Easynote TK83-SB - blown AMD chipset replacement gone wrong - the nitty gritty.

[Refrigerator]

I had this PackardBell Easynote TK83-SB for a while.
When i got it the laptop did power up but would only display a single "_" at the top left corner and power cycle after half a minute or so.

On the rare occasion it would display a few error messages and stop power cycling, pressing anything would result in a loud beep and a copy of the same error messages to appear.
I googled one of the error messages and it said that it tried booting to LAN.

Then i found that the main CPU chipset was in fact blown, a burned trace could be seen on the BGA going straight to the silicon chip.

I ordered a used chipset from eBay for 5 bucks, i reckoned chipsets don't fail that often, a used one should be fine, plus this one came with the balls already on.

I guess i took a gamble and lost because after the chipset replacement the laptop does not power on anymore. 

When i press the power button the fan goes whoosh for a few seconds and then happens what appears to be a rather violent shut-down, maybe indicating an over current protection ? 

The time between power-on and shut-off is about the same as what would be the time between power-on and display-on previously.

Also worth taking note is that before with the blown chipset, with a single 2GB ram stick it would do as mentioned above but with a single 4GB ram stick, instead of the 2GB one, it would show absolutely no signs of life ( no fan spin, no lights - nothing ).
While on with the old chipset it would get rather hot, considering it was just sitting there doing nothing.
And also the heatsink had a carpet of dust on it, so the original source of the issue was overheating, WASD area under the heyboard shows no signs of abuse so the laptop was not used for gaming.
Cleaning the heatsink was the first thing i did when i got the laptop.

Even with a defective chipset i expected it to do more than to hard shut-down after a few seconds of power-on, makes me think what else could be bad on that board.

Edit: pics are up.

[Refrigerator]

[Whales]

Is that a pic of the bad one?  It's earthquake fractured on the lower-right.

Could be anything.  Bad soldering of the balls,  another part being fried, mounting the chipset the wrong way around, bad part from eBay.

Perhaps the chipsets were different revs?  How does the old part look to the new part, side by side?

[Refrigerator]

I added a pic of the "new" chipset, i can upload more detailed close-ups of the chips you you need.

[Refrigerator]

A closeup of the blown chipset.


And found more blown parts !


This part here is the ESD protection diode on the USB port, which, as you can see, has released the magic smoke !
Did someone plug a USB killer into this thing, how the hell do you blow an ESD protection diode on the data line of the USB this hard ?

Maybe this is why the laptop refuses to stay powered on ? 
Also, the chipset could not have possibly blown that diode or else the choke next to the diode would be blown as well, and it does check out fine.
But what doesn't check out fine are both of the data lines on the USB, which are shorted to ground, with a 0.034 and 0.015 voltage drop to ground.

[Refrigerator]

Removing the blown diode didn't help much, but i noticed that my chipset was not perfectly flat parallel to the board so i decided to reflow the board in the oven one more time.
And what do you know, the damn thing works !
But the colors on the screen look funny, seems like it can't display red color, all colors are either blue or green.
And it won't work with an external monitor, i've yet to install the AMD graphics drivers, maybe that's why ?

[wraper]

But the colors on the screen look funny, seems like it can't display red color, all colors are either blue or green.

Either some balls have not made a contact, balls shorted together or chip is damaged.

And it won't work with an external monitor, i've yet to install the AMD graphics drivers, maybe that's why ?

Might be, but quiet likely read above.

[Refrigerator]

So i installed the graphics drivers and the colors are still off, haven't checked it with an external monitor yet.
I need to get this color crap fixed because it is genuinely making my eyes hurt.

Otherwise it runs well.

[Refrigerator]

Maybe it's my mangled LCD cable that's at fault as i did have to fight to make it fit back into the case and to close back up.
I do have a spare cable to test out and i also have a scope to probe the LCD connector, if there are indeed separate data lines for each color then i can probe each one to see it there's any signal.
This can't be a fault in the GPU balls because i did do a pretty hot reflow on the second run to make sure it all goes well and if it was enough to solder in the chipset then it should have been enough to solder the minute solder fractures that occur under the GPU, which is also in close proximity to the chipset.
Here's what i'm looking at:


Also this was written on the same machine that i just brought back to life. 

[Refrigerator]

So just for the heck of it i ran Unigine Valley benchmark and it did score 328 points, not bad.
During the benchmark CPU temp topped out at 70^oC and GPU topped out at 63^oC, which is really good.
While browsing i noticed that the laptop is quite quick at it, so it would be a shame if i couldn't fix it.
Also i tried watching some 1080p 60fps and the GPU barely handles it, there is occasional stutter and noticeable V-Sync issues.

[Refrigerator]

quick update.
On the schematic: http://kythuatphancung.vn/uploads/download/d36fd_Acer_Aspire_7741G_Wistron_JE70-CP_Rev-1M.pdf
On page 24 there is the LCD connector and i can already see that there's three distinct double pairs of data lines that are referred to as A/B/C 0/1/2, which would indicate R/G/B if i'm not mistaken.

[Refrigerator]

Oh sorry, i meant there's 0/1/2 instead of A/B/C.
Maybe TXA is for the internal monitor and TXB is for external monitor ?

[Refrigerator]

I found out that i need to press Fn+F4 to output to external monitor.
Voila !

Well it only supports one monitor by the looks of it, but that's fine and as you can see the colors look like they should on the external monitor. 

To anyone, who's wondering why there's an acer logo on a PackardBell laptop is because i bought the laptop without a screen and i put the screen of my long dead acer in, which was a direct drop-in fit.

[Refrigerator]

So i tried a different LCD cable but the colors are still wrong.

[Refrigerator]

I probed around the LCD connector with my scope and found out some interesting things.
pin 15 and 16 -> green
pin 17 and 18 -> blue
pin 19 and 20 -> red
How do i know, you ask ? Simple.
Short 15 with 16 screen goes green.
short 17 with 18 screen goes blue.
short 19 with 20 screen goes black, and i know why.
Both R, G and B come in as pairs, of which one measured at ground level and the other measured at about 3.16 V_rms.
But the red one does not, both pin 19 and 20 show basically the same waveform.

Now i have to find out why is my red signal bad and this repair will be done.

[Refrigerator]

Well, i'm a bit lost at this point at this point, to be honest.
I tried poking and measuring all kinds of signals but could not come up with a solution to fix my no red issue.
when i shorted the other color signals they would go to ground level, manually pulling 19 and 20 down would result in the same black image.

I'm almost starting to think that my LCD could be bad, i don't really have any other LCD to test that out with.
Whatever i do it just refuses to display the red color.

[PA0PBZ]

I'm almost starting to think that my LCD could be bad, i don't really have any other LCD to test that out with.

If you can swap the red signal with either the blue or green you will know for sure 

[Refrigerator]

i'll have to try that.

[amyk]

18bpp single-channel LVDS, which this laptop seems to be using, is arranged like this: http://www.ti.com/ods/images/SLLSEC1G/fig4_flatlink_LLSEC1.gif

From the schematic it seems TXBOUT0 contains all the red channel. Is it a discrete or integrated GPU? The schematic has both options shown. Check the signals at RN49 (integrated) or RN47 (discrete) --- note that TXBOUT1 also passes through the same resistor network.

Swapping the signals will not work, since LVDS is not actually divided into R,G,B,HS,VS; instead, here there are 3 lanes, the first contains all 6 bits of red and the lowest bit of green; the second contains 5 bits of green and 2 bits of blue; and the third contains the rest of the blue as well as HS, VS, and DE. You may try duplicating the second or third lane onto the first and seeing if any red shows up, although I'm not too sure whether the signal integrity will be affected too much for it to work --- LVDS is current-mode and requires a termination resistance of 100R.

[Refrigerator]

It has an ATI HD7400M series GPU according to dxdiag.
I'll take screenshots of the signals at pins 15 to 20 and post them here later.
Maybe that will clear things out, but the signals do not look like digital ones at all, well atleast not any that i know of.
There is a ~1kHz square wave signal on one of the other pins, but that will most likely be the PWM for the backlight.

I think that at this point it's safe to assume that the motherboard is fine, since the external output works perfectly fine, unless i'm mistaken.

[amyk]

They won't look very "digital" if your scope doesn't have enough bandwidth. 1600x900 at 60Hz needs a pixel clock of ~86MHz, but the other LVDS lanes will be switching at up to 7x that or ~600MHz.

[Refrigerator]

my scope is only 25MHz, but it's a 1366x768 display on the laptop.
Nevertheless it still wouldn't show the signal properly, i suppose.
But i will take screenshots tomorrow just to show how things are looking over there.
Maybe there could be slight issue with compatibility, since if i try googling red images they show up as completely black on screen, indicating that the red pixels are not being enabled when the blue/green are disabled. 
Very weird how this happens but it's cool to learn all this new stuff.
I think there could be a possibility to splice the red into blue/green lines, you'll see once i get the screenshots, i have a spare cable so it's worth a shot.

[Refrigerator]

Ok so i took screenshots of pins 15 to 20.
Can't really make the scope show anything more than that but maybe it will be enough to give some ideas.
Also you can see some flat spots in the signal in some screenshots, these flat spots are present on all pins that show a signal but are hard to catch for a screenshot.
15 - LCD_TXBOUT2-
16 - LCD_TXBOUT2+
17 - LCD_TXBOUT1+
18 - LCD_TXBOUT1-
19 - LCD_TXBOUT0+
20 - LCD_TXBOUT0-

[Refrigerator]

Also a pic to better illustrate the color issue.

[Refrigerator]

I think i was right about the incompatibility of the LCD/motherboard.
I found the old dead acer motherboard, which used the LCD that is now in the packard bell.
here's the schematic:
http://www.s-manuals.com/pdf/motherboard/compal/compal_la-6582p_r0.2_schematics.pdf
On page 22 you can find that the pin-out is different from what's on the packard bell motherboard.
Hopefully i can rewire the cable and fix the issue.   

[amyk]

Ok so i took screenshots of pins 15 to 20.
Can't really make the scope show anything more than that but maybe it will be enough to give some ideas.
Also you can see some flat spots in the signal in some screenshots, these flat spots are present on all pins that show a signal but are hard to catch for a screenshot.
15 - LCD_TXBOUT2-
16 - LCD_TXBOUT2+
17 - LCD_TXBOUT1+
18 - LCD_TXBOUT1-
19 - LCD_TXBOUT0+
20 - LCD_TXBOUT0-

500us/div?

These signals are in the hundreds of MHz range (i.e. edge-to-edge times are only a few ns), even 500ns/div is too slow to see much of anything...

If your two schematics are the right ones, I'm surprised you even got it to display anything at all because the pinouts are almost completely different, and the LVDS lanes don't even match (not even the clock.)

Code: [Select]

13 CLK+     OUT0-
14 CLK-     OUT0+
15 OUT2-    GND
16 OUT2+    OUT1-
17 OUT1+    OUT1+
18 OUT1-    GND
19 OUT0+    OUT2-
20 OUT0-    OUT2+
21 GND      GND
22 USBDM    CLK-
23 USBDP    CLK+
And I'm pretty sure if it had a webcam it's fried now, with the full system DCBATOUT being applied to the 3V and USB lines...

[Refrigerator]

I already knew that my scope was not good enough to show the LVDS signal, but i took the screenshots anyway just to show that there is something.
Well, the screenshots are not relevant anymore anyway since i found my main problem.

[Refrigerator]

I rewired most of the connector and after plugging it in i saw absolutely nothing.
Then when i started buzzing out connections with my DMM i found out the the ground connections are not right on the PCB.
Turns out i had found the wrong schematic. 
Here, this one should be the right one, page 23. 
https://www.rom.by/files/acer_5551g-5552g_la-5911p_www.mycomp.su_.pdf
Well, i only rewired the data lines, and not the power lines, so no real damage was done.

By the looks of it the only difference between the two is that on the packard bell the LOCAL_DIM and COLOR_ENG_EN pins are marked as inputs rather than outputs.