In a display where you have several LED/pixel you would normally look at driving them in series to allow a higher voltage and lower current. A big problem with large signs is the very high current. If you have 240x16 LED (a quite ordinary sign) you have almost 4k diodes. If each diode gets an average current of 20mA when the display is at max intensity, then the required current with all LED turned on (remember that the diodes are controlled individually) would be 76.8A. The power to the diodes would be about 150W and if you get 5V from the supply you would have 225W to burn in driver chips and series resistors.

If one LED/dot isn't enough, it wouldn't be too funny to continue and scale this drive current furhter. With four LED/pixel in parallell you would get 300A. If driven in series, you could stay with a similar amount of power loss in the drive electronics and just dimension for four times the voltage to the LEDs - 600W to LEDs and 225W to the driver electronics.


You still haven't got the multiplexing. A multiplexed LED shining at only a fraction of the time can still give the same amount of light as a LED that is driven continuously and be 100% fulfilling the datasheet for the LED. The multiplexing is used to reduce the amount of drive electronics - not as a way to reduce the display intensity. It is only if the multiplexing factor passes the quota between max pulse current and max average current supported by the diode that the multiplexing sign will be required to never reach max intensity of the individual diodes.


Correct. But a driver chip with 32 simultaneous outputs will have 8 times more diodes to control, so the power loss in a 32-output driver will be 8 times higher. But the capsule of a 32-output driver will normally not have 8 times lower thermal resistance so there will be harder resistrictions on max current/output - either directly written in the datasheet, or as a result of the allowed Pd.

An Allegro A6277 8-output chip has a max operating temperature of +85°C. The 20-pin SOICW package can handle up to 122mA concurrently on all 8 outputs when designed with a 1V internal voltage drop. The package has 90°C/W and at 50°C will allow 1.2W dissipation. At 75°C max dissipation will be 0.8W. So max drive current/LED is directly affected by the ambient temperature since the capacity of the driver is a linear function that directly follows from the 90°C/W thermal resistance.

Let's switch to the 16-output sibling Allegro A6276. The socket has increased from a 20-pin to a 24-pin SOICW. It has 85°C/W which is slightly better and caused by a 20% larger surface area of the 24-pin package. The datasheet gives max 1.2W at 50°C (same as the 8-output variant) and about 0.85W at 75°C.

As you can see, the total power loss if the 16-output variant is almost identical - but in this case you have twice as many outputs. So if used in an multilexed outdoor sign (and probably just about all multiplexed indoor signs too), you will either have to use only half of the 16-pin outputs (which would be silly) or you will have to reduce the current required from each pin (reduce the LED current) or reduce the voltage drop in the chip (which would affect the ability to regulate the current).

This is a real design constraint. So as a professional developer, you must not recommend people to go for 16-bit or 32-bit driver chips unless (!) they are run at low output currents (such as in a DC-driven display where you need hardly any power at all from each output) or for an indoor display where the ambient temperature will be much lower and the required light intensity from the diodes need not be as strong.

Yes, you can go for big 32-bit shift registers with external buffer chips. But what would be the gain? The whole point of the 16-bit or 32-bit LED drivers is not that you may look at external buffers, but to keep down the surface area needed for the driver. The cheapest possible 8-bit latched shift register with drive capability would be a better alternative to a longer shift regsiter feeding multiple driver chips or a huge number of discrete transistors.

But you have more factors to think about. For the argument, let's take a EL15-21VYC surface-mounted
LED. Probably lousy for a billboard but it should suffice as an example.
Vf 2V
Ifavg 30mA
Ifpeak 160mA (1/10 @ 1kHz)
1206 form factor

In a DC-driven display, you may feed 20mA / output from each output of your Allegro chip (well within spec of the diode that can handle 30mA). Hardly any current at all, so no problem with a 16-output chip.

In a multiplexed dislay, you do not multiplex between the pins of the driver, so a 16-ouput driver do not alternate between 16 LED - on on each output. It always uses all 16 outputs, but multiplexes which of (in this case 8 to 10 LED) the LED connected to each pin that will be lit.

With a 8/1 multiplexing factor, you would then use 8*20 = 160mA instead. This is still within spec of the diode and the power to each diode will be the same as in the DC-driven display. You would not see any difference in display intensity (if we assume everything is linear). But suddenly, this 16-output Allegro will no longer have to source 16*20 = 320mA total current (or 320mW if 1V drop in the driver). Instead, it will have to source 2560mA which is 2.56W at 1V voltage drop. Even with cooling, the chip will be used way outside the datasheet maximas.

Now it should be obvious why a 16-ouput driver chip can not feed a multiplexed display just as well as four 4-output drivers. In this world, the 16-output drivers are not built into packages with half the thermal resistance of their 8-output siblings.

But there are more factors to think about when designing a display. The LED I mentioned is in 1206 form factor. Let's say you have 4mm pitch distance. 16 LED beside each other would consume 64mm in side. A 16x16 panel would be 64x64 = 4096 mm2. Nothing strange there.

A 8/1 multiplexed display would need 2 x 16-ouput chips, but as can be seen from the above, the Allegro A6276 would go up in smoke. The 8-output little brother would manage if you keep track of the ambient temperature. You would then need 4x 8-ouput chips. The outline of the 20pin SOICW is 10.3 x 12.8 mm or 131.8 mm2 so four chips would require 527 mm2. A lot of space left for the pads, traces, capacitors, resistors, ...

Switch to a DC-driven display and you would need eight times as many driver outputs - 16x16 = 256 outputs within your 4096mm2 board area. You would need 32 8-output chips for the 16x16 module. That would add up to 4218 mm2 which is larger than the PCB area for the diodes. If the 16x16 panel is expected to be stackable in height and width, then you would have to build daughter boards if you used the Allegro A6277. Let's go back and look at the 16-output sibling again. You would need 16 x A6276, and the outline of the 24-pin SOICW (excluding solder areas) is 10.3 x 15.4 mm = 158.6mm2 which would accumulate to 2538mm2 for 16 chips.

So as you can see, there is a market for 16-output or 32-output LED drivers. But they are only marginally usable in multiplexed designs. So if you worry about professionalism and proper design, I have to conclude that you should avoid recommending large-output drivers in a thread specifically about multiplexed displays.

But from your previous posts, I kind of think that you have still missed about about the multiplexing. You wrote in another post:

This indicates that you seem to think a 16-output driver in a multiplexed design will only drive one LED at a time. All 16 outputs will be used to drive one LED each at m times higher current than in a non-multiplexed sign. Then all 16 outputs will be used to drive a second set of 16 LEDs (also at m times higher current than in a non-multiplexed sign). A 16-output SIPO will drive twice as many diodes concurrently as a 8-output SIPO, resulting in twice as much power loss in a package that does not have half the thermal resistance.


If I can afford to have "extra" outputs in the design, then the design is incorrect. If I have 16 outputs from a driver, then I would use all 16 all the time. Either to control 16 diodes in a DC-driven sign, or maybe 64 diodes in a 4/1 multiplexed sign. But I would never ever start to multiplex which of the outputs from the shift register that gets used. That would just introduce high-current pulses without any gain. The multiplexing is to let few outputs drive many LED, not to let many outputs regularly rest.

The shift register is not (!) usd for the scan lines. It is for the columns so each output may be connected to multiple diodes, but each output will only drive one LED at a time, but possibly at m times higher current than the LED is used with in a DC-driven sign. The other side of the diodes are however connected to row-drive transistors, where each row-drive transistor is responsible for a scan-line. In this case, the row-drive transistor may handle 80 LEDs but for real displays you normally split the display into horisontally stackable modules and let each module contain its own row-drive transistors (just as it contains its own column-drive shift registers/buffers).


Which is why a designer selects drivers with few outputs in multiplexed and/or outdoor signs. It is a direct result of the designer reading the datasheets before designing.


I have never limited myself to constant-current drivers. I did write in another post that shift-register + buffer + resistor or shift-register with buffer + resistor or shift-register with constant-current driver will all work. Some solutions a bit better, and some a bit worse. But the designer must always limit himself to physical properties. One important factor is the amount of board space available for the drivers. Another factor is the amount of power too cool off the driver chips. These are controlling design criteria. A design that doesn't fit in the box or can't do 10+ years in the required operating environment is a failure.