Cloned ICs for a display driver?

[mikeselectricstuff]

I don't recall the OP mentioning greyscale

[amyk]

Here's another option.  Sticking with the MAX7219 approach, Austria Microsystems (AMS) also has their AS1106 and AS1107 equivalents of the MAX7219 and MAX7221 respectively.  They're not as cheap as the MAX7219 clones/counterfeits, but I found them to be about half the price of the Maxim parts at Mouser last time I bought a couple.

Incidentally, AMS also has one of the cheapest IC prototyping services, if you ever want a custom IC... http://cmp.imag.fr/IMG/pdf/cmp_prices_schedule_sept-16.pdf

[technix]

Video data bandwidth = framerate * frame height * frame width * bits per pixel. For example a 320x240 screen at 1bps (black/white only, no grayscale) at 30fps spells more than 2Mbps pixel bandwidth. 1080p at 24bpp and 60fps (a common computer screen) comes close to 3Gbps. Can your signal source handle that data bandwidth?

If I were tasked to design such a system, I would use an HDMI receiver followed by an FPGA to confine all high speed dataflow inside the FPGA, then the FPGA subdivides 1080/60p input into small blocks, say, 36 sub blocks, each handles 480px*120px*24bpp*60fps=83Mbps, then each sub block data stream goes into a small and cheap CPLD, to further subdivide into 36 channels of 40*40 matrix, which can be then used by LED matrix, implemented by either shift logic, CPLD or LED driver ASIC. The longest and most complicated wiring system would be the last stage, which carries only 2.3Mbps, so if terminated properly, even using single ended logic level, it should be still good enough.

How do you keep all screen update in sync then? Will the propagation delay bite?

[technix]

I don't recall the OP mentioning greyscale

Even without grayscale (1bpp) pixel data still build up very fast when the display resolution and framerate increase.

[Fungus]

I don't recall the OP mentioning greyscale

...or a frame rate.

He might be showing cricket scores for all we know.

[mikeselectricstuff]

I don't recall the OP mentioning greyscale

...or a frame rate.

He might be showing cricket scores for all we know.

You still need a decent refresh rate for a multiplexed display - that's the limiting factor bandwidth-wise.
With  greyscaling that's going to increase by  significant factor depending on number of greyscales

[technix]

I don't recall the OP mentioning greyscale

...or a frame rate.

He might be showing cricket scores for all we know.

You still need a decent refresh rate for a multiplexed display - that's the limiting factor bandwidth-wise.
With  greyscaling that's going to increase by  significant factor depending on number of greyscales

He can use distributed VRAM, this means each chunk of frames have its own VRAM and scanning chip (like MAX7219,) but then it comes the data synchronization issue. Keeping data in sync is tedious and easy to make mistakes.

[MobileWill]

I went through all this when I designed my led matrix links. Most of the max7219 are fake. There is a post from tronix about things to look for. IMO the MAX7219 is worth it and makes your life easier. That said the fakes do seem to work, there are lots of products that use them and projects without issue. For my product I ended up switching to the AMS1107 which is a tad cheaper and has a wider voltage range. It's a drop in replacement and the same code works. If you want an example of mine it is at https://friedcircuits.us along with design files.

[skytoastar]

Well the scope of the project certainly grew a lot over the weekend! HDMI. FPGAs. Custom ICs. Frame buffers. Not really what I had in mind but that's okay.

He might be showing cricket scores for all we know.

That's not too far off base from what I had in mind for these displays. Each one would display a alpha-numeric character. There might be some animations with the characters but they would be relatively slow and simple. No grayscale but there would need to be PWM dimming.

I got quotes back from a couple of PCBA houses and they both said $1.40 each in quantities of 100 for a MAX7219CWG. They certainly would make things simpler than shift registers in that I wouldn't need to push bits constantly whether my display data was updated or not (unless I use so many registers that I don't need multiplexing). But shift registers aren't out of the question. I'll certainly look into the AS1106/7 option, also.

[viperidae]

How is a tlc59281 a better fit for this than a max7219?
The tlc can only drive 16 leds without another chip doing multiplexing. The max does 64 leds.

[mariush]

@viperidae
He wants alphanumeric digits made out of 5x7 pixels, max7219 seems to be 8 digits x 7 segment (+dot) numeric only (with some intensity control).
He wants big digits, which means if he goes with max7129 every "segment" would probably need multiple leds and the total current of all those segments in a digit may exceed the limits of the chip, making some digits brighter than others, depending on how many segments are reached.
He wants lots of alphanumeric elements, not just 8 or less. So there's going to be multiple of these units, whether they're 8 digits or 1..n alphanumeric blocks of 5x7 pixels , so besides this max 7219 or some other chip, he's still gonna a microcontroller for each or something like that.
 

[ebclr]

TM1638

Easy to use and very very cheap

[viperidae]

@viperidae
He wants alphanumeric digits made out of 5x7 pixels, max7219 seems to be 8 digits x 7 segment (+dot) numeric only (with some intensity control).

max7219's can be configured to drive an 8x8 led matrix.