LED current mirror

[deSitter]

Hello, first time poster, long time lurker.

I'm working on a LED lighting solution and am looking for some feedback on my schematic. Due to certain constraints I've chosen to go with a single 1400mA constant current driver with suitable voltage to supply 8 parallel strings of 8 COB LED's each with ~0.175 A of current. I'm finding my biggest hurdle at the moment is dealing with current imbalances/thermal runaway in each string. I'm attempting to use a current mirror to resolve this problem, but this is the first time I've made a current mirror, so my confidence in the design is not that great. Does anyone have any thoughts or improvements on this design (see attachment image)? My electronics knowledge is limited to high-school physics, so please keep that in mind.

I've simulated this schematic and everything appears to be working as intended under stable conditions. Though I am aware that a failure in all but the reference string will result in over-driving the remaining strings. Also, as for the resistors R1-R8, I'm not entirely convinced of their purpose. I somewhat understand their use for stability (thermal changes in the transistor?) with regards to DC biasing an amplifier circuit, but I'm not sure if the effect is the same in this case, nor am I sure of their value if they were to help.

Thank you,

[f5r5e5d]

yes the R are local negative feedback and definitely help balance the currents vs transistor heating and other parameter differences

for this circuit an estimate of the amount of feedback has an interesting shortcut: feedback factor ~ V dropped by the R divided by 26 mV

so sizing the R to drop > 260 mV makes the circuit very independent of the transistors - but do use all the same model number Q and stay within their power ratings

[David Hess]

It should work fine but I wonder if it is worth it.  The reference transistor will drop a Vbe of 0.6 volts and about 0.3 volts for a reasonable R.  So if an average of 0.9 volts drop for each current sink is assumed, would just replacing them with appropriate power resistors be good enough?

Grading the transistors for Vbe will noticeably improve the matching with such a low emitter resistance but matters less as the value of R increases.