Have an ongoing LED project that has reached the stage of the first breadboard based prototype. Looking for review/suggestions on achieving next project goal.

Short disclaimer... I am not an electronics guy. I can however, read a schematic, and solder adequately to construct a functioning circuit. I know just enough theory to be dangerous, and to hack someone else's design into a format that fits my needs (if I am lucky). That being said, here is a short outline of the project requirements and current status.

Requirements:
1 Basic design to include 15ea 40000 mcd LEDs operating on a supply voltage of 12-15 Vdc.
2. Available space for components = ~2 cubic inches. LED shells above the shoulder will be external to the constrained space.
3. Operating current <125 mA
4. Optional "dusk to dawn" control
5. Total electronic component cost less than $5.00/unit in 100 unit qtys

Status:
Prototype v0.01 constructed on breadboard using Cree 503D-WAN LEDs (20mA at 3.2Vf) with LM317 Linear regulator for constant current in the following configuration. Calculated total current for array 104 mA.

Prototype working as expected. Need to add optional "dusk to dawn" photo sensor circuit. Cruising the intertube looking for a solution, I found this setup used to run a single red LED using a LTR-4206E photo-transistor and a 2N3904. (Links point to the spec sheets)

My question to y'all at this point is, since the the 2N3904 is rated at 200mA continuous, and my array pulls 104 mA, would something like the following work?

At this point I am beyond my learning curve, and may be making an obvious bonehead error. The simple circuit setup appeals due to the low cost and number of components . Easy to stuff into my small allowable space.

Any comments, suggestions, variations on a theme, or general hilarity at my errors are appreciated.

TIA
Cap' Couillon

 

It is generally a terrible idea to run LEDs directly in parallel. It can cause premature failure of all the LEDs. I explain why here...

LEDs, 555s, Flashers, and Light Chasers

 

While I understand the possible failure (and the reasons for) inherent in the array as designed, I have been unable to arrive at a better alternative given the space and cost constraints I am attempting. The prototype board was constructed using 15 random LEDs selected from the 100 ordered. The 3 Led strings are showing an average difference in Vf of +/- of around 8 mV or a total of 1/2%. I was surprised at the consistency of the Vf given what I have read. Unless I can arrive at a better scheme, I will probably just have to take my chances with Cree's qc...

Bigger question at this point is suitability of the photo-transistor circuit. Opinion or alternative?

Tks again for the input.

 

Adding something to the input may not be such a good idea. Right now the regulator needs about 1.75 volts to regulate and the 12 ohm resistor needs another 1.2 volts and your LEDs 9.6. 12.55 volts without +/-. Let me think about a better way.

 

Welcome to AAC.

As Bill mentioned, connecting LEDs in parallel as you've shown is not ideal. The LEDs will fight each other for current and you'll end up with some getting too much current and eventually burning out and others not getting enough. Visually, you'll see some bright and some dim.

If it is absolutely imperative the LEDs are always at the same brightness, then you'll need to add a LM317 to each series strand of LEDs. Being that is not practical for most applications, you can skip the LM317 and just use a current-limiting resistor for each series strand. As long as the input voltage does not change, the LEDs will remain at the same brightness.

I have practically no experience with light-detecting circuits, but based on the circuit you provided which I assume came from here, I think it will work for your application. I suggest using a PN2222A transistor instead since it will safely handle more current than the 2N3904. I also suggest using the transistor in a common emitter configuration which is better for switching applications.

I've put together a schematic I believe will work. R6 limits the base current to the transistor when the photo transistor is not conducting. Hopefully others will point out any faults if they see them.

 

I'd switch to a MOSFET for the switch transistor since there is plenty of base voltage available and the MOSFET will not require a base current. I assume power saving is a concern?

 

I'd briefly considered a MOSFET, but the OP seemed to stress cost, so I figured a BJT would be cheaper than a MOSFET. Based strictly on the BS170 and PN2222A, the MOSFET is a little over double that of the BJT. This can vary depending on the MOSFET used of course and power considerations should be considered, especially if using a battery.

I've modified my original schematic to show one with a MOSFET. Not sure what R6 should be, so I randomly choose 1kΩ. Again, open to criticism if anyone sees errors.

 

Not sure what R6 should be, so I randomly choose 1kΩ.

If power saving is crucial, I'd go to 5-100kΩ. The datasheet for the transistor says the dark current is less than 0.1µA. The selection of R6 is a trade-off between switching speed and power saving.

 

The selection of R6 is a trade-off between switching speed and power saving.

Forgive my ignorance but I'm trying to learn more about using transistors in general - so increasing R6 results in less current (good for power-saving), but also decreases response time of the MOSFET (presumably only a concern for fast switching operations like PWM). Is this correct?

 

Right, R6 resistance combines with the capacitance of the MOSFET's gate to make an RC timing element. The MOSFET is conducting with some resistance when the gate voltage is in between 0V and saturation at 10V or so. This means heat is being generated and you want to avoid this with rapid switching, especially if the switching is recurring. You want the percent of time spent making heat to be as low as possible. For infrequent switching as in this application, some slowness can be tolerated.

That said, there is some light level in this application that is "perfect" in the sense of holding the switch in the worst possible intermediate position. The MOSFET needs to be rated to handle that worst case scenario. I'd have to think about how to identify that case. Using a comparator would eliminate the problem. Maybe someone can show us how to add a little hysteresis to the transistor circuit.

 

...
Maybe someone can show us how to add a little hysteresis to the transistor circuit.

Sure, here is some hysteresis added, using just one small NPN transistor and two resistors.

The 22k resistor sets the hysteresis, and generally is a value many times higher than the sensor voltage divider (1k). You may need to "tweak" the 22k value to get the best "snap" switching.

I've used this system on many occasions in light sensor and temperature sensor circuits, so it is proven.

 

If you didn't say you were going to build 100 of these I would not say anything, but since its quite a few......
You probably have LEDs all from one bin as the entire spec is like 3 to 4 volts for Vf. All's cool if you know the bin number so the next time you buy some you can order the same bin number.
Since your supply voltage goes to 15 volts you should probably design around 20 ma at 15 volts. This would make the resistor 270 ohms. This may make it necessary to reconsider the constant current idea as there could be a pretty big change in brightness.

 

Updated schematic to include Roman's hysteresis circuit.

 

Thanks! That's a bit neater than my scribble.