30W SEPIC DC-DC converter review (5-35V in, 0.8-35V out)

[station240]

This product had such promise, but fails due to poor design.
And I mean fails, smoke came out.

ebay links (for those who want to blow one up themselves)
http://www.ebay.com/itm/141737411043 HK seller
http://www.ebay.com/itm/221711984831 UK seller *
Note there are only two sellers for the 30W version, which is obviously for a reason.



There is also a 25W version of this, with smaller PCB/inductors. Virtually the same parts.
The russian review linked further down is of this one.

Seller specifications:

Size: 53x38x12mm
Input Voltage: DC 5-35V
Output Voltage: DC 0.8V-35V
Output Current: 3A max
Output Power: 20W, peak 30W (input voltage and output voltage related)
Ripple: 150mV
Conversion efficiency: 90% (with input and output parameters)
Input reverse polarity protection: Yes, reverse current is 0
Output short circuit protection: Yes
Output over-current, over-voltage protection: Yes

reality (sucks)

Size: 53x38x12mm
Input Voltage: DC 5-35V
Output Voltage: DC 0.5V-52.8V from 12V supply (note output capacitor only 35V)
Output Current: not able to test
Output Power: not able to test
Ripple: 150mV
Conversion efficiency: terrible to 90% (some combinations of input and output voltages cause mosfet failure)
Input reverse polarity protection: Yes, reverse current is 0
Output short circuit protection: No
Output over-current, over-voltage protection: No
Thermal overload protection: No, but really needed

Circuitry

Reverse polarity protection
Q1 D4184A 40V 50A N-channel mosfet

Input/output filtering
C2, C3 330uF 35V random capacitors (different brands!)

DC-DC converter power stage
Q2 20N06L 60V 20A N-channel power mosfet
D1, D2 SS36 60V 3A Schottky diode
L1, L2 33uH inductor
C3, C4 unknown MLCC capacitors

DC-DC converter control
IC1 FP5139 PWM controller, top sanded off chip data sheet
D3 SMD zener diode ?
Q3 S8050 NPN transistor J3Y sot-23
R1 682  6.8Kohm
R2 R22  0.22ohm
R3 331  330 ohm
R4 102? 1Kohm
C5-C8 unknown SMD ceramic
10K multiturn trimmer

Circuit from 25W version, some minor differences.


Testing
First I did the usual test anyone does with something with a volume/voltage knob, see how low/high it goes.
But this was with only a multimeter as a load. 12V in, 0.5V to 52.8V out (actually better than it's ebay listing).

Next was a voltage regulation test, I set it to output 12V, and tried it with two different input voltages.
5V  input 12.00V out
12V input 12.05V out +0.41%

Basic standby (no load) current measurement
Note the blue LED on the output side would draw most of this current.
12V in, 12V out 8.45mA

Super capacitor test
I used the converter to boost 5V to charge a 5.5V 0.33F super cap.
It does work, but did prove my old super capacitor has very high internal resistance.

LiIon charger (crude)
12V input to converter, 4.2V output to recharge 3.7V LiIon laptop cells.
This is a pack I rebuild with salvaged cells from a NOS (new old stock) pack.
The old cells still needed a charge, and the repaired pack needed a jumpstart voltage to get the BMS to wake up.

However my laptop managed to damage the pack somehow, discharging some cells more than others. I suspect it's charger is faulty
I again used the SEPIC converter to recharge the cells, when I got to the last cell pair smoke came out during the charging.

There is a TO220 version of the now dead 20N06L mosfet, I didn't have one on hand. The replacements I found with virtually the same stats didn't work.

Why it failed
Russian review of a 25W version of this product
Has a different mosfet, inductors. But also a circuit, which I'm using.
http://mysku.ru/blog/china-stores/35161.html

google translation

During the tests it was observed that the higher the input voltage, the lower the efficiency.
For example, when the output 15V input efficiency up to 20 Volts 80%, and for 26 Volt only 62%.
The higher the output, the efficiency is even lower. With 20 Volt output I easily got the input current is more than 2 Amps and an efficiency below 40%.
Then I remembered that some of the transistor was a small drop of solder, which was not up to the breakdown and the output voltage after the last of the experiment was 25 volts, and I and the inlet clocked almost 30, he had not even squeak.
Those. it turns out that the transistor literally "has had it." This is due more to the fact that the inductance began to enter in saturation mode.
SEPIC course can work in a wide voltage range, but the optimal range is still tied to the use of components and can not cover everything.

So worse efficiency as the input voltage climbs higher than output voltage, which gets converted to heat. The output inductor will run hot. As the mosfet only has a small die and no heatsink, it simply fails.

As there is no current measurement anywhere in this design, there is no way to prevent failure due to poor efficiency, or output short circuits.
The claimed output short circuit protection, over-current, over-voltage protection is a total lie. It just doesn't exist. The input reverse polarity protection consists of a mosfet wired as a diode, that part does appear to work.

To work properly this really should have:
1. Current sense resistor to monitor the current through the mosfet/output
2. Basic thermal cutout.
3. Better mosfet, which means a proper mosfet driver
4. Heatsinking for the mosfets and inductors.
5. A user manual

* as one seller calls this a CuK converter (it isn't), here is how to tell the difference.

If the diode and mosfet are directly connected its a CuK, if not it's a SEPIC.
The other way is to measure between the + out, and - in. If you get a negative reading it's a CuK.

[Kilrah]

LiIon charger (crude)
12V input to converter, 4.2V output to recharge 3.7V LiIon laptop cells.
I again used the SEPIC converter to recharge the cells, when I got to the last cell pair smoke came out during the charging.

What was used to limit the charge current? 

[station240]

LiIon charger (crude)
12V input to converter, 4.2V output to recharge 3.7V LiIon laptop cells.
I again used the SEPIC converter to recharge the cells, when I got to the last cell pair smoke came out during the charging.

What was used to limit the charge current? 

I monitored it with a multimeter, starts at 400mA and drops during charging. The pack's BMS was still in circuit anyway.

The recommended max charge current for these cells is 5A, as I have two in parallel.
As the converter cannot supply more than 3A it's not possible to overcharge.

[chris_leyson]

I was curious as to how the short circuit protection works, so I had a look at the datasheet. https://www.insidegadgets.com/wp-content/uploads/2015/07/FP5139.pdf Under heavy load conditions the voltage at the feedback pin drops and the error amplifier starts to charge the soft start capacitor until the SCP threshold is reached (typ 0.8V). It's very crude, there is no current sense, it relies on sensing the output voltage.

The low efficiency you were getting at higher input voltages is possibly due to very slow mosfet switching times, datasheet says minimum source current 20mA, typical 40mA, minimum sink current 30mA typical 40mA.

Hot inductors at high input volts, not surprised really, they're a bit on the small side for a 25W converter given that the switching frequency is limited by the poor drive, large volt time product, small number of turns and small core area equals high flux density.

The FP5139 is not a very well designed controller and whoever designed copied the circuit from the datasheet isn't a very good designer. Any fool can copy a circuit from a datasheet and not have a clue.

Avoid these converters.

[station240]

I was curious as to how the short circuit protection works, so I had a look at the datasheet. https://www.insidegadgets.com/wp-content/uploads/2015/07/FP5139.pdf Under heavy load conditions the voltage at the feedback pin drops and the error amplifier starts to charge the soft start capacitor until the SCP threshold is reached (typ 0.8V). It's very crude, there is no current sense, it relies on sensing the output voltage.

The low efficiency you were getting at higher input voltages is possibly due to very slow mosfet switching times, datasheet says minimum source current 20mA, typical 40mA, minimum sink current 30mA typical 40mA.

Thanks for the explanation, unless their is a near dead short then it's got no over current protection.
Yes poor drive current, which makes slow on/off ramp times. Designer failed to take into account mosfet heating during high duty cycles.

Hot inductors at high input volts, not surprised really, they're a bit on the small side for a 25W converter given that the switching frequency is limited by the poor drive, large volt time product, small number of turns and small core area equals high flux density.

I measured 100kHz on the gate drive, after I removed the dead mosfet.
The inductors pictured are from the 30W (actually 20W) converter. The 25W version has much smaller inductors.

The FP5139 is not a very well designed controller and whoever designed copied the circuit from the datasheet isn't a very good designer. Any fool can copy a circuit from a datasheet and not have a clue.

It's not even a copy, as the totem pole transistors have been left out to save a few cents.
Freeling Technology don't think much of their FP5139, as it's not on their list of products anymore.
The Anitrashpress sellers have this IC at 10 for US$5

Yes someone could add circuitry to add proper current limiting, but I think using a better IC that has that already has pins for monitoring would be easier.

[Kilrah]

As the converter cannot supply more than 3A it's not possible to overcharge.

That's a common mistake and is the reason I made my post.
The PSU is rated at 3A i.e. can only be used safely (in general, not talking of this particular... thing) up to that current, but unless there is an actual current limit (rare) it will certainly supply more if the load "asks" for it until it burns up, which in the case of a Lithium cell could happen after the load does.

[Wolfram]

In a SEPIC converter, the peak switch voltage is the sum of the input and output voltage. The 40 V rating on the MOSFET used in this converter is way too low for the specified input and output voltage ranges. This is likely the reason for the low efficiency at high input and/or output voltages, where the MOSFET will go into avalanche breakdown and dissipate part of the energy stored in the inductors.

A second design fault is the use of uncoupled inductors together with ceramic series capacitors. This often leads to parasitic LC oscillation between these components.

[iso14000]

moreover, current in the transistor is the sum of both input and output currents ! transistor in this kind of topology is .... sensible.

I have no idea about the the relationship between Coupled/UnCoupled inductors against LC parasitics oscillation, but coupled inductor allow the ripple to be split between both windings and separate inductor can be tuned according the average current (they are different for input and ouput).

A++