From this explanation it seems like the circuit should work with a higher
voltage LED.  Has anyone tried to power a blue LED from a single cell?

snip%<-----------------------------------------------------
>>> On 2/2/98 Pasi T Mustalahti <ptmusta@UTU.FI> wrote:
>>>
>>> -------8<--------
>>> >
>>> >        ------------------------------------- +U (0.6..1.55 V)
>>> >        |                    |
>>> >        R20K              || <
>>> >        |                 || <
>>> >        ----------------> || <
>>> >        |               > || <
>>> >        = 10nF          |    /------>|--|
>>> >        |               ----| BC337     |
>>> >        |                    \          |
>>> >        |                    |          |
>>> >        --------------------------------------
>>> >
>
snip %< ----------------------------------------------------
>The circuit is best described as a self-oscillating flyback converter.
>Each time the transistor turns on, it charges the transformer with
>current until it saturates.  When it saturates, the voltage induced in
>the base winding decreases, causing the transistor to turn off.  The
>energy stored in the transformer is then dumped to the load (LED in this
>case).
>
>For a given frequency of operation and transformer core, the circuit will
>deliver approximately constant *power* to the load, regardless of the
>load voltage.  The power is distributed in pulses having the energy that
>the transformer core can hold before it saturates.  Since it always
>charges until it saturates, the output power doesn't depend on the input
>voltage, if (and this is a big "if"), the frequency doesn't vary.
>
>If the DC supplied to the transistor base circuit is adequate, the
>circuit will oscillate continuously at the highes practical frequency
>(depends on the input voltage and the inductance of the winding).  More
>likely, Pasi's circuit is running in "relaxation mode".  The base current
>required by each cycle discharges the capacitor somewhat, to less than
>Vbe so the transistor doesn't turn on right away after the transformer
>voltage reaches zero (all energy having been delivered to the LED).
>There is a delay during which the resistor has to charge up the capacitor
>to start the transistor conducting again.  So the frequency probably
>decreases significantly with input voltage as the current available thru
>the resistor decreases.  Varying the resistor should vary the frequency,
>and thus the brightness of the LED.  A diode in parallel with the
>transistor base (to keep it from going too far negative) would supply
>current to the capacitor while the transistor is off, and probably get
>the circuit to oscillate continuously rather than in relaxation mode.
>However, the resulting output power would likely be too high, and
>difficult to control.
>
>However, the experiment shows that it had satisfactory performance over
>the life of the battery.  The self-regulating properties of the circuit
>are apparently working well enough.
>
snip%<----------------------------------------------------------------
------------------------------------------------------
Wireless Link Corp.
wllink.com          web site
bryan@wllink.com    email
(408)739-5465 x103  ph
(408)739-5483       fax