BLiVIT

BLiVIT Project Description

One of the things I have noticed missing, for a fairly long time, is a good *safe* lithium cell charger that simultaneously supplies a load with a good efficient regulator. My answer is the BLiVIT. This project is a moderate-sized board that incorporates charging from either USB or a 5V DC “wall wart” adapter, efficiently manages a single Li+ or LiFePO4 cell, has a power-on/off soft (using an elevated Powah-1 module, also described here), a fully isolated load supply (no need to blow up my scope when the USB is connected and my probe accidentally touches something else also earth-ground referenced), and an efficient boost/buck regulator that supplies either 3.3V or 5V.
Elsewhere in the open source hardware blogosphere, I have been seeing other LIFePO4 chargers. Most of the designs out there have been intended for standalone charging function only. Most often these were for use on cell form-factors similar to AA and C type alkaline batteries.
The deficiencies of this design are serious, and I will enumerate some of them here.
The common AA “rechargable” cell chemistry does not allow for a great deal of charge density. There are also a lot of thermal problems, mostly having to do with the shape of the cells. Another problem is that the typical charger device has about 4 cell slots, and the charging circuitry simply is not designed properly to balance charging cycles in multiple-cell geometries.
That is why I have elected to support a single cell in the BLiVIT project. Many characteristics and compositions of cell chemistry are now available. In my prototype I have chosen a cell with 950mAH charge capacity, while taking up only about 37 x 50 x 7 mm volume.
The next major deficiency of typical charger systems is that they rely on physically isolating the cell from the power load during the charging cycle. Aside from design life issues with mechanical connectors there are safety issues with handling a fully charged (or worse, a recently charged) lithium cell. That is why for the BLiVIT I required the capability of supporting the power load during the charging cycle while also striving for preserving cell working life. There are numerous articles and books published which warn very strongly that ineffective charging cycle management will have severe effect on the working life of the cell or present a fire hazard.
The necessity of overcoming some of these limitations of previous charger designs required much research. I did not want to use jellybean parts alone, so I opted to just start at the “source end” of the project and work my way forward to the “business end” of the supply.

I exhaustively worked over many datasheets for power management devices and found many that "almost" fit my needs. The first criterion for the device was to have the outside circuitry be as simple as possible. The second being that at least one of a DC adapter or USB power be permissible for sourcing the charger. I finally came across a part from Maxim, the MAX1874. Happily, the MAX1874 can be used with both sources, the DC power taking precedence when connected.
My next requirement was to be able to drive the load or charge the cell with automatic switch-over. It was nice to find out that the MAX1874 could do that as well, as well as drive the load circuit.
Thermal management with lithium cells is important. Lithium ion cells generate a lot of heat, which can be slow to dissipate depending on the geometry of the cell. The MAX1874 has a thermistor sensor feature that I have chosen to employ with an NTC thermistor that has resistance 10K at 25 C. The combination of constant current sensing and temperature sensing means the charging cycles can be constantly monitored for best efficiency.
That accounts for most of the front-end of the BLiVIT. Next we must look at downstream requirements.
Even if the power load is miniscule it is not zero, so a soft-toggle switch is necessary. For noise reasons I also wanted the switch to be significantly debounced, and physically elevated from the BliVIT. The standalone module called the Powah-1 will do these things, and provide physical safety due to the fact that the operating switch is kept well away from the Li+ cell. The Powah-1 is simply a pair of Schmitt inverters with a delay circuit on one of the gates to ensure consistent startup state at power-on. The enabling gate turns on a FDN302 PFET, with high enough Vgs of 20V to adequately carry the expected load of 100mA constant current. The standby drain of the Powah-1 is lower than the 20uA current minimum of my instruments.

Downstream of the Powah-1 I use an SN6501 isolation transformer driver chip which directs the load supply through a transformer with 1:1 winding ratio.

After the isolation stage I use a ADP2503 switching boost/buck regulator to convert any of the input voltage to a locked output voltage. I am using an adjustable regulator so that a jumper is used to select 3.3V or 5V output, which is then available at a header or a screw terminal block.

Theory of Operation

The ferrite bead inductor L3, with C13 is the AC decoupling filter consistent with USB 1.1 / 2.0 requirements. The PFET Q1 is the DC Power enable switch activated by /DCOK from U1, the MAX1874. The PFET Q2 is the USB power enable switch activated by /UOK from U1. The PFET Q3 is the power output enable switch controlled by PON. This connects the battery terminal to the load when PON is low, indicating the external sources are disconnected. TR1 and R7 uses the output of the built-in precision reference pin REF to measure the ambient termperature of the lithium cell. Jumper J7 directs either bypass voltage or ground to the USEL pin, which sets the constant current load of the input power to either 100mA (BYP) or 500mA (GND). The lithium cell (or external source) power is directed to the switching pad SW_IN. This connects to the input of the Powah-1 module, which is elevated on long KK-style pins for safety and form-factor reasons. The Powah-1 output is connected via SW_OUT to the isolation driver U4, which performs the low-power oscillation and switching functions needed to drive the primary winding of the isolation transformer L2. The secondary is connected through protection diodes D3 and D4 to feed the input of the ADP2503 switching regulator U3. U3 output voltage is selected by directing the feedback signal through either of the two resistors R12 and R14, in combination with R13, to the localized ground plane. The inductor L1 provides the switching storage function, and filter capacitors C9 and C10 are normal components of a switching regulator necessary to filter out the induced 1.2MHz ripple from the converter, which then feeds J3 and J4.  Inductors L3 and L4 are ferrite beads for providing downstream EMI protection to both the regulator and the components upstream of the isolation point. Note that at no point is the ground plane of the regulation section connected prior to the transformer secondary.

Powah-1

On the module U1A and U1B are both Schmitt inverters connected in a latch configuration. The RC circuit R2 and C1 ensure that the input to U1A has a slight delay reaching the schmitt activation level than that for U1B. This ensures that the output of U1A is set high right away, but U1B is held off a little bit, and thereby U1A takes precedence. The gate voltage set to logic high turns off the PFET Q1 until the RC circuit is discharged through the NO tact switch S1, which reverses the logic levels on the latch, and thereby dropping the gate voltage and turning on Q1.  In another article I will delve further into what the magic of the "Qi" header means.

P01-109 BLiVIT Assembly

P01-109 BLiVIT Project

BLiVIT Project Enclosure

Powah Module

BLiVIT Project Wireless Receiver

BLiVIT Project Wireless Transmitter

20130607 Update

I have made a slight modification.  Instead of the power output going directly to the barrier terminals I terminate it at one half of a four-post header.  The adjacent two pins of the header then go to the barrier terminals.  This allows me to have add-on boardlets that can do things like draw off and invert the power to a charge-pump circuit.  Then I can feed that back out to form up a split-rail power supply.  I think I can easily have a 1cm^2 boardlet that creates a +9/-9 supply.  To make that happen I would reduce the supported current load to about 50mA.