Ghosting in columns and other troubleshooting adventures

[TimW]

I'm playing around with one of Frank's PCB Ledtriks panels at the moment (16x16 leds)

Sigh... it was so easy to put this one together... I remember the pegboard days!

Anyway, have just come across an interesting problem with my setup. Not sure anyone else will see exactly the same thing but just in case I thought it might be worth mentioning.

Looking at the points below I'm not trying to put anyone off Ledtriks - just sharing how I worked though this latest problem with the help of the forum posts here. Of course not every board/ panel will have these problems!! But folks here have been good over the last year or so figuring any difficulties out and a summary might help.

There are various posts on the forum about troubleshooting Ledtriks boards/panels. As I understand it there seem to be several problems people have run into...

1) Cabling / soldering faults. Yes, if you wire it up wrong it doesn't work... this includes putting leds in the wrong way, putting chips in the wrong way, bad solder joints and the harder to chase down - occasional RJ45 crimping and socket issues. we all know there is a lot of soldering with this particular project!!!

2) Grounding/ power config faults - remember that when using Ledtriks with a parallel port you need to run a separate ground to the supply

3) Parallel port issues... same principle as the Grinch here... the MBI chips need 0.8 of full voltage present to recognise a '1'.. if the parallel port isn't up to it (or if the supply voltage is a bit high) data might start getting screwy which will result in a variety of symptoms (clock pulses missing, latching problems etc). Note that the Allegro chips only need 0.7 of full voltage which is why they seem to work better... most of the time!

One of the recent successes people have had addressing the voltage issue is putting a diode in the supply line to drop Vsupply a bit. this appears to solve some problems (but if you drop the voltage to the whole ledtriks board it dims the leds a bit as well! (The more complex solution would be to provide full voltage to the ULN chips supplying the Anodes of the Leds and drop the voltage to the MBI/Allegro chips).

(or you could try a TRIKS-C .. or a different power supply?)

4) Bad Leds in a column
Sometimes people see a column of LEDs dimly lit all the time (usually with one led out). Replacing the led that is out -or working through the column replacing led until you find the bad one seems to fix this.

5) Faulty MBI/ Allegro.
I've had a couple of chips die completely (must have been something I did ) Others can fail with a single column 'leaking' that doesn't turn fully off. This gives a symptom like 4 above but it moves columns if you swap the position of the MBI or allegro chips.

6) STP chips.
Some of us tried the STP chips when Allegro/MBI not available - I had a problem where the rows went dim if more than 8 or so leds were on in a row. Turned out to be not enough decoupling caps on the board - problem disappeared with bigger caps.

There are probably more variants and experiences out there. But the new one I have been trying to get rid of on this latest rig is 'column ghosting'. What I see if I turn on a single led in a column is an 'echo' - one (or occasionally 2) of the other leds in the column lights dimly only while the actual led is on. the effect tracks across columns and is fairly consistent - and annoying - on my panel.

After trying various things over the last couple of days to fix it I decided that it must be due to some capacitance on the cathode line that continues to sink a small amount of current even after the MBI ouput is turned off. In the case of my board it was enough to allow the next row of the column to switch on a bit even when it was not meant to (remembering that the sequence of the rows is not 1-16).

I could not find where the capacitance was or why it should be different to my other panels. So I fixed this problem by tieing each of the cathode rows to +5V via a 470 ohm resistor. Now the panel is rock solid - ghosting is gone!

I think this works because it stops the cathode line 'floating' when the MBI goes high impedance... and instead pulls the cathode voltage to the Anode voltage level so there is no way any leds in the column can turn on . Downside is another 10ma or so for the chip to sink when column is on... but its only on for 1/16 of the time anyway because of how Ledtriks addressing works. I guess I could compensate by changing the current limiting resistor.... but for now it still seems bright enough!

The 16 resistors mount neatly onto Frank's PCB into one of the hole sets provided at the edge of the boards. Neat.

Not sure that anyone should add the resistors unless its a problem -(even if my diagnosis of my problem is right your capacitance may vary!) but its another potential solution to add to the list if you're stuck!

Thoughts anyone? Anything I missed?

Tim

[kostyun]

Wants to post a quick followup to this!

My Ledtriks have been plagued for 3 years with ghosting, etc with no real solution working - except this one - using 470ohm resistors.

I had had the issue with the Allegro's, and most recently the MBI chips.

I added the resistors as 470ohm 9 pin SIP resistors (6 on each board) as shown below.

[XmasInGalt]

Wants to post a quick followup to this!

I added the resistors as 470ohm 9 pin SIP resistors (6 on each board) as shown below.

Frank, do you have a mouser # on the 9 pin SIP resistor that you used? I want to make sure I get the right one. Thanks.

[Aurbo99]

Playing around with my Triks-C this evening I moved the original Trick board jumper from 1 through to 4 and each time it displayed columns or rows of lit leds.

The reason I tried this was because the very bottom row, columns 37-48 refused to come on when powered up with the shipped programmed chip, I dont recall what CO-OP it was for the -C with chip.

The board reads TRIKS-C with 0-2 when powered up. changing the jumper on the original Triks board shows the offending LED's do in fact work.

I've replaced 1 bad led over in column 6, it wasn't anywhere near the others.

No solder bridges, no bad joints seen, running power to each row/column with also shows every led functions.

Using a computer PSU with ATX adapter board for power. Power goes into the Triks-C board and jumps over to the original Triks board.

Did anyone else encounter this?

[budude]

The stored image for the 2313 does not have those bits on for some reason. If you have a programmer you can change it.

EDIT - here are the changes required:

INITFRAME:
.db 255 , 255 , 255 , 255 , 255 , 255
.db 1 , 0 , 0 , 0 , 0 , 128
.db 249 , 30 , 39 , 243 , 0 , 143
.db 33 , 34 , 34 , 9 , 128 , 128
.db 33 , 34 , 226 , 8 , 128 , 128
.db 33 , 30 , 98 , 112 , 156 , 128
.db 33 , 6 , 226 , 128 , 128 , 128
.db 33 , 10 , 34 , 129 , 128 , 128
.db 33 , 50 , 39 , 123 , 0 , 143
.db 1 , 0 , 0 , 0 , 0 , 128
.db 1 , 0 , 0 , 0 , 224 , 184
.db 1 , 0 , 0 , 0 , 160 , 160
.db 1 , 0 , 0 , 0 , 160 , 186
.db 1 , 0 , 0 , 0 , 160 , 136
.db 1 , 0 , 0 , 0 , 224 , 184
.db 255 , 255 , 255 , 255 , 15 , 128

Change that very last db statement to;
.db 255 , 255 , 255 , 255 , 255, 255

That will fix the issue of the missing columns in the last row.

[Aurbo99]

Thanks, I'll give it a try once I build my AVR Programmer

[budude]

I should have mentioned - there is a tool somewhere in the forum that presents a 16x48 model of the panels and you simply click on the pattern you want then click on a button and it spits out the appropriate .db output you can cut/paste there instead of the standard one. This way you can customize the start-up to anything you want. If I can find the link I'll update this post.

Edit - - Here is the thread for the tool... http://www.doityourselfchristmas.com...ead.php?t=4618

In addition, there are multiple frames of the same format which perform the self-test patterns - you can replace all of these with your own also. It's pretty obvious where they go once you peak at the code. I think there are 8 frames or so.

Brian

[Matt_Edwards]

My daughter and I have just finished 2 Trix panels.
One works faultlessly, the other has issues. by swapping the modules out and have traced the fault to the Triks boards and not the LED panels (peg boards) or the Trix-c boards.

This is the fault: columns 1, 17 and 33 flash really fast. Column 17 is very faint. all other columns have no output.

The funny thing was both worked OK to begin with. when I swapped the Trix-c boards over, the fault disappeared for 20 hours.

I intend to swap out each IC on the Triks board and try to isolate the problem.

If anyone has noticed this sort of fault, what did you do?

I will keep you posted

Matt

[Matt_Edwards]

I replaced all the ICs one at a time and found it panel works for a bit the then goes into the mad high speed flashing again. I ended up with the original ICs back in and it still runs for a short time and then faults.

BUT i also noticed the Diag LED on the Trix-c stops (some time hard on , sometime off) when it board goes into fault mode.

Time to retire to bed - I will look at it in ht morning with the scope on the power supply.
Cheers
Matt

[budude]

Almost sounds like the Triks-C is not sending the strobe signal (or it's a bad level) to the Triks controller board to clock out the bits ? Have to ask but did you try another cable between the two? Did you check the voltage to all the boards? I tuned my PS down to 4.7v to clean up the worst of the stray 'noise' on the display.