16-channel "RGB" LED drivers - am I being dumb?

[HwAoRrDk]

There are 16-channel LED driver ICs from several manufacturers that describe themselves as being for driving RGB LEDs. ICs such as the PCA9532, LP3943, CAT9532.

Unless I'm completely missing something, it appears to me that these do not actually appear to be capable of driving RGB LEDs with different PWM duty cycles per R, G and B channels.

Each output channel can be set to choose between one of four states: off (hi-z), on, on at PWM rate 0, on at PWM rate 1. These two PWM rates 0/1 are global.

So how the hell are you supposed to set an RGB LED that's hooked up to 3 output channels to, for example, 50% red, 93% green, and 10% blue? You're covered for the first two colour channels, but the third has to be the same as one of the other two!?

Please tell me I'm being dumb and have missed something important, as otherwise these chips seems useless for one of the claimed headline uses.

[mikeselectricstuff]

These look like they're designed for backlights etc. where there is only one colour generated overall.
There's no shortage of drivers that can drive individual channels seperately

[S. Petrukhin]

It looks like in these chips you can choose which LEDs are turned on, but the PWM on them is shared.
You have the selection registers LS0-LS3 that control the switching on of the output type, the PWM is fed to all at once at the same time using the registers PWM0-PWM1.

[HwAoRrDk]

These look like they're designed for backlights etc. where there is only one colour generated overall.

Even in that application, how are you supposed to generate that colour when you only have the ability to specify the PWM value of 2 out of the R, G, and B triplet?

There's no shortage of drivers that can drive individual channels seperately

Of that I don't doubt. I just found it stupefying that these chips can't do something they claim to be useful for.

[S. Petrukhin]

Of that I don't doubt. I just found it stupefying that these chips can't do something they claim to be useful for.

These are chips for lighting lamps - there is no need to generate a rainbow, but you need one or two combinations: warm/cold.

[ajb]

Using only two colors out of RGB with variable brightness at a time gives you access to all of the full saturation colors (the circumference of the color wheel or circumference of cylindrical color space), which is probably fine for a lot of indicator or accent lighting applications.  You just won't have any desaturated colors (like pastels), but RGB LEDs don't tend to do those very well anyway unless they have good diffusion/mixing optics. 

The datasheet for the PCA9532 also makes it pretty clear that color control is really a bodge based on adapting blink functionality designed for indicators.  I guess realized that if they just crank up the blink rate, hey presto, it's got PWM dimming!  But the PWM frequency is only 152Hz and the resolution is only 8 bit, which is pretty abysmal if you care about smooth fades and flicker-free dimming.  You generally won't notice flicker with the naked eye under static conditions, but if the LED is moving rapidly within your field of view (like you turn your head rapidly) it can be visible, and it will absolutely show up on camera.

[mikeselectricstuff]

Interestingly these chips have essentially the same pinout and register set  - This suggests to me they're clones of each other for a high volume market that cares about second sources.

[Gibson486]

Linear Tech (or Analog Devices now) has a few ICs that can do what you want, although I am not sure they go to 16 channels, so you will need multiple ICs. With their devices, you provide at actual input signal, though. I have used them for multiple designs successfully.

[Benta]

The way I read the datasheet is, that you have 1 IC for 16 x R LEDs, 1 IC for 16 x G LEDs and 1 IC for 16 x B LEDs.
Whether the LEDs are combined or discrete is of no significance.

So yes, you can do colour mixing.