Thursday, 29 November 2018

High efficiency LED drive + energy feedback; n > 100%

the switching circuit shown below in Fig 1. operates at approx 130 kHz, in 2 half-cycles a) & b):-

Fig. 1

a) the transistor switches on, its collector current illuminates the output LED(s) and also stores energy in the coil field and electrolytic cap; when the coil saturates then the transistor turns off;  enough energy remains in the electrolytic cap to supply the output LED(s) until the next cycle (the cap voltage is DC with minimal ripple)

b) current supplied from the energy stored in the collapsing coil-field and the electrolytic cap is fed back to the battery (while the transistor is off) via the feedback LED - IFF the feedback LED is illuminated then the battery is receiving charge in this half-cycle (in this case, equivalent to approximately one-fifth of the output current)

(the transformer here is approx 1cm^3 volume of ferrite, solenoidal wound, almost completely enclosed by the ferrite;  the wire is 2x45x0.45cm diam. insulated copper, 5:5 turns)

for these tests, the battery is an LIR providing 4.15V;  supply current is 5.4mA

pk voltage of feedback pulse across feedback LED + battery is approx 7V;  approx 20% duty cycle

Fig 2. Voltage pulse across Feedback LED

so-called 'conventional' current flows OUT of the positive terminal of the battery in half-cycle a) and INTO that terminal in half-cycle b)

voltage across 1ohm Current Sensing Resistor in +ve supply line shows approx same peak magnitude current flows, into and out of the battery, matching the duty cycle for the respective average current flows

Fig 3. Voltage across 1ohm CSR in +ve supply line

(NB.  all values measured using a True RMS meter and confirmed in order of magnitude using a 'scope)

feedback current:  1.18mA (True RMS)
feedback LED load:  approx 2.6V * 1.18mA = 3mW approx

main LED branch current:  6.5mA (True RMS)
main LED branch load:  4.15V * 6.5mA = 26.98mW

voltage across electrolytic//main LEDs is 2.7V DC
main LED load:  2.7V * 6.5mA = 17.55 mW

switching cct load:  26.98 - 17.55 = 9.43mW

total LED load:  17.55 + 3 = 20.55mW

total power load:  20.55 + 9.43 = 29.98mW

supply current:  5.4mA (True RMS)
total power supply:  4.15 * 5.4mA = 22.41mW 

Efficiency  n  =  (total load / supply)  =  (29.98 / 22.41) = 134%

results shown are instanteous power, mW (these are proportional to the energy being converted, mWh)

Thursday, 20 July 2017

The Green Lantern

Purpose of Latest Tests

This is an experiment with looping energy back to the source, using a blocking oscillator as a self-driving switch

the approach used here builds on two previous areas of test

a) my first investigation goes back about 9 years - it looked at an anomaly in conventional charge accounting - the test circuit was a simple switched-capacitor setup, transferring charge from one large capacitor (or battery) to another, using a much smaller capacitor, and an inductor, as a charge transfer method

b) in my second set of tests, just over a year ago, i looked at the use of blocking oscillators as a building-block for LED flashlight and battery-charging applications

my interest in thread (a) was revived recently, when i finally found someone who could give a practical explanation for the phenomenon seen when switching large proportions of 'charge' between capacitors (which many of us had documented as an approx 50% gain in total 'charge', stored in circuit capacitors, in stand-alone circuits, when transferring charge from input to output and using many small 'packets' of charge)

John Decker, a physicist, explains on his website:

the deep-rooted misuse of the word 'charge' when applied to the storage of energy in capacitors and batteries - he suggests the use of the word 'gorge' instead - charge is very definitely conserved - gorge'(as we've discovered, practically) is very definitely not conserved

if gorge is not conserved, this raises the question; "is there some way we can we re-cycle gorge to do more work?"  (my question, not Decker's!)

my latest investigation draws together all these threads, as building-blocks ...attempting to use 'synergy' to get more use from 'energy' 

my original circuit arrangement transferred gorge from input storage to a separate output storage, increasing total gorge as a result

my new circuit configuration returns what was the original output back to be part of the input - to see if gorge can be re-cycled (looped-back) and re-used

Approach used

these are the steps involved:-

first create a series-connected stack with 2 parts A & B - use a battery for store A, above, and either another battery or large capacitance for store B, below

this stack will be used to transfer gorge to a smaller store, Csw (eg. a lower-value capacitance) via a series inductor L1

gorge will then be switched in short pulses from Csw to store B (B2 or C1)

using a blocking oscillator as the switch means that we can arrange for some gorge to return to store B, as we charge the main inductor in the oscillator,

and then we can arrange for its coil-field collapse energy to be directed to return some gorge to store A

so, as an overview of the cycle, we allow a small amount of gorge to transfer, with low loss, from store (A+B) to Csw, then we transfer another small amount of gorge from Csw back to store (A & B) and repeat

if we comply with the transfer requirements, as shown by Decker, then we should be able to increase the total gorge in the system

Circuit operation

a switching capacitor C2 (the Csw), is filter-charged, via an inductor L1, by two similar batteries in series

the action of the blocking oscillator causes the collector winding of T1 to charge as current flows from C2 into B2, and the resultant voltage across the base winding of T1 re-inforces the turn-on of the transistor switch Q1 via C1

when the oscillator reaches saturation point the base voltage drops and Q1 switches off; the collapsing field in the collector winding is able to find a current path via the base winding and the LEDs, this current spike is preferentially applied to B1 (in a lower impedance path than L1, C2)

so B1 & B2 supply energy to the circuit, charging C2, and a proportion of this this energy is then distributed back to B1 & B2, with some being dissipated in the LEDs (and a small proportion of the energy being dissipated as losses in the circuit components, as heat)

Initial Results

The main and flyback windings are each 20 turns of 0.45mm magnet wire on a 20mm diam. x 10mm high ferrite toroid

with a Lantern/Flashlight type setup, using 4x 5mm Hibrite LEDs (green, on this occasion) as a parallel load to the blocking oscillator, and 2x 6V 10AH gel lead-acid batteries as B1 & B2,  the current draw of the active circuit, measured between L1 and C2, is approx 15mA

the base bias resistor VR1 has been adjusted to maximum resistance for this test, but still produces a bright output from the LEDs - the photo (showing the view from above) gives a good indication of the somewhat obscuring effect on the eyes, caused by the circuit in operation

in my ongoing test, both B1 & B2 started with low 'charge', having a combined on-load voltage of approx. 11.7V (both batteries are 'on load' since they are connected in series across L1 and C2, which then supplies current to the blocking oscillator)

initial results are very promising:  IF the battery experienced a current draw of 15mA, each Battery terminal voltage should decrease by approximately 1.5mV per hour - so the expected voltage drop for the combined stack of 2 batteries, over a period of 10 hours, should be in the region of 2 x 1.5 x 10 = 30mV

the combined starting voltage (Vb1+Vb2) for this initial test is 11.73V, therefore we would expect that voltage to decrease to approximately 11.70V after 10 hours continuous operation

the graph below shows the datalog results from the initial test:  the 10hr voltage drop is certainly nowhere near 30mV, it is not even 10% of the expected drop

in fact, at the moment, it is difficult to determine visually from the datalog whether the average combined voltage has dropped at all !   

[Update:  the average rate of combined voltage drop over the last 18 hours is 1mV/hr approx. - this represents a current drain from the batteries which is about 1/3 of that which is being supplied to the blocking oscillator.  If correct, this suggests that the circuit is able to perform at least 50% more work for the same energy  ie. 60mW being supplied --> 90mW being used**]

so it appears that the looping-back of the gorge has had a significant effect on the net current draw from the batteries!   i will continue to monitor the combined terminal voltage of B1 + B2 so that a more accurate value of the actual load on the batteries can be determined

Input power: (Vb1+Vb2) x battery drain current  = 12V x 5mA approx.
Power converted: ((Vb1+Vb2) - Vb2 ) x  circuit supply current = 6V x 15mA approx.]

Monday, 11 April 2016

Accidental discovery!

Last night i put together a quick blocking oscillator circuit to test the feasiblity of it being supplied by Peltier junction devices, stacked and connected in series to develop a higher voltage from hand heat.  The stack certainly worked as i'd hoped - but - the very low impedance of the stack wasn't a good match for the oscillator (even with a 1F buffer capacitor) 

Although my intended test wasn't fruitful, i noticed that when the oscillator was allowed to discharge the large buffer capacitor, on disconnection of the battery (2 very depleted 750mAh NiMHs in series, just under 2.4V total) the circuit sometimes seemed to enter a sustained regenerative feedback mode, after the circuit had reached its low-voltage supply cut-off point

The feedback appeared to be caused by the piezo buzzer component i was using as the capacitive coupling from the flyback output winding to the base of the switching transistor

I replaced a few components with different values and then found that the circuit pulsed at around 20 Hz and the battery terminal voltage was slowly increasing. Red trace below is the AC voltage pulse on the battery terminal, blue trace is the flyback output voltage to the LED

The circuit has been running now for nearly 3 days and the battery terminal voltage has continued to increase slowly, see datalog of on-load battery terminal voltage below (this is NOT the typical 'battery relaxation' effect after previous heavy loading of the battery followed by a low-current loading - the cells had been left depleted and unused for a week or so before this test, and the loading at all times has been at a similar level)

i've tried different output configurations and the best arrangement, so far, seems to be a full-bridge rectifier with red LEDs in the positive output positions

The transformer core is a split ferrite toroid used for reduction of EM interference pickup via cables (toroid is approx 25mm OD, 10mm high); windings use 0.45mm magnet wire

Q1 is a high-gain, low-power device (eg. BC327); C1 is currently 1000uF; D2 & D3 are BAT42 Schottky diodes (not suitable for use as D1);  L1 is around 2mH, using approx 60 turns of 0.45mm magnet wire on a 12mm OD ferrite tube, approx 35mm long

The pnp device is biased 'On' using the reverse-leakage current from D1, which can be either a Germanium or suitable Schottky diode, and the device is switched 'Off' by the inverted output signal fed back via the piezo element;  pulse width is approx 20us

i've used a pnp part only because that is the first high-gain device i happened to pick up in my spares tin - the circuit should achieve the same *interesting* behaviour if it is re-arranged, polarity-wise, to suit an npn device (eg. BC547)


Tests with the Flyback PSU setup continue - and they're giving very interesting results - Watch this space!

Saturday, 9 January 2016

Charging cells using Flyback PSU type circuits

experiments during the last year have focussed on pulse motor operation adapting brushless fan motors, and also battery charging using flyback switchmode power supply type circuits (Boost Converters)

an interesting development has occurred with one particular variant of these test circits:

the circuit was being used to charge a battery of 2 NiMH AAA cells using a similar battery as input; the initial offload voltages of the i/p & o/p pairs were 2.62V & 2.63V (ie. both batteries approximately 50% charged)

after the first test, the end voltages were 2.3V (i/p) & 2.9V (o/p), as might be expected, input nearly fully discharged, output fully charged; the batteries were then swapped and the test repeated

observing the in-circuit terminal voltages as the test progressed the run was interrupted at an appropriate point to determine the 'crossover' voltage where the two batteries become equally charged (after a rest offline)

this voltage was found to be approximately 2.65V (within a few millivolts) - higher than both the original offline battery voltages

it appears that this circuit/battery combination is able to charge its output batteries slightly faster than it discharges its input batteries

more tests are underway to investigate this behaviour - the next test certainly appears to confirm that first observation:

this test circuit is similar to a small flashlight - it has a single white LED which is bright enough to cast a clear shadow on a white surface approx 7' away (>2 metres) in a darkened room; the circuit is powered by a single AAA NiMH (750mAh) and it is also charging a second similar cell on its output

the in-circuit voltage of the input cell (B1) dropped by 3mV (holding around approx 1.3V), while the output cell (B2) has increased by 30mV (from approx 1.29V)

i'll upload some of the graphical output from the datalogging PC, showing the voltage traces, as soon as i can prepare images & co-ordinate file transfers between my Windows & Linux systems

Update: the graph here shows the in-circuit terminal voltages logged for the input & output batteries, B1 & B2.  It can be seen that B2 charges whilst B1 discharges, as the LED is illuminated. After the flashlight has been operated for a while, switch S1 can be used to swap the 2 batteries between i/p & o/p for the next time that the flashlight is used.  In this way, the operation of the flashlight can be extended from the original charge of the NiMHs.  My application here has been for a small flashlight, using 1 run battery, 1 charging battery and a single LED lamp;  the same principle could be extended to use more cells per battery and a larger number of LEDs. VR1 can be used to alter the intensity of the light output (which will alter the discharge/charge rates accordingly) Switch S1 has a central 'Off' position, in addition to the the 2 'On' positions which select the current input battery

Summary, 2015/16...

A quick summary of the state of the 'experiments' listed below:-

Charge Anomaly - i've now learnt that what appeared to be an increase in  total 'charge' stored in a circuit's capacitors, in one of my earliest experiments, is in fact an increase in 'gorge' (as defined by John Denker at - unlike charge & energy, gorge is not conserved and my experiment clearly shows an increase in this quantity within test runs

my DIY cells (those which haven't been cannibalised) are still able to power their circuits; some circuits (eg. digital clocks) have been powered by a more liquid version of the electrolyte in sealed capsules, rather than the original 'gel' version sandwiched between the electrode plates

my spring pendulum worked well for about a year using three DIY cells with liquid electrolyte, but at present the circuit has latched-up pulling down the supply to 0.6V - i think i need to review the circuit and possibly return to the earlier single transistor with transformer version, rather than the 2 transistor driver with inductor

Monday, 3 February 2014

Related developments in 2013/14...

What's been happening?

i decided to try and apply the DIY cell 'technology' to something slightly more demanding than a flashing LED load circuit

over the past year, i've powered several LCD-based clock & weather-station units, using the same basic approach as the cells described here

the higher current-drive requirements of these devices has been achieved by increasing the water content - the electrode area has been reduced

normally, a galvanic cell arrangement such as these, using a few ml of water as a single electrolyte, would have polarised within a few months - needing the water to be changed and probably requiring the electrodes to be cleaned

these devices have been operating for nearly a year now, without attention

at the start of 2014, i wanted to move up to powering something more than just electronics, so i've started an experiment using the basic LED flasher circuit to energise a coil with just enough power to maintain a spring pendulum

it uses a dual transistor pulse circuit, based on the drive part of my original 3 transistor LED flasher; the coil is air-cored to prevent drag on the magnetic 'bob' of the spring pendulum;
the device is powered by 3 DIY cells in series providing a total of approximately 1V on continuous load (current draw is slightly less than 10uA)

Spring Pendulum drive circuit

it retains the energy capture and feedback from the coil field-collapse, via an LED, back to the supply (as featured in the earlier circuits) - and since the coil/magnet of the spring pendulum is similar to a 'Shake Flashlight' type arrangement, some kinetic energy is being converted back to electrical energy and returned to the supply, too

the system is contained within a clear plastic 'bell' cover to reduce effects from any ambient air movement

the LED flash period is approximately 20 seconds, the pendulum period is approximately 1 second (full-cycle):

(apologies for the creaky sound-effects - must be my knees!)

in the video, the DVM is displaying the supply voltage of the relaxation oscillator, which triggers at approx 1.2V and drops to approx 0.75V on pulsing the coil; the oscillator supply then re-charges from the battery supply
the components for the circuit aren't critical

C2, C3 and R1 form the timing of the relaxation oscillator - i selected values for C2 (10uF) and C3 (300uF) which produced sufficient pulse width to give an acceptable 'kick' to the pendulum, and then i selected R1 (150K) to give a repetition period of approx 20 seconds

a certain amount of feedback in the circuit triggers each pulse at a regular point in the pendulum movement (which can be either a vertical linear path, or an arc)

Q1 and Q2 are general-purpose, low-power, high-gain transistors with hFE >= 400

diode D1 is Schottky (for low reverse-leakage/forward-voltage), D2 is Germanium (to provide some leakage current)

C1 needs to be a suitable value to buffer the high impedance of the DIY cells - i have had the system operate with a value as low as 300uF, but 1000uF seems more reliable in the long term

the battery output voltage, on-load, is less than 2V, so low-voltage capacitors can be used, to reduce physical size 

LED1 is a hi-brite white type

the drive coil was hand-wound onto a card spool;

DC resistance is approx. 20 ohm
(a few hundred turns of multi-strand insulated copper wire, 7/0.09mm, air-core);
approx 30mm diameter, 10mm high, 10mm diameter air-gap
the battery consists of 3 DIY cells, 1 strip each of 15mm-wide copper and 5mm-wide zinc, approx 35mm submerged in about 2mL of 1:1 honey:tap-water solution, well-mixed; reasonably-well sealed with a plastic cap

the copper foil is bent into a 'U' shape to partly overlap the zinc strip on either side; a piece of sponge/foam is used to keep the two electrodes separate at the bottom of each cell

Friday, 13 July 2012

Cell #2: now 1+ years operation, voltage increased

Cell stack #2 has now been operating continuously for nearly 17 months, on load to the same LED flasher circuit

the graph below shows that the average terminal voltage for the cell stack, as indicated by the trend of the data, has been increasing since March 2011 (approx. 1.3 years so far, as at this post)

the voltage data still shows a strong inverse correlation with the day-to-day ambient temperature readings