Driving 7 Segment LED Display with High Efficiency

[Zero999]

Can you elaborate? Low voltage red LED's (AlInGaP 2V) are still at the bottom for efficiency and brightness compared to the 3V club green, blue.
InGaN phosphor-converted amber, red LED's have taken over when you need highest efficiency and brightness but they are basically blue >3V.

You're talking lighting LEDs, which is complete different story, running at 350 or 700 mA or even higher.
For low-power LEDs, red or green are the most efficient, due to low forward voltage. Red: ~1.8 V, green ~2.0 V, as well as current/light conversion ratio. Red leads here.
Example: https://www.vishay.com/docs/83180/tdcx10x0m.pdf

Green has an advantage in "eye efiiciency". The human eye is most sensitive to green light, followed by red (and at the bottom: blue).

Hope this clarifies my opinion.

Actually no, green LEDs, especially the type with the low voltage drop have pitiful efficiency, even taking into account the human eye's increased sensitivity to green, which they also miss, because they're yellowish-green.

Decent, modern red LEDs are more efficient than blue ones, in terms of power in vs power out, but many 7 segment displays use old, inefficient red LEDs.

Modern pure green LEDs, with a higher voltage drop are more efficient, than the old yellow-green LEDs, but still not as much as red and blue, although they do look brighter, because the eye is more sensitive to green.

[Siwastaja]

Ignore all advice regarding how LEDs made in some specific colors are more efficient than others; just check the actual numbers. Which will take us directly to the point:

LED choice is by far the most important! Look at as many distributors and datasheets as you can find, sort for the best luminous intensity. There might be easily 50x difference between the worst and best LEDs, in such case it is totally secondary if you can increase your driver efficiency from 55% (resistive dropper) to say 90% (very well designed mini-buck per segment).

Now you can ignore some colors due to the sensitivity curve of the eye, but this is wasted effort because eye sensitivity curve is already baked in the lumen and candela numbers! Following wrong forum advice you might end up ignoring a really efficient 625nm red LED just because it's "red", so that eye sensitivity curve gets taken into account twice by accident, first by the manufacturer then by you again.

Also, is there really a reason to run the microcontroller at 3.3V? Most low-power MCUs run at lower than this, and while doing so, also consume less power. 2 AA cells range from 2V..3.4V; many MCUs can run directly off this supply. With average battery voltage being a tad below 3.0V or so, you would already gain extra efficiency in the simple resistor solution; some 20% improvement from lower drop over the resistors, and another 20% from skipping the boost . Low-end 2.0V is still enough for red LEDs with Vf around 1.6V; while multiplexing anyway, you could increase the duty cycle (on-time proportion) based on measured battery voltage; which you likely measure anyway with the MCU - which again would be easier if you did not have an unnecessary boost converter, as most MCUs can easily measure their own supply voltage.

If the sensor requires 3.3V to operate, I would look for a replacement part which operates over the 2V..3.4V range, getting rid of the boost converter would be a huge help. If this is impossible, then running only the sensor off the boosted supply, and level-shifting the communication to sensor (if necessary) is possible.

TLDR; invest your design time in:
* Finding highest luminous intensity LED display,
* Finding an MCU which operates directly of 2xAA
* Finding a replacement sensor which does the same, if unable, limit the boosted supply to sensor only and level-shift communication

[Kane]

Thanks, Siwastaja, Great advice.

Since starting the post I did find a sensor without the 3.3V requirement which as you said allowed me to eliminate the boost converter from my design and run directly off batteries.

I am quite happy with where the designs at. Everything running direct off batteries, Red 7 segments with low forward voltage and high brightness which while using a constant current sink will run across the whole batteries voltage range from 3.2V fresh down to the forward voltage of 1.8V.

I'm currently looking into making my own 7 segment display from discrete LED's as a lot of manufacturers have extremely bright 7 segment displays but are not available anywhere and the modern discrete LED's often have luminous intensity's 10x higher than the 7 segment displays available.

[Peabody]

Or you might find a really efficient single digit, and get four of them.  I'm curious how the Vishay TDSR1350 and 1360 compare to your current display if you can get them in the "K" Luminous Instensity Factor classification.  The datasheet says they produce 1800-3300 ucd at 1mA.  I did a video back in the day explaining multiplexing by segment, and used these displays.  Of course I was working much brighter than you will be, but they seemed to be quite bright on little current.  The average current to display "88" was 4mA at 3.3V, which was a lot better than the girls I had dated up to then.  Discussion of the displays is at about 10:40 into the video.

[Zero999]

IT also depends on the light level. In the dark, the peak response of the human eye shifts towards the blue end of the spectrum, making cyan LEDs the most efficient choice, purely on the response of the human eye.

[bsdphk]

When HP made the first pocket calculator, LED display power drain was a major issue for them.

There is a quite detailed article about it somewhere, the gist of which is that our eyes register the peak intensity rather than the average, so their solution was to drive the display at very low duty-cycle something like 1µs 100 times per second (= 1:10000), but at a current which was only something like 10-100 times higher, thus saving a LOT of power.

The limiting factor is the temperature of the LED dies in the display.

[Siwastaja]

IT also depends on the light level. In the dark, the peak response of the human eye shifts towards the blue end of the spectrum, making cyan LEDs the most efficient choice, purely on the response of the human eye.

This shift isn't massive. It would come into play if you had to choose e.g. between yellow LED rated at 1000 mcd and cyan-green LED rated at 900 mcd - AFAIK lm and cd are defined per the day vision sensitivity curves, and the difference you mention would swap the order of their perceived brightness. With that being said: in extreme enough darkness for the shift you mention does happen, it doesn't matter because a ridiculously low amount of light would suffice anyway, say just tens of uA. If your device ever has to be operated within anything more than complete darkness (even just room lighting), so that the day sensitivity curve applies, then you are obviously going to choose the most efficient LEDs for the day vision curve; which is already baked in the lumen and candela ratings as I said, so no need to overthink it.

And any lens angle is baked into the candela rating so it's probably the best metric available.

[Siwastaja]

There is a quite detailed article about it somewhere, the gist of which is that our eyes register the peak intensity rather than the average, so their solution was to drive the display at very low duty-cycle something like 1µs 100 times per second (= 1:10000), but at a current which was only something like 10-100 times higher, thus saving a LOT of power.

IMHO this is a total myth. I have at least tried to test it myself and don't see it working, unless you go slow enough that it starts really hurting your eyes. Make your product blink brightly at, say, 30Hz in darkness, and sure it won't go unnoticed - but I mean in a bad way.

The real reason for fast blinking is just multiplexing - or that some ancient LED technology getting better efficiency at higher current.

One thing is sure: on all modern high-brightness LEDs, driving short pulses at large current is going to decrease efficiency compared to driving at the same average current, DC. Two things happen when current increases: fewer photons are generated per coulomb, and Vf rises, both lowering the efficiency.

This is the opposite of some ancient LED technologies which would have sweet spot for current. LEDs have changed a lot during early 2000's, a lot of old advice (such as efficiency estimates based on color like "green is bad"), or this pulsing thing, simply do not apply today.

[Phil1977]

When HP made the first pocket calculator, LED display power drain was a major issue for them.

There is a quite detailed article about it somewhere, the gist of which is that our eyes register the peak intensity rather than the average, so their solution was to drive the display at very low duty-cycle something like 1µs 100 times per second (= 1:10000), but at a current which was only something like 10-100 times higher, thus saving a LOT of power.

The limiting factor is the temperature of the LED dies in the display.

Unfortunately this "amplification" of LED brightness only works for frequencies that are at the threshold of flickering and it´s also limited to signalling applications - its not working for lighting.

With 100Hz or more the eye effectively averages any modulation in the perceived brightness.

But if the author of this thread does not mind a slight flickering sensation, he can use e.g. 50Hz with 5%PWM. Then the display looks brighter than it is.

[Siwastaja]

Unfortunately this "amplification" of LED brightness only works for frequencies that are at the threshold of flickering and it´s also limited to signalling applications

Yeah, more specifically the "eyes register the peak intensity rather than the average" part I replied to is, IMO, incorrect. Within integration period of the eye (which is of course a huge simplification, in reality it's flexible and not like in a camera, but this simplification is good enough for this discussion), eye pretty much just counts photons like a camera sensor does. This is: average.

Now to point out that there's no magic, consider this: obviously if you shine a bright light for a minute and then keep it off for nine minutes, you are able to see well during that minute, and have reduced average power consumption to one tenth. You can keep shortening this period, and as long as the time of bright period is within the eye's integration period, you will see it as a brighter light, because it is. But you will also be seeing the off-time.

Once you bring up the frequency to the point of minimal or no flicker, your eyes are just averaging the light.

[Kane]

Anybody know of any IC's that contain 8 basic current sinks? I've looked for a while but all the LED driver sinks are waaay overcomplicated for my needs and often consume many milliamps. I just want something akin to the circuit below but 8 of them in a single chip, each channel can have the same set current, they don't need to be independently controllable but it would be ideal if each sink had it's own enable.

It must also work across the supply voltage from 1.8 to 3.2V, as soon as I put this filter in the Current sinks categories of Digikey nearly all of the results disappear so I'm not sure if I'm searching for the right thing or in the right place.

[mariush]

Have you considered ICs like ULN2003V12 / ULN2003F12  - it's mosfet versions of classic ULN2003A

The ULN2003F12 has 4 channel,  ULN2003V12 has 7 channels ... minimum 1.65v required to turn on a channel, each channel does 100mA with 3.3v on the io pin, less at lower voltages... 500mA in total on all channels.

ULN2003F12 : https://www.digikey.com/short/rv7n32cv

ULN2003V12 : https://www.digikey.com/short/v1jtw4hh

You'd have to add resistors on each channel to limit the current, but you can get resistor networks that use very little space (ex 7 resistors in a 1206 footprint)

Toshiba makes some chips with 8 channels but they usually need at least 3v .. they have some p-channel based ones for high-side switching that go down to 2v


If you can relax your input voltage requirements a bit, there's drivers like IS31FL3242 that can control 12 leds and consumes around 0.5mA :

IS31FL3242 https://www.digikey.com/en/products/detail/lumissil-microsystems/IS31FL3242-QFLS4-TR/14640385

Works with min 2.7v to 5.5v  - maybe a basic charge pump could boost 1.8v to 3.6v just for the driver - and no resistors to set max current, each channel can be set between 0 and 48mA digitally through commands (256 steps for 0 ..8 ma, 256 steps for 8..48mA) and then there's pwm 4096 steps for each channel.

The leds can be powered directly from battery, ex 2v+ for red leds... the chip needs around 0.25v (max 0.3v) above led forward voltage to properly limit the current on that channel.