BT82x

[Rudolph]

The picture is nice and stable, yes, I started to implement the new commands for the BT820. 

But I currently have an issue that I can not explain.
First I noticed that I need to place additional VERTEX_FORMAT(0) in my display list, somehow the setting is not persistent.
Then I noticed that the back-to-back CMD_TEXT in the snippet I am using with CMD_APPEND no longer work, only the first line is displayed.
And when checking with the logic analyzer I found that a single CMD_TEXT is keeping the command-co-pro busy for an unusual amount of time, like there is some issue.
When I interleave the CMD_TEXT with COLOR_RGB then all the lines are displayed, the co-pro still is busy with this for quite a while though.

Choosing the less complicated CMD_NUMBER I get the same:

EVE_cmd_dlstart();
EVE_color_rgb(BLACK);
EVE_cmd_number(10, 515, 26, 0, 12345);
EVE_cmd_number(10, 500, 26, 0, 12345);
EVE_execute_cmd();

Followed by:
num_dl_static = EVE_memRead16(REG_CMD_DL);
EVE_cmd_memcpy(MEM_DL_STATIC, EVE_RAM_DL, num_dl_static);

The result right now is that the second line is not completely displayed.
When I switch the order of the lines, the other one gets incorrectly displayed.

When I throw in an extra EVE_color_rgb(BLACK); between the two CMD_NUMBER, then both lines are displayed.
When I use a NOP one of the lines still get corrupted.

The resulting sequence for EVE_cmd_number(10, 515, 26, 0, 12345); is:
0xff 0x01 0x00 0x00 - write + REG_CMDB_WRITE
0x2a 0xff 0xff 0xff - CMD_NUMBER
0x0a 0x00 0x03 0x02 - x = 10, y = 515
0x1a 0x00 0x00 0x00 - font = 26, options = 0
0x39 0x30 0x00 0x00 - number = 12345

I still need to read out the resulting display list snippet and analyze it, but right now I do not see anything wrong with my code.

Hmm, I added the lines with x=110 to the main display list and also made the list use separate commands instead of using DMA,
the two lines are displayed just fine.
Then I checked my implementation of CMD_APPEND, nothing there, changed the target address for the snippet, no change.

Looks like the co-processor does not like to put commands in a display list anymore that is not meant to be displayed on it's own?
There is no CLEAR_COLOR_RGB or CLEAR for the snippet, as this is in the main display list.


Edit: I just modified my code to use CMD_NEWLIST/CMD_ENDLIST/CMD_CALLLIST which just stores the co-processor command instead of executing them.
And this works just fine, when executed in the context of the display list, both lines for CMD_NUMBER are displayed correctly.

After restoring the rest of my original code everything is executed as it should, I even can remove the two extra VERTEX_FORMAT(0) lines that
were necessary before.
I noticed in ESE that both CMD_NUMBER and CMD_TEXT switch to use SAVE_CONTEXT / VERTEX_FORMAT(2) when going beyond 511 on the B817/BT818.
And with the BT820 SAVE_CONTEXT is replaced with REGION.
Oh wait, that is the issue, the destination in REGION is absolute, not relative, so with CMD_APPEND destinations are pointing to the wrong address in the display list.

At least there is a workaround, although it is exclusive to BT817/BT818/BT820.
Makes me again wonder why VERTEX_FORMAT(0) is not just the default.

Additionally I found that CMD_NEWLIST is not finishing on it's own, REG_CMDB_SPACE never indicates that the FIFO is empty.
Only after I removed the busy check things worked.
This behaviour is not documented and something I should try on the BT817.

[Rudolph]

I just implemented CMD_RESULT and had an idea.

The register region has room for over 2100 registers in the BT820.
Documented are 155 right now and regardless of how many register are undocumented, it should be highly unlikely that
there are not plenty of unused slots.

Can we please reserve a bunch and name them like REG_USER0 to REG_USER9?

Why?
Because CMD_RESULT(REG_USER0) would be nicer than needing to find places in the memory map over and over again.

[Rudolph]

I just added the functions for CMD_RESTORECONTEXT and CMD_SAVECONTEXT.
And I do not get it, what was the point of adding these as commands to the co-processor?

SAVE_CONTEXT and RESTORE_CONTEXT are display-list commands that were introduced with the FT800, IDs 0x22 and 0x23.
And executing these with the co-processor, like the rest of the display-list commands, never was an issue.

So instead of having the co-processor just copy over the commands to the display-list it now has to exectute the
command to write the same values to the display-list.
This really strikes me as weird.
And then these were the only display-list commands that got that privilege to be elevated to co-processor commands.

[Rudolph]

I am trying out the SD card functions now and just successfully used CMD_FSDIR.

CMD_FSDIR "only" returns a bunch of names.
How can I find out which of these names are directories?

Edit: I implemented CMD_FSSIZE and the number returned for directories is 0xffffffff.
Well, yes, that matches the description of CMD_FSSIZE, that entry is not a file, so no file of that name was found.
And consequently CMD_FSREAD and CMD_FSSOURCE return 2 on folders for file-not-found.

So when the name is returned by CMD_FSDIR and CMD_FSSOURCE returns 2, then it is a folder.

[BRT Community]

Hi Rudolph,

Thanks for your suggestion, we'll pass that back to our R&D team and get their feedback,

Best Regards, BRT Community

I just implemented CMD_RESULT and had an idea.

The register region has room for over 2100 registers in the BT820.
Documented are 155 right now and regardless of how many register are undocumented, it should be highly unlikely that
there are not plenty of unused slots.

Can we please reserve a bunch and name them like REG_USER0 to REG_USER9?

Why?
Because CMD_RESULT(REG_USER0) would be nicer than needing to find places in the memory map over and over again.

[BRT Community]

I am trying out the SD card functions now and just successfully used CMD_FSDIR.

CMD_FSDIR "only" returns a bunch of names.
How can I find out which of these names are directories?

Hi Rudolph,

One way to check is to perform a cmd_fssize.  Directories will return -1 for the size.
We'll add that as a note in the document too,

Hope that helps,

Best Regards, BRT Community

[Rudolph]

One way to check is to perform a cmd_fssize.  Directories will return -1 for the size.

Well, yes and no, technically. :-)
The size is documented to be a uint32_t.

And I got a couple of steps further along now, my test code is using CMD_FSSOURCE, CMD_LOADIMAGE with OPT_FS and CMD_GETIMAGE
to load one of the files returned by CMD_FSDIR and then display it.
I also found OPT_TRUECOLOR during this.

[Rudolph]

Another day, another screen of progress. :-)

I just added CMD_LOADASSET and tried it with two fonts converted with EAB 3.0.
Loading from SD card is really nice.

I am wondering what is driving up the display-list so much.
EVE_cmd_cgradient(EVE_EDGE_ZERO, 300, 200, 200, 200, WHITE, PURPLE); - 140 bytes
EVE_cmd_arc(400, 300, 100, 110, 0, 16384); - 144 bytes
debug-output of 8 hex-numbers - 736 bytes

Ok then, most of it is for all the text.

[BRT Community]

Nice work Rudolph! Your application looks great and good to see you've got the SD card running,
Yes, the longer RAM_DL is a nice thing to have on the BT820B to support larger amounts of text etc.

Best Regards, BRT Community

[Rudolph]

I implemented a couple more commands and got to CMD_WATCHDOG.

The watchdog command enables the watchdog timer and sets the watchdog reset intervals in clocks.
The watchdog flag in REG_BOOT_CFG register must be set.

C prototype
void cmd_watchdog ( uint32_t init_val );

Command layout
+0 CMD_WATCHDOG(0xFFFF FF83)
+4 init_val

Parameters
init_val
Watchdog timeout initialization value, in clocks. Must be a sensible value to prevent
watchdog being triggered prematurely.

Examples
To set the watchdog timeout every 72000000 clocks, which is 1 second on a 72 MHz system:
cmd_watchdog(72000000);

Adds watchdog timer to facilitate automatic correction.

The watchdog timer uses a 32-bit counter and is driven by the system clock. Hence, at default system
clock of 72 MHz, the maximum timeout period supported is 59.6s.
The watchdog timer begins and resets as it counts down from a user-defined initial value. Upon reaching
a count value of zero, the watchdog timer generates a system reset. The timer needs to be periodically
reset to the initial value to prevent the system reset from happening.
Refer to BRT_AN_086 BT82X Series Programming Guide for details on how to program the Watchdog
Timer using cmd_watchdog

That is practically all if it.
So yes, the watchdog system needs to be enabled during boot.
And it does a system reset - whatever "automatic correction" this is good for.
Then there is CMD_WATCHDOG which is documented to enable the watchdog and configure the timeout.
But there is no command like CMD_TRIGGER_WATCHDOG, so CMD_WATCHDOG is resetting the timer?

And if CMD_WATCHDOG really is the only command, does a CMD_WATCHDOG(0) disable the watchdog?

What safety requirements are covered by this?
If the EVE chip would crash hard, and I never encountered this, would the watchdog timer really not be affected?
And how would a reset help in this scenario? The host controller had to detect either way that the EVE chip is not answering anymore.
Ok, different scenario, the host controller crashes, how would resetting the EVE chip help? The host controller would still be dead.
So when the watchdog is not triggered by CMD_WATCHDOG, the screen goes dark?

What am I missing here?
If the EVE chip would be able to reset the host controller, that would be different, that would make the EVE chip an external watchdog controller and that could  be viewed as positive for the system safety.
But I would use a window-watchdog with challenge-response for such a task, preferably in a SBC.

[Rudolph]

I implemented CMD_LOADWAV and in order to test it I enable audio in BOOTCFG.
The screen stays black though.

The host command is correct, it changes from 0xff 0xe8 0xc0 0x00 0x00 to 0xff 0xe8 0xe0 0x00 0x00.
Things look normal on the logic analyzer, "only" the screen stays black.

I am not writing to any of the audio registers, including not using CMD_LOADWAV.
I even moved the buffers for the swapchains.

Is it possible that the audio engine is messing with the clock tree?

[Rudolph]

Some might have noticed it already, I uploaded a new library on Github: https://github.com/RudolphRiedel/EmbeddedVideoEngine
I had the idea of renaming my library for a while, "FT800-FT813" was just wrong for a while now.
However, renaming seemed to be a bit disruptive, so I created a new repository for V6 of my library and the name should be future-proof now. :-)
I was already using the name for a while now in the library.json for PlatformIO.

Yesterday I tested the Arduino example, made sure it compiles for all tagets (although a few are non functional like the one for BBC Microbit.
And then I tested ESP32-S2, Arduno UNO V3, Teensy 4.1, Raspberry Pi Pico and Metro M4 with both a RVT50H (BT817) and the VM820C.

Right now I have the VM820C with the 7" 1024x600 TFT up and running with an ATSAME51 after a fix to the DMA code, and I already checked it with a BT817.
As usual, no warnings with -Wall -Wextra -pedantic.

Before I was using a STM32F407, so most targets are covered by now, at least are compiling or actually are tested.

While the lib is feature complete for the BT820, this is not finished.
There are still a couple of things missing like configuring the memory size for the BT820, the LVDS output clock is merely working by chance now, Audio does not seem to be working at all, video input is unexplored...

Unfortunately there are no modules for sale with a BT820, as far as I know. And only one was anounced.

I also learned again that sourcing LCDs is still a major pain, these things are really not commodities, speaking of course for not buying in bulk from a company account, trying to aquire single units.

[Rudolph]

Time for some torture-testing of the BT820. :-)

The attached image shows this code:

Code Select Expand


        int16_t xc0 = 0;
        int16_t yc0 = 0;
        uint16_t length = 3;
        static uint16_t increase = 0;
        uint32_t colors[] = {RED, ORANGE, GREEN, BLUE, YELLOW, MAGENTA, PURPLE, BLACK, 0xc0c0c0, BLUE_1};

        for(uint8_t arms = 0; arms < 12; arms++)
        {
            EVE_color_rgb(colors[arms]);
            EVE_begin(EVE_LINE_STRIP);
            for(uint8_t steps = 0; steps < 135; steps++)
            {
                EVE_polar_cartesian(length*steps, arms*30+steps*increase, &xc0, &yc0);
                EVE_vertex2f(xc0 + 450, yc0 + 320);
            }
            EVE_end();
        }

        static uint8_t delay = 0;

        delay ++;

        if(delay > 2)
        {
            delay = 0;
            increase += 3;
        }


12 Linestrips with 135 segments each, 6916 bytes in the SPI buffe, a display-list of >8100 bytes, takes 1.3ms to preprare the buffer.
And this runs continously now, changing constantly.

All I noticed was that it started to flicker when I increased the "steps" in the loop.
Not for every value of "increase" though.
The flickering goes away when I decrease the maximum "steps" or increase the time between display-list updates.
Right now I am running this as posted and refresh the display list every 75ms, so 13 changes per second.
So this put the render-engine into stress-mode, sort of.

Nice, this stupid level of useless complexity should be way beyond what a normal application can put out to render. 

[Rudolph]

Any ETA on an updated BT82x programming guide?

A few answers to the more crucial questions I posted would be very nice beforehand.
Like how REG_LVDSTX_PLLCFG actually works - if the guide is wrong there and I am not using the register incorrectly...


And I have a new question.
Is there a way to check if rendering of a display list is done?
Well yes, I could read REG_RE_RENDERS and check for a change.
But this takes a bit of caution and requires to not into a deadlock.
Or I guess using CMD_GRAPHICSFINISH could work, reading REG_CMDB_SPACE after this should indicate that the co-processor is waiting for the render engine to finish.
Hmm, sounds like a plan, I maybe should always end the BT82x display list with CMD_GRAPHICSFINISH and not send a new display list when the co-processor is still busy.

Well yes, there might be other things to do for the co-processor while the render-engine is busy, like loading assets or using CMD_TEXTDIM, in that case ending the display-list with CMD_GRAPHICSFINISH might be a bit inefficent, perhaps, especially when the render engine needs a bit more time than usual.

[BRT Community]

Hello,

On the programmers guide point, this is currently going through the final review stages and we hope to have it released shortly.

I will follow up with the R&D team for your other queries.

Best Regards,
BRT Community