PEF-301 phase selector switch from Novatek-Electro

Not long ago, I was asked to review an interesting, if rather specialized, device: a phase-selection relay. Today I’m not only reviewing the relay itself, but also adding a small detail that I think is quite significant.
As usual, though, it’s best to read the full review, which includes a description, a closer look, and a teardown.


To start with, this really is a specialized device. I’ll do my best to explain everything in plain language, but a review like this assumes at least some basic knowledge of the subject—for example, what three-phase power is and how it differs from a circuit breaker.

And, of course, I’ll explain what this thing is and what it’s for.
Main application - Suppose you have a single-phase load whose shutdown could have serious consequences: a heating boiler, a large freezer, a server. If your home has a three-phase supply, installing this relay can keep the load running as long as at least one phase is working. This is especially important if the equipment runs unattended for long periods or is located far away. For a boiler or refrigerator, the relay can work on its own; for equipment where even a brief interruption is critical, it can be used together with a UPS.
How it works: connect three phases (or at least two) to this little box. If there’s a problem with the first phase, it switches the output to the second or third. If all three phases have a problem, it disconnects the load, protecting it from damage.

Additional application - This kind of relay can also be used to put together a simple automatic transfer switch (ATS), with a few limitations. For example, I use relays like these to power a server room and other loads from two independent incoming supplies fed by different substations. But it’s best to consult an electrician to find out whether this is suitable for your situation.
How it works: one incoming supply is connected to the first input, and the other to the second. If power fails on the first supply, the relay switches to the second. While it’s switching, the equipment is powered by a UPS. The advantage is that you don’t need a high-capacity UPS; it only needs to keep things running during the switchover.

Another option - A standard voltage protection relay that protects against overvoltage and undervoltage, with an adjustable delay and even an option to prevent it from switching back on after it trips.

I bought the first relay around 2010, if I remember correctly, and it’s still working properly as an ATS with a load current of about 10–12 A.


The title gives the price in hryvnias and dollars. If you need the price in rubles, this relay is easy to find on Russian websites, for example here or here, although the price there may be a little higher.

The packaging is branded, and the box is clearly oversized, but here’s what’s interesting (and this is the extra detail I mentioned): the old packaging looks completely different.
Yes, you read that right: today’s review really does cover two relays. The first is brand new, delivered just a week ago; the second is already 5 years old.
The explanation is simple. When I told a friend I was preparing a review of a phase switch, he said we already had one of these. We bought a lot of them in early 2015, and a couple were left over—not as spares, but because the technical requirements changed. I’d completely forgotten about them. So I retook some of the photos, and now I can even compare the two devices.


The point of the comparison is that manufacturers often change a device’s internals, and sometimes even its appearance, over its lifetime. Sometimes these changes aren’t for the better—what they call “optimization”—and the manufacturer cites the following:
The manufacturer reserves the right to make changes to the circuit that do not impair the device’s functional performance.

So far, the changes are purely cosmetic: the product label is slightly different, and the sticker has moved from the side to the bottom (the new device is on the left, the old one on the right, here and below). You can also see the plastic clip that secures the device to the DIN rail. It has no spring; it’s a standard sliding clip.


I’ll put photos of the manuals and a comparison of the new and old versions under the spoiler. You can download the manual here.


Finding the technical specifications in Russian turned out to be harder than I expected.

As you can see from the manual, the maximum switched current is 16 A. Naturally, that’s for a resistive load. Even so, this is enough current for quite a wide range of applications. You can also connect external contactors, but more on that later.


The enclosure is standard for devices like this: it mounts on a DIN rail and is 4 modules wide (70 mm).


The device hasn’t changed at all on the outside in 5 years; the only difference is the color of the jumpers and terminal blocks.


The height of the protruding part is roughly the same as that of a standard circuit breaker (meaning the height of the opening in the panel), though a standard breaker is a little shorter. The device fits perfectly in the panel, though—I’ve checked this many times.

The input terminals are on top; connect the three incoming phases and neutral here. They must be connected through circuit breakers; the manufacturer recommends B16 breakers.
The output terminal block is at the bottom. It provides connections to all three internal relays, as well as the input for protection against external contactor contacts sticking.


There are two wiring diagrams. The first is for use without contactors; in this configuration, the current is limited to 16 A and the power to 3.5 kW. The second uses external contactors, in which case the current is limited only by the contactor rating. With the second configuration, you must connect the input for protection against stuck contacts (terminal 12).

It’s strange that the diagrams don’t show the protective circuit breakers mentioned above. In my opinion, that’s a significant omission. For example, I had to dig through the text to find out which breaker to use.


The terminal blocks are good quality, elevator-style: the wire is pressed upward by a U-shaped clamping mechanism, which is noticeably better than a standard clamp.
Objectively, the only difference between the old and new versions is the color. Subjectively, I prefer the plastic on the old (green) version.


The relay is configured from the factory to work without contactors, so two jumpers are already installed. They look like 1.5 mm² solid wire. The terminal block is rated for 2.5 mm²; you could probably try fitting 4 mm² wire, but it’s better not to.
The terminal block is rated for up to 300 V and 16 A. You might ask, “What about 380 V if it’s only rated for 300?” It’s simple: if you look at the photo above, you’ll see that some of the input terminals are left empty—in other words, some contacts aren’t connected to anything. This allows the permissible voltage to be increased.




All the controls and indicators are on the front panel. There’s no digital voltage display, but in my opinion it’s not really needed. Besides, leaving it out reduces both the device’s cost and the requirements for its built-in power supply.

Indicators
Three green LEDs show which relay is currently active. There are four possible states: 1, 2, 3, and 0.
One red LED indicates an input fault and counts down the delay before returning to normal once the voltage has recovered.

The controls consist of four potentiometers:
1. Sets the minimum acceptable input voltage threshold; adjustment range: 160–210 V.
2. Sets the maximum acceptable input voltage threshold; adjustment range: 230–280 V.
3. Delay before the output turns on after an emergency shutdown. This is needed if the relay is connected to equipment that must not be powered immediately after the fault clears, such as refrigerators. Adjustment range: 1 s–10 min.
4. Delay before switching back to the previous phase. For example, if the relay switches from phase 1 to phase 2, it won’t switch back as soon as the voltage on phase 1 returns to normal; it will wait for the set delay. At the far-right setting (infinity), it won’t switch back automatically. Adjustment range: 5–200 s + infinity.



The manufacturer has a good video explaining the settings.


The relay is very easy to take apart: remove the four screws and lift off the cover.


It’s built around two circuit boards—a power board and a control board—connected by a ribbon cable. Another small difference is noticeable here: the new version has rubber sleeves fitted over the resistor shafts.
The board revisions are identical, and the board was originally designed to accommodate five variable resistors, with the fifth shorted by a jumper.

On both units, the upper board sits slightly crooked relative to the lower one, which makes putting the cover on a bit awkward. That said, users have no reason to remove it.


At first I thought the rubber sleeve was meant to seal the joint between the case and the shaft, but it doesn't even reach that far, so its actual purpose remains a mystery to me.


A couple of comparison photos of the new and old relays.



The upper board is held in place by four long screws and four plastic standoffs. The design is basically fine, but connecting the boards with a flexible ribbon cable doesn't seem very production-friendly to me, even though it's more reliable than a pin header. By the way, the cable is secured at both ends with hot glue, which deserves a plus in my book: it's often the little details that shape your impression of a manufacturer's products.


Three power jumpers made from stranded wire, probably 2.5 mm², run across the board. As a result, the dropper capacitors sit crooked, so perfectionists may want to skip this photo.



1. The unit uses Fujitsu K1CK024W relays, rated for currents up to 16 A.
2. The power supply is a circuit with three dropper capacitors, a three-phase rectifier, and a limiter made up of three Zener diodes. The filter capacitor is 220 µF, 35 V; the old version had a 470 µF capacitor.
3, 4. The ballast capacitors are better in the new version: 310 V versus 275 V. Their capacitance is the same in both cases, 0.68 µF.

By the way, I do have an issue with the capacitors—or, more precisely, how they're mounted. Apparently the board was originally designed for a different type, so the ones shown in the photo are sitting at odd angles, which detracts somewhat from the otherwise quite decent assembly.


Since the relays are a very important component that effectively determines the product's service life, I tracked down their datasheet. That's where I found out who makes them.


But I found a small discrepancy here. The device specifications list the relay life as:
under a 16 A load, at least 100,000 operations
under a 5 A load, at least 1 million operations

The first figure is limited by the contacts' electrical endurance; the second is essentially the life of the mechanism itself. But the datasheet gives just 50,000 operations for this series, while 100,000 is specified for the K1AK series.

The datasheet seemed a little odd to me in general. For one thing, it says the relay can switch up to 300 V DC, which is physically impossible for mechanical relays of this size.


The datasheet also specifies contact resistance of up to 100 mΩ, which seemed a bit high to me, so I decided to check it.
As specified in the datasheet, I measured the resistance at a current of 1 A. The voltage drop across the relay contacts themselves was 1.5 mV, and the total drop across the entire device was 6.3 mV. That gives us resistances of 1.5 and 6.3 mΩ, respectively. At maximum current, the power losses would be 0.4 and 1.6 W, which is very good.


The board is well made, clean and neat throughout.


Power reaches the ballast capacitors through 75 Ω current-limiting resistors and narrow traces that serve as fuses.
All the power traces are reinforced with plenty of solder, especially around the relay contacts.


The device has apparently been in production for a very long time, judging by the Atmega8 microcontroller. On the other hand, it has plenty of pins, the price is affordable, so why not?
At the top of the board, there's a footprint for a connector, presumably for programming the controller, though for some reason there are more contacts than expected. In any case, the user has no need for this connector.


The boards differ slightly: on the old one, some of the labels overlapped, while the new one has no such issue.


There are other differences too. For example, it looks like the old board was coated with protective lacquer, or at least it certainly looks that way. The new board is clean.


The new board's soldering and component placement are also much neater, more like machine assembly. The old one was clearly soldered entirely by hand.


To get a better idea of the circuit design, I redrew the schematic.

The input has a multiphase rectifier with ballast capacitors. The voltage is limited to around 22–24 V by a chain of three Zener diodes, followed by three relays with switching transistors and protection diodes.
There are five voltage dividers: three measure the input voltage, and two are part of the anti-sticking protection circuit. Their divider ratios differ.

There aren't many components on the control board, but it's good to see protection diodes on all the measurement inputs. The transistor drive circuit is interesting: the logic signal from the Atmega output goes to the transistor base through a 4.3 kΩ resistor and an LED connected in series. This saves three resistors without affecting functionality. By the way, judging by appearances, all the resistors are precision types.
There are also ceramic decoupling capacitors, which is good too.

In short, it's a simple circuit with no frills, but it's quite reliable. The upper resistors in the dividers are high-power types, not because they dissipate a lot of power, but because resistors of this type are usually rated for higher voltages—which is the right choice.



That's it for the outside and inside; now on to the tests.
Unfortunately, I don't have three-phase power at home, so all measurements and tests were done using a single phase.
In normal operation, one of the green LEDs is lit. If I start changing the overvoltage trip threshold, I can easily trigger the protection with an input of 232 V, since the threshold starts at 230 V. The relay then switches to the second input, then the third, and finally enters “fault” mode. After that, I raise the trip threshold, and it switches from the third input to the second, then to the first, though sometimes it jumps straight from the third to the first.
I suspect there was a slight difference in the divider values, which could make it read different voltages even when they were the same. In any case, the difference was so small that it can be ignored.



Next, I measured its own power consumption and, consequently, the heat it generates.
The relay was powered through all three inputs, but from a single phase.
Right after power-up, the relay consumed about 2.1 W. After roughly an hour of thermal warm-up, consumption rose to 2.2 W, but this wattmeter is very inaccurate at such low power levels and with this kind of load. As a reminder, the manufacturer specifies its own consumption as 1 VA.


Component temperatures with the case closed, after 20 minutes of warm-up:
1. Right side: the hottest components were the Zener diodes, at 68 degrees.
2. The relays run cooler—or rather, only the active relay heats up, to 37 degrees.
3. On the left, essentially only the divider resistors and the side of the relay case heat up, to around 40 degrees.


After another 40 minutes of warm-up, the temperatures had risen slightly:
1. Zener diodes — 73 degrees.
2. Relays — 42 degrees.
3. Side of the relay case — 46 degrees.

Temperatures will be higher inside the electrical panel, but there's still a fairly generous margin. If we take 100 degrees as the maximum temperature for the Zener diodes, they would reach that temperature when the ambient air is around 50–55 degrees, which is within the limits specified by the manufacturer.



Checking the accuracy of the undervoltage trip threshold adjustment proved somewhat difficult because of the “progressive” response algorithm for faults, which I'll describe later.
Another issue is that it looks like the adjustment here isn't continuous but stepped. In other words, you can't set, say, 182 V or 188 V; it'll be more like 180 or 190. The front-panel photo shows marks around the controls. For voltage adjustment, they correspond to 5 V increments, and that's roughly how you set it. The principle is similar to connecting a digital keypad using analog inputs, where each button corresponds to a specific voltage—except here, variable resistors do the job.
The same goes for setting the time, so it's easier to use the marks as approximate values rather than trying to tweak it “just a little.”

Example of the response to low voltage:
1, 2. 160 V setting, shuts off at 156, turns on at 166 V
3, 4. Set to around 180–185 V, shuts off at 178, turns on at 193.


There is no delay when responding to an overvoltage, so it is easier to measure the thresholds.
1, 2. Threshold set to 250 V, shuts off at 252, turns on at 247.5
3, 4. Threshold set to 255 V, shuts off at 255, turns on at 250.5.

Maybe it just seemed that way to me, but the hysteresis here is smaller, although that could be due to different fault-handling algorithms.



Its response to voltage dropping below the setpoint is different.
For this test, I powered the relay through a low-power isolation transformer and connected a light bulb as the load. So, while the bulb is disconnected, the relay input voltage is normal; when the bulb is connected, the voltage drops to 165–170 V.
If you set the adjustment to 160–170 V in this mode, nothing happens: the relay keeps the first input switched on steadily.
But if you set it to around 195–210 V (each mark roughly corresponds to 5 V), then when the voltage drops to 170 V, the relay disconnects the load instantly. The voltage rises, the relay counts down the set delay (I set it to the minimum to make this easier to see), and reconnects the load. The cycle repeats. The relay behaves the same way when the voltage exceeds the setpoint.


But if you set the adjustment to 170–195 V, the relay waits 12 seconds before disconnecting the load when the voltage drops. This mode is especially useful with “heavy” loads on the mains: their startup initially causes the voltage to drop, but it then returns to normal, helping reduce false trips and unnecessary switching.

As you can see, the voltage drops when the load is connected. The relay waits for it to return to normal—which doesn’t happen in my case—and only then disconnects the load. The load is reconnected immediately because the delay control is set to minimum.

A very, very long time ago, I suggested to a friend that he add a “progressive” algorithm to his overvoltage protection relay, so that the response time would depend on how far the voltage exceeded the limit. For example (hypothetically), with the limit set to 250 V, it would trip after 5 seconds at 255 V, after 2 seconds at 260 V, after 1 second at 265 V, and instantly at 270 V. This device implements the same principle in simplified form, but it works when the voltage drops and has only one threshold.


Response to the voltage disappearing and then returning at inputs 1 and 2. The circuit breakers on the left and right are connected to inputs 1 and 2, respectively; input 3 is connected directly.


More or less the same thing, but with different combinations of voltage loss and return.


Testing the response to a gradual change in voltage. You can also see here that the device starts up at around 47–50 V and reaches its normal operating mode at 50–55 V, which I think gives it plenty of headroom.


Oscillograms of the device’s transient response and power-rail ripple.
1, 2. Switching when voltage is lost at input 1, followed by switching to input 2; two examples. Total time is about 100 ms.
3, 4. Switching from a working input 2 to a working input 1, initiated by the device itself; two examples. Total time is about 30 ms.
5. Ripple on the relay coil, DC input mode.
6. Ripple on the power rail, AC input mode. Some may think it’s a lot, but remember, it was powered from a single phase; with three phases, the ripple will be much lower.


And one last test: checking whether the enclosure plastic is self-extinguishing. This property isn’t explicitly specified or claimed on the manufacturer’s website, but I did a simple test. Of course, it would have been better to use larger pieces, but I really didn’t feel like cutting the enclosure into noodles.
In the end, the plastic catches fire and burns vigorously while in the flame, but goes out once the heat source is removed. The area that burned chars, but there is practically none of the characteristic smell of burning plastic.

If this test isn’t enough, I can of course break off a larger piece, but I think this property is fine.


I burned the piece on the right on both sides; the one on the left is shown in the GIF.



Conclusions.
Sometimes you’re preparing a review and, with a pile of drawbacks in front of you, you look for the good points; other times, you look for drawbacks among the good points. Today is the second kind of case. The thing is, the drawbacks here are mostly minor, though they are there.
Build quality. I wouldn’t call it top-notch, but it’s quite decent. My main complaint is the somewhat sloppy PCB routing—or, more precisely, the relative placement of three things: the ballast capacitors, jumpers, and standoffs for the upper board. It’s obvious they get in each other’s way.
Circuit design and components. Yes, overall, this one is pretty good too: quality relays, protection for the microcontroller’s measurement inputs, protection against contacts sticking, and the option to connect external contactors.

Now, a few general thoughts about the device itself.
In practice, it’s a protection relay with extra features and no voltage display. You could certainly use it as a protection relay, and it would work just fine, but without external contactors it’s limited to 16 A, and it’s expensive for a basic protection relay.
It has all the necessary settings for thresholds and switching times, as well as switching to the priority phase. As a bonus, it delays switching if the voltage drops by a relatively small amount.

What I liked most, though, is that over at least the past 5 years, the manufacturer hasn’t simplified the device—in fact, they’ve improved its quality. The only change for the worse is the smaller filter capacitor. At the same time, the ballast capacitors have been replaced with higher-voltage ones, and the component assembly quality has clearly improved.

In short, I’ve been using these exact relays myself for almost 10 years (if I haven’t lost count). I also still have several bought 5 years ago in service, and I have no complaints about those either. By the way, the manufacturer advertises a 10-year warranty, though I haven’t had a reason to test it yet.

Suggestions:
1. Add the circuit diagram shown in the manual, including the circuit breakers.
2. Improve the PCB layout. Massive components (in this case, the capacitors) should be mounted either directly on the board or on spacers, rather than being supported only by the stiffness of their leads.



That’s all from me. I hope this was useful. If you have suggestions for other tests, or any general questions, I’ll be happy to answer. I also now have the opportunity to test other devices, so feel free to send in your suggestions.