X6X-series LED backlight conversion kit howtos

[85101]

I just installed a modded fancyboost in my X61s with the IDTech N121X5-L06 display (42T0343). About a vertical third of my screen is darker than the rest and tinted yellowish. It's not a sharp transition (no "lines"). Could this be bad LEDs on the strip or another problem?

Not bad LEDs-- that would cause left-to-right unevenness along the bottom. I think your LED strip has simply rotated out of position. The yellow tint is because of the kapton tape on the strip-- peel it off. It serves no useful purpose and gets in the way.

I removed the kapton tape and readjusted the position of the LED strip. The dark areas were fixed but the yellowing persists. It is only in a fuzzy band about one inch wide just to the right of the LCD cable. Any ideas?

Also, the ctrl and alt keys on my keyboard no longer work together after reinstalling my keyboard. They only produce a beep sound and have no effect. This is the third keyboard that I have managed to break in a similar way (simply removing and reinstalling), and I have no idea why.

[khseal]

Yes, the CCFL tubes are very fragile! But if only some rubber gets stuck, you can push it out with a thin dowel, or even just some q-tips with the cotton cut off. The tubes are fragile, but the screen matrix, when still assembled, is not. Just be sure not to use something like a metal rod with a sharp lip; it might gouge the polished edge of the waveguide, or rip up the insulating mylar coating in the bracket.

Actually, the matrix pieces are still fairly strong when disassembled, just be careful of the flex cables where they bond to the glass. It is, however, a real pain to figure out how to get it back together smoothly if you've not done it before Also, fingerprints on the rear polarizer....

I have experience in this matter. Display long ago assembled. There are a couple of dust, but overall I was expecting the worst. It was difficult to insert the polarizer in the lower the iron part.
I tried to pull the rubber band for a long time. As a result, there is yet another object stuck into rubber =)

The LEDs being sold in these kits are often too low a color temp and far too green. It is why I have been binning LEDs by hand the past few weeks...

There is another problem with LEDs. I think the new LED lights change color spectrum after a few hours. Maybe I'm wrong. I have no measuring devices.

[xiphmont]

I removed the kapton tape and readjusted the position of the LED strip. The dark areas were fixed but the yellowing persists. It is only in a fuzzy band about one inch wide just to the right of the LCD cable. Any ideas?

probably still positioning, just more likely the up/down position. On those side-facing LEDs (if they're the ones that came with my Fancyboost), the edges of the emission pattern are very yellow. The strips are too flexible and too small to sit in a consistent position and can sit too high/low (as well as too far back, or rotated, etc). I'd bet if you pull the lid bezel off or run the screen outside the lid, you can poke at the strip and watch the dramatic differences in the illumination pattern from small tweaks.

Also, the ctrl and alt keys on my keyboard no longer work together after reinstalling my keyboard. They only produce a beep sound and have no effect. This is the third keyboard that I have managed to break in a similar way (simply removing and reinstalling), and I have no idea why.

Oh, man. I've never had that problem on a X6x despite handling close to 100 keyboards at this point. I wonder if there's something inadvertantly terrifying about your disassembly technique (or maybe... an intermittent problem on the mobo that only looks like different keyboards?)

The only slightly fragile part of the keyboard is at the top where the ribbon cable disappears behind a metal screwed-on bracket. That's actually a sandwich connection that mechanically clamps the external flex cable to the matching contacts on the internal membrane. It can get pulled slightly out of place, and I think the modifier keys are the traces at the edges that would get misaligned first. Perhaps that's what happening? It's easy to disassemble and put back together. The parts themselves are tough, they can just get misaligned is all.

[xiphmont]

I have experience in this matter. Display long ago assembled. There are a couple of dust, but overall I was expecting the worst. It was difficult to insert the polarizer in the lower the iron part.

That's the waveguide. I use a thin piece of stiff polyester as a 'shoehorn' but these days I avoid disassembling that far if there's any way not to...
[the rear polarizer is the silvery layer on the back of the glass. Very easy to scratch, picks up dust and fingerprints like crazy].

There is another problem with LEDs. I think the new LED lights change color spectrum after a few hours. Maybe I'm wrong. I have no measuring devices.

LEDs do change color as they heat up. Some change color more than others. I haven't tested those to see if they're worse than usual, though I have some so I could....

[85101]

probably still positioning, just more likely the up/down position. On those side-facing LEDs (if they're the ones that came with my Fancyboost), the edges of the emission pattern are very yellow. The strips are too flexible and too small to sit in a consistent position and can sit too high/low (as well as too far back, or rotated, etc). I'd bet if you pull the lid bezel off or run the screen outside the lid, you can poke at the strip and watch the dramatic differences in the illumination pattern from small tweaks.

The first time the misalignment was caused by the wires twisting when I connected the LED strip to the inverter. This time I made sure that the end of the strip was properly oriented and taped the wires to the back of the panel to prevent them from twisting during installation, but it also doesn't seem to be completely effective. Do you know of any compatible LED strips that can be more easily placed?

Oh, man. I've never had that problem on a X6x despite handling close to 100 keyboards at this point. I wonder if there's something inadvertantly terrifying about your disassembly technique (or maybe... an intermittent problem on the mobo that only looks like different keyboards?)

The only slightly fragile part of the keyboard is at the top where the ribbon cable disappears behind a metal screwed-on bracket. That's actually a sandwich connection that mechanically clamps the external flex cable to the matching contacts on the internal membrane. It can get pulled slightly out of place, and I think the modifier keys are the traces at the edges that would get misaligned first. Perhaps that's what happening? It's easy to disassemble and put back together. The parts themselves are tough, they can just get misaligned is all.

I'll probably take a look by disassembling one of my other broken keyboards. Of my previous two keyboards, one had the 1, 2, 3, 4 keys stop working and the other both alt keys.

[wileE]

Did you read about what xiphmont calls shimming here: https://people.xiph.org/~xiphmont/think ... anel.shtml
You will have to try what works best for you.

Have had broken keyboards too. The traces in the ribbon cable break in the fold at the plug. They are not made to be plugged and unplugged for 100+ times.

[khseal]

Hello!
I calibrated my display.
before
https://dl.dropboxusercontent.com/u/30741962/11.png
https://dl.dropboxusercontent.com/u/30741962/22.png
after
https://dl.dropboxusercontent.com/u/30741962/111.png
https://dl.dropboxusercontent.com/u/30741962/222.png
The result was not very good. Display brightness dropped and the display itself shows the colors are not very good.

[Qing Dao]

Hello!
I calibrated my display.
before
https://dl.dropboxusercontent.com/u/30741962/11.png
https://dl.dropboxusercontent.com/u/30741962/22.png
after
https://dl.dropboxusercontent.com/u/30741962/111.png
https://dl.dropboxusercontent.com/u/30741962/222.png
The result was not very good. Display brightness dropped and the display itself shows the colors are not very good.

So you aren't happy with the results of your screen calibration? When I changed my T500's display to WSXGA+ and combined it with my T500's original LED parts, the colors were really bad. After calibrating the display, some things got better and some things got worse. Calibration can't perform miracles, so if it was originally way off before, there is only so much that it can do. After calibration, whites looked whiter and less yellow than before, and colors were more balanced. But the problem was that blacks got blacker and I lost some detail in dark images.

What did you use to calibrate your screen? That looks very professional.

[xiphmont]

Hello!
I calibrated my display.
before
https://dl.dropboxusercontent.com/u/30741962/11.png
https://dl.dropboxusercontent.com/u/30741962/22.png
after
https://dl.dropboxusercontent.com/u/30741962/111.png
https://dl.dropboxusercontent.com/u/30741962/222.png
The result was not very good. Display brightness dropped and the display itself shows the colors are not very good.

Keep in mind that you need a spectrophotometer based calibrator to calibrate arbitrary backlights, not just a colorimiter-based calibrator. Or, if you use a colorimeter, it needs to know the exact backlight in use because it's only sampling at three frequency bands, and you end up with an undersampling error that must be corrected for. Your data seems to show the calibrator thought it had done a good job...

Also, gamut of these screens is not high. It's just not going to be able to saturate colors very highly without going to an RGB-style backlight.

(BTW, your and WileE's driver boards went in the mail last thursday, expect them 'eventually'. I need to make up a little pic diagramming the connections, not all of the silkscreen came out perfectly clear)

[khseal]

Keep in mind that you need a spectrophotometer based calibrator to calibrate arbitrary backlights, not just a colorimiter-based calibrator. Or, if you use a colorimeter, it needs to know the exact backlight in use because it's only sampling at three frequency bands, and you end up with an undersampling error that must be corrected for. Your data seems to show the calibrator thought it had done a good job...

My calibrator support LED backlight, or everything is more complicated? Chinese LEDs more special?
Color after calibration become more balanced.
Check out the other IPS display LED backlight everything is good.
Here are the results https://dl.dropboxusercontent.com/u/30741962/333.png

What did you use to calibrate your screen? That looks very professional.

I caliber standard program. Check the calibration program HCFR Сalibration.

'eventually'. I need to make up a little pic diagramming the connections, not all of the silkscreen came out perfectly clear)

Thank you for the driver, when it comes I will report the results.

[xiphmont]

My calibrator support LED backlight, or everything is more complicated? Chinese LEDs more special?

Colorimeters need to know the backlight, and if the software is automatically guessing the backlight correction based on the panel EDID, it's probably going to calibrate it as if it had a CCFL.

Check out the other IPS display LED backlight everything is good.
Here are the results https://dl.dropboxusercontent.com/u/30741962/333.png

That's from a panel that came with LED backlighting stock?
When you calibrated, did you have to tell the software what kind of panel it was, or was it automatic?

[khseal]

Colorimeters need to know the backlight, and if the software is automatically guessing the backlight correction based on the panel EDID, it's probably going to calibrate it as if it had a CCFL.

It is quite possible, but I doubt it. Interesting to see the original display parameters. I looked three orginal Files ICC Profile wileE, which is on the German site. All files has a shift of blue color.

That's from a panel that came with LED backlighting stock?
When you calibrated, did you have to tell the software what kind of panel it was, or was it automatic?

Yes, it's another IPS display with LED backlight stock. The program has a choice only between CRT and LCD monitors.
There is also an interesting option "LCD native".
The new version of the calibrator Spyder 4 has a choice of type backlight ...
http://digital-photography-school.com/w ... tion-5.jpg
But I make the measurement in another program, there is no choice of the type of lighting, there is only the option refresh and no refresh the display.
People write on the forums that the Spyder 3(Description Manufacturer: Support monitors with extended color gamut, LED backlight, glossy finish.) can be applied to the LED backlight, but the calibration data may not be accurate.

[wileE]

Got a very nice parcel from Monty today. With two versions of LED boards to play with.

Both work very well with a LED strip from ccfloffer. It has 48 LEDs like the fancyboost kits, but with groups of three LED.
Colors are better with that strip, noticeably more neutral.

The boards fit perfectly on a X61 inverter board. Max brightness is about the same as with the fancyboost kits before the PWM mod. Around 240 nits in a SXGA+ X61.

Dimming range is completely different.
Measured the steps with my Spyder3: http://imgur.com/9EIj00e 77KB pic warning.

I have to admit that I like the steps with the fancyboost better.
Anything below 70 nits is pretty useless for me, and that step between 13 and 14 is much too big. More steps in the 120 to 210 nits brightness range would be nice.

[xiphmont]

Dimming range is completely different.
Measured the steps with my Spyder3: http://imgur.com/9EIj00e 77KB pic warning.

I have to admit that I like the steps with the fancyboost better.
Anything below 70 nits is pretty useless for me, and that step between 13 and 14 is much too big. More steps in the 120 to 210 nits brightness range would be nice.

I'd intentionally made the step spacing wider at high brightness due to non-linear perception. 10x actual brightness change roughly equals 2x apparent brightness change. Neither of my drivers implement a particularly accurate log scale, but they're much closer than a linear ramp. (And the step from 14 to 15 is actually supposed to be the biggest, but the Richtek controller chip I used in design 2 overshoots the top of the adjustment range by a bit.)

But I think I should have asked ahead of time what brightness range you wanted since it's changeable via resistors RA and RB. I defaulted to the really wide brightness range so that folks who were used to using the machine at night with the min brightness setting would still have a min brightness roughly equivalent to what was offered via a CCFL. Perhaps I overcompensated!

You can also remove R3 to completely disable the non-linear portion of the brightness circuit, which should also have the effect of increasing the min brightness without affecting max brightness. I believe the range should be about 8:1 without it. diagram: https://people.xiph.org/~xiphmont/thinkpad/tld2.png <-- BIG PIC WARNING

[khseal]

The boards fit perfectly on a X61 inverter board. Max brightness is about the same as with the fancyboost kits before the PWM mod. Around 240 nits in a SXGA+ X61.

You have received a high brightness. I got the maximum brightness of 120 nits, but I glued protective glass and other display model HV121P01-100 and driver. Perhaps, I have not as bright LEDs.
I have not received the parcel. Mail my country is not as fast)

[wileE]

You have received a high brightness. I got the maximum brightness of 120 nits, but I glued protective glass and other display model HV121P01-100. Perhaps, I have not as bright LEDs.

You removed the metal bracket from the waveguide? Doing that costs you half of the achievable brightness of your LED strip.
That is something that should always be avoided, especially with any kind of IPS display. It is impossible to put everything together again as it was before.

Removed R3 as suggested. Results: http://i.imgur.com/nAA1c5M.jpg pic warning.

Feels weird with these steps, but it is what I asked for. I think this will work better for me.

[xiphmont]

Removed R3 as suggested. Results: http://i.imgur.com/nAA1c5M.jpg pic warning.
Feels weird with these steps, but it is what I asked for. I think this will work better for me.

The PWM steps are not themselves perfectly evenly spaced, which is a bit weird (known property). The min brightness in linear mode is a bit lower than I thought it would be, but that otherwise seems to be the spacing I expect.

Perhaps I should compromise between the two (linear and log) spacings....

[wileE]

Perhaps I should compromise between the two (linear and log) spacings....

Yes. Having more than 10 steps below 100 nits does not make much sense.

Added ver.1 measurements: http://i.imgur.com/j5xTxfa.jpg pic warning

[khseal]

You removed the metal bracket from the waveguide? Doing that costs you half of the achievable brightness of your LED strip.
That is something that should always be avoided, especially with any kind of IPS display. It is impossible to put everything together again as it was before.

Yes, i removed metal bracket and install a back. I've done this before with other displays, but do not notice the deterioration of brightness. I even remade one display with LED on CCFL =)
My display HV121P01-100 was stock low brightness(Here write about it http://forum.thinkpads.com/viewtopic.php?t=109746 ). I had a display HV121P01-101, it seems it was not so dark. I think it is necessary to press densely the strip LED.

[khseal]

del

[wileE]

Topic brightness steps again.

Screen is now in my X60s, which has a Core Duo running a 32 Bit Ubuntu Gnome 14.04. Testing yesterday was done done with a 64 Bit board.

There is, and has been for some years, a stupid bug with the brightness steps in some distros where the desktop is Gnome 3 or based on that. My 32 Bit Ubuntu Gnome is one of those.

It has only 4 steps default. When I add

Code: Select all

echo "N" > /sys/module/video/parameters/brightness_switch_enabled

to my rc.local, I get 7 steps.

This works fine with the ver.2 board with R3 removed. But the brightness steps with the ver.1 board are unuseable.
http://i.imgur.com/Zg6GgEA.jpg

Is there a way to fix ver.1 like the R3 removal? Otherwise no bugs found.

[khseal]

Received a parcel from Xiphmont. Thank you very much. Everything works fine. There is an increase in brightness.
Xiphmont sorry for criticism, but your soldering is not very good =) It seemed to me there are in some places the cold soldering. Especially on large sites. You are using lead-free solder?

[xiphmont]

Perhaps I should compromise between the two (linear and log) spacings....

Yes. Having more than 10 steps below 100 nits does not make much sense.

Personally, I thought it kind of silly that the top half of the steps were so close together they were practically useless. Perhaps reasonable people can simply have different desires here... I can see how linear spacing makes calibration to 120 nits easier...

Regarding TLD1, the brightness setup is not quite as flexible as the TLD2 design. I can make the scale more liner, but there's no simple change to board 1 like removing R3.

There is, and has been for some years, a stupid bug with the brightness steps in some distros where the desktop is Gnome 3 or based on that. My 32 Bit Ubuntu Gnome is one of those.

I don't think it's Gnome 3, because I use Gnome 3 on all my machines (running Fedora) and I have 15 steps of brightness, with roughly linearly spaced PWM duty cycles. Perhaps Ubuntu came to the same conclusion I did--- all those upper steps being so close together was silly, so they spaced them out more nonlinearly. Unfortunately if so, that would interact badly with my driver hardware, which also implements a nonlinear scale. I agree the scale you measured is unusable.

[xiphmont]

Received a parcel from Xiphmont. Thank you very much. Everything works fine. There is an increase in brightness.
Xiphmont sorry for criticism, but your soldering is not very good =) It seemed to me there are in some places the cold soldering. Especially on large sites. You are using lead-free solder?

K100LD lead-free eutectic solder yes. I think I know what you're talking about and those shouldn't actually be cold-solder joints. Since I was soldering by hand, I over-flowed the large areas (to get sufficient heat transfer), then used a vacuum tool to remove the excess bead. That leaves a surface that looks like a bit like a cold solder joint, especially after cleaning off the flux with Novec. I considered adding a warning and decided against it figuring no one would be looking that closely

[khseal]

K100LD lead-free eutectic solder yes.

Lead-free solder requires higher temperature soldering. Many repairers do not like a lead-free solder. Often even make reballing BGA chips on lead solder to the soldering not to overheat the chip.

[wileE]

Personally, I thought it kind of silly that the top half of the steps were so close together they were practically useless.

Yes, steps 11 to 15 are useless. It is the rest of the steps I like better. Having more steps in the range from 100 to 200 nits was important to me.

I am thinking about doing the LED mod on my T500. 107 nits is not enough anymore on sunny days. Can I use the ver.1 board for strips with 64 LEDs?

EDIT
Tried Fedora-Live-Workstation-i686-21-5 on X61 T7300 = 16 steps. On X60s L2400 = 7 steps.

[xiphmont]

Lead-free solder requires higher temperature soldering. Many repairers do not like a lead-free solder.

Sure, there's no way that any lead-free solder compares to the former glory of leaded alloys. K100LD is the best I've found, but it's no 60-40.

OTOH, I have high lead levels (probably from a lifetime of working with lead, not just solder), and one of my boys also ended up with clinical lead poisoning as a toddler. Also, I have pets and my workshop is at home... I just won't use leaded solder (or any kind of lead) for projects anymore.

[xiphmont]

Yes, steps 11 to 15 are useless. It is the rest of the steps I like better. Having more steps in the range from 100 to 200 nits was important to me.

Ah, and you never use the machine in near darkness I'd guess? I was going for a minimum brightness similar to the CCFL, but I think now that's just too wide. It also complicated tuning the design of board 1. Starting at 15 nits instead of 6 would make for a much bigger/easier tuning envelope, could probably get some more efficiency out of it too.

I am thinking about doing the LED mod on my T500. 107 nits is not enough anymore on sunny days. Can I use the ver.1 board for strips with 64 LEDs?

I think so, but I've not tested the PWM signal on the T500 and I don't know the inverter connection points. Worst case is it won't work right (brightness levels could be off).
Design 1 is intended and tested for up to 500mA output (just under 5W); you can either use it as is (335mA), or replace the current sense resistors to hike the output to 500. I can give you the values if you like.

Oh wait, 64 LEDs? That's not groups of 3. Is it groups of 2, 4, 8...? groups of 2 would still work, groups of 4 would be OK on machines that use 4/8 cell batteries but not 3/6/9.

EDIT
Tried Fedora-Live-Workstation-i686-21-5 on X61 T7300 = 16 steps. On X60s L2400 = 7 steps.

Ah! So is the difference the machine, not the distro? Dang, I have a X60s for testing, but it's at home not with me here. I never noticed it having a different number of brightness steps.
The number of steps on the X60s is still 7 even when not in the OS, eg, when sitting in the BIOS?

I'll spend some time re-jiggering the brightness range resistor values on boards 1 and 2 to take some of this new information into account. It shouldn't require any design change.

[wileE]

Ah, and you never use the machine in near darkness I'd guess?

Rarely. Even in complete darkness I would not consider anything below 60nits to be useable.

Oh wait, 64 LEDs? That's not groups of 3.

64 is just the number with the fancyboost kits. I do not know what I would get when ordering a strip with the same lenght somewhere else.

Ah! So is the difference the machine, not the distro?

Yes. My mistake. It is the hardware. The X60s only has 7 steps.
The difference in the distros is that Fedora has the 7 steps without the modification to the rc.local.

[xiphmont]

64 is just the number with the fancyboost kits. I do not know what I would get when ordering a strip with the same lenght somewhere else.

Oh! OK. Yes, both boards will drive longer strips just fine, so long as they're groups of 3, and can be configured to put out more current for larger panels if desired. Board 1 is good to about 5W (limiting factor is the inductor), board 2's components can handle 10W easily, but I've not tested it above 5.