Why not go with a 54v power supply or 48V supply adjusted to 53V (e.g LRS-350-48), 5x LDD-H 700mA drivers ($25 vs $16 for the 4 LDD-L) and not have to spend hours trying to get circuitry working that will end up costing you extra in components and effort?
I undervalue my free time at ~$10 and there are still too many instances where coming up with extravagant ways to save money ends up with no net cost savings. It looks like you've been debating about this project for nearly 3 weeks, and a conservative 1 hour per day of pondering stacks up to $200 in effort. Is it really worth it to save $50 or so over that time period?
This is a hobby. Part of the satisfaction comes from an efficient well designed system, hopefully better than what is available.
I purchased the LDD-L drivers last year, specifically for their analog dimming capability. This is a multi-channel build that I specifically wanted a
number of pots for independently adjusting channels. Eventually I will move to arduino based PWM dimming, but there is value in ability to switch to analog controls.
Additionally, having a 36V supply w/ 30v avail from driver is better sized for my supplemental color channels. I can fit 20x 3W leds on a LDD-1200L. (Incidentally, the epi colors come in 10x allotments.) If I went with a higher voltage supply and LDD-H, I would only be able to run 14x color leds per channel (which I don't think is enough, meaning extra drivers with unused capacity), and I wouldn't have enough power for all the channels.
4x 1500 Whites= 6 Amp
3x 1200 Colors = 3.6 Amp
Power supply has 9.7 Amps.
This is the best possible design that encompasses all my requirements. Sure I could buy another power supply, buy more LDD-H drivers just to avoid using a current mirror, but in the grand scheme of things, building a current mirror into each channel is negligible. I also learned some new things and find it interesting.
This seems to be where it gets complicated. The voltage drop across transistor and resistor is 6.5V in my simulation. This can be adjusted by lowering the 680 ohm resistor. BUT when you lower the 680 ohm resistor you reduce the effectiveness of the protection provided by the BC337.
Conversely the diode protection circuit if you don't use a resistor on the base pullup the drop is less than 2V. You also need to add the drop across the additional transistor/diode which brings it up to just under 4V.
Lol, this is not going to waste 4-6.5v. Thats 20% of the power. The resistors/transistors would go up in smoke if that was the case. Your assesment doesn't pass the "sniff test".
Worst case scenario if the current mirror pushes over the voltage limit... Each group of six cxa1304 are on the same heat sink so temperature should be shared and prevent thermal runaway. Testing might show they can be run without current mirror. However, I can still use the simplified current mirror with just two BD139. It wont protect the other leds if one burns out.
Alternatively, I could run them as a 2x3 (or 3x3) array instead of a 3x2. It may be that I'm running them at < 50%, so running them at 500mA would be no big deal.
A 3x3 array has built in protection. If one string goes out, the other two are still capable of handing the current.
There are a number of avenues for this to be a success.