Three-phase automatic transfer switch controller PEF-321AVR from Novatek-Electro
It currently costs 5700 UAH, or about $200, but there was a sale not long ago when it was available for 5130 UAH ($185), so it may be worth negotiating when you buy one.
In Russia, it costs 14800 RUB, and in Belarus 489 BYN.
An ATS can actually be a very simple device. In the most basic case, even a dedicated switch can do the job. There are also simple circuits with a pair of contactors, and for those who don’t want to build one themselves or spend a lot of money, Chinese manufacturers have started making inexpensive mechanical ATS units.

With some caveats, I’ve used the PEF-301 phase selector as a single-phase ATS. It has three inputs with priority selection and upper and lower voltage limits. The caveat is that, if you want to do the job properly, when powering from two different substations you need to switch the neutral conductor as well, and monitor the voltage on each input independently.
I tried to explain it in simple terms, but it’s not really working, so I’ll move on to the device under review and explain things as I go.
So, as usual with Novatek, we have a neat little box with the device name and the manufacturer’s address on it.

Inside are the ATS controller, a USB cable, and a fairly substantial manual.

Besides the technical specifications and descriptions of the operating modes and settings, the manual also describes the MODBUS commands. As with many of its products, the manufacturer provides a 10-year warranty, according to the manual.
The included cable is a standard USB-to-microUSB cable.

You can download the full manual here. As usual, I’ll quote some excerpts from it, starting with the technical specifications.

Let’s take a look at the controller.

The enclosure is fairly large, 9 modules wide. It looks like the same enclosure used for the RPM-416 data logger I showed earlier. Well, standardization is useful.
There’s a serial number sticker and a spring-loaded latch on the bottom.

Overall dimensions of the device.

There are 25 terminals on the bottom for the output signals and relay contacts. The RS485 interface connections are also brought out here.
1–3. Control of input 1 (changeover contacts)
4–6. Control of input 2 (changeover contacts)
7–11. Not used;
12, 13. Voltage status on input 1;
14, 15. Voltage status on input 2;
16, 17. Voltage status on both inputs;
18, 19. Generator operating status;
20–22. Generator control (changeover contacts)
23–25. RS-485
It’s interesting that terminals 12–19 are only numbered on the device enclosure, with no icons to identify them. This can be very inconvenient in some situations (if something goes wrong, you have to find the manual, which, as always, is in the farthest drawer).
As with Novatek’s other devices, it uses high-quality “elevator” type terminals.

The diagram shows the information output modes for terminals 12–19. These are handy for connecting indicators, for example.

There are 15 terminals on top for measuring the input voltage, measuring the output voltage after the contactors, and connecting a battery.
26–28. Phases 1–3 of input 1
29. Neutral of input 1
30–32. Phases 1–3 of input 2 or generator connection
33. Neutral of input 2, for powering the device or connecting a generator;
34–36. Phases 1–3 for measuring the load voltage
37. Not used;
38. External button for powering the device from a battery;
39. “+” 12 V battery supply;
40. “−” 12 V battery supply.

Here’s the complete wiring diagram. I should point out that the device is designed to switch contactors with five contact groups: four power contacts that close the circuit and one auxiliary contact that opens it.
I decided to draw attention to this because when I tried to find these contactors for sale, it turned out that buying them ready-made is quite difficult, and the problem is specifically the fourth normally open power contact.
To reiterate, the neutral needs to be switched if you have two separate neutrals. If the neutral is shared, you can use 4-pole contactors (3+1).
In this case, the contactor's normally closed contact is an emergency contact, used to block the other contactor if the ATS relay sticks.

The front panel has two indicators, one for each input, a cluster of LEDs, three control buttons, and a USB port for connecting to a computer.

Controls and indicators.

Before testing it and describing how it works, let's take a look inside this little box. The cover comes off very easily: there are two clips on the sides, which I simply released.

The device consists of five circuit boards connected by connectors.

The layout is very compact. Keep in mind that the board carries not only high voltage, but also two independent inputs and measurement circuits, so safe clearances have to be maintained.

Some of the vertically mounted components are insulated with heat-shrink tubing.

The top board is secured with plastic screws through spacers and electrically connected to three vertically mounted boards.

The vertical boards have a fairly complex shape, also dictated by the compact layout.

The top board carries the microcontroller that handles all control functions, the indicators and buttons, the battery disconnect relay, and several voltage regulators.
An Atmega32A is used as the main microcontroller.

Two boards measure the mains parameters and send the data to the main microcontroller.

The small side board can't be removed, and I didn't desolder it.

The device fully disassembled.

The bottom board carries the input voltage dividers, power supplies, a bunch of varistors, and the relays that control the contactors and generator.

There are two power supplies, each powered from its own input. Their main outputs are connected in parallel, providing fully independent power.
Each power supply has two input capacitors: 10 µF, 400 V, connected in series, giving 5 µF at 800 V.
They are built around TNY266P controllers. Since the device is designed for a higher input voltage, voltage limiters are installed ahead of them. Each power supply has three outputs: 12 V for the device's electronics, 24 V for charging the battery, and a separate output, galvanically isolated from the others, to power the measurement module for the corresponding channel.

1. The battery circuit has a 1 A fuse. Next to it, on the underside, is a simple 60 mA current regulator; the voltage is limited only by the power supply.
2. The board has a bunch of varistors rated for 460 V (Vrms). You can also see resistors covered with heat-shrink tubing nearby. Just to be sure, I checked that the tubing insulates not only the outside, but also protects the resistor's internal vertical lead.
3. There are also six X capacitors, three for each input, and more resistors covered with heat-shrink tubing. Together with the 75 and 150 Ω resistors, the X capacitors form an RCR filter that protects the inputs of the internal switching power supplies.
4. In some places the layout is so cramped that a few components are mounted diagonally.

As I mentioned, one of the auxiliary boards is soldered onto the bottom board. In my opinion, this makes repairs less convenient, especially since the board carries chips that communicate with the “outside world”—and, as we know, they sometimes burn out.
1. On the left is an RS-485 transceiver and some simple logic.
2. On the right is a UART-to-USB converter based on the CH340G.
Galvanic isolation is present, but not between the “microcontroller and converter chip”; instead, it’s between the measurement section and the main microcontroller. With such a compact layout, though, it would be safer to have isolation between the microcontroller and the converter chip.

1. Auxiliary signals are switched using a pair of PVT412 PhotoMOS relays with MOSFET outputs (up to 400 V AC/DC) and a pair of MOC3063 optotriacs. They switch AC only and include zero-cross detection.
2. A pair of PE014012 relays made by TE Connectivity controls the contactor outputs. The same relays are used in the battery circuit and the generator-start output. The contacts of the relays controlling the contactors and generator are simply routed to the output terminals and aren’t connected to any other circuits, which is convenient.

There are plenty of components on the underside too, but they’re laid out more spaciously.

Interestingly, the documentation says terminals 7–11 are unused, but they are used on the board. Judging by the resistor and (presumably) TVS diode protection near the terminals, and the RCR filter near the microcontroller, these appear to be low-voltage digital inputs.

There are plenty of creepage clearances, both around the terminal blocks and beneath the divider resistors and Y capacitors.

But the SMD soldering quality... leaves a lot to be desired. I got the impression that the through-hole components were installed and soldered in an oven first, and the SMD parts were then soldered by hand, although it’s usually done the other way around. The joints are solid, and they didn’t skimp on solder, but the finish is a bit rough.

I didn’t draw a full schematic, but I sketched a block diagram to show the device’s general operating principle.
Inputs 1 and 2 are on the left; on the right is the “output”—or, more precisely, the terminals where the voltage is supplied after the contactors, i.e. the load.
The device contains two independent switch-mode power supplies and two independent measurement boards, each powered by a separate transformer winding that is isolated from the power supply for the monitoring and control section. Data from the measurement modules is sent to the controller through optocoupler-based galvanic isolation, fully isolating the measurement circuits connected to the high-voltage circuits from the monitoring circuit.

How it works, plus a few tests.
To start, I connected one input, with all the phase input terminals connected together.
Naturally, the controller immediately gave me a telling-off, periodically displaying the input voltage: the first channel showed around 220–230, while the second showed zero.
Power consumption from the mains in this setup is around 2–2.5 W, but that’s without a battery connected. Charging the battery will require about another 1 W, keeping the total within the stated 4 W consumption.
At startup, the displays show “Sta”, followed by the current phase voltage for phase 1 of inputs 1 and 2, respectively. The left/right buttons let you choose between displaying phase voltage (photo 4) or line voltage for each phase on both inputs simultaneously. You can also display the frequency (photo 5) for both inputs.
There’s also a second display mode, enabled in the menu, where phase or line voltages alternate automatically on the displays every 5 seconds. The buttons can be used to show all the values on demand, but after 15 seconds the device returns to automatic mode.
The LED indicators at the top show:
L1–L3: lit green when voltage is present.
I: lit blue when the output is active.
Triangle icon: lit red when there’s an input fault (for example, voltage outside the set limits) or an error warning.

The display modes are described below. The previous state isn’t saved: whenever power is applied, the device starts in the first mode. You can then cycle through the modes using the buttons.

The device can be controlled either through the software or from the front panel. The latter is noticeably less convenient, of course, but it’s perfectly usable, and some of the abbreviations are intuitive.
1–2. To enter the menu, press and hold the OK button for a few seconds. The device will then ask for a password; the default is 123.
3. The first menu item lets you select the input operating mode. The parameter appears on the left display, and its value on the right.
4. Next, you’re prompted to change the password, though in most cases there’s not much need to.
5–8. Some of the abbreviations aren’t very clear, which is a limitation of the simple seven-segment display.
9–16. But settings such as the voltage thresholds and under/overvoltage protection are easy to understand. The input number and minimum or maximum voltage are selected. In this case, the options shown are for setting phase and line voltage, respectively.

Since there are so many menu items, rather than cluttering the article with unnecessary photos, I’ll just attach the table from the manual listing all the settings and their default values.
Probably the most confusing item is the parameter for the number of automatic reclose attempts (automatic recovery), which can be set from 0 to 5. During testing, I ran into an “automatic reclose count error” several times and couldn’t figure out what it meant for quite a while.
From the description, you might think it changes the reclose time multiplier, since the parameter is configurable. But it turns out to be the number of fault-triggered switching attempts. If they occur more frequently than the reclose interval, an error message is displayed and the relay won’t switch back.
In other words, if an input is unstable, you can have the device make several attempts to switch to the priority input; if the problem persists, it will stay on the backup input. 0 disables this function.
If the next trip occurs after an interval longer than the reclose setting (60 seconds by default), the automatic reclose counter is reset.
I think the description would be clearer if the word “time” were removed from the automatic reclose count parameter, since it refers to the number of attempts, not the time.

For the demonstration, I set the count to 3 and adjusted input 1 (selected as the priority input). The overvoltage protection tripped, and I brought the voltage back to normal before it switched to input 2 (to speed things up). On the third trip, the relay forcibly switched to input 2 and stayed there.
If the interval between trips is longer than the reclose time setting, the counter resets and the relay won’t forcibly switch to the other input.

At first I wanted to buy suitable contactors with 230 V coils, but when I started looking for them, I realized it would be much easier to use ordinary relays with three contact groups.
The thing is, I don’t need to switch the neutral for this test, and I don’t need cross-interlocking protection either.
In the end, I used relays with three changeover contact groups and 12 V coils, powered by a separate power supply.
Once connected, this whole Medusa’s-head tangle was wired to the controller according to the diagram in the manual.

For the first test, the input of channel 2 was connected to the system output—that is, where the load is supposed to be connected. So the right-hand display shows the output voltage.
At first, the device complained about “phase overlap” because it was powered from a single phase, so I had to disable monitoring for this fault in the settings.
The manual says the device’s operating range starts at 100 V. In reality, it started up at around 35–37 V, but didn’t measure the voltage. When the input voltage reached 43 V, it began displaying 48 V, and only started measuring correctly once the input was around 47–50 V.
Raising the voltage further, up to 250 V, revealed a small measurement error of about 1 V.
At this point, I’d say the results are pretty good, since the input range is wider than specified. However, the range may be a little narrower when a battery is connected, since it would then need to be charged.

So far, I’ve tested phase voltage measurement, but I also wanted to check how it handles line voltage. In this case, line voltage was applied across two phase inputs.
1–5. Everything worked correctly here too, although the voltage measurement error was around 1–2 V.
6. As expected, the device complained about “voltage imbalance” and “phase loss.”
7–9. As mentioned earlier, the display can show either phase or line voltage for the inputs. Below are examples of what it displays when one phase is “lost,” with phases 1, 2, and 3 selected in turn.
As you can see in the photos, the relay doesn’t connect the output in any of these situations. You can disable monitoring for these faults too, which may be useful when having two phases is better than having none.

There were also a few “oddities” that, in my opinion, could be considered bugs.
1, 2. The input is at 180 V; we raise it to 190 V, and the relay switches on correctly, responding to the undervoltage protection.
3, 4. But I found it strange that the relay also connects the output for a few seconds when the voltage is below the lower limit. In other words, the relay is de-energized, we apply 100 V, the relay starts up and connects the output, then “sees” that the voltage is too low and disconnects the output.

Demonstration of auto-start at low input voltage. You can see that the input voltage is even below the specified minimum of 100 V, yet the device still energizes the contactor, which trips a few seconds later due to a "low voltage" fault.
Obviously, the last situation is unlikely at very low voltages, since a standard contactor simply won't energize, but it's still a bit odd. If you apply a higher input voltage, this doesn't happen and everything works correctly.

For the generator control system to work correctly—more specifically, for its startup and pause functions to operate properly—a battery must be connected. The manufacturer intends for the controller itself to charge it. When the battery is connected with no mains power available, the device can be switched on manually using the button between contacts 38 and 39.
In principle, everything works as it should, but there are a few caveats.
1. After connecting the battery with the device "cold," I found that it still draws a small current from it, around 3 mA. That's not much, but as always, there are a few "buts."
2. When power is applied, the battery starts charging at around 30–35 mA. The manual specifies "up to 60 mA." 30–35 mA is within that limit, of course, but the difference is rather large. That may actually be a good thing, though, because there's a second "but."
3. The second caveat is that only the current is limited; the charge termination voltage is almost equal to the power supply voltage, i.e. 24–25 V. Here, the low charging current is an advantage: you simply can't overcharge the battery at this current.
4. Until the generator starts, the device draws 85–90 mA. Once power is applied to input 2 (from the generator), the battery starts charging at the same current as before, i.e. 30–35 mA.
Now for the caveats.
The charger uses a primitive circuit with no battery voltage monitoring, and the current regulator is linear.
Since the current regulator is linear, it dissipates a noticeable amount of heat even at 35 mA: about 0.4 W under normal conditions and up to 0.87 W under abnormal conditions (if the battery develops an internal short circuit).
The device's power consumption is low, and it only needs to supply a small amount of power until the generator starts. A 100 mAh battery would be more than enough, but the high current draw while the device is off will discharge it quickly.
So, if I were the manufacturer, I'd make a few improvements:
1. Offer a battery as a DIN-rail module, with the battery itself, along with its protection and charging circuitry, inside the enclosure.
2. Design it from the outset to work with LiFePO4 batteries rather than lead-acid ones—or, better yet, with supercapacitors, especially since the required capacity is small, say 200 mAh.
3. Possibly revise the circuit so that after the battery is discharged, a relay completely disconnects it from the circuit.

A relay handles the battery connection. The charger appears to be based on an LM317 current regulator. There's reverse-polarity protection using a separate transistor, very similar to the familiar circuit with a MOSFET.
A while ago, I showed how to convert a standard switching power supply into a simple UPS. I suggested a circuit using a TL431 and a relay for over-discharge and reverse-polarity protection, with the relay coil powered from the battery through a diode. Here's an excerpt from that circuit; the section on the right shows this exact setup. I think this approach is even simpler.

I had no complaints about the device's thermal performance. It gets hottest when the battery is connected; the first photo shows the temperature after an hour. The second photo shows the temperature with the battery disconnected.

Inside, the battery charging current regulator gets hottest, reaching about 60°C, followed by the relay at 52–55°C.

There are four operating modes; I'll try to demonstrate them.
Mode 0: equivalent inputs.

If the power fails or goes outside the set limits, the device switches to the other input if it's working, and stays on it regardless of what happens on the first input.
For the demonstration: the voltage on input 2 is stable, while I simulate an overvoltage on input 1. If the voltage returns to normal before the switch occurs, the device switches back to input 1; but if it has already switched, it does not switch back.

Modes 1 and 2: input priority.

Here you can set input 1 or 2 as the priority input. In normal operation, the load will always be connected to the selected input. If that input fails, the device switches to the backup input, then switches back once the fault is cleared.
If the mains parameters on the backup input are also outside the configured limits, the load is simply disconnected.
For the demonstration: input 1 is the priority input. When its voltage exceeds the limit, the device automatically switches to input 2. Once the voltage returns to normal and the delay has elapsed, the relay switches back to input 1.

Mode 3: generator operation.

For this mode to work correctly, a battery must be connected to the ATS. The generator must also always be connected to input 2.
Here’s an example of what happens when the voltage on input 1 fails: the generator is connected to input 2, and its control relay supplies voltage to input 2 to simulate the generator. To save time, I reduced the generator startup and shutdown delays to the minimum of 5 seconds.
If input 1 fails, the ATS sends a command to start the generator and counts down the delay for the generator to reach operating mode (5–900 sec), after which the load is connected to it. The switch back works similarly: after the usual return delay, the ATS switches to input 1, then counts down the generator shutdown delay (5–900 sec), and finally shuts down the generator.
I should mention the well-thought-out time settings: the increments are 5 sec at first (up to 60 sec), then 30 sec (up to 90 sec).

The overvoltage response follows the same algorithm. The same applies to other mains faults—phase imbalance, phase loss, phase reversal, etc.—if the corresponding responses are enabled in the settings.

Modes 4 and 5: operation with the selected input only.
There is no automatic switching in these modes. You select the input you need, and the ATS only disconnects the output if the mains parameters on that input go outside the specified limits.
Example: operating with input 1 only. Its voltage is too high, so the relay disconnects the output but does not switch to input 2, where the power is normal. Once the power on input 1 returns to normal, the load is connected again.

For easier device control and monitoring, the manufacturer offers a software program. After installing it, I got a bit of a scolding from Windows and had to give it permission to write to its installation folder.

You can connect via USB or RS485, though in practice both work through converters. In my case, COM6 was USB, while COM9 was RS485 via a USB converter. You can also select a TCP connection, though as I understand it, that requires the appropriate adapter.
One immediate drawback: the software not only fails to search for its devices automatically, it also doesn’t remember the last device connected. That’s a bit annoying, since you have to open the settings every time.

These are the settings I saw on first connection—the defaults. Since only one input was connected, and it was only a single phase, there were naturally a whole bunch of errors.
While writing this review, I noticed that the top right shows “voltage at the output terminals” and a green icon for phase 1. But as far as I remember, nothing was connected there; perhaps it was picking up some interference.

An important section: the settings. You can configure everything here that you can from the front panel, but it’s much more convenient. There’s also an extra feature: you can set intermediate parameter values. For example, on the device’s front panel you can set a delay of 5 or 10 seconds, while the software lets you set any whole-second value. This is done this way because settings have to be entered by stepping through them on the panel, while the software allows you to enter values directly.
One minor drawback: when you open the settings menu, you have to wait a few seconds for the software to load the current settings from the device and display the corresponding window.

The settings are the same as on the panel. A useful addition is a tooltip for each item showing the setting range and default value.

In its full view (as much as possible with a single-phase supply), the window is quite informative. It lets you monitor the phase and line voltages on both inputs, as well as the output status, battery voltage, frequency, and current faults.
Since the device applies various time delays when switching, a progress bar at the bottom shows the countdown.

Some fault conditions can be reset, such as the repeated automatic transfer switching fault.

During testing, I noticed that the device cannot detect the mains frequency when the input voltage is below 100 V. Interestingly, though, it doesn't trigger a frequency fault in that case.

While the device is connected to a PC, all incoming data is logged automatically. The system was complaining about an error writing this file.
The data is viewed using the same viewer as the one included with the RPM-416 recorder software.

There are a lot of parameters to choose from, but only six can be displayed at once.

Different parameters are shown in different colors.

But, as with the RPM-416, there is one inconvenience: there are only two scales, horizontal and vertical, and parameters displayed on the vertical scale all use the same units. This is less critical here than with the RPM-416, but it would still be nice to fix the awkward autoscaling, which doesn't let you display values in rounded increments (for example, 50, 100, 150, 200 instead of 51.3, 102.3, 154.7).
This screenshot also shows a “forest” of voltage dips around the middle of the graph. In reality, the voltage was fine; the device was communicating over RS485 at the time. After switching to USB, everything looked normal. I suspect the problem was caused by a cheap converter; I'll try another one later.

The device worked reliably during testing, and overall I had no serious complaints—if it weren't for one big “but.”
The problem is that although the measurement and control circuits are supposed to be galvanically isolated, in reality things turned out to be much more complicated.
Galvanic isolation can mean different things. Sometimes all you need is to avoid a direct connection and prevent a “ground loop”; this is known as functional isolation. But in this case, isolation must provide not only separation, but also a certain level of electrical safety.
This protects against electric shock if any parts are exposed outside the enclosure, and ensures safe operation if those parts are connected to other devices.
Just as a reminder, the breakdown voltage is usually estimated at 1 kV per 1 mm. To provide an acceptable level of safety, the distance between the “hot” and “cold” sides should be at least 5 mm (sometimes more), with protective slots added.
The slots prevent arcing across the surface of the circuit board. Someone might ask why you need this level of safety when the outlet is only 220 V. The point is that this applies under normal conditions, but there can also be humidity, dust, static discharges, high-voltage pulses, and so on.
A close inspection of the board revealed a rather grim picture.
1. The photo shows a vertically mounted board with a trace running underneath it.
2. The problem is that the trace on the main board is at mains potential, while the ground plane on the bottom of the vertical board is connected to the control board ground. The distance between them is about 1 mm. On the other side of the vertical board, 2 mm from the live trace, is the USB converter's power input.
3. The upper board, with a pair of optocouplers; traces run off to the left toward the measurement boards.
4. But beneath the optocouplers, the distance between the “cold” and “hot” sides is about the same 1 mm, not to mention how tightly the traces are packed above them.

There are similar spots on the lower board. Blue marks the “cold” side, red marks circuits directly connected to the input voltage, and orange marks the neutral line when connected correctly. There are actually more spots like these; I marked the ones that are immediately visible. I was really surprised by the developers’ strange logic in routing the high-voltage dividers so far away from the measurement modules and placing them near the device’s “cold” side...
This means the device would fail an electrical safety test using a hipot tester and does not meet its stated electrical safety class.
Just to be clear, solder mask is not an insulator in this case, unlike conformal coating.

This is a classic mistake. Here’s an example of another device whose developers even added protective slots near the relay contacts, but forgot that there are narrow clearances elsewhere, too. It also has a USB connector sticking out, and the large copper pour is the connector’s GND.

Things like this can have serious consequences. The photo below shows a power supply that was damaged while I was preparing this review. The grounding wire connects to the right of the Y capacitor.

But in reality, things turned out much worse. I was running tests, with everything connected as shown in the diagram. The Variac wasn’t connected quite correctly, but it was set so that, in practice, it had no effect.
Here’s what happened:
The ATS was powered on, and I was adjusting the voltage with the Variac. After a few adjustments, the ATS reported an “reclosing attempt count” error and switched off the contactors.
I decided to check how the error reset worked. I set the Variac to neutral, connected a USB cable to the hub, launched the software, and connected to the device.
Everything worked correctly. I selected “Clear faults” and confirmed. The contactor clicked, then almost simultaneously the lights went out and the circuit breaker for that line tripped.
My first thought, of course, was that I’d messed something up, but everything had been working fine until then.
I checked the circuits again; everything looked fine. I reset the breaker and powered up the ATS. Nothing.
Then I thought I should turn on the computer (it’s on the same line, so it had shut down too). But pressing the power button did nothing.
That’s when I realized that something had clearly gone wrong.
Here’s what failed:
1. The USB hub was completely dead, but its power supply was fine.
2. The Bluetooth adapter had failed and overheated, since the hub’s power supply kept running.
3. The VFD clock, powered from that hub, had a burned-out microcontroller.
4. The mini PC had a shorted processor on the USB data lines and a short on the 3.3 V rail. The three SSDs, 16 GB of RAM, and USB SSD survived, and its power supply was fine too.
5. The monitor’s power supply failed, but the monitor survived.
6. Probably the 3D printer controller, but I haven’t checked since I don’t use it. It was also connected to the mini PC on the same pair of ports.

I contacted technical support and spoke with one of the engineers, explained what had happened, and also mentioned an oddity I’d noticed right at the beginning.
When I assembled the circuit with the relays (which are essentially contactors), the device complained about a stuck contactor when I powered it on. It turned out that with the relays off, terminal 35 (L2 load) had 220 V, while terminals 34 and 36 had about 50 V. I connected a small load to that line, and the relay switched on. I wanted to take a photo of this strange behavior, but when I tried again a couple of hours later, everything worked perfectly.
I asked about galvanic isolation. They confirmed that the circuits were isolated and that even if I connected the live and neutral wires the wrong way around, nothing would happen (there was a chance I’d mixed them up).
They suspected one of the divider resistors might have been lying against the board. It was connected to the very line where I’d measured the phantom 220 V.
Yes, it’s theoretically possible, but I checked a photo taken right at the start of disassembly—that is, showing how it was originally—and sure enough, the resistor was pressed against one of the regulators.

The odd thing is that when I took it apart, I bent it back into the correct position. I’m not certain, since I didn’t take a photo when I opened it again, but I decided to check.
1. This is how the resistor is supposed to sit.
2. This is how it might have ended up if it had been pressed against the board.

And even if it were pressed against the board, it would still be over a “bare” area of the board, with at least 2 mm of clearance to the nearest conductive parts.
Based on my analysis, the failure most likely occurred on a data line or, less likely, on the 5 V rail. But there’s a resistor on the board in the 5 V circuit, and it was still intact.
I later found that the converter chip had failed with a short across its power pins. After I disconnected it from the power circuit, the ATS started working normally, and RS485 worked too. The engineer was surprised, since it shares a power supply with the USB converter.

There are several ways to make the device safe:
1. The difficult option: reroute the traces on the top and bottom boards.
2. The easy option: add separate galvanic isolation for the USB and RS485 circuits.
Personally, I think the second option is more acceptable and technically easier to implement, since both interfaces are on the same small board and only that board needs to be rerouted. The only potentially dangerous point would be the output for connecting the battery, but we could try to address that by adding protective slots and perhaps inserting an insulator. By the way, isolating the interfaces would make the clearance requirements less stringent, including in places where the clearances are currently adequate.
The process engineer may also want to consider the vertically mounted resistor separately, so that it either cannot be bent at all or has more reliable insulation. I also suggested rotating the resistors 180 degrees so that the pin sticking up is not the one connected to the phase, but the one connected to the lower leg of the divider.
To be fair, I was concerned about a breakdown between this lead and the resistor body, but under the shared heat-shrink tubing I found additional insulation on this lead.

Galvanic isolation in this case is usually implemented either with a dedicated ADUM chip or with a pair of optocouplers.
Optocoupler-based solutions are very common even in Chinese devices. They also have some safety concerns, but on the other hand, the chance of connecting them to high-voltage circuits is considerably lower.
1. The USB converter board from the DPS and DPH series.
2, 3. RS485 is more complicated. Here, too, you can use a pair of optocouplers, as in the Sousim load, but the transceiver also needs a power supply, so a small isolated DC-DC converter is used.
4. The DC-DC converter should have greater spacing between its input and output pins, for example, like the B0505S-2W, but even these are rated for only up to 1500 V, while 2500 V is required.

And now, the conclusions.
Strange as it may seem, I liked the device overall, for several reasons.
1. Stable operation
2. Decent functionality, including contactor status monitoring.
3. A large number of settings, which means plenty of flexibility in adjusting it
4. Remote control via dedicated software
5. MODBUS protocol support. I haven't tested this, but the manual provides a fairly detailed description, so perhaps I'll test it later.
6. Logger mode when connected to a computer
So, looking at the device as a whole, it's genuinely quite good. I even liked the design, as well as the use of quality relays, varistors, and capacitors. The soldering could be a little better, but in this case that's more of an aesthetic issue.
But, as I described at the end of the review, there's a pretty big catch: the insulation isn't very reliable. That doesn't matter for the device itself, but it's critical when connecting it to a computer—and for the user, since the USB connector can be touched. I'd be interested to hear what other readers think. Maybe I'm wrong somewhere and things aren't actually that bad? In other words, the two classic questions: who's to blame, and what should be done?
For now, as a safety measure, I'd recommend connecting it through a USB galvanic isolator. This will improve safety, but it won't solve the problem completely.
In short, it's functional, works well, and is quite thoughtfully designed and reliable in its own way. But when connecting external interfaces, you should treat them as connected to the mains and take the appropriate safety precautions.
The device has now been sent back to the manufacturer for repair and analysis of the problems described. Most likely, they'll send a replacement, so if you have any questions about the software or how it works, I can check them out.
That's all for now. I hope this was useful.
плато авр на 3 фазы конденсаторы, 6, подключение контроллера к авр, 1, схемой шкафа авр пэф-301 схема, коммутатор переключатель фаз авр, https://www.kirich.blog/obzory/drugie-obzory/998-kontroller-trehfaznogo-avr-pef-321avr-ot-novatek-elektro, как подключить пэф321авр, новатек пэф 321 подключение, схема подключения 3 фазной сети в доме с авр пэф321, пэф321 обзор, 3 фазы контроллер авр
Questions and answers
What can the PEF-321AVR controller do?
It controls switching between two inputs, can control a generator and provides RS-485 data connections. It also connects to a computer via USB and supports MODBUS.
What safety issue did the review find with the PEF-321AVR USB connection?
The author found the isolation inadequate, meaning the USB interface may be connected to the mains. A USB galvanic isolator is recommended to improve safety.
What contactors does the PEF-321AVR require?
The wiring diagram calls for contactors with five contact groups: four normally open power contacts and one auxiliary normally closed contact. With a shared neutral, the author says four groups (3+1) may be used.
Discounted products
You may be interested in
Discounted products
All the content on the site and the advice are free, but we’d be grateful for your support of the project and the channel!
Only registered readers can comment. It takes a minute to sign up, and you can log in with the same account you use on the forum.
Log inRegister



