UPS on the cheap: Ecoflow River 2 versus River 3 (non-plus)

This blog post compares UPS transfer times between Ecoflow River 2 and River 3 (non-plus). That's it. That's the intro.

Test objectives

I'm running an Ubiquiti network setup with two separate segments:

• The "upstairs" segment running ~70W of equipment (Starlink, Ubiquiti Dream Machine, a switch, an AP).
• The "downstairs" segment running a few switches and APs for a total of ~40W.

Living in rural Colorado (kind of), I'm facing regular power outages lasting from a few seconds to a few hours. Those 2-second outages are the worst: the power comes back before you realize what's happening, but the network stack goes down (along with Starlink and a lot of IoT devices I built for home automation), and the whole system takes ~20 minutes to reboot and stabilize. Anyway, I want to power both segments during power outages for up to ~3 hours, and I don't want to lose any network functionality when transitioning to or from battery power.

My late-2024 research found no name-brand devices (Eaton, Vertiv etc.) with pure sine wave output below $500. However, as an avid camper, I've been using Ecoflow power stations for quite some time, and I've heard good things about them being used in EPS (emergency power supply) and UPS (uninterruptible power supply) scenarios. I'm not going to go into the differences between EPS and UPS here, but it has to do with how quickly the device can switch (transfer) from grid power to battery power.

My use case has no exact transfer time requirements, all I need is for my devices to stay alive during grid/backup switchovers. Your average PCs or network devices will be happy with a <30ms transfer time, which is approximately two YAC cycles (Yankee Alternating Current, 120V/60Hz).

Ecoflow River 2 and River 3 (non-plus)

I already own multiple River 2 power stations to support my camping trips, and I just purchased two River 3 (non-plus) we'll use for comparison here. River 2s advertise <30ms transfer time, whereas River 3 (non-plus) power stations claim <20ms switchovers. River 2 and River 3 power stations both provide pure sine wave AC output, and they use a "bypass" approach for switching between grid and battery power:

• When the power stations is connected to the grid, AC outputs pull directly from the grid bypassing the battery. The inverter stays on to help provide faster transfer times.
• Whenever the power station switches between grid and battery power, it doesn't do so seamlessly. The time needed to flip power relays (or whatever it is they use) creates a gap in the output AC waveform. This is commonly referred to as the "transfer time".

I will skip over the theory behind power supply topologies and why for some of them the transfer time matters more than for others. The bottom line is that, for a 60Hz AC wave, you skip one cycle every ~17ms (1000/60), and most low-power electronics are perfectly capable of skipping a couple of cycles thanks to the capacitance of the device's power supply. The point of this post isn't to argue about the right/correct/minimum acceptable transfer time for any specific device, I just want to show the differences between two generations of Ecoflow devices I own.

Last but not least, River 3 power stations claim to improve upon River 2s in more than just the transfer time. River 3s use GaN circuitry as opposed to River 2's old-school silicon stuff, which reduces heat waste and increases inverter efficiency. I'll briefly touch upon the practical differences later in this post, but it's not my main focus here.

The test

The test setup is as follows:

• I'm comparing Ecoflow River 2 Max (512Wh) to River 3 non-plus (245Wh).
• The power station is connected to the grid through a power strip. The power strip's rocker switch is used to turn grid power on and off (to simulate power loss and recovery).
• The power station is under a constant 170W resistive load during transfer tests.
• I use a Siglent SDS1202X-E oscilloscope with a Micsig DP10013 differential probe to observe the output waveform and measure transfer time. 
• I repeat each transfer test two times. That's four readings for each power station (two grid->battery tets, then two battery->grid tests).

Without further ado, here are the results:

Ecoflow River 2

When switching from grid power to battery power, the River 2 delivers a ~16ms transfer time:

When switching form battery to grid power, we're around ~20ms:

Ecoflow River 3

When switching from grid power to battery power, River 3 performs significantly better at ~9ms transfer time:

When switching form battery to grid power, we're at ~7ms:

For context, transfer times under 10ms are well within the traditional UPS territory. Pretty much the only way to improve from here is to use an "online UPS", which is a type of device where the output power comes from the battery at all times, effectively creating a transfer time of zero milliseconds. However,  you'd be hard-pressed to find a consumer product that won't tolerate a ~10ms transfer time (including most PCs and CPAP machines), so the results we've seen from River 3 are more than adequate for household use.

Inverter efficiency

I also ran a simple efficiency test like so:

1. Fully charge the power station.
2. Discharge at 30W through the DC output (the car lighter socket). All outputs are lossy, but the car socket is likely the least lossy one. Record discharged capacity in Wh.
3. Fully charge the power station again.
4. Discharge at 30W through the AC output.

This test isn't exactly correct since we're comparing power stations with different capacities. The larger device (River 2) spends more time with its AC output on, which means we're incurring inverter losses over a longer time period. It is what it is, I don't have a regular (non-Max) River 2 to test ¯\_(ツ)_/¯. 

Note: I've had the River 2 for a while, whereas the River 3 is almost new. Still, I'm estimating that the River 2 has less than 100 cycles on it, which should count as "almost new" for a LFP pack.

Here are the numbers:

• River 2, rated at 512Wh, delivered 425Wh via the DC output (83% of rated cap.) and 330Wh via the AC output (64% of rated cap.). This means that the inverter delivered ~78% of the energy we pulled from the DC output.
• River 3, rated at 245Wh, delivered 223Wh via the DC output (91% of rated cap.) and 177Wh via the AC output (72% of rated cap.). This means that the inverter delivered ~79% of the energy we saw at DC.

This is... pretty underwhelming. To be fair, Ecoflow only advertises higher inverter efficiency at super low load levels (single-digit watts). This makes sense because for any fixed amount of overhead, lowering the load increases the relative overhead as a proportion of total load. For example, a 3W overhead with a 300W load is nothing, but a 3W overhead with a 3W load is half the total load.

I don't currently have a use case for continuous loads in the single-digit watts, so I'm not going to test that right now. Let me know in the comments if you've done any tests below 10 watts, I'd be curious to hear about your findings.