Concept & Design of 3.3V Parallel 16-bit VGA Boards

[Arlet]

Well, the board works again, and it was not the FPGA that was cooked. Instead there was a ~90 Ohm short between the PROGRAM trace and ground, somewhere in the board itself. I removed both PROM ICs, and I lifted the PROGRAM pin on the FPGA, and something was still pulling down the PROGRAM signal. I ended up cutting the trace on the back of the board, near R1, where it goes to the PROMs, and that fixed it.

[Arlet]

Anyway, I checked in some line drawing experimental code. It's still limited, but seems to work. Only a single line from 0,0 to w,h is supported, and w > h. Also, there's no end condition yet, so the line will continue all the way until the bottom of the screen, wrapping along the X-axis. Part of the code has been copied from my sprite code, and it still contains some references to sprites.

[ElEctric_EyE]

Excellent! I'll re-check the v1.0h design, as I have not tried using it yet, and compare it to my functioning v1.0g board. I really don't remember changing anything related to the JTAG programming, except for the fix to the MUX.

One question though: It was working before, why would it have stopped working all of a sudden?

Haha, you're tackling lines?! Awesome!

[Arlet]

ElEctric_EyE wrote:

One question though: It was working before, why would it have stopped working all of a sudden?

No idea. I had both PROMs removed, and the FPGA pin lifted, and there was still a connection between ground and the PROGRAM signal. After everything worked, I tried to investigate some more, but then the "short" disappeared.

I looked at the entire length of the trace with a magnifying glass, but couldn't see anything suspicious. One possible explanation is that some flux was left on the board, and it started conducting, but I don't see any place where the trace is exposed and close to ground. Another explanation could be a board manufacturing error where one of the vias is touching the ground in an inner layer.

[ElEctric_EyE]

Arlet wrote:

... I ended up cutting the trace on the back of the board, near R1, where it goes to the PROMs, and that fixed it.

That is the /Program signal going to the PROMs. So now you have a /Program signal going to the FPGA from the pushbutton, but no /Program signal going to the PROMs? Are you back to programming the FPGA direct only?

[Arlet]

ElEctric_EyE wrote:

Arlet wrote:

... I ended up cutting the trace on the back of the board, near R1, where it goes to the PROMs, and that fixed it.

That is the /Program signal going to the PROMs. So now you have a /Program signal going to the FPGA from the pushbutton, but no /Program signal going to the PROMs? Are you back to programming the FPGA direct only?

Yes, but I can still do some measurements to make sure the short is gone, and restore the broken trace. Right now, I'll just program the FPGA. It's quicker anyway.

[Arlet]

I've added support for drawing multiple lines on the fly, like in a vector display. Sources on github. Here's a still image, but the slope of the lines is adjusted every frame, so they move on the screen.

I noticed that the bottom line sometimes flashes on. It looks like this happens when a line crosses the bottom right pixel. Also, on the very right, you can see some stray pixels (not in this picture, but only at some line angles)

0003.png (6.63 KiB) Viewed 2058 times

[ElEctric_EyE]

I've tested for shorts on the /Program line on the 2 bare boards I have and found none.

I happened to download your newest github post about 30 min's ago. Works good, except when dy>dx. Also the white line looks 2 pixels thick. As you mentioned, I've also noticed the stray pixels on the very right edge, and flashing horizontal line on the very bottom.

[Arlet]

Yes, the "dotted" line when dy > dx is a known issue. It's not hard to fix, but I decided to work on other stuff first. The other bugs I have no idea yet what's causing them. The problem is that it takes a while to show up, and simulating more than a few frames is a pain, so I need to set the initial conditions such that the bug appears right away.

Of course, I also need to provide proper start and endpoints, instead of extending the lines all the way across the screen. I didn't notice anything exceptional in the white line, but I'll take a closer look.

[ElEctric_EyE]

If you add x <= x - 1 after line 107 of line.v module, negative slopes and dy>dx seem to work.

[Arlet]

If you look closely, that results in stairstep patterns. Also I'm not supposed to have negative slopes yet

I've checked in a fix for the dy > dx case. I'll look at the other bugs after this.

vertical.png (2.5 KiB) Viewed 2047 times

[ElEctric_EyE]

Interesting to note also, the "random" pixels on the right are gone. Maybe a clue there even though the x <= x - 1 is wrong...

[Arlet]

Fixed the flashing bottom scanline. I also don't see the stray pixels on the right anymore.

Edit: added support for negative dx, and added some more lines in the demo.

0004.png (3.03 KiB) Viewed 2036 times

Done for tonight, and I won't have any time tomorrow.

[ElEctric_EyE]

I was trying to add more lines too in order to see if it affected speed, I'll compare your code. Thanks!

I also traded out the DLL_SP for a PLL_BASE, even though the tools optimized it to a PLL_ADV. With the 7 lines the system can run @133MHz with 640x480 @25MHz pixel clock.

I got the resolution up to 1024x768 @71MHz pixel rate, but the system clock max with 7 lines is about 88MHz. It was a quickie test and it ran, but without switching all other parameters in your code, you can imagine it looks like a mess, but I could see the general behavior well.

I will try your new code now and observe top speed limitations at both resolutions.

[ElEctric_EyE]

Ah sweet! I kept the settings for 1024x768 and it still passes synthesis with no timing issues with the 16 lines.