Much time since i worked with LM317. But for what i remember, although specs usually advice to let 3V for the possible drop in the regulator, when you measure it often is around 1.5V. You know, there is wide tolerances and manufacturers obviously give safe margins of operation.

Anyway, i think the better current regulation worth using the LM317. Resistors arnt adequate to drive high power leds. Doable, but seeing the maths on my last post, is easy to understand why current control is advisable. 10%, and still 20% of variation on the current target is not a problem. But with resistors the variation may go way far from that.

I dont think transients are a practical problem. Although any manufacturers recommend to avoid them, there are many systems working without compensating them and work fine. So its preferable, and life test are done avoiding them, so its difficult to know if not compensating them shorten leds life. But on large strings i dont think its a mayor problem. But i dont know for sure, i use smooth regulators.

The Alumina and the Silver works pretty similar. In the practice, no difference. Any of them have low heat transmission compared to a metal. The key with them is to keep layer thin if you want it dont penalize heat transfer too much. I like the Crees because the electrically isolated bottom slug, as it allows to forget concerns about derivations if the adhesive layer not cover fully the slug.

I like to apply the adhesive directly to the slug, ensuring all is covered and then place it and apply some pressure with the hand (i use a nut to avoid pressing the dome). I let the adhesive cure enough to keep LED in place and then i put a book over the module (a nut over each led) for all the night. Using anodized aluminum is a secondary way of avoiding electrical derivations (the aluminum oxide layer is not electrically conductive).

You can try too to add kapton tape to the conductive slug. It works very well isolating. But then the problem is to glue the led itself to the board. But i think it would be possible to add adhesive (and not requiring to use thermal conductive one) on the perimeter of the base.

I think the main reason of many people to use stars is ensuring that there is no problems with conductive bottom slugs (and beware with that, ive found stars which indeed have the bottom electrically connected)

AfricanAlien, it will work perfectly as far as the constant current device is able to give 3*6.84V=20.5V @700mA. From 24V input, it should, still if its a linear regulator. But be ensure about that.

I have an idea that I have also been working on in order to get more growth in Z-plane. This would also help if you are wanting to grow taller plants.

I have a design for a vertical plug-in module that drops down below the top of the canopy. The problem is with cooling. Heat sinks block too much light, plus they are dissipating heat onto the plants.
My design low profile water cooled module, where I am mounting emitters right onto a piece of U bent copper tubing about 6 inch long. I have only built one and it seems to do okay. I don't know how ganging up mulitple units will work.
Heat from a heatsink shouldnt be any problem for plants. All thermal issues of LED arrays are about themselves perfomance and duration, not about plants. If you design well the system, the heatsink shouldnt be over 40ºC. A max, you should feel it slightly warm at the touch. That temp wont be any problem for plants (its similar to T12 fluorescents walls temperature).

Heatsinks on side walls shouldnt block any large amount of light either. But intracanopy, light losses may be noticeable, i agree with that.

I think there are two ways to fighting it. By using low power densities arrays intracanopy, able to be mounted on flat heatsinks, flat white painted (i use a highly reflective DIY mix, by adding barite to latex paint).

Or by mounting the leds on tubes (preferably, at least with some flat area, or directly squared) and aircooling them.

NASA intracanopy modules used this last option: each module has its own driver on a edge with a fan sucking thought a tube where LEDs are mounted.

Another issue is that you would have to reduce top lighting in order to keep within your watts/sq. ft. parameter. So the question that needs to be answered is "where will a watt of LED power need to be positioned to get the best results?" On the top?, On the side? or a combo?
Yep, this question is very important. I believe nobody along the Mj forums has studied this topic. And not only about LED lighting, but for any type of side lighting.

On the group buy im doing on a local forum, most people is going to use side and/or intracanopy lighting (mostly, side because cabs are small, mostly microgrow areas).

The initial rule of thumb we are using to size LED modules is to use 300uE/m2 from top and 180uE/m2 for lower areas. We think that each module should cover one cubic feet.

So we put 27uE for each upper cubic feet, so the distance covered for each LED module is about 1ft. And 16 uE on each bottom cubic feet.

But as the system is modular, at any moment each grower may change the configuration and put more or less power to up or sides.

We expect that we may learn the best power distributions from this experience, as there are 20 growers participating.

The initial rule of thumb has been extracted from the light distribution of a 400W HPS in 1 sq meter (about 11 sqft) for plants 75cm tall (2 1/2 ft). But translating some more light to bottom areas than the HPS does, and less directly on top (about 25%), so keeping light density way more even, and just about 60% higher on the top of the plants than on the bottom.

I hope that the experience of 20 growers using different styles of growing allows us to learn fast the best way of distributing light along the 3 dimensions of a grow.