The perils of being a tech innovator, working with a tech innovator

Here’s the detail of what we found, and what we’ve done, from Phil Steele, our Future Technologies Evangelist.

My Conducted Emissions are painful

Any electrical device can create something called a Conducted Emission and in turn be susceptible to Conducted Emissions.

If we look at how a motor works for example, the central rotor spins due to a magnetic field created by passing a current through coils of wires wrapped round the axle which then interacts with permanent magnets mounted on the outside to force it to rotate. These coils are fixed on the rotating part so to get the electricity to them there’s a set of brushes that connect to a set of small pads (commutators) on the axle. As the axle rotates these brushes come into contact with a different set of pads each time before then connecting with the next set and so on. As these brushes connect and disconnect, like you might have occasionally seen a bit of a spark when the toaster pops, you get a similar jump each time. With the right equipment connected to the supply you can see this jump on the electricity supply and as it’s not particularly desirable it’s referred to as noise.

I’ve massively simplified here but this is the general concept and sufficient for the rest of this article.

If the axle is rotating say 10 times per second and there’s just one pair of pads, then this noise has a frequency of 10 Hertz and a graph would show 10 jumps of noise each second.

Speed that up and you’ll get faster frequency noise - say 100s or thousands per second.

This noise is called Conducted Emissions and if there’s too much, it can cause other things to behave differently than they were designed to - in fact a good example is an old transistor radio; these are susceptible to noisy supplies and you can actually hear a buzz sometimes.

Because you can never wholly eliminate noise, there are a range of industry standards( <150kHz range as per IEC 61000-2-2) that dictate how much noise a device is allowed to create and also how much noise a device should be able to cope with. In this way, electronics are designed to both minimise the amount of noise they generate (you can never completely eliminate it) and also to cope with noise.

I’ve used the example of a motor, as the spinning axle is the easiest way to explain how noise is generated and how the frequency it occurs at can vary massively from 10s of Hertz to 1000s of Hertz (speed it’s rotating and if there’s multiple coils on the rotor and therefore several pairs of pads). But power supplies that convert 220v AC to 12, 24 or 48v DC also cause noise; your laptop, mobile phone, the one built into your TV, even those in dishwashers and washing machines. Vacuum cleaners have high powered motors so can be problematic too. The electronics are powered by Direct Current (DC) so have to be converted from the mains supply Alternating Current (as well as being reduced from 220 volts to either 6, 12, 24 or 48 volts). This conversion is done using a rectifying circuit and over a century ago this used to be done with a rotating mechanical device.

So there’s a similarity here with my motor example; as the AC is converted to DC in a modern power supply it too may put noise onto the electricity supply.

Rectifying circuits also have a lot of advanced circuitry to further smooth the DC output, as well as ensure they’re immune to any noise (those microchips want as flat and precise a voltage as possible). Although mains supply 220v AC runs at 50Hz the advanced complexity of smoothing to DC means it’s possible for noise to be generated at all sorts of frequency ranges as the rectifying circuit modifies the DC output.

Aside: one day with all our lights being LED, all our devices running on 12v or 24v DC (TVs, etc) we really should run our homes on DC - that’s a topic for another day.

As you might imagine a device that has higher power demand risks generating a higher noise, so for example the AC to DC conversion for home battery storage requires a more substantial rectifier and greater potential for noise generation.

And this is where we’ve recently seen issues with Tesla Powerwall batteries. During lock-down we worked with Tesla to install a Powerwall in one of our partner’s labs in response to a small number of customers reporting issues. Like your regular AA battery, all home storage batteries work on DC too so have an inverter (a type of rectifier) that changes the energy between AC and DC.

As we’re the only energy supplier to use the half hour smart meter measurements and also the only energy supplier to make these directly available to customers it’s possible to see the effect on the account dashboard and to compare that to the equivalent readings from the Tesla Powerwall app.

Without being an expert just seeing a graph with 0.25kWh consumption every half hour raised warning flags when I first started to see these. This particular customer is charging their EV overnight and has a Powerwall storing solar generated energy and through the day the majority of the home consumption is supplied by the solar and in the evening the Powerwall. This screenshot is for a day the family were out so there should’ve been no grid import other than the EV charging and yet we can see that 0.25kWh every half hour. The Tesla Powerwall app readings were showing zero.

As Greg’s referred to elsewhere working with smart meters is a learning curve but we are dedicated to getting right. Without smart meters we wouldn't be able to offer award winning tariffs such as Agile, Go, Outgoing and our Tesla tariff.

In the lab tests and when we visited a customer’s property we were seeing high frequency noise on the mains supply originating from the Powerwall under certain conditions (usually when idle, but not when charging/discharging even at low load). But when we started looking at it and running tests this noise is well within the limits imposed by the Conducted Emissions standards and therefore conforms to the sub 150kHz range as per IEC 61000-2-2.

The reports I received from customers concerned L+G smart meters so we ran a set of tests in the lab with L+G and a range of other smart meter manufacturers for comparison and interestingly we were only able to replicate the issue with one specific L+G smart meter model.

L+G spent time investigating this and in fact myself, Tesla and L+G visited a customer’s property to see the issue in situ.

As well as the Conducted Emissions standards there’s also a set of standards for how smart meters should measure energy referred to as being MID compliant (Measuring Instruments Directive). This mandates that all smart meter manufacturers go through stringent test and verification processes to ensure they meet these standards.

EC Directive 2014/32/EU

So both the Powerwall and the L+G meter have passed their respective tests.

If we look at the graph from the tests we carried out with the high frequency imposed on the AC sine wave, depending if you zoom in or out it appears either as solid thick line or you can zoom in and see the fast cycles. If I want to measure the overall sine wave what should I do?

A noise with a high frequency just looks like a very thick line and I can’t work out if there’s noise there or not. In the same way a smart meter measuring that is effectively fooled by the noise and so measures the effect of the noise as though the actual signal is higher. And this is exactly what’s happening between the Tesla powerwall and the specific L+G smart meter model.

It’s an extremely rare scenario to have a Powerwall, an L+G smart meter and nothing actually connected to the meter (in fact I’d go as far as saying this is only something you’d see in the lab) but it makes you pause and think if this set of conditions causes an error is that why we’re seeing customer issues.

Lockdown has hindered us getting to grips with this particular issue given we couldn’t visit customer homes to monitor the issue (the one home we visited was an employee) and had to wait until we were able to install equipment into a lab and run tests.

Our conclusion is that both devices fully comply withtheir respective standards,however there is an incompatibility between them which is only detectable under the rare conditions when the L+G meter doesn’t have a load and is exposed to the noise from the Powerwall . We have concluded that it’s just the one L+G meter model and so far only with the Tesla Powerwall. We’re yet to find any other battery system experience this issue (we’ve tested one other manufacturer in the lab too) although this may be because Tesla is simply the most popular and well established nor is there an issue with any other smart meters (we’ve tested 6 others in the lab) or other L+G models.

Now we’ve solved the riddle we’re changing Octopus Energy customers that have a Tesla Powerwall and the L+G smart meter by changing their L+G meter for an equivalent L+G model.

We’re keen to stress this issue is affecting a very small number of customers and is only related to the combination of the Tesla Powerwall and the specific L+G smart meter model but unfortunately there’s been a fair amount of public discussion more generally referring to all home battery systems and all SMETS2 smart meters so we’re sharing our investigation in full and what action we’re taking.

If you think you’re affected by this feel free to contact me directly via agile@octopus.energy or find me on Twitter as @agile_phil and we’ll take a look at your Octopus account (we need to look at the half hour data over a few weeks) and investigate and resolve it if you are affected.