to monitor the main ring mains and the immersion more accurately.

Exciting times ahead, as with 0% VAT on battery I am finally adding storage to my system to hopefully work with Solar and increase self consumption of green energy.

This means I will need extra circuits in the garage to be monitored, plural as during the work my existing inverter will be moved from the loft space into the garage (to stop overheating in summer and prolong its longevity) as well as the battery being added.

I have an EmonTX3 I can add to do this, although I need to buy another AC adapter (so it can read negative and positive values) as they seem to expire after about 4-5 years (been through 2 already).

An option is to move the EmonTX4 to the garage, and a TX3 to the house instead. I understand the EmonTX4 is more accurate? Measuring large loads like 5kW solar, 7kW EVSE or 5kW battery though I am wondering if that refinement is lost and the EmonTX3 is perfectly capable?

Another option, as there is electrical work happening, is to install some Easton meters into the fusebox, I think there are just enough ways to add in front of Solar and Battery and use RS485 to monitor. When I last looked at this many years ago this was in its infancy with EmonCMS but I see they are for sale in the shop and threads on here and modules in GitHub so it’s now well supported?

Head in a spin as multiple permutations here, can anyone give any insight into accuracy between the 3 options that might help me narrow it down bit?

Hello Paul

In terms of worst case component tolerances including importantly the tolerance of the older emonTx3 CT sensors and ACAC adapters the error could have been up to +11.25% -10.33% Sources of Error in the emonTx Voltage and Current Inputs — OpenEnergyMonitor 0.0.1 documentation In practice the performance we were seeing in comparisons with class 1 meters was much better, especially when well calibrated e.g closer to the 1-3% range.

The big step forward for the emonTx4 was to choose higher quality CT sensors and to develop the higher precision emonVs, bringing that worst case error to 1.2% Overview — OpenEnergyMonitor 0.0.1 documentation

In practice e.g on my long running test here the emonTx4 20A CT sensor for my heat pump is currently showing 0.5% difference to an SDM120 over 18 months and 0.3% over the last 6 months. 0.24% over the last month.

I found a PM message from 2019 with a test result for the emonTx3 vs an Elster A100C class 1 meter that I was using at the time, I’ve summarised the main points from that message here:

The total consumption on the Elster was 809.1 kWh and EmonTx3 792.4 kWh, which means the EmonTx3 was under reporting by 2.1%. The error varied through the year depending on how much the heatpump was running. From Jan to Apr 2019 the difference was 508.3 vs 501.2 a 1.4% difference and again in the last 11 days of this window, with a lot of heat pump heating again 55.48 vs 54.7 (1.4%)

The error was much larger for the period of very low power consumption where the heat pump was basically on standby for 5 months (6.3%). I think this was pre installing the DHW cylinder. This is consistent with expected lower accuracy at low power levels expected from the EmonTx3 hardware. The power consumption on standby is 11W!

In mid January 2019 I ran the heat pump continuously at ~514W electric input for 80 hours. The Elster meter reported 41.7 kWh of consumption and the EmonTx3 41.4 kWh a 0.72% difference.

The EmonTx3 was running the stock EmonTxCM firmware with the default calibration, given the tolerance on the sensors and components the error is well within range.

In tests that we’ve made between continuous monitoring and discreet firmware on an EmonTx3 the difference has been 0.45% vs 1.48% or 1.73% vs 2.24% respectively for two different systems monitoring house consumption.

The person I was discussing the above with was running the discreet sampling firmware and had an error of 3.2% over ~ 100 kWh and I recommended trying the newer continuous monitoring firmware at the time.

The emonTx3 CT sensors have a tolerance of ±3% and ACAC adapter ±3%. Whilst the newer CT sensors for the emonTx4 & emonPi2 have a tolerance of ± 0.5% and emonVs ±0.5%

Thanks Trystan,

I think on reflection I am going to add my spare EmonTX3 in the garage and use that for battery storage and the relocated inverter. I believe the powerwall has an integration so I can calibrate the sensor to the powerwall’s reading.

I did this for my solar, which is currently on the EmonPi Power2, as I can see in the emonhub.conf I have a scale value of 1.095 (derived from comparing the SMA reading via bluetooth to the EmonPi CT reading and adjusting to align).

I have a Easton MID meter on my EVSE, but I have never connected the RS485, I will have to add a Pi to the garage so I can read that and compare the CT reading.

With the add on board the shop sorted me out with I can then monitor more loads in the house more accurately (adding a clamp for Fridge and Oven has been something I always wanted to explore.

1 Like