# Controlling the current through multiple LEDs linearily and efficiently

I have around 100 infrared LEDs, each with an adjacent phototransistor to detect small differences in the distance of several objects (QRE1113).

As the objects materials differ, I want to be able to calibrate the intensity of every emitter separately to compensate for this. As each phototransistor's output is measured with a very high sample rate (~50ksps), I cannot use a PWM controller, as those typically operate around 20kHz, and so need to directly control the current flow through each IR emitter.

As I would like to do this calibration phase over I2C, my idea was to use a digital potentiometer such as the MCP4451 as a rheostat. It has 4 channels and so I thought of using it as the current limiting resistor for 4 LEDs. While this is appealing on the one hand (low price, easy setup), I see two main problems:

1. All affordable digitial potentiometers have a range of at least 5kΩ. I need a range of around 400Ω. By using a parallel resistor I could adjust the range, but the resulting range would is not be linear anymore. While perfect linearity is not required, I want to be able to do reasonable adjustments in all regions, and not just on the low amperage end.

2. Efficiency. I am driving all these from a 5V source, and without calibration hardware I was able to put 3 LEDs in series with a resistor. Therefore, the average power consumption per LED was around 30mW, which is great. With this approach, each LED would stand on its own with its resistor, resulting in additional 100mW heat for LED operation at 30mW.

In the following graph you can see the resulting relation from wiper setting of the MCP4415 to the resulting total resistance and LED forward current. The calculations are for using a parallel 680Ω resistor and a 68Ω in series (to set the maximum):

Related circuit:

simulate this circuit – Schematic created using CircuitLab

My main problem is definitely the non-linear wiper to resistance relation, the efficiency is only a secondary consideration. I'm open to any solutions, but I cannot afford a dedicated LED controller for each LED, as my budget is limited, which is why the MCP4415 appeared as an appealing solution to me (~25€ to control all LEDs).

EDIT: I want to clarify some aspects:

• I would like to be able to control the current for each LED separately, and linearly, from 10mA to 30mA in at least 64 steps. I don't need absolute precision for any of these values, just a approximately uniform way to step through this range to adjust the brightness for the LED linearily.
• Efficiency is not my main goal. Functionality and cost are the most important factors.
• I was hoping to be able to use something similar to the MCP4415, as it is cheap and it does not need to be actively interfaced the whole time (it retains its value at least until shutdown).
• My calibration phase occurrs once at startup, where i will iteratively approximate the correct current by using the feedback from the phototransistor.
• The 5V network will mainly power these LEDs, and a few logic ICs. If required, I can switch to another voltage.

Firstly, please let me clarify that this is not a commercial project, but just a personal project. Also, I am a computer scientist and not an electronics engineer, so please bear that in mind. (Therefore, there is no spec I need to follow) - I'm open to any suggestions.

Show application. And what distance and range of reflectivity?

Objects are made of wood, and cannot be painted or modified. Therefore the reflectance varies over different objects (dark spots, etc.).

Closest possible is around 3mm (+- 1mm), furthest is around 13mm (+-2mm). So the travel distance will be around 9-10mm for each object. Everything is wood, therefore all measurements have these high tolerances. The PCB is already at the furthest point from the objects. I probably can't change these distances (maybe -+ 1mm).

I have 12-bit ADCs measuring the phototransistors, and I just want to waste as little precision as possible. Therefore, I want to adjust the IR LED so that the near point (3mm) has roughly the same ADC value for all objects. Roughly equal is all I need - I just don't want one ADC to measure 3000 at the near point and another measure 3500. The far points are not important.

Fundamentally you have made some poor interpretations of datasheet, assumptions and thus poor choices. hFE=CTR (effective) = IcON/If varies from 0.2% to 0.9% at 10mA @ 1mm using a polished alum mirror.. What is your spec for error tolerance ? And what variation in Chip height? 0.1mm max? 0.05 mm?

I see, this is also the reason why I want to be able to do calibration. All tolerance information I could gather is listed above.

Regarding chip height: I have to hand-solder all chips, so assume a pessimistic 1 mm.

You say 3 IRs in series but then 1 R per IR LED ??? Showing none in series ?

Sorry if I have been unclear. I meant one resistor per 3 LEDs in series, this is the current design (below). What I meant was that I can't put them in series if I want to control them separately.

In total I have 90 sensors, so this pattern repeats 30 times.

• Comments are not for extended discussion; this conversation has been moved to chat. Any conclusions reached should be edited back into the question and/or any answer(s). Commented Jul 1, 2019 at 16:36
• If you are using CMOS to drive LED’s ( hi or low side) Req of the CMOS can vary from 20 to each 60 Ohms depending on Vdd of ARM chip or Microchip etc (Vol/Iol=Req) @ Vdd varies with Vdd Commented Jul 1, 2019 at 16:42
• Roughly equal is all I need - I just don't want one ADC to measure 3000 at the near point and another measure 3500. - I know this is not what you asked, but it might be simpler to make all LEDs the same brightness from a well-regulated power supply, and then simply calibrate the ADC readings on startup and post-process their readings to compensate for what you measured during calibration. You might lose a little bit of precision that way for some objects, but I doubt that's the biggest source of imprecision anyway. Only you can make this call, but it's worth considering :) Commented Jul 1, 2019 at 17:57
• @marcelm Well, I did ask for any suggestions ;) - So yeah, I can do that. There is just one problem which would persist, and that is that I can do nothing if my values clip, i.e. my brightness is to high for one object. But as there is no other viable solution right now I think I don't have a real choice here, do I? Commented Jul 1, 2019 at 18:11

How about an OPAMP current source fed by a modulated variable voltage from PWM?

simulate this circuit – Schematic created using CircuitLab

The idea is to make the OA1 regulate the current in a way that makes the voltage on Rsense equal to Vin, thus making current though the LED proportional to Vin. OA2 acts as a buffer amplifier for that variable voltage which can be generated by PWM and filtered to quasi-DC.

OA2 provides a way to modulate Vin with your digital signal Vmod. The modulation is inverted: When Vmod is high, Vin drops to zero, while Vmod=0 results in Vin being set proportionally to the PWM duty cycle.

• I really appreciate your answer, but this seems to me like a good solution only for a few LEDs, as I would need to constantly provide 100 different PWM signals in my case - one to each of these circuits per LED. If I don't want to do this from a µC, I would have to additionally get a timer IC for each of these and use the digital potentiometer that I have mentioned, which retains the programmed value. All in all this seems to become very complicated and expensive at this scale. I was hoping to be able to do it with just a few additional components. Commented Jul 1, 2019 at 14:17
• @nyronium you can still use your digital pots instead of the PWM and low-pass filter, using trivial voltage dividers. I don't think you'll find a solution using substantially less components, unless there's a specific IC doing exactly what you want. You need modulation which you cannot do without an active component. You could drop OA1 and connect Q1 directly to Vin, letting the resistor act as a poor man's voltage-controlled currents source, but that's about all the simplifications you can do. Commented Jul 2, 2019 at 8:58
• Initially i thought that there might be a really easy way, which I wasn't aware of. But this seems to be as simple as it gets and while this design is too big for my use case, I think it answers the question. Thank you. Commented Jul 2, 2019 at 9:21
• @nyronium Note that you don't have to accept my answer if it doesn't really help. While I'm not aware of a better way for doing this, someone else might. Not having an accepted answer might draw an extra bit of attention to your question, though in my experience few user go through a list of question with no accepted answer when they look for a new challenge. Sadly, most people (including me) usually go though HNQ list to and another useless answer to a question which was already answered 10+ times. Commented Jul 2, 2019 at 9:39
• Well, I did accept your answer because you did answer the question I asked. I'm aware that this will reduce the likeliness of other answers, but while discussing I noticed that people get lost in details that don't matter, because I think I have provided too much unrelated information. I guess my question would have been better received, if I had asked something along the lines of "Simplest current source for 10-30mA regulated by 5k potentiometer". That would remove all the unnecessary followup questions. But i guess it is too late for that, as it would immediately be flagged a duplicate. Commented Jul 2, 2019 at 11:34