Note: This is not meant to be a complete guide to building a SEPIC based DC-DC converter from scratch. This is just a gallery and some notes I encountered when I built one.
I needed various DC-DC converters for many applications so I designed and had boards made for a converter module that can be configured for SEPIC or Boost topology converters. I also incorporated output current limiting for constant current operation. Applications I had in mind are the usual voltage converters, chargers for large battery packs and high power LED drivers.
Top PCB layout.
Bottom PCB layout.
The basic circuit uses the very old but still widely used TL494 since I had a bunch of those. The two internal comparators are both used for current and voltage feedback. A separate circuit using a MAX4172 is used for output current sensing. It can also be eliminated and a ground side resistor used if high side current sensing / direct connection of input ground and output ground is not necessary.
The prototype used all Panasonic FM low impedance caps due to the high ripple currents inherent with switchmode supplies. The SEPIC inductors are wound on a common T106 Hi-Flux core. 9 turns x2 with litz wire made from 20 strands of #31 enameled wire. The board can also be used in boost converter mode by simply using a single winding coil and linking two pads.
Some problems I encountered during debugging of the prototype:
First application I have for it is a 15Ah LiFePO4 battery pack charger which needs about CC at 3A and CV at 14.60V.
I had a really hard time stabilizing the circuit. The CV mode is fine, transient response works well with a small overshoot and there's a small lag at no load to full load response.
In CC mode, current is maintained but the circuit goes crazy and makes buzzing noises.
The point when the circuit goes crazy is at moderate loads. It occurs well below the point where the circuit goes from discontinuous mode to continuous mode but I tried another coil with more turns and still occurs at approx same current level.
At first, when I wound the inductor using single #18 wire per winding but after switching to 20 strands of #31 wire per winding, it made things much better but the oscillation in current mode persisted but much less so than using the single #18 wire. Turn counts were the same. Switching frequency is around 56kHz. I couldn't explain the mechanics of it. It must have something to do with better current flow due to the litz wire at high freq and high current but don't know much more.
I later noticed that the instability occurs at about >2.7A input current. If input voltage is increased or output current limit threshold is lowered to maintain 2.7A input current.
I worked on and off on this thing for about two weeks until I stumbled upon some application notes I found in the internet regarding the complexities of SEPIC loop control. Ended up having to add a series RC network (3K + 1uF) across R32 to greatly slow down the current feedback.
Once it was slowed down, circuit behaved perfectly. It's actually fine since the load does not need a fast response since in battery charging, the voltage and current does not change quickly. One minor problem I have is that if the circuit is powered before the battery is connected, once battery makes contact, there is a huge current spike (approx 13A peak - calculated) at the output as it takes a bit before the current limiting action takes effect. If the battery is connected before power applied to the circuit, there is no problem as the circuit has soft start implemented on the TL494 via C4, R7 and R8.
Although the PCB also has options to be used as a boost converter, I don't have immediate application for that so it is not yet tested for that use. I will update this page once I test a boost converter version of this circuit.
Here are some pics of the prototype unit used as a LiFePO4 battery pack charger:
Professionally made PCB.
Different views of the completed board.
Bottom side of converter board (some resist is left off to add solder to improve current carrying capacity of the PCB tracks.
The two trimmers that set output voltage and current limiting. I didn't have two 20K multiturn trimmers so I used a single turn trimmer in the current adjust. (bodge circuit is also visible - needed to stabilize the complex transfer function of the SEPIC topology during CC mode).
Output current display - I chopped off the display driver part of the DIY RTA board for a quickie display module.
High side current sensor.
Current indicator in full. Max indication is for currents >800mA output.
Completed converter/charger module.
Top cover showing milled slots for cooling.
The battery pack I made this charger for.
13 May 2013:
I built three new ones using the same PCB for the boost converter versions. Each are going to be used to power a 28W LED array using 28pcs Cree XP-E in 7S4P config. Open circuit voltage is adjustable from 16V to 30V and constant current is adjustable from 20mA to 1.8A. These did not use the MAX4172 current sensors but instead used a 0.27Ohm shunt resistor to the - return path. The two PCBs are smaller since the controller section was cut off and mounted vertically. The bigger PCB was the initial one used for debugging and fine tuning.
15 June 2013:
The COB LED strips have arrived so I started work on the LED lighting that is powered by the Alternative power system at home. I used an 18" lenght of aluminum C channel and mounted the LED strips on it. The LED strips I used are two of this in series powered by the boost converter above. It actually turned out to be too bright. Much brighter than the 4 foot 40watt flouro tube even at only 1A driving current that I had to turn it down to 550mA and still it was brighter than the flouro tube.
Some detail shots of the assembly. I left the flouro lamp on so that you can see how the entire assembly was made.
LED hanged in the ceiling.
View of the room (no flash and mains flouro lamp turned off).
Another test setup I made using a different LED module. I made a mistake in the order and chose a warm white instead of cool white.
Shot of the room. This one is installed in the workshop. This one runs at about 700mA.
04 Aug 2013:
Here's another iteration of the design. The two boards are configured as boost converters then the current sensors (MAX4172) are fixed to current limit at 4A. One is configured as slave so that the switching frequency is synced and reduce unwanted EMI. The two converters share the current pretty well once the output voltage is properly set.
The parallel boost is used in the alternative power home system
Another picture from the side.
Page created and copyright R.Quan © 19 Nov 2012.