Why wast majority of LED driving IC's are current sinks instead of source?

[jimon]

I never really asked myself this question until I ended up in need for current source IC because I want to drive a HDSP-2000 LED display (more or less museum piece similar was featured on eevblog youtube channel not so long ago) you can find a datasheet here. Basically it's plain stupid array of LED's with shift register ONLY for cathodes, aka all column anodes are wired together and exposed as 5 pins .. in high brightness each this pin can sink like 20-30 mA, so 5 pins and total can be 100 mA per display, and mul. it by 3+ displays - viola we have half an amp!

Because I'm lazy and don't want to spend 5 pins of MCU with a buffer array IC in between, I was looking for something like serial interface + current sources in one IC, but almost everything I found was serial + current sinks! Like this one:


So I was thinking, why is that? Is making high power NMOS is that much easier/cheaper than PMOS? Or is there something else? Or is it just basic pragmatic practicality of IC companies?

PS. Is there non-obscure i2c + current source IC out there?

[mikeselectricstuff]

Probably easier to make, and then it becomes the convention.
It does have some disadvantages, like making it harder to use lower supply voltage for red in big muxed displays to reduce heat.

Once company has a "revolutionary" common-cathode drive system for use in video walls :
http://www.silicon-core.com/common-cathode

https://youtu.be/Jm0aVF6_y_c

[shteii01]

Because sophisticated miniature circuits and high currents don't play well together.

[r4tch3t]

NMOS is easier to manufacture, I did a quick search and found this crazy chipL T3595 - 16 Channel Buck Mode LED Driver
The pca9685 can source 10mA power channel if that's enough

Sent from my ALE-L02 using Tapatalk

[danadak]

If you look at a "typical" CMOS output it is an NMOS and PMOS structure
between the rails. One of the principal issues in a simple non bootstrapped
output stage is source bulk effect on P devices (that connects output to Vdd
rail) ability to turn on hard.

So the NMOS is easiest to turn on hard, which connects the output to ground,
Vss.

Look at typical device output source and sink current ratings, you will see this.

To offset this issue typically the PMOS device is sized with greater channel capability.

This is more than you want to know -

http://bwrcs.eecs.berkeley.edu/Classes/icdesign/ee141_f01/Notes/chapter5.pdf


Regards, Dana.

[djococaud]

Simply because you can then add more leds in series on one output with a higher supply voltage...(within limits...)

[jimon]

danadak, wow thanks! just checked datasheets for some parts and indeed - current rating for LO state are like 100 mA, while HI are like 5-10 mA ...

[bson]

It's also much easier to design circuits around a common ground rather than common supplies.  Even the example shows dual supplies - 3.3V for the logic and 5V for the current source; with a common supply the device would either need to operate at 5V or you'd need multiple offset grounds.  Supplies are easier to have many of since they only need to source; ground needs to both source and sink, which is more complex, and makes it easier to only have one of.

[brabus]

The problem with PMOS is also the lower electron mobility in the P-silicon, if I well remember.
In other words, a PMOS has to be 3x larger to perform as the equivalent NMOS.

You could say: why don't we use an NMOS to switch the positive rail? That would be good, but the gate driver voltage would have to swing over the supply... a real PITA, basically only done on switching power electronics applications, using a bootstrap capacitor.

[danadak]

The current per output, sink, is one of two ratings you have to pay attention to.

The other is total current of all outputs flowing in supply rail. This tends to be a greater
problem if driving to Vdd vs sink which is driving to Vss. Reason is where the current
is concentrated, in later case substrate. Vdd rail more sensitive because of the general
way its routed and the large number of drops to circuits off the rail.

So look for the total supply pin allowed rating to manage the number of outputs that
can carry max current. Pay attention also to special notes, does the rating include overall
chip load or is it just I/O loads.

Regards, Dana.

[mikeselectricstuff]

The current per output, sink, is one of two ratings you have to pay attention to.

Three things - you also need to watch total power dissipation. Many 16 or more channel drivers can't deliver their rated maximum output on all channels if the drop across them is much more than the dropout voltage, though they do typically shut down or reduce current rather than melting.