That's not quite current and correct on Cree's highest efficiency as they have up around 254 Lumen's per watt.    At least that's what they've had roughly in the lab for over a year.  In production units the last I checked are around 180 Lumens per watt.  I haven't checked all their models but that was the last I heard a few months ago on the XM-L.  I've got one of those in a U-bin (their highest BIN) and it is extremely bright when running around 9 or 10 watts.  It's easily like a 60 watt bulb equivalent. 

And a comment about efficiency of this latest overall circuit Lasersaber has.  The actual LED's in these units are likely XM-L models unless they are using multiple LED's.  Nearly all high power LED's like these run on roughly 3.5 to 4 volts.   The circuitry inside those bulbs needs to take 120 VAC down to that much lower DC voltage and so in that process you lose some efficiency.  Which is to say when you take that into account this is even more amazing to see that much light out of this setup. 

I think this latest circuit from LS is definitely worth looking more at.   Have you put a scope on the output to the batteries being charged?  Just curious what that output looks like.  Apologize if this is shown in your vid and as I didn't have time to watch it all right now. 

@E2matrix, ty for your comments.

I hadn't seen that Cree's R@D is up to 250Lm/W in the lab.  You are off a bit on Cree though.  The XM-L are 100Lumens per watt and the highest currently available for purchase.  http://ledsupply.com/creexml.php
Their next improvement, which got the US Govt. award for exceeding 150 Lumens per watt.  The MG-T is just above this, but is NOT available for purchase.  This was listed on their site, first, in June of 2011. 
  This newest claim they have for 256 lumens per watt was added to their site, appx, 19 days ago.  I apologize for not seeing this most recent news.  However, since we still can't buy the MG-T, it will be a couple years before the SC technology hits the market. 

I fully understand that these bulbs have an internal circuit and am using that to an ADVANTAGE;  rather than a loss.  This is similar for the Choke, L3.  As for the LED's in the buld, I would HIGHLY doubt they are XM-L's;   as Cree bulbs are a little salty and not as reliable as Philips  (walmart of LEDS).   This is why they are not preferred where spectrum, Lux and PAR are most important;  Marine SPS corals.


Last here, I wanted to note, that I separated this circuit design from LS as it bears NO similarity;  in operation or functionality.  The only remaining item is a single connection.  LS, circuit is designed to run without ANY other parts and only self-oscillate.   Adding a base bias, as I'd first done,  is NOT a new circuit;  so still LS design (this is like a paint job for the circuit).  However, at this point, Requiring the caps, L3, charge circuit, bias, and NOT ever being "self-oscillative" (no base resistor);  this cannot be considered the same. 
 
IDK, I just wanted to put this out for others to possibly utilize.  I had to fight with my friends to even list this;  being patentable and marketable n' such.  I've stated before I'm not a business man and don't care about making money, so I'm just offering this up.

Hope this helps,
Thanks

Oh, the bulbs were 10% off at lowes, so 17.99 ea.  Walmart 7.5w LED's are cheaper (around 15) but I haven't tried "tuning" with those.
 

EDIT: the LED's in these bulbs are around 65 Lumens per watt, another confirmation of NOT being Cree's.  7.5w@490Lu = 65 Lumens per watt.  Adding in circuit loss the LED's are probably 70 Lumens per watt.  So, this circuit is getting 2x rated LED efficiency.  IF the LED circuit was a LOSS then it would be less than 70lpw.

  If we apply the same assumption as current and assuming I can "tune" to the new bulbs;  A bulb of 250 Lumens per watt, should provide 500 Lumens per watt;  if added to this circuit.   I'll be anxious when they get these better bulbs out in a year or two for sure!

@e2matrix, I see too, your reference of efficacy.  I'm with ya' on watching LED's as we should have more efficient conversion of electricity to light by now;  lol, were due! 

I posted a couple videos that are related to this circuit, from a design Pov.;  also an explanation of operation.  The last video, explanation, should probably be put in lasersabers' Joule Ringer thread;  but it's not necessary to be posting multiple times for these smaller bits.  But the explanation video DOES talk more with the Lasersaber Super Joule Ringer 2.0;  though I cover SJRC aspects.

These two (sorry for being long) I assume can give someone the insight of how to "tune" the overall SJRC circuit and get similar results. 
http://www.youtube.com/watch?v=LMtqqmEMtw8&feature=plcp
First is with  Model T.  It's long, but the whole time shown is 12v and appx .5A (slightly Less at end); so 6w.  I show how to get as much of that to the "load" as possible by tuning.  This is essential to "tuning" of the SJRC with the caps and L3 coil.
http://www.youtube.com/watch?v=YH9IZFbqaAM&feature=plcp
This is the explanation of circuit operation, again Mostly Lasersaber related;  but shows operation, biasing, LED on 3v, comparison of 1 LED (over full bright) @ 12v with Lasersabers Joule Ringer 2.0 schema. (w/base mod / paint). 

Last, I have to help a friend (I believe) with car troubles; but I am looking for another coil for this circuit.  It's a bit more tricky for the SJRC, but I have a few thoughts; so I'll update with those as I get a chance to test another setup. 

Also, I suppose a note, since the DC output / charger is decoupled and filtered (for the most part).  Yes, you can feed appx. 40ma back to the source supply instead of charge; but that is terribly small to input.  Yes, you can run a string of the 10mm white LEDs at full brightness, but again we already have 2800 lumens;  so this is petty too.   I'm still needing to acquire batts for higher charge testing, but the highest volts / ma ratio will be best;  as this will net the most power (watts) through the DC out.  E.g. in video was around 170mw @3.4v and 370mw @9v.  If we could get 25ma through 36v that would then be .9w; all while not losing light (most technically increasing slightly).

Thanks

First, I wanted to say good job on those air cores!  Very nice construction, I'd have to say I like the larger diameter the best .  Also, I suppose, I'm not an expert either;  but time has offered opportunity to become well-rounded at least.

I'll see if I can help answer your questions to help if I can.

1)
Heating transistor = Hard switching = greater Pd (potential dissipation) inside the transistor.  This can be related to 3 things primarily.  First, a leading current phase relative to the voltage phase.  Second, Improper phase of base signal relative to collector (instantaneous V/A values upon switching).  Or Third, the junction capacitance's of the transistor are "coupled" into a series / parallel LCR; thus including current "Circulation" in the junction. 
  I will note, running a stock Super Joule Ringer as shown in that explanation video @ 12v and one bulb;  WILL heat the transistor.  Now it won't fry or anything, but I wouldn't imagine it would last long without a heatsink.  However, the SJRC video, running all the load (and a bit  more power);  the transistor does NOT heat at all.  It does not even require a heatsink.  My best reference here, is with the starting Model T coil rambling;  as the first setup shown dissipates a greater % of power through the switcher (eg- transistor).  For another reference here.  If I use the stock model T and a 3055 transistor Instead of the "built-in" one;  @ 12v 1A, the transistor will fry within a couple minutes. 
   In your case specifically, with the higher operational frequency and low inductance;  your most likely losing most from a Junction LCR inclusion and a bit off on the phase of the base signal. 

2) Air core, the best and worst of both worlds, lol.  They are nice, since we don't have core loss and yes;  most technically air would be the best.  The trick with air cores would be three fold.  First, since (as related in explanation) the energy output is related to "charge stored" in L2 or the secondary;  we need a high inductance.  However, I pointed on Lasersabers' thread this still needs accomplished with Q in mind.  Second, we don't want the Fo of the circuit to be anywhere close to transmission for loss of power there.  If doing the SJRC type system, the frequency should be lower for better energy transfer and power output (e.g. lower Z of L2 = lower Z of output = more theoretical power can be had).  Third, then is related to coupling; as we need higher coupling for best energy transfer (yes, dampened oscillations). 
 
  This is a bit of a side note, but a good time to point out.  The SJRC, then DOES have a "bandwidth" of efficient operation; since we are dealing with Z values.   Whereas the Super Joule Ringer, we do not have a bandwidth, UNLESS (key) it will self oscillate.  If the circuit will self oscillate without a base tie, you will notice bandwidth related effects.  However, of  course, the SJRC cannot self -oscillate and does have bandwidth. 

So, for an air core; you'd probably want opposing coil halves for both increased inductance and self capacitance (with lower R of course).  I suppose this could be pancake (like tesla), helical or solenoidal.  The primary will have greatest coupling to the secondary if wound around these secondaries.  I would recommend even, winding opposing half coil primaries over each secondary half (center probably of two pancakes if they are used).  Again for best coupling here, we want the primary to envelop a large portion of the secondary surface area.  It will still take a fair bit of wire to accomplish this with enough Inductance and the wire should be as large as possible. 

Another side note here, I suppose, but it is related to ALL of these designs.  The voltage ratio of primary to secondary,( note V of secondary is not a direct relation of N1/N2 or V1/V2 or L1/L2 alone specifically) should not be overly great.  Simply here, there is no reason to "step-up" in voltage much beyond the 120v load targets;  as this will lead to other negative effects and less power is potentially available. 

3)
Capacitor values, now there is where we start to create "tank" circuits and "bind" the primary, secondary and load.  I tried to show this with the Model T coil and there are many things that could be taken.  Note, though that we did not get a purple (V/A phase alignment / PF) spark until BOTH the primary and secondary were tuned to transfer the most power to the secondary and, of course, through the load. 
  Can we use standard LCR formulas?  well, I would hesitate against that to some degree.  Though all the formulas are obviously available, the math would be complex;  if we considered the variable XL, coupling, Junctions of transistors, etc.  The variation in just capacitance, resistance, voltage, hfe, switching time etc from the transistor is quite complex in this design;  an in most cases the transistor is altering the wave form from these dynamic changes.  Also, I cannot stress this enough, if you DO make a "resonant" (e.g. phase aligned) L1 or L2 the current IN the coil would be highest and is not necessarily what we want;  if say we want power to circulate elsewhere (e.g. model T coil in the spark / load).  I would recommend (since again we have an advantage of audibility) tuning with V/A readings, lumen meter and your ear.  E.G a nice clean smooth sound is a better interaction with the L1 L2 and transistor.  Whereas a rough sound can be a beat of two frequencies, improper phasing (more notable with darlington and its' larger phase delay in some cases), or even core saturation in certain cases.

  At the end of the model T video I refer to a Quasi-resonance, of sorts.  Since all these coils are highly dampened, we cannot have free oscillation.  Since these are below transmission, we are not dealing with antenna.  Since, the Velocity factor of these designs is not altered in any great degree;  we are dealing with lumped mode inductors and not a slow- wave resonance (Corum).  Since we are dealing with Lumped mode inductors, all formulas will say current is even across coil;  however I do show (with the T) "moving" the relative current in the secondary with various setups. 
  To apply the same "tuning" I did with the model T to more of this circuit we can easily get the info we need.  Dissipation at switch = Transistor heat.  Light of bulb in lumens = Power of spark.   Lower power in @ same nominal lumens = better PF between the, now bound;  primary, secondary and load.  Of course the smooth sound applies the same to both.

  I'll hesitate to provide exactly what I was looking for with tuning, to not take enjoyment out of playing .  But, between the obvious "odd" capacitor ratio shown in schematic and model T video;  I assume one could easily know what to do.   

I hope this helps answer your questions and doesn't sound too nuts.  Let me know if I need to explain something further or differently.

Thanks PB

@Peanutbutter:

Thank you very much for taking the time to answer in such a helpful way. I will try to integrate your suggestions into my existing and also into new air core coils.

The other builders (whom I see on YouTube and in the forums) seem to go for ferrite core transformers (for good reasons) and therefore we will have many results from this side. I kind of like air core coils because they have the aura of 19th century Frankenstein technology and therefore will try to go down this route as far as it takes me.

Greetings, Conrad

@Peanutbutter29:

In order to make sure that I understood your air core coil advice correctly I made a drawing of a coil with opposing coil halves.

The drawing is for a helically wound air core coil (e.g. wound onto a plastic pipe). I would like to ask you to check that this is what you meant?

I have no idea how it could be done with pancake type coils?

Greetings, Conrad

After the above edit I thought I should add the video links showing other oscillator shots.

First, is a standard Lasersaber joule ringer
http://www.youtube.com/watch?v=Gl3h5D5vpBA
Shots start close to 2:00

Second, is a nice video that shows 3 scope shots;  A positive, 2 transistor positive, and lasersaber Super Joule Ringer
http://www.youtube.com/watch?v=T9hvkr2G9Lw
Shots through out.

Makes a good comparison of all the designs that way, an two Super Joule Ringer shots.  There's a pretty large difference between those and the SJRC shots above. 

Thanks, PB