Low power LED flasher 3

Microampere power consumption, 5 years expected battery life (CR2032)

1. Introduction

This is another low power LED flasher, this time built again only with discrete parts, with current consumption of less than 5 µA (3-4 µA typical). The LED flashes once every 2 to 3 seconds.

This circuit is somewhat similar to the previous low power LED flasher 2, but designed from scratch and optimized using SPICE, and without charge pumping (so it can not run on 1.5 volts without modifications, unlike the previous circuit).

A consumption of 4 µA should give 50000 hours (5.7 years) of operating life on a single CR2032 cell (210-230 mAh typical, assuming usable capacity of 200 mAh).

Potential uses: fake alarm, visual effect, "device placement indicator" in dark, etc.

Internal model designation: TFEL-LPFL3-v1.

Circuit board (the battery holder is on the other side) (click for full resolution)

2. Schematic diagram and operation principle description

Schematic diagram (click for full resolution)

2.1 Working principle

Assuming V_BE-on≈0.4-0.5 V at low current, otherwise 0.6-0.7 V, V_SUPPLY=3 V, V_CE-sat=0.2 V.

1. The capacitor C2 (1 µF) gets charged through resistors R5, R3 (470 kΩ). The time constant of R3+R5, C2 is an order of magnitude lower than that of R1, C1.
2. C1 (1 µF) gets slowly charged though R1 (10 MΩ).
3. Voltage on the base of the NPN transistor Q1 rises, let's assume V_BE≈0.4~0.5 V. Voltage on R2+R3 also slightly rises, pulling the emitter of Q1 a bit higher (through C2 the voltage on the emitter of Q2 also very slightly rises). Assume that C2 is almost fully charged now.
4. Q1 (BC847B, same as Q3) starts to pull the base of Q2 (PNP, BC857B) low through R4 (4.7 kΩ). R6 (1 MΩ) sets the minimum Q1 collector current for Q2 to start turning on and limits effects of leakage currents - enough voltage drop has to be created on R6.
5. Q2 starts to turn on, voltage (referenced to the ground) on the emitter of Q1 starts to fall (pulse from E of Q2 is transferred to E of Q1 through C2), which causes V_BE and I_B to increase - positive feedback further fully turns on Q1 by increasing (-)V_BE and (-)I_B), turning both Q1 and Q2 fully on. R2, R4 (4.7 kΩ) limit the currents through (set the discharge time of) C1, C2. When Q2 is turned on and enough current flows through it, Q3 gets turned on. The minimum current to turn on Q3 is set using R7.
6. Voltage on R3 falls to -(V_supply-V_{BE_Q3}-V_{CE-sat_Q2}).
7. Some charge is transferred between C1 and C2 through the B-E junction of Q1, with R2 limiting the current through the emitter of Q1 (thus limiting currents through both the B-E junction of Q1 and through the base of Q2).
   • Assume C2 is fully charged (3 V) and will get discharged to V_{BE_Q1}+V_{BE_Q3}-V__{CE-sat_Q2}≈1.2 V → resulting ΔV_C2≈-1.8 V
   • Assume C1 voltage ≈ 0.5 V
   • Assume approx. 1/3 of the charge in C2 gets transferred into C1 & C1 gets charged "in reverse" (the capacitances are the same): V_C1 drops: 0.5→-0.1 V. The charge transfer happens with C1/C2 essentially in series - R2 sets the time by setting the current, but R3 is much higher - and thus can be neglected here.
8. As the charge transfer gets finished, currents fall, until Q1&Q2 turn off. This is also assisted by the positive feedback through C2. During the pulse when Q1 and Q2 are on, the LED is driven through Q3.
9. The cycle repeats.

Rough estimate of times (assuming C1=C2=C, R4=R2): $t_{off}$ = $k_1 \cdot R_1 \cdot C$, respectively $t_{on} = k_2 \cdot R_2 \cdot C$ ($k_1 \approx \textrm{0.22}$; $k_2 \approx \textrm{0.65}$), during $t_{off}$ the LED is off and C1 is being charged and vice versa.. These constants were determined based on simulation results. The time constant of R3+R5 and C2 should be about 5-10 times lower than that of C1 and R1. Bigger changes of those part values (<50%; >200%) or changes to other part values should be verified by simulation or testing.

2.2 Additional notes

R9, C3 can be modified to limit power consumption. In the prototype, 220 Ω and 10 µF were used. LED1 should be a super high brightness red LED with a voltage drop of less than 2 V. TVSs D1, D2 are not necessary, but footprints are present on the board in case the battery and LED are connected externally with long wires in a scenario with a risk of strong ESD. If using a strong supply, like a CR123 cell, add a small fuse in line with the board (100-200 mA).

The circuit was simulated with a wide range of β and according to simulation, should work with all of the BC847/BC857 series (BC847A/BC847B/BC847C, BC857A/BC857B/BC857C), but this was not tested in practice and the behavior (including power consumption and timing) of the circuit changes to some extent. BC547B/BC557B is the through-hole equivalent of BC847B/BC857B and should be usable as well.

Due to very low currents and very high resistances in the circuit, contamination/moisture can affect operation. The circuit board should be cleaned well after assembly.

C3 and C4 play the same role in the schematic and populating only one is OK. They are both there for PCB layout purposes (so two different footprints could be used).

The prototype was built using mostly 0603 SMD parts, but part sizes/packages are not critical.

2.3 Simulation results:

Note: values will vary slightly with changes of simulation settings (timestep, etc). The peak LED current is generally in the order of several mA.

2.4 Additional images:

Flash (single video frame, lower quality) (click for full resolution)

3. Downloads

Archive - KiCad project, LTspice simulation. Note, the simulation parameters will need to be adjusted for reasonable results (max. timestep <100 µs, recommended 10 µs).