Powerful and Unusual Universal Converter Based on the LTC3780

I've finally gotten around to looking at an unusual converter using SEPIC technology (or so I thought at first).
It's unusual, to say the least, because it uses just one inductor, whereas a conventional SEPIC always uses two (or a coupled inductor).

One user often asked in the comments on my SEPIC reviews, “Why not Linear?”
Well, here it is: Linear—or, more precisely, a converter that uses an LTC3780 PWM controller made by Linear Technology.

Linear Technology chips usually stand out for two things: high-quality performance and a high price. There is one small upside, though: the company sends out product samples for free. Not always, and not to everyone, but it's worth a try.

This converter caught my eye as soon as I saw it. I ordered it without any particular practical purpose, just to study it myself and show others what it's all about.
In this review, I'll talk a little about the features of this type of converter and the board as a whole.

First, the shipping. The store pleasantly surprised me by packing it as if it were made of glass—they really went to town with the wrapping :)


But I'll start my story with something other than the converter. I ordered these terminals along with it. At first, I had an idea for what to use them for, but the idea went away and the terminals stayed.
Since they were also sent to me for review, I'll tell you about them too. As it happens, while writing this review, I came up with an idea for what to use them for.
The terminals are available here in the store, priced at $2.29.


These are spring-loaded terminals. Besides the terminals themselves, the kit includes two plastic washers, a lug, a lock washer, and a nut.


The terminals have a standard M4 thread and come as a red-and-black pair.
The plastic washer is designed to let you mount a terminal on a metal surface without worrying about it shorting to the chassis. The washer is quite thick, which is a plus.
All the parts (except the plastic washers) have a gold-colored finish. I doubt there's any actual gold in there, but it looks nice.


They're quite heavy, and yet the right tail is longer the black set weighs more.
Overall length: 40 mm; diameter: 12.5 mm; threaded section length: 16 mm.


The spring is quite stiff, so it clamps really well.
At first, I wanted to use these terminals for a power supply, but then changed my mind. I don't think uninsulated terminals belong on a power supply.
I'm now thinking of using them in an electronic load (possibly a future one) designed for high currents.


That's enough about the terminals; now we can move on to the converter board.
The board doesn't look very large, although it's bigger than the converter board in this review.
The board measures 77 mm long, 46 mm wide, and 13 mm high overall.
The board's height can be reduced by 1.5–2 mm by lowering the solder joints on the connector pins and inductor.


Specifications:
Input voltage: DC5-32V
Output voltage: DC1V-30V, adjustable
Output current: 8A continuous, 10A peak
Output power: 80W continuous, 130W peak; additional cooling is required above 80W
Output ripple: 50mV
Operating temperature range: -45 ~ + 85 ℃

The board's build quality is a little disappointing: the connectors are soldered on crookedly, and there are some smudges and flux residue. This doesn't really affect how it works; it just looks bad.
The power input connector is on the left.
Just to the right is an LED indicating a fault—or, more precisely, that the input voltage has dropped below the acceptable level.
There are three trimmer potentiometers in the center of the board. I really had to think about why there were three. It turned out to be simple.
The first trimmer sets the input undervoltage cutoff threshold. This can be used as a protective measure to prevent the converter from operating when the input voltage is low, since the lower the input voltage, the harder the converter has to work. Although people online say this can be useful when working with solar panels.
The second trimmer sets the maximum output current limit. When this threshold is reached, the converter switches to CC mode.
The third trimmer adjusts the output voltage.
Just to the right is the OUT LED, which indicates that voltage is present at the output.
And on the right there's a terminal block for connecting the load.


I included this photo for two reasons. I originally found it while trying to figure out what the unknown trimmer potentiometer was for. But the page also had some useful information, which I'll mention as I go through the review.
The photo shows what the trimmer potentiometers do, but it also shows the second (or, more precisely, the first) version of this board.
This information is also on the store page, but it's less clear, so I'll explain.
The converter comes in two versions:
1. With a heatsink on the bottom and a red inductor winding
2. With heatsinks on top and a yellow inductor winding.
The store ships the converters at random, but I think the board with the heatsink on the bottom is better, since its heatsink is slightly larger. However, there are no vias beneath the power components, so perhaps they're equivalent.
But on the board with heatsinks on top, they're glued in place, while on the first type the heatsink is screwed on, which may be more reliable.


Since there's a version of the board with a heatsink on the bottom, the underside of the board is almost empty. It only has the inductor and connector pads.


My complaint was about the soldering quality on the inductor itself, and especially on the connectors. It looks like they were soldered with a cold iron. If yours looks like this, it's best to reflow the joints, just in case.


As I mentioned above, the board is only 13mm thick; the tallest components are the capacitors.


One unusual feature is the fuse, which, in my opinion, has a rather odd form factor that will definitely make it harder to replace. Still, it survived my tests :)
There are also contacts nearby for controlling the output voltage. I tried shorting them, and the output voltage shuts off completely, but that's a consequence of this converter's circuit design.


At the input, there are two 330 µF, 50 V aluminum capacitors, along with a ceramic capacitor connected in parallel with them.
The converter operates at a high frequency, around 400 kHz, and at these frequencies the PWM controller manufacturer recommends using a ceramic capacitor.


The output is almost the same, except there are three capacitors here instead of two.


The inductor is located between the heatsinks for the power components. Note that it doesn't have two windings; it's simply wound with two wires.
I mention this because there is a SEPIC converter topology that uses a single inductor with two windings instead of two identical inductors.
The store page says:
Inductors using Sendust, 0.8 wound bifilar, heat generating was less. I won't argue about the core material being Sendust, since I can't verify that.
But I will dispute the claim that it has a bifilar winding. The bifilar winding method is specifically intended to eliminate inductance. Without going into detail, I'll just say that this is most likely a bad translation, and originally it meant that the winding is made with two wires, that's all.
Correction: a bifilar coil is wound with two wires placed close together, so yes, this is a bifilar coil.

In my opinion, the inductor is a little undersized, and they could have used thicker wire—or better yet, wound it with 3-4 wires like these.


The whole conversion process is “run” by the LTC3780 PWM controller made by Linear Technology.
This PWM controller doesn't have powerful output stages, so it needs external MOSFETs to operate.
The board clearly hasn't been cleaned very well; the controller in the photo has already been partially cleaned.


I'll try to explain this controller and the converter's clever topology in a little more detail.

I didn't redraw the circuit for this board, since that would be rather inconvenient. But I can say that I found no fundamental differences from the datasheet, so you can use the manufacturer's documentation as a reference.

Surprisingly, despite the rather convoluted connections, the circuit isn't that complicated.
It's easy to pick out the following on the schematic:
The controller
Four MOSFETs
Two diodes, each connected in parallel with one of two of the four MOSFETs.
A 10 mΩ shunt. Incidentally, this is one of the differences: the board uses a 7 mΩ shunt.
The “supporting” components, the feedback compensation network, the feedback itself, and the supply circuits for the “high-side” transistors.


To make things clearer, I've drawn a simplified block diagram. I'll use it to explain how this type of converter works.

I've already reviewed converters of various types and described how they work. In this case, we're looking at a hybrid of a snake and a hedgehog—a step-down and step-up converter.
Yes. At first, I thought this converter worked like a regular SEPIC, but it turned out to be both more complicated and simpler at the same time.

The converter topology resembles a bridge, but it works a little differently.
The circuit looks symmetrical, but there are differences. For example, diodes are connected in parallel with two of the MOSFETs, but the diode is at the bottom on the left and at the top on the right.


First, let me explain why the diodes are needed. To improve efficiency, the converter uses synchronous rectification, but diodes are still necessary because the MOSFET body diodes have poorer characteristics, and in this circuit the diode is part of the topology. In other words, the diode conducts first, and then the transistor takes over.
What is a synchronous rectifier?
When rectifying with a diode, we have losses across the diode junction, and at high currents they are significant even with Schottky diodes. To reduce losses, a MOSFET is connected in parallel with the diode; it has lower losses. But you can't always simply shunt the diode.
The thing is, a diode stops conducting when the polarity reverses, while a MOSFET conducts in both directions when it is on. So, to prevent current from flowing back into the source after the inductor has delivered all its energy to the load, a control circuit is added to monitor this process, and the shunt plays an important role here.

In full-wave circuits, though, synchronous rectification can be implemented more simply: you just need to add another pair of windings to the transformer to turn on the transistors, as in the diagram below, for example.


As I wrote above, the circuit resembles a bridge. At first, I thought the converter always operated in one mode, “charging” the inductor by turning on transistors S1 and S4, then “dumping” the energy into the load by turning on transistors S2 and S3. But it turned out to work differently.
I'll briefly describe the operating cycles according to the datasheet. There are formally four, but cycles 2 and 3 are very similar, so I'll describe three.
1. The input voltage is higher than the output voltage.
Transistor S3 is permanently on, while S4 is permanently off. Transistors S1 and S2, diode VD1, and the inductor form a classic step-down converter, with the PWM signal on the Buck pins.
2. The input voltage is lower than the output voltage.
The circuit operates in the opposite way (as I explained in one of my reviews, all these circuits use the same components, just connected differently).
Transistor S1 turns fully on, S2 turns off, and transistors S3 and S4, diode VD2, and the inductor form an equally classic step-up converter, with the PWM signal on the Boost pins.
3. The input and output voltages are close.
This is the most complex mode. I won't get too deep into the weeds, but in this mode the PWM signal is applied to all outputs, and formally the circuit continuously “transfers” energy by turning on the transistors diagonally: first S1 and S4, then S2 and S4. The only difference is the on-time: for step-up operation, S1 and S4 are on longer; for step-down operation, S2 and S4 are on longer. At least, that's how I understand this converter's operation.

Based on all of the above, I can conclude that, strictly speaking, this circuit is not a pure SEPIC converter, but a hybrid step-up/step-down circuit whose operating modes switch dynamically depending on the difference between the input and output voltages.


A simplified block diagram to help explain how it works.
So, we get: step-down + step-up + synchronous rectifier = LTC3780


Under the heatsinks are the two halves of the converter's power stage, each with a pair of transistors and a diode.
The diode is SS54, although judging by its size, I think it's more likely an SS34, which is less capable.
The transistors are AOD4184A. These MOSFETs are rated for up to 40 V and 50 A, have an on-resistance of 7–9.5 mΩ, and a gate capacitance of 1.5 nF. The converter's performance can be improved by replacing the transistors with equivalents that have better characteristics. You could also install a proper heatsink then, since the glued-on heatsink is very ineffective.


There is also a 5 V regulator on the board.
Next to it are the Fault LED and a corresponding trimmer for adjusting the trip threshold. The word “Fault” has somehow drifted all the way over to the fuse, but it refers to the LED.


The fault signal comes from the PWM controller.
There is also a brief transition when both LEDs—the Fault and Run LEDs—are lit.


The board also has an LM358 op-amp, which handles the current-limiting circuit and the converter's fault shutdown.
As you can see in the photo, the board has two 7 mΩ shunts. They are connected in series, not in parallel. The first handles the PWM controller's overcurrent protection; the second works with the op-amp and the maximum-current limiting circuit.
The LED on the right lights up when voltage is present at the output.


The circuit for the maximum-current limiting stage and for adjusting current and voltage, along with the connections between them.
By the way, voltage and current adjust very smoothly—in fact, linearly. I had no trouble setting the voltage to either 1 V or 30 V.


This is the part of the circuit that adjusts the trip threshold for protection against operating at low input voltage. For example, you can set the threshold to 10 V. If the input voltage drops below that, the converter will shut down; once the voltage rises again, it will resume operation.
But since the circuit has no hysteresis, the voltage drops gradually, and the converter keeps oscillating between on and off.
That is what causes the transition mode, when both the Fault and Run LEDs are lit.


That’s about it; time to move on to the tests.
For testing the converter, I put together a small setup consisting of:
Adjustable power supply
Electronic load
Oscilloscope
Multimeter
Non-contact thermometer
A pen and paper.

The converter was initially set to an output voltage of 12 V—or, more precisely, 12.07 V. I left it as it was.


The datasheet says the IC can operate from as low as 5 V, but it also includes a table with information on efficiency and heat dissipation.
The graph in that table starts at around 6 V, so I started testing at 6 V as well.


First, I checked the ripple level and how the device performed with an input voltage of 6 V, an output voltage of 12 V, and load currents of 1.25 A (25% of the maximum) and 4.5 A. According to the datasheet, the output current in this mode is specified as 5 A, but my converter went into protection, with the input current reaching 12–13 A.
It's hard to see in the photos (all photos in my reviews are clickable), so I'll explain as I go and point out the features and shortcomings I noticed.
1. Input voltage: 6 V; output: 12 V; load current: 1.25 A. Everything works normally, although the ripple is higher than the specified Output ripple: 50mV, measuring about 70–75 mV.
2. Input voltage: 6 V (at the board terminals); output: 12.48 V; load current: 4.5 A. The converter could deliver 5 A, but only briefly. Ripple, however, climbed all the way to 160 mV.

A brief digression on output voltage regulation accuracy.
Above, I wrote that the voltage at a current of 4.5 A was 12.48 V. The converter regulates the voltage very well, but only while it's cold. The thing is, the board uses ordinary resistors, not precision ones.
Precision resistors differ from ordinary 5% resistors not only in resistance accuracy, but also in how much their resistance changes with temperature. The TCR (temperature coefficient of resistance) of precision resistors is noticeably lower than that of ordinary ones, so as the board warmed up, the output voltage rose. Even blowing on the board made the voltage start to drop immediately.
You can improve accuracy either by improving the cooling or by replacing the feedback resistors with precision ones. The trimmer resistor will still be there, and it too can drift as it heats up, but regulation will still be better.


1. Input voltage: 12 V; output: 12 V; load current: 8 A. Everything worked beautifully here, with ripple measuring just 150 mV.
2. I ran this test purely out of curiosity.
I set the output to the lowest possible voltage, about 0.85 V, and loaded it to 6.7 A. This test isn't included in the summary table; the board didn't get very hot, and everything worked great. Even the ripple level was within spec.


Now let's test the converter in step-down mode.
Input: 30 V; output: 12 V. The load current was initially 8 A, but after a few minutes the converter shut off the output voltage (I noticed when the electronic load's alarm went off).
After that, I had to reduce the output current to 6.7 A, and it worked normally at that level. Ripple was only about 60 mV, not much higher than the specified 50 mV.


1. Step down from 12 V to 5 V at a load current of 8 A. The converter performs excellently.
2. Step up from 12 V to 30 V at a load current of 3 A. I'd say the converter performed excellently, if not for the high ripple and the gradual rise in output voltage as it warmed up.


1. Input voltage: 6 V; output: 30 V. I initially set the load current to 1.7 A, but the converter didn't like that very much: it started getting very hot and became unstable.
2. After reducing the load current to 1.5 A, things were almost normal. It still got quite hot, and the ripple level left a lot to be desired.


But besides the voltage drifting due to heat, another problem showed up.
The photo shows an experiment stepping the voltage down from 30 to 5 V.
As you can see, the output voltage is rock steady at exactly 5 V, and the load current is also unchanged at 8 A.
But the input current is 1.4 A in the first photo and 1.7 A in the second.
This showed up a short time after startup. I heard an unusual sound from the converter (with the load's fans running) and immediately noticed this behavior.
When I increased the load to 7.5 A, the converter would “break into” self-oscillation and start drawing more current. I could easily set the output to 8 A, as you can see in the photo, but as you can imagine, it started heating up pretty quickly because an “extra” 9 W was being dissipated.
The waveform turned into something incomprehensible (I didn't capture it separately, so it's only visible in the photo).
When I reduced the load current to 5.7 A, the self-oscillation stopped and the converter returned to normal operation. I could safely raise the current to 7 A, and it ran perfectly steadily.


All the test results were compiled in a table with the following columns:
VT_L — Temperature of the left pair of transistors (near the input connector)
Cвх — Input capacitor temperature
Inductor — Inductor temperature.
VT_R — Temperature of the right pair of transistors
Cвых — Output capacitor temperature
Uвх — Rounded input voltage.
Iвх — Input current according to the power supply
Uвых — Output voltage according to the multimeter
I вых — Output current according to the electronic load (rounded)
Рвх — Input power according to the power supply
Рвых — Output power according to the electronic load
Efficiency — Calculated efficiency, based on the difference between input and output power

At low currents, the power supply's actual output current is lower than the set value, but above 2 A it behaves quite accurately. That's why the calculated efficiency differs somewhat from the actual figure. If you need a more accurate measurement, I can test it at the required settings—for example, measure the efficiency at a specific input voltage, output voltage, and load current.
The inductor gets very hot in some tests. I can't say for sure what it's made of, but powdered-iron cores generally operate without issue at temperatures up to 200 degrees, although during testing I suspected it was overheating. More precisely, the converter's behavior resembled that of one with an overheated inductor wound on a ferrite core.


Thermal image of the board in its hottest operating mode.
During testing, the board simply sat on the table with the heatsinks facing up. This is far from the best position for heat dissipation; it would have worked better upright, with the heatsinks facing up, but testing it that way would have been inconvenient and not very representative.
The table also shows that depending on the operating mode (step-up/step-down), either the left or right pair of transistors gets hotter.


Now we can analyze this converter and identify its pros and cons.
Pros
The converter works, which is already a good start.
It really does provide adjustable output voltage from 1 to 30 V with an input voltage of 6–30 V.
The converter is fairly powerful, although it falls short of the claimed specifications.
The efficiency is quite respectable for such a wide operating range.

Cons
The build quality could be better.
The feedback circuit uses standard resistors rather than precision ones. As a result, the output voltage rises as they heat up.
Although the device's efficiency is decent, poor heat dissipation from the transistors can cause overheating at maximum load.
The ripple is clearly higher than claimed.

What do I think of this converter?
In my opinion, it's definitely an interesting device, although it's not quite a finished product and is more of a kit. Its performance can be improved by upgrading the cooling and replacing a few resistors with precision ones.
In my opinion, the best operating point is with a 12 V input. In other words, you can connect a 12 V power supply and get an output voltage from 1 to 30 V. However, I wouldn't recommend using this board as the basis for an adjustable power supply, since the maximum current varies between the minimum and maximum output voltages. A maximum power limit would be useful; one option would be to limit the maximum input current.
I mentioned the ripple above. Ripple is undesirable, but at 400 kHz it can be reduced relatively easily with an LC filter.
Testing showed that in some operating modes, the converter really can deliver high output power for short periods. This can be useful with loads that normally draw little power but occasionally need a lot of current.

What can this converter be used for effectively?
Charging batteries or powering various devices from solar panels.
Powering various loads in a car. For example, you can safely power a small computer that runs on 12 V without worrying about it shutting down while the starter is running or being exposed to overvoltage while the engine is running. You can also adjust it to power a laptop, or change the settings the other way and charge a tablet.

In other words, this converter really comes into its own when you need a specific output voltage and the input voltage may be either higher or lower than the output. If the output voltage is always either lower or higher than the input, a converter with a fixed operating mode is a better choice and will be more efficient.

That's all. I've tried to test the converter's main operating modes, but with an input range of 6–30 V and an output range of 1–30 V, there are a lot of possible combinations.
As always, I look forward to your questions, additions, corrections, and just plain comments :)

The product was provided by the store for review.