WS2811 and PIC16F1824 trouble!!!

[maxikys]

Hi everyone. I'm trying to develop a device which converts the protocol of DMX-512 control protocol 20 pixels based on chip WS2811.
I was faced with the difficulty of correctly display colors. Code does not work.
Forum looked through all the threads about the WS-2811. But doing something wrong.Compiler Version 5.15.
Here is the code that should extinguish and light all the LEDs:

[ckielstra]

I never used the WS2811 chips before so I did a quick study:
The chips like the WS2811 do clock in the data when the data line goes from high to low, this means there is no separate clock line needed. The chips see the difference between a '0' and a '1' by the difference in time the data line is high.

Because of the tight timing requirements you can use a hardware module like SPI or UART to generate the bit stream. You have chosen SPI. Great.

A '0' is sent as 0.5us high and 2.0us low. (+/- 150ns on each)
A '1' is sent as 1.2us high and 1.3us low. (+/- 150ns on each)

As SPI sends 8 bits at a time and all with equal timing in the bits you will have to come up with a trick: send data at 8 times the required speed and define two different patterns that will create these timing sets.
bit 0: 11000000 will create 0,625us high (just within spec for max. 0,65us high)
bit 1: 11110000 will create 1,250us high

First problem in your code is that you have defined a SPI baud rate of 1.6M, this equals to a bit frequency of 1.6M/8 = 200kHz. This is too low. Only allowed bit rates are 400kHz or 800kHz. Change SPI to 3.2Mbaud.

Another possible problem is that for the '0' you have chosen a bit pattern '10000000' which equals a high time of 0,325us, just a tiny bit out of specifications. Better change it to '11000000'.

[maxikys]

[maxikys]

Using the example of the link http://www.insomnialighting.com/code/ws2811.c
I have the color change with the first test of WS2811. Other modules do not respond, the data on the DO pin persist.

[bkamen]

ckielstra already covered the big issues --- so next is:

Do you have a scope to monitor the serial stream?

(I have working code for the WS2812 -- so I can tell you it works)

Monitoring what you're doing is crucial when tight timing is involved.
_________________
Dazed and confused? I don't think so. Just "plain lost" will do. :D

[ckielstra]

When the first few modules work but fail for the modules later in the chain this is an indication for timing problems.

A Bkamen indicated it would help a lot when you can do actual timing measurements. Right now we are guessing. In the end we might get lucky and guess right but it is not a professional approach.
Do you have access to a scope?

One luring problem is the use of CCS software SPI library being mixed with hardware SPI. The website where you copied the code from already has this problem, but you deleted the line:

[maxikys]

Thanks for the tips guys, I pulled the string

[Ttelmah]

Time, time, time.....

First thing. With SPI on the PIC, there are two _separate_ ways of handling things. The original 'hardware only' mode, using setup_spi, and spi_read/spi_write. Then the newer 'hardware or software' mode, which uses #use spi, and spi_xfer. These should not be mixed. The manual does not make this clear, but if you look at the #use spi entry, it refers you to the spi_xfer entry, and _not_ to the spi_read and spi_write entries. Similarly, if you look at the setup_spi entry, this refers you to the spi_read, and spi_write entries, _not_ to spi_xfer. They are distinctly different libraries, and should be kept apart....
Ckielstra has pointed this out.

Unfortunately, the #use spi library, will not tell you if it cannot achieve the timings you are asking for. Hence if timing is critical, you need to check these yourself.

Now, that having been said, the code you started with, used the setup_spi form. The timings are more complex than it may at first appear, since the clock rate is not the SPI rate. The way this chip is handled, it uses two different pulse widths to generate 0 or 1. The SPI interface is being 'cheated' to generate this, by using eight SPI bits for each bit in the stream to the chip. It sends either a single '1' at the SPI rate, or four 1's at this rate, to give a 4:1 ratio in the times. 250nSec, or 1uSec at the original clock rate. The data sheet does not list the timing errors accepted at the high rate (only giving them for the low clock rate), and while the '0' time is spot on, the '1' time is rather long.

Now the problem is that at half the clock rate, you cannot generate this timing. The prescalers from the master clock for the SPI, are 4, 16, and 64 (steps of 4*), and you are changing the clock rate by 2*. So, you need to step back and see how to generate the required rate. Unfortunately on your chip at 32MHz, you can't generate this timing.

The 3.2MHz rate being suggested, would be ideal, but can't be generated by your hardware. Even switching to using Timer2, the available divisions, start at /16.

So, 'yes', your problem with the code, is the change to 32Mhz. However you may be able to get it working, by aiming to use the interface at half the rate. Unfortunately, at your clock rate, the ideal timing of 3.2MHz, is not available. I'd suggest trying the code at the lower rate, exactly as originally posted, with the SPI clocking off /16, _except_ changing the define:

#define ws2811_one 0b11110000

to

#define ws2811_one 0b11100000

This then gives 500uSec for the '0', and 1.5uSec for the '1'. Still 'out of spec', but closer.

You could also try switching to running your chip at 16MHz, and using SPI_CLK_DIV_4, which would then give the original timings. I'd still do the same change to the define, but this might well work.

Best Wishes

[bkamen]

[stinky]

Not sure if OP is still interested but I came across this thread while trying to get the 16LF1829 to run the WS2812b. I found the comments here very helpful. Hope this helps.
Edit: CCS PCM V4.141

[ressas]

Yes, this code works properly, but only one led is lit, my module consists of 9 leds. How can we do this

[stinky]

This code example only communicates with one WS2812b module to show proof of concept of the function. You just need to keep calling that function with more data. And it looks like you'd then call the WS_RESET macro.


You could iterate over an array of rgb values and call

[temtronic]

Since you're using the internal oscillator, you may need to adjust/alter/trim the timing of the WS_ONE(), WS_ZERO() macros.
Always good to see the pulses so you can measure them, then adjust code if required. If the timing is off, you'll see that the LEDs at the end of the line will not function properly.That's the nature of a long serial datastream.

edit: I dld/ read both datasheets,those PICs internal HF oscillator is +-2%. Do the math to see if it's within range of the WS2811 timing needs...
Jay

[Ttelmah]

The chip allows timing errors up to just over 10%, so the internal
oscillator should be fine.

[bkamen]