LED backlighting suggestions (4 in CC config, 3.2-3.4V 20mA)

[Nominal Animal]

I'm looking for ideas on how to do better backlighting for my 2.8" 240x320 IPS panels (datasheet). 

There are four white diodes in common cathode setup – so I have access to the anodes of each (LEDA1 through LEDA4 in the datasheet diagrams), but they're internally connected together at the cathode (LEDK1(-) in the datasheet diagrams).

The forward voltage of each white LED is 3.2V to 3.4V, and maximum current is 20mA per LED (80 mA total), although I'd prefer to use at a slightly lower current (and brightness).  I'd also like to have MCU brightness control, maybe via PWM or something (although I'll only use it for smooth on-off transitions).

I'm using a Teensy 4.x, so all of my logic levels are 3.3V.  I do have 5V available (USB), but I'd prefer to avoid that.  During experimentation, I've just used a common current-limiting resistor (a few in parallel to get a suitable resistance around 25 ohm or so) and a 5V supply.

A charge pump LED driver would let me use the 3.3V, and there are chips that would suit me perfectly (TI LM27951 for example), but they're either out of stock (chipageddon...) or have BGA footprints which is on the outer edge of my electronics skills.  It is particularly annoying that the forward voltage is right around the supply voltage level, which makes controlling the current difficult.

I could use a single transistor from 5V in constant-current mode, and PWM it using a high-side (P-channel) or low-side (N-channel) MOSFET.  This will give me approx. 66% efficiency (3.3V out of 5V dropped in the LEDs), with just a few components; quite acceptable.  Is this the best I can do?

So, any suggestions, advice?  Especially, any known reliable designs and parts you'd recommend?

(I am just a hobbyist, but I do intend to do a carrier board (for Teensy) and this display panel in EasyEda, and license it under CC0-1.0, i.e. OSHW, as far as anything I'm contributing.  My other stuff is there as well.)

[Zero999]

Use a voltage doubler? In this instance you'll need to run it off some sort of buffer, as an MCU output won't be powerful enough.

[Benta]

What's the problem with USB power? By far the simplest solution: five resistors, four transistors and two diodes (without PWM) for the current control. Perhaps 30 cents all in all.

[Nominal Animal]

Use a voltage doubler?

As shown,

would immediately burn out the LEDs due to overcurrent.

If the supply voltage is higher than the forward voltage of the LED, then that circuit fails to limit the current in any way.

And if I just stick a current-limiting resistor in there, I'm back to where I am right now – except that I'm down to 50% efficiency.

What's the problem with USB power? By far the simplest solution: five resistors, four transistors and two diodes (without PWM) for the current control. Perhaps 30 cents all in all.

I did ask whether it was the best I could do!  Remember, I'm a hobbyist with very little real world experience in electronics, and I'd like to hear some words of experience.

I shall draw up the one I think I should use for illustration and comments.  The issue is that my PWM output can only swing between 0 and 3.3V, and higher voltages could fry the I/O pin, and I'd also want the LEDs to be dark if the PWM output floats (during bootup).  Apologies, I should have done this to begin with.

[MikeK]

What about a 555 voltage doubler?

[Zero999]

Use a voltage doubler?

As shown,

would immediately burn out the LEDs due to overcurrent.

If the supply voltage is higher than the forward voltage of the LED, then that circuit fails to limit the current in any way.

And if I just stick a current-limiting resistor in there, I'm back to where I am right now – except that I'm down to 50% efficiency.

I thought your power supply voltage was 3.3V? The diode in series with the LEDs would give a minimum forward voltage of 3.8V, so very little current will flow through the LEDs.

[Nominal Animal]

Use a voltage doubler?

As shown,

would immediately burn out the LEDs due to overcurrent.

If the supply voltage is higher than the forward voltage of the LED, then that circuit fails to limit the current in any way.

And if I just stick a current-limiting resistor in there, I'm back to where I am right now – except that I'm down to 50% efficiency.

I thought your power supply voltage was 3.3V? The diode in series with the LEDs would give a minimum forward voltage of 3.8V, so very little current will flow through the LEDs.

At 75mA?  (Checks datasheets.. yes, about 0.5V forward voltage for many diodes.)  I was thinking of a Schottky diode there, it seems; apologies.

Yes, I do have 3.3V supply, and also a 5V USB VCC supply available.  Routing both to the carrier board increases a risk of someone like me accidentally connecting the 5V to an I/O pin, and Teensy 4.x pins are not 5V tolerant.  (Yes, they do have ESD protection, so with a 4k7 or higher resistance current-limiting resistor, they can be considered to be.. but I don't need to cut corners here, this is a hobby project and not a commercial product where every cent counts.)

However, I don't see any current limiting in your schematic; that must be controlled by software, keeping the PWM duty cycle within acceptable limits?

[Benta]

This how I'd do it using the +5 V from the USB.
In normal times, cost will be around 15...20 cents, in today's strange market it's probably 50 cents.
Current to each LED is not just limited, but actually regulated, which should give you a nice even backlight.

Cheers.

EDIT: as MK14 pointed out, there's a typo in the schematic which is corrected now.
An additional comment: depending on PWM frequency, a "bleed" resistor across D1/D2 of ~100k could be a good idea.

[MK14]

R0 at 47k, would seem to be too high, to give enough current to the current regulating transistors, to reliably be able to give 15 milliamps.  Perhaps 2k2 or less, would be more like it?

[Benta]

R0 at 47k, would seem to be too high, to give enough current to the current regulating transistors, to reliably be able to give 15 milliamps.

Good catch, it's a typo. It should be 4.7k. Thanks.

EDIT: schematic's been corrected.

[MK14]

Yes, I do have 3.3V supply, and also a 5V USB VCC supply available.  Routing both to the carrier board increases a risk of someone like me accidentally connecting the 5V to an I/O pin, and Teensy 4.x pins are not 5V tolerant.

Although on the one hand you are the customer and boss of this project and can specify what you like.

5V is a pretty standard thing to be lying around, and I'd prefer to see your hobbyist techniques, grow/improve a tiny bit, and be brave enough to cope with it.  While at the same time, hoping to NOT sound too patronizing here.

Basically, if something WON'T work from 3V3 directly, which seems to be the case.  You have to either give it 5V (with necessary current limiting circuits or resistors, after the 5V, to protect the LEDs), boost the 3V3 voltage up, live with any backlight brightness limit, that 3V3 gives you, or choose another display, which works from no more than 3V3.

Boosting the voltage, is undesirable, because it can introduce noise, especially if implemented by someone, who is increasing their electronics knowledge, which can worsen A2D readings, and other issues.

Can't you use a polarized connector (which DOESN'T allow it to be connected the wrong way round), or headers with a blank pin, which is physically filled in, so it can't be put in the wrong way ?

PC motherboards have similar problems, and they have come up with various polarized connectors and/or ingenuity as to how the signals are arranged, so misconnecting it, hopefully doesn't damage anything.

[Zero999]

This how I'd do it using the +5 V from the USB.
In normal times, cost will be around 15...20 cents, in today's strange market it's probably 50 cents.
Current to each LED is not just limited, but actually regulated, which should give you a nice even backlight.

Cheers.

EDIT: as MK14 pointed out, there's a typo in the schematic which is corrected now.
An additional comment: depending on PWM frequency, a "bleed" resistor across D1/D2 of ~100k could be a good idea.

Since the 3.3V PWM is regulated, then how about using it as the reference, rather than the diodes?

[Benta]

You could also use the LM2660 voltage doubler from TI.
The constant current driver circuit doesn't care whether it's fed from 5.0 or 6.6 V.

[Benta]

Since the 3.3V PWM is regulated, then how about using it as the reference, rather than the diodes?

Intriguing idea, but with some serious drawbacks.
First, the 3.3 and 5.0 V supplies would need to track perfectly. This would complicate the power supply unnecessary.
Second, the temperature-compensating effect of the diodes is gone.
Not worth it to save 2 cents IMO.

[Zero999]

Since the 3.3V PWM is regulated, then how about using it as the reference, rather than the diodes?

Intriguing idea, but with some serious drawbacks.
First, the 3.3 and 5.0 V supplies would need to track perfectly. This would complicate the power supply unnecessary.

No they don't. The current is purely set by the 3.3V supply. Changing the 5V supply will make little difference.

Second, the temperature-compensating effect of the diodes is gone.
Not worth it to save 2 cents IMO.

Each circuit has its advantages and disadvantages. Yours is more temperature stable. Mine has better stability with regards to the 5V power supply variability.

[Benta]

No they don't. The current is purely set by the 3.3V supply. Changing the 5V supply will make little difference.

Each circuit has its advantages and disadvantages. Yours is more temperature stable. Mine has better stability with regards to the 5V power supply variability.

My bad. I didn't look properly at your circuit (dinner time) and thought it was my own repeated for reference. Sorry. I see your point now.
So your circuit assumes that the logic output of the MCU is 3.3 V. Probably is, unless that output is loaded otherwise or has other voltage specs. I don't know the MCU in question.
The last comment I don't understand 100%, unless you're referring to (very small) changes in the diode forward voltages due to QPWM current varying a bit with supply voltage.
But I'm flattered that you bring improvements and other suggestions to my idea.

[Nominal Animal]

This is exactly what I was hoping for!  Thanks, everyone.  I owe ya.

I did notice in QUCS simulations of this circuit that a 10 kOhm resistor across the diode pair (D1 and D2) helps quash the trailing edge; but still, even a short 1µs pulse causes the LED currents to rise to their maximum in a few microseconds, limiting the operating range to audible frequencies.  I assume this is due to the inherent capacitances in the transistors.


I doubt something like 1 kHz dimming would actually cause audible noise in any of the components at these currents, but DACs like Microchip MCP48 and MCP49 series are cheap and still available, and therefore:

How could I keep the per-LED maximum current adjustment, but scale them all using a control voltage?  Replacing PWM with analog dimming, in other words.

(I don't expect hand-holding; I'm studying this in a simulator that does have BC847C and BC857C models as those are cheap and easily available in SOT-23.  But further advice would be greatly appreciated, before I start burning out components on a breadboard.)

The other alternative I'm looking at is making four Howland voltage-controlled current sources, using one or four DACs with a suitable voltage shunt reference.  Absolute precision is not important here, stability is, and that it can go down to zero (but I'll use a separate P-channel MOSFET as a high side shut-off switch anyway).  Everything happens at human-observable intervals, too.

[MK14]

With backlights, it is not always that simple.  The colour rendition quality, may only hold up when the LEDs receive a significant amount of current (for arguments sake, let's stick with the 15 milliamps, previously discussed).  So, if you PWM the LEDs, the on part of the cycle is always at the same 15 milliamps, which can help give the best colour rendition (colour gamut).
But, as nice as it might sound, if you linearly controlled the current, it may not give the full/correct colour balance (gamut), at lower current levels.

This depends on the LEDs they have used, and your sensitivity/importance to this possible effect.  So it may not matter.

I suppose you could breadboard both options, and see if it matters.

[Zero999]

How about simply moving the resistor from the collector to the emitter? Now a base resistor is unnecessary, the circuit will be less sensitive to variations in the 5V supply and it will draw less current from the MCU's output pin.

[Nominal Animal]

With backlights, it is not always that simple.  The colour rendition quality, may only hold up when the LEDs receive a significant amount of current (for arguments sake, let's stick with the 15 milliamps, previously discussed).  So, if you PWM the LEDs, the on part of the cycle is always at the same 15 milliamps, which can help give the best colour rendition (colour gamut).
But, as nice as it might sound, if you linearly controlled the current, it may not give the full/correct colour balance (gamut), at lower current levels.

This is absolutely true, and I was aware of this.

During operation, I want the LEDs to be driven at a constant current (somewhere between 10mA and 20mA each, haven't decided yet; need to test in different lighting conditions to decide).  However, I want both turn-on and turn-off to be smooth; so between zero and the max, there is only a ramp (that is not linear either) that takes about half a second or so.  I can easily generate the desired PWM or PDM pattern, and adjust the max current by tuning the resistors or replacing them with pots, so you could say that the fixed version of Benta's circuit is the solution I was looking for.

There probably is a way to do the ramping in hardware as well, but I prefer to have the option to just switch the backlight on and off at command, with the ramping optional and configurable.  And yes, I am aware of that I could also do this in software, using either the gamma control (available on ILI9341 and similar display controllers I'm using) or by just blending in the framebuffer from black, but that loses color resolution.  I'm not after quick-and-easy; I want to know the different ways of doing this stuff right, but without going completely Blabbo the Verbose Overboard on it.

Having more than one tool to do the job is always a plus.  Because this is a hobby project, I'm using the opportunity to find out the different ways of doing this kind of stuff, and exploring them in simulators (QUCS since easy UI) and EasyEDA (as actual schematics and boards).

While the upper half of the current range is used for steady state operation, it'd still be nice to be able to get the LEDs down to really dim, so that the ramp does actually feel and seem like a ramp, and not like a light stabilizing after being turned on.  With Benta's circuit, the minimum maximum-intensity current pulse across the LEDs takes about 20µs, which means that at 250 Hz PWM, the minimum effective duty cycle is 0.5%; at 1 kHz, 2%; at 5 kHz, 10%.  So, for my purposes, I'm really limited to PWM frequencies between 250 Hz or so (because I hate flicker, and can definitely see 100Hz flicker) and 1000 Hz (because otherwise I lose the resolution at the low end).

With current control, especially with say 20mA/256 steps = 0.078125mA per step (or 12.8 steps per mA, or 0.4% of full range per step), the minimum current would be around 0.1mA or so, which is definitely dim enough to not be perceived as "fully on".  And even dual and quad 12-bit DACs with internal bandgaps (MCP47CVB24, MCP48CVB24) (giving a resolution of about 5µA at the upper half, for balancing the LED currents digitally) are within realm of possibility since they are actually somewhat available still, although they would double the price of the panel (from about $10).

[Benta]

Actually, you should look at Zero999's modifed version of my original circuit (Reply #11).
The more I look at it, the better I like it. It has the potential for much better/faster switching characteristics, and if you want softer switching edges for noise considerations, adding a little RC time constant to the base of QPWM will do that.

[Nominal Animal]

How about simply moving the resistor from the collector to the emitter? Now a base resistor is unnecessary, the circuit will be less sensitive to variations in the 5V supply and it will draw less current from the MCU's output pin.

Excellent!

I ran a simulation (with 4×BC875C + 1×BC847C, R1=25 (for max. 19mA), R2=30 (for max.16mA), R3=37 (for max.13mA), R4=50 (for max.10mA), R0=1k, RB=2k, and PWM input from 1.7V to 3.3V.  I got perfect results!

The current-voltage relationship is perfectly linear from around 2.2V to 3.3V, and between 1.7V and 2.2V curves exactly as desired (i.e. starting slow and smooth), so that I have the extra precision at low current range.

Now, I can stick a $1 SPI or I2C DAC there with a 3.0V reference (say, LM4040-N-3.0) from USB 5V, giving me some 110 effective steps (since 0=0.0V, 145=1.7V, 255=3.0V); and I can calibrate the maximum current per LED by shorting the reference to the DAC output (before soldering it in).  That lets me use the DAC at 3.3V VDD, leaving 10% room for VDD sagging a bit.  The 19mA-10mA current range then corresponds to 35 to 17 Ohms.  I shall draw up the exact schematic for comments forthwith.

[MK14]

I shall draw up the exact schematic for comments forthwith.

Great, I look forward to seeing the new schematic.

[Benta]

Ok, third proposal, partly based on Zero999's modification.
This one does not rely on specific input logic voltage levels for the PWM.
Also, it should be very fast for high-frequency PWM, as all transistors operate in linear mode (no turn-on, turn-off, storage-time problems). In fact, you may have to tame it a bit with a cap across D1/D2.
Some interesting points:
QPWM acts as a level shifter, so supply voltage to the LEDs is unimportant as long as it's higher than V_LED + 1V. Even the cathode voltage to the LEDs doesn't matter.
The transistors and diodes compensate well for temperature.
With a higher RB, higher logic level inputs can also be accommodated (eg, 12 V).

Cheers.

[MK14]

Ok, third proposal, partly based on Zero999's modification.
This one does not rely on specific input logic voltage levels for the PWM.
Also, it should be very fast for high-frequency PWM, as all transistors operate in linear mode (no turn-on, turn-off, storage-time problems). In fact, you may have to tame it a bit with a cap across D1/D2.
Some interesting points:
QPWM acts as a level shifter, so supply voltage to the LEDs is unimportant as long as it's higher than V_LED + 1V. Even the cathode voltage to the LEDs doesn't matter.
The transistors and diodes compensate well for temperature.
With a higher RB, higher logic level inputs can also be accommodated (eg, 12 V).

Cheers.

I'm not convinced by some parts of that circuit, and don't want to spend time extensively analyzing/simulating it, to make sure by criticisms are solid/valid.  But anyway.

RB, 15k, at 2.5V drive (from your schematic), would only attempt to give D1 and D2, 0.166mA.  My understanding is that if you want to use diodes like that, for a somewhat stable reference voltage.  It needs a reasonable amount of current, of perhaps a milliamp.  Otherwise, in some cases, the diodes, will have a significantly lower voltage, than intended.

Secondly, RC, 620 Ohms.  Is going to potentially have significant amounts of the current, stolen by the 4 transistors (bases), Q1 through to Q4, as they will tend to lose Hfe, because the 15mA to 20mA per transistor, is perhaps a safe value of maybe 50 to 100 (Hfe), depending on how you interpret the graphs and safety margins.  They seem to be specified for a minimum of 125 at 2mA.

https://assets.nexperia.com/documents/data-sheet/BC856_BC857_BC858.pdf

So for arguments sake, let's say it happened to be exactly 100 (Hfe) at 20mA (typical values would give a higher Hfe, but less device to device safety margin).  Then 4 x 0.2mA per transistor = 0.8mA, which would tend to leave RC with around 1.2mA (in this example), potentially leading to inaccurate current levels.

In the typical case, the diodes would tend to work with the values shown, and the transistors would probably have a more generous Hfe in practice.  I.e. It would tend to work in a 1-off situation, but not be a good idea for production or if high reliability was desired.