Tim said:
Schematics do not just trip lightly from one's fingers -- they take
work. And work ain't free.
An off-line switcher circuit, modified to provide constant current to a
string of LEDs, would be one candidate for 'best', although not for
'cheapest'.
The biggest part of any 'best' design would be part selection -- most of
the schematics are probably OK if you don't use the cheapest of Chinese
parts.
The FIRST thing to do is firm-up your requirements.
The most important issues are not specified.
Do you want to learn about technology or provide LIGHT?
You can learn most of what you need without building anything.
Assuming you want light...
Do you care what it costs?
If money is no object, go to Home Depot and buy LED lights.
If cost is an issue, go to Home Depot and buy CFL lights.
How much are you willing to invest in a few percentage points of
efficiency? Do the math.
Are we done??? Guess not...
So, you're determined to build something...
If your leds are all the same and their reverse voltage rating is at least
as high as the forward drop, you can make two strings, hook them in
reverse parallel. Put that in series with a capacitor sized for the peak
current and frequency you want/have. Voila...light.
(note that if you buy unknown leds from an unknown vendor with unknown
specs, you have no guarantee that all the lights in the batch will
have the same characteristics. 'Bout all you can count on is that
they were probably swept off the floor of the "reject devices" room
with the same broom...maybe...)
The first design has serious problems with line spikes. Anything
that increases the dV/dT of the AC can pop the string in a heartbeat.
So, don't turn off your blender.
You can mitigate this with some series resistance. You trade simplicity
and efficiency for reliability.
This light will flicker. If that is a problem for you, you're gonna have to
go to DC or higher frequency AC Drive.
White leds have a phosphor with some persistence. I've never tried to
measure the light to see how much that helps.
For DC, you already stated the plan. Split the parallel strings and use
a resistor for each string and put them in parallel (not reversed).
Use a cap and a rectifier. You fixed the flicker and made the cap or the
rectifier the casualty of line spikes. More series resistance on the
input side.
By now, you have a working light with poor efficiency, poor reliability,
poor safety...the list of "poor" goes on.
Go to eetmag.com.
Browse the pages and you'll find advertising for different kinds of
chips to power LED lights. Go to the vendor sites and get reference
designs.
You can probably build a breadboard of a pretty nice light.
But it has problems....
How are you gonna package it so that it's safe, (doesn't electrocute
your kids
or burn down your house) doesn't radiate noise that shows up on your TV or
causes your X10 devices to malfunction, fits in the place you want to
put it,
dissipates the waste heat safely?
The other problem is where are you gonna get parts? Want 10,000/month,
no problem. IF you want 1 part, even if you can get it, by the time
you add shipping and handling for three different components
from different places, you're gonna spend more than the cost
of a REAL commercial LED light.
If you want to have an interesting discussion, call up your fire insurance
carrier and explain how you're gonna fill the place with untested,
unapproved, uninspected
lights tied into the electrical grid. Ask if that affects your fire
coverage.
Lights of America has some interesting devices. Can get 'em at Walmart
or Costco for about $6 each. They're 2 Watts 110 lumens or so.
Stock varies by store, so call ahead. You'll find lots of reliability
complaints for LOA stuff, but my experience has been good.
Last year's design has 20 old-style axial leaded LEDs pointing straight up.
This year's design has a stalk in the center with 4 vertical strips of
surface mount leds along the stalk pointing sideways and 3 more on top
pointing up. Somewhat more light output, and much better light
dispersion for a typical incandescent replacement.
It also has a heat sink in the base suggesting that it's
actively managing the power.
The light output is probably as good as you're gonna get with a DIY
design. Also worth the price to take one apart and see how they do it.
I just use 'em, so don't know or care what's inside. I'll take one
apart when it quits working.
I have three of them burning 24/7 to light my house. My neighbors call
me the "mole", but I find I can easily get from my computer/tv chair to
the kitchen for coffee and the bathroom to dump coffee. Only have to
turn on the CFL if I want to read something on paper.
Bottom line, building a custom LED to learn about it can be a fun
time waster...aka hobby.
Building a custom LED lamp or two to get light is a fool's errand.
As always, YMMV.
Are we having fun yet?