Discrete LM3909: Green LED at 1.15 V

The green-LED discrete LM3909 is still flashing, even with its AA NiMH cells burned down to 1.15 V:

LM3909 green LED - 1.15 V NiMH
LM3909 green LED – 1.15 V NiMH

If the truth be known, one of the cells is now reverse-charged to 200 mV, so that’s a bit beyond as low as it can go.

The flash period has stretched to 8.7 s:

LM3909 green LED - 1.17 V - 8.7s period
LM3909 green LED – 1.17 V – 8.7s period

The circuit boosts the battery by 800 mV to put 1.94 V across the green LED at the start of each flash:

LM3909 green LED - 1.15 V - V LED
LM3909 green LED – 1.15 V – V LED

Admittedly, the LED isn’t particularly bright at 2.8 mA:

LM3909 green LED - 1.15 V - LED current
LM3909 green LED – 1.15 V – LED current

But it’s still flashing!

Swapping the cells into the LM3909 with a blue LED doesn’t produce any blinking, which is about what the earlier tests showed.

More AAA-to-AA Alkaline Adapters

Having a handful of not-dead-yet AAA alkalines and a bunch of LED blinkies built for AA alkalines, a pair of adapters seemed in order:

AAA-to-AA Alkaline Adapters - installed
AAA-to-AA Alkaline Adapters – installed

The blinkies need a somewhat wider base than they’d get from a pair of AAA alkalines, so it’s not quite as dumb as it may seem.

In any event, the positive terminal comes from a brass rod:

AAA-to-AA Alkaline Adapters - brass terminal
AAA-to-AA Alkaline Adapters – brass terminal

Nobody will ever see the fancy Hilbert Curve infill around the brass:

AAA-to-AA Alkaline Adapters - end view
AAA-to-AA Alkaline Adapters – end view

In this application, they’ll go from not-dead-yet to oh-it’s-dead faster than AA cells, so I can watch how the blinkies work with lower voltages.

Discrete LM3909: Blue LED Radome

Dropping a simplified ping-pong ball radome for a Piranha RGB LED atop a discrete LM3909 on the AA alkaline cell holder:

Discrete LM3909 Radome - AA alkaline
Discrete LM3909 Radome – AA alkaline

The solid model has screw holes for the lid and the revised LED spider:

Astable Multivibrator - Alkaline AA Base - radome - solid model
Astable Multivibrator – Alkaline AA Base – radome – solid model

The RGB LED needs only two wires, as the LM3909 circuit can blink only one LED. I tried all three colors, but only blue and green justify the LM3909 hairball; red can get along with the astable circuit.

The LED wires connect across a 1 MΩ resistor serving as a mechanical strut between the 9.1 kΩ resistor on the left and the 10 Ω ballast resistor on the right.

Fresh alkaline cells at 3.0 V put 3.3 V across the blue LED with a 37 mA peak current. Older cells at 2.3 V produce 2.9 V at 15 mA. Dead cells at 1.9 V still fire the LED with 2.7 V at 4.2 mA, although the flash is barely visible in ordinary room light.

The lovely blue ball looks better in person!

The OpenSCAD source code as a GitHub Gist:

// Astable Multivibrator
// Holder for Alkaline cells
// Ed Nisley KE4ZNU August 2020
// 2020-09 add LED radome
/* [Layout options] */
Layout = "Build"; // [Build,Show,Lid,Spider]
/* [Hidden] */
CellName = "AA"; // [AA] -- does not work with anything else
NumCells = 2; // [2] -- likewise
Struts = -1; // [0:None, -1:Dual, 1:Quad] -- Quad is dead
// Extrusion parameters
/* [Hidden] */
ThreadThick = 0.25;
ThreadWidth = 0.40;
HoleWindage = 0.2;
function IntegerMultiple(Size,Unit) = Unit * ceil(Size / Unit);
function IntegerLessMultiple(Size,Unit) = Unit * floor(Size / Unit);
Protrusion = 0.1; // make holes end cleanly
inch = 25.4;
//- Basic dimensions
WallThick = IntegerMultiple(3.0,ThreadWidth);
CornerRadius = WallThick/2;
FloorThick = IntegerMultiple(3.0,ThreadThick);
TopThick = IntegerMultiple(2.0,ThreadThick);
WireOD = 1.5; // battery & LED wiring
WireOC = 4;
Gap = 5.0;
// Cylindrical cell sizes
// https://en.wikipedia.org/wiki/List_of_battery_sizes#Cylindrical_batteries
CELL_OD = 1;
// FIXME search() needs special-casing to properly find AAA and AAAA
// Which is why CellName is limited to AA
CellData = [
["18650",18.8,65.2], // bare 18650 with button end
["18650Prot",19.0,70.0], // protected 18650 = 19670 plus a bit
CellIndex = search([CellName],CellData,1,0)[0];
echo(str("Cell index: ",CellIndex," = ",CellData[CellIndex][CELL_NAME]));
//- Contact dimensions
CONTACT_THICK = 3; // plate thickness
CONTACT_TIP = 4; // tip to rear face
CONTACT_TAB = 5; // solder tab width
ContactData = [
["AA+",12.2,12.2,0.3,1.7,3.5], // pos bump
["AA-",12.2,12.2,0.3,5.0,3.5], // half-compressed neg spring
["AA+-",28.2,12.2,0.3,5.0,0], // pos-neg bridge
function ConDat(name,dim) = ContactData[search([name],ContactData,1,0)[0]][dim];
ContactRecess = 2*ConDat(str(CellName,"+"),CONTACT_THICK);
ContactOC = CellData[CellIndex][CELL_OD];
WireBay = 6.0; // room for wiring to contacts
//- Wire struts
StrutDia = 1.6; // AWG 14 = 1.6 mm
StrutSides = 3*4;
ID = 0;
OD = 1;
StrutBase = [StrutDia,StrutDia + 2*5*ThreadWidth, // ID = wire, OD = buildable
FloorThick + CellData[CellIndex][CELL_OD]]; // LENGTH = base is flush with cell top
//- Holder dimensions
BatterySize = [CellData[CellIndex][CELL_OAL] + // cell
ConDat(str(CellName,"+"),CONTACT_TIP) + // pos contact
ConDat(str(CellName,"-"),CONTACT_TIP) - // neg contact
2*ContactRecess, // sink into wall
echo(str("Battery space: ",BatterySize));
CaseSize = [3*WallThick + // end walls + wiring partition
BatterySize.x + // cell
WireBay, // wiring bay
2*WallThick + BatterySize.y,
FloorThick + BatterySize.z
BatteryOffset = (CaseSize.x - (2*WallThick +
CellData[CellIndex][CELL_OAL] +
) /2 ;
ThumbRadius = 0.75 * CaseSize.z;
StrutOC = [IntegerLessMultiple(CaseSize.x - 2*CornerRadius -2*StrutBase[OD],5.0),
IntegerMultiple(CaseSize.y + StrutBase[OD],5.0)];
StrutAngle = atan(StrutOC.y/StrutOC.x);
echo(str("Strut OC: ",StrutOC));
LidSize = [2*WallThick + WireBay + ConDat(str(CellName,"+"),CONTACT_THICK), CaseSize.y, FloorThick/2];
LidScrew = [2.0,3.8,7.0]; // M2 pan head screw (LENGTH = threaded)
LidScrewOC = CaseSize.y/2 - CornerRadius - LidScrew[OD]; // allow space around screw head
//- Piranha LEDs
PiranhaBody = [8.0,8.0,8.0]; // Z = heatsink fins + body + lens height
PiranhaPin = 0.0; // trimmed pin length beyond heatsink
PiranhaPinsOC = [5.0,5.0]; // pin XY distance
PiranhaRecess = PiranhaBody.z + PiranhaPin/2; // minimum LED recess depth
BallOD = 40.0; // radome sphere
BallSides = 4*StrutSides; // nice smoothness
BallPillar = [norm([PiranhaBody.x,PiranhaBody.y]), // ID
norm([PiranhaBody.x,PiranhaBody.y]) + 3*WallThick, // OD
StrutBase[OD] + PiranhaBody.z]; // height to base of chord
echo(str("Pillar OD: ",BallPillar[OD]));
BallChordM = BallOD/2 - sqrt(pow(BallOD/2,2) - (pow(BallPillar[OD],2))/4);
echo(str("Ball chord depth: ",BallChordM));
// Useful routines
module PolyCyl(Dia,Height,ForceSides=0) { // based on nophead's polyholes
Sides = (ForceSides != 0) ? ForceSides : (ceil(Dia) + 2);
FixDia = Dia / cos(180/Sides);
cylinder(r=(FixDia + HoleWindage)/2,h=Height,$fn=Sides);
// Spider for single LED atop struts, with the ball
module DualSpider() {
difference() {
union() {
for (j=[-1,1]) {
translate([0,0,StrutBase[OD]/4]) // connecting bars
for (j=[-1,1]) // strut wires
for (n=[-1,1]) // LED wiring
translate([0,0,BallOD/2 + BallPillar[LENGTH] - BallChordM]) // ball inset
translate([0,0,BallPillar.z - PiranhaRecess + BallPillar.z/2]) // LED inset
cube(PiranhaBody + [HoleWindage,HoleWindage,BallPillar.z],center=true); // XY clearance
translate([0,0,StrutBase[OD]/2 + WireOD/2 + 0*Protrusion]) // wire channels
cube([WireOD,BallPillar[OD] + 2*WallThick,WireOD],center=true);
//-- Overall case with origin at battery center
module Case() {
union() {
difference() {
union() {
for (i=[-1,1], j=[-1,1])
translate([i*(CaseSize.x/2 - CornerRadius),
j*(CaseSize.y/2 - CornerRadius),
cylinder(r=CornerRadius/cos(180/8),h=CaseSize.z,$fn=8); // cos() fixes undersize spheres!
if (Struts)
for (i = (Struts == 1) ? [-1,1] : -1) { // strut bases
for (j=[-1,1])
cube([2*StrutBase[OD],StrutOC.y,StrutBase[LENGTH]],center=true); // blocks for fairing
for (j=[-1,1]) // hemisphere caps
translate([BatteryOffset,0,BatterySize.z/2 + FloorThick]) // cells
cube(BatterySize + [0,0,Protrusion],center=true);
translate([BatterySize.x/2 + BatteryOffset + ContactRecess/2 - Protrusion/2, // contacts
BatterySize.z/2 + FloorThick])
cube([ContactRecess + Protrusion,
translate([-(BatterySize.x/2 - BatteryOffset + ContactRecess/2 - Protrusion/2),
BatterySize.z/2 + FloorThick])
cube([ContactRecess + Protrusion,
translate([-(BatterySize.x/2 - BatteryOffset + ContactRecess/2 - Protrusion/2),
BatterySize.z/2 + FloorThick])
cube([ContactRecess + Protrusion,
translate([-CaseSize.x/2 + WireBay/2 + WallThick, // wire bay with screw bosses
BatterySize.z/2 + FloorThick + Protrusion/2])
2*LidScrewOC - LidScrew[ID] - 2*4*ThreadWidth,
BatterySize.z + Protrusion
for (j=[-1,1]) // screw holes
translate([-CaseSize.x/2 + WireBay/2 + WallThick,
CaseSize.z - LidScrew[LENGTH] + Protrusion])
for (j=[-1,1])
translate([-(BatterySize.x/2 - BatteryOffset + WallThick/2), // contact tabs
BatterySize.z + FloorThick - Protrusion])
(BatterySize.z - ConDat(str(CellName,"+"),CONTACT_HIGH))
if (false)
translate([0,0,CaseSize.z]) // finger cutout
translate([0,0,ThreadThick - Protrusion]) // recess around name
if (Struts)
for (i2 = (Struts == 1) ? [-1,1] : -1) { // strut wire holes and fairing
for (j=[-1,1])
for (i=[-1,1], j=[-1,1]) // fairing cutaways
translate([i*StrutBase[OD] + (i2*StrutOC.x/2),
PolyCyl(StrutBase[OD],StrutBase[LENGTH] + 2*Protrusion,StrutSides);
linear_extrude(height=2*ThreadThick + Protrusion,convexity=10)
module Lid() {
difference() {
for (i=[-1,1], j=[-1,1], k=[-1,1])
translate([i*(LidSize.x/2 - CornerRadius),
j*(LidSize.y/2 - CornerRadius),
k*(LidSize.z - CornerRadius)]) // double thickness for flat bottom
translate([0,0,-LidSize.z]) // remove bottom
cube([(LidSize.x + 2*Protrusion),(LidSize.y + 2*Protrusion),2*LidSize.z],center=true);
for (j=[-1,1]) // wire holes
for (j=[-1,1])
// Build it!
if (Layout == "Case")
if (Layout == "Lid")
if (Layout == "Spider")
if (Struts == -1)
if (Layout == "Build") {
translate([0,-(CaseSize.x/2 + LidSize.x/2 + Gap),0])
if (Struts == -1)
if (Layout == "Show") {
translate([-CaseSize.x/2 + LidSize.x/2,0,(CaseSize.z + Gap)])

Discrete LM3909: Blue LED Waveforms

The circuitry and instrumentation is essentially the same discrete LM3909 as before:

LM3909 - blue - test setup
LM3909 – blue – test setup

With a few minor tweaks:

  • Blue LED, forward voltage 2.56 to 2.97 V
  • 24 Ω R1
  • One Q2 current mirror transistor driving Q3

With a pair of fresh AA alkaline cells producing 3.1 V (not the NiMH Duracells you see in the picture), the blue LED blinks brightly.

The 610 mV peak voltage across R1 shows the LED starts at 25.4 mA:

LM3909 blue - 3.1 V - R1 24 ohm
LM3909 blue – 3.1 V – R1 24 ohm

The capacitor reaches 1 V, then goes about 150 mV into reverse charge during the flash (note the different horizontal scales):

LM3909 blue - 3.1 V - C1 V
LM3909 blue – 3.1 V – C1 V

The Darlington version of Q1 seems to do a decent job of keeping the cap out of reverse charge. A Shottky diode would add a few hundred mV, but I doubt there’s anything nasty going on inside the cap as it stands.

The blue LED has a forward drop of 2.97 V at 20 mA, so I’m surprised the voltage across it hits 3.1 V at 25 mA:

LM3909 blue - 3.1 V - LED V
LM3909 blue – 3.1 V – LED V

Very little of the voltage appears across Q3, the driver transistor:

LM3909 blue - 3.1 V - Q3 coll
LM3909 blue – 3.1 V – Q3 coll

With a pair of nearly dead alkaline cells for a 2.0 V supply, the LED current peak drops to 4.6 mA:

LM3909 blue - 2.0 V - R1 24 ohm
LM3909 blue – 2.0 V – R1 24 ohm

The LED lights brightly, then fades away exactly like you’d expect from that waveform.

The cap still charges to about 1 V and stays well above 0 V during the (much longer) flash:

LM3909 blue - 2.0 V - C1 voltage
LM3909 blue – 2.0 V – C1 voltage

The voltage across the LED now reaches only 2.7 V, which is substantially higher than the 2.0 V battery supply and exactly why the LM3909 existed:

LM3909 blue - 2.0 V - LED voltage
LM3909 blue – 2.0 V – LED voltage

Q3 continues to saturate, although you can see the effect of the decreased base drive during the flash:

LM3909 blue - 2.0 V - Q3 coll
LM3909 blue – 2.0 V – Q3 coll

The blue LED won’t light at 1.3 V, but still gives out a weak flash at 1.7 V, so I’d say the tweaked LM3909 circuitry works reasonably well.

Alkaline AA Astable vs. RGB and Yellow LEDs

A fresh pair of alkaline AA cells at 3.2(-ish) V can’t light a Vf = 3 V blue LED with any authority, but I laid out an astable multivibrator circuit with a Piranha RGB LED to see how the colors looked:

Astable AA Alkaline - build test
Astable AA Alkaline – build test

Lighting all three LEDs at once doesn’t make much sense, although I did try it just for the amusement. Spoiler: red wins, even with more-or-less equal currents.

Mounting the hairball on an AA alkaline holder looks better:

Astable AA Alkaline - RGB LED test
Astable AA Alkaline – RGB LED test

Red being the only LED color making any kind of sense meant the Piranha was overqualified for the job, so I replaced all that clutter with a simple 5 mm yellow LED:

Astable AA Alkaline - yellow
Astable AA Alkaline – yellow

It’s shatteringly bright at 20 mA from fresh alkalines at 3.2 V and remains visible down to 1.8 V.

The original circuit schematic / layout doodle:

Astable wiring layout - Piranha RGB test
Astable wiring layout – Piranha RGB test

No surprise in any of this, as it’s why the discrete LM3909 circuitry happened, but it’s nice to have a simple LED atop some alkalines for show-n-tell. If, of course, show-n-tell events ever happen again …

Discrete LM3909 vs. DSO150 Scope

Although I’m a big fan of multi-channel scopes and Hall-effect current probes, a dirt-cheap single-trace oscilloscope can get you quite a ways to the goal:

LM3909 - DSO150 test setup
LM3909 – DSO150 test setup

That’s a genuine JYETech DSO150 powered by an 18650 lithium cell and a boost converter set to 9 V. Make sure you get a genuine DSO150 from an authorized seller, rather than one of the myriad knockoffs; it doesn’t cost much more and tends to reward the right folks.

Anyhow, battery power means you can connect it directly across components to measure what would otherwise be a differential voltage:

LM3909 - Darl Q1 3x Q2 - 1.5 V - R1 V - DSO150
LM3909 – Darl Q1 3x Q2 – 1.5 V – R1 V – DSO150

That’s the voltage across R1, the 39 Ω LED ballast resistor in the discrete LM3909 circuit running from a 1.5 V supply. Divide the 314 mV peak by 39 Ω to get 8 mA of LED current.

The voltage across C1, the timing and boost capacitor, looks like this:

LM3909 - Darl Q1 3x Q2 - 1.5 V - C1 V - DSO150
LM3909 – Darl Q1 3x Q2 – 1.5 V – C1 V – DSO150

So the cap adds half a volt to the supply in order to put 2.0 V across the LED, which accounts for the relatively low current; the green LED has a forward drop of about 2.2 V at 20 mA and 1.9 V at µA-level current.

For completeness, the voltage across the LED:

LM3909 - Darl Q1 3x Q2 - 1.5 V - Green LED V - DSO150
LM3909 – Darl Q1 3x Q2 – 1.5 V – Green LED V – DSO150

So, yup, the LED really does see 2.0 V. I love it when the numbers work out.

Crank the supply to 3 V and see this across R1:

LM3909 - Darl Q1 3x Q2 - 3.2 V - R1 V - DSO150
LM3909 – Darl Q1 3x Q2 – 3.2 V – R1 V – DSO150

The LED current is now 1.23 V / 39 Ω = 33 mA.

The capacitor just barely enters reverse charge:

LM3909 - Darl Q1 3x Q2 - 3.2 V - C1 V - DSO150
LM3909 – Darl Q1 3x Q2 – 3.2 V – C1 V – DSO150

Pop quiz: what voltage to you expect to see across the LED?

I’ll leave further investigation to your imagination, but for low-frequency analog work, you can do worse than a DSO150.

Discrete LM3909: First Light

Another entry in the “The bigger the blob, the better the job” soldering contest:

LM3909 - Darl Q1 3x Q2 mirror - installed
LM3909 – Darl Q1 3x Q2 mirror – installed

It’s a discrete-transistor version of the LM3909 atop the alkaline AA cell holder, with a little PTC fuse for that good safety vibe. The overall layout follows this doodle:

LM3909 wiring layout - Darl Q1 - 3x Q2 gain
LM3909 wiring layout – Darl Q1 – 3x Q2 gain

The four transistors across the bottom row let me test the simulation suggesting there’s no need for the 3× current gain mentioned in the App Note. Spoiler: future LM3909 circuits have the usual two-transistor mirror.

Adding some instrumentation required a bit of unsoldering and clip-lead action: to get the Tek current probe around the LED wiring:

LM3909 - Darl Q1 3x Q2 mirror - test setup
LM3909 – Darl Q1 3x Q2 mirror – test setup

The voltage probe is across the LED, although you’ll also see the voltage across the capacitor and differential voltages measured properly with the common clip leads on the battery negative terminal. I unsoldered two of the mirror transistors after verifying a single mirror transistor can saturate Q3.

Removing the AA cells and feeding it with 3 V from a bench supply:

LM3909 - Darl Q1 1x Q2 - V on C1 - I 3V VCC 10 mA-div
LM3909 – Darl Q1 1x Q2 – V on C1 – I 3V VCC 10 mA-div

The yellow trace is the voltage at the collector of Q3 = positive terminal of C1. The purple trace is the voltage at the LED cathode = negative terminal of C1. The fuzzy white trace is the difference of those two, showing C1 charges to about 1 V at the start of the LED flash. The white wedge over on the left marks the 0 V level, confirming the cap doesn’t enter reverse-charge territory during the flash.

The green LED produces a bright flash starting at 30 mA (bottom trace, 10 mA/div) for 15 ms. With 1 V on the cap, the LED + 39 Ω ballast resistor see nearly 4 V at the start of the pulse, because Q3 saturates around 20 mV.

Reducing the supply voltage to 1.5 V flattens the current and lengthens the flash to 35 ms:

LM3909 - Darl Q1 1x Q2 - V on C1 - I 1.5V VCC 10 mA-div
LM3909 – Darl Q1 1x Q2 – V on C1 – I 1.5V VCC 10 mA-div

The cap still charges to 1 V between on-times, but the lower supply puts barely 2.5 V across the LED + 39 Ω resistor and the current peaks at 10 mA. The increased duration turns the flash into a blink.

It’s good enough, so AA alkalines should last quite a while.