X6X-series LED backlight conversion kit howtos

[khseal]

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.

I also have a workshop at home. I am engaged in electronics for over 10 years and using lead solder. I have a cat who once fell asleep on the soldering iron. Cat has more than 15 years.
I'm more worried about the harmfulness of flux.

How are your experimenting with LEDs?

[xiphmont]

I also have a workshop at home. I am engaged in electronics for over 10 years and using lead solder. I have a cat who once fell asleep on the soldering iron. Cat has more than 15 years.

It's harmless if you don't accidentally ingest some, and it's not even all that likely solder was the culprit. But I have indeed ingested some amount of lead (and so has my little one), so I've decided not to take any more chances with it. It only takes some errant dust from a desoldering pump, and I do have the bad habit of sometimes biting my fingernails...

How are your experimenting with LEDs?

I will hopefully finish binning this afternoon! I did not find any LEDs that were a good match to use with filters, but I think this batch that I'm messing about with right now will produce correct colors without filtering. I just wish I had a source for more (preferrably pre-binned). I don't... so once these are all in strips, they're gone. Unfortunately, I think that's the best I can do.

This is only an issue for the AFFS screens as far as I know. Normal 6500K LEDs seem to be perfectly suitable for TN screens.

[wileE]

I will hopefully finish binning this afternoon!

You are certainly determined to to the best possible LED mod. Sorting through a pile of LEDs does not look like much fun.

Did you produce your own PCB for the LEDs? Same number of LEDs as the chinese strips, or more with tighter spacing?

[xiphmont]

I will hopefully finish binning this afternoon!

You are certainly determined to to the best possible LED mod. Sorting through a pile of LEDs does not look like much fun.

I did not quite finish, but I'm close. Only 275 left to go. It does not strike me as any less fun than, say, piecing a quilt, and a glass of wine definitely makes it go faster

Did you produce your own PCB for the LEDs? Same number of LEDs as the chinese strips, or more with tighter spacing?

I'm reusing PCBs from a number of the Chinese strips I have. A few of the suppliers used really quite nice LEDs (the ones that came with the Simpleboosts were quite good), but most of them were a mediocre grade and I don't mind wasting the LEDs to reclaim the strips. 48-51 LEDs for a 12.1" screen is about optimal really. I wouldn't want to pump in more than about 3W of lighting power anyway.

That said, testing found several midpower 1406 sized LEDs that seem to work pretty well for TN screens (like the Osram Duris series) and are rated for 50-65mA apiece at 3V. Those would make for an incredibly bright screen, not sure if the power dissipation would melt anything! Unfortunately, I couldn't find any colder than 7k, so they're not suitable for the AFFS screens. You could also run them at low power to get the higher efficiency and heat tolerance.

Man, every year LEDs just keep getting better and the improvements aren't slowing down!

[wileE]

Can you recommend a specific type of LED for a strip in the X6x?
I have a side emmitting strip with a fancyboost driver. It looks like it was made of leftovers. I think I could try putting new LEDs on that, without having to worry too much about causing any damage with my soldering.

That fancyboost I sent you, any idea yet why it misbehaves?

[wileE]

Info for whoever might want to to the LED mod on a T61 or T500.

Here a picture of a T61 inverterboard with a fancyboost: http://i.imgur.com/0JgAyN1.jpg Large pic warning. 550KB

You can see where the T61 board was modified. The T500 board is larger, but the layout is nearly identical. So you can use the T61 board as a guide.

[xiphmont]

Can you recommend a specific type of LED for a strip in the X6x?
I have a side emmitting strip with a fancyboost driver. It looks like it was made of leftovers. I think I could try putting new LEDs on that, without having to worry too much about causing any damage with my soldering.

I have sample strips I made of just about every retail 0805 through 1406 cold white LED I could get my hands on in the US. I'll go measure a few of the ones I remember as promising.
Which screen should I check specifically?

That fancyboost I sent you, any idea yet why it misbehaves?

I spent more time with it today. I cannot get it to do anything the least bit flickery. One thing I noticed though... my dimming range is only about 4:1, just like the new ones. You had mentioned when it was misbehaving it was more like 8:1?

I have a stupid question, please don't be offended (it's the kind of thing I'd do): Did you perhaps reverse the DIM and ENA connections? That would also give the behavior you describe. The ENA input is fast enough to work like a PWM input, but it would add a startup delay that would appear to extend the brightness range...

[wileE]

Did you perhaps reverse the DIM and ENA connections?

Yes. I wrote it down in the wrong order on a picture I made in 2012. And obviously made that mistake every time I used that particular picture.

Thank you. And I sincerely apologize for wasting your time with my stupid mistake.

Dimming range is different now. Lowest brightness was 5 nits with the wrong connections, now its 33 nits. Max. brightness is 5 nits lower.

Which screen should I check specifically?

It is just the HV121P01-101 that I consider worth the effort.

[xiphmont]

Did you perhaps reverse the DIM and ENA connections?

Yes. I wrote it down in the wrong order on a picture I made in 2012. And obviously made that mistake every time I used that particular picture.

Thank you. And I sincerely apologize for wasting your time with my stupid mistake.

Not at all, I'm happy the mystery is solved!

Which screen should I check specifically?

It is just the HV121P01-101 that I consider worth the effort.

Hm, OK. Everything except the 'accidental' LEDs I found from the Chinese seller who won't tell me where they came from (and can't sell me more) are either too low temp or too green for the AFFS screens, but I'll go back and see what was closest.
The LEDs from the Simpleboost strips (with the dark blue/black PCB strips) are good for AFFS actually. I wish I could determine the source of those. They're not as good as the weird chip LEDs I've been binning, but they're a close second.

[wileE]

The LEDs from the Simpleboost strips (with the dark blue/black PCB strips) are good for AFFS actually.

Then I hope I get those with the shipment on its way from creatall.

[xiphmont]

The LEDs from the Simpleboost strips (with the dark blue/black PCB strips) are good for AFFS actually.

Then I hope I get those with the shipment on its way from creatall.

If you do get Simpleboost strips, they'll work well with the Fancyboost so long as you match up +/-. BTW, do you want your fancyboost back?

[wileE]

According to the pictures it should be Unboosty. I will see when they get here. Maybe it will be something else alltogether.

BTW, do you want your fancyboost back?

Only if you do not have any use for it. If you do send it back, choose the cheapest snail-mail available.

[wileE]

Kits from ebay seller creatall88 got here today. With unboosty boards. Strip for T500 has 66 LEDs top-emitting.

I put your version 1 board on my T500 inverterboard. Triple checked the dim connection was right.

No problems yet on battery (will check with empty batterys tomorrow). Brightness on battery is the same as on AC. 256 nits max.

But I have a big problem with the brightness steps. http://imgur.com/ycYNTfn That is only 3 steps I am likely to ever use. And step 13 only very rarely.

Anything I could do to get a few more steps in the useable range?

Positioning the strip in the bracket is also proving to be difficult. Pressed against the waveguide, as I did with all the X6x screens, results in distorted mirror images of the LEDs in the lower 1.5 cm of the screen.
With the strip on the bottom of the bracket it is much better. Tried double sided ahesive tape and pressing it down with the help of platic strips. But it does not stay there when the LEDs get warm. I am reluctant to use glue.

_____

No problems with your board when the batteries are nearly empty. Tested with a 6 and 9 cell.

But I cannot recommend the LED strips from creatall88. In the LG WSXGA+ the resulting yellow-green tinge is ugly.

[xiphmont]

No problems yet on battery (will check with empty batterys tomorrow). Brightness on battery is the same as on AC. 256 nits max.

Good to know, especially your empty-battery update.

But I have a big problem with the brightness steps. http://imgur.com/ycYNTfn That is only 3 steps I am likely to ever use. And step 13 only very rarely.

Anything I could do to get a few more steps in the useable range?

Thanks for the scale, that will help. Yes, there's something to do, you'll need to replace a resistor with a different value. Let me try it here today on one of my scratch boards to get the right ballpark. You said you're unlikely to ever want less than 60 nits under any circumstances, right?

Positioning the strip in the bracket is also proving to be difficult. Pressed against the waveguide, as I did with all the X6x screens, results in distorted mirror images of the LEDs in the lower 1.5 cm of the screen.
With the strip on the bottom of the bracket it is much better. Tried double sided ahesive tape and pressing it down with the help of platic strips. But it does not stay there when the LEDs get warm. I am reluctant to use glue.

Hm, this screen must use a very weak diffuser to increase brightness, or has no bottom diffuser. The 'best' way to fix that problem is to slip a stronger diffuser layer between the waveguide and prism layer, but it requires popping the screen frame open. Good diffuser sheets should be ~ readily available but I don't know the internal dimensions to cut for WSXGA. I could tell you for the X60/61 screens.

You could also get some inexpensive sheets of clear PET plastic, slice thin strips on a paper cutter, and use them as spacers in front of the LED strip. Farther away will probably reduce brightness more than using a stronger diffuser.

(It is hard for me to tell what a normal diffuser setup in each screen is because I've bought lots of reclaimed screens and it appears each and every unit has been tampered with-- none have what appear to be stock diffuser layers in them. On the other hand, I have played quite a bit with the screen internals.)

But I cannot recommend the LED strips from creatall88. In the LG WSXGA+ the resulting yellow-green tinge is ugly.

Yes, that was my experience too, along with brightness variations between the LEDs on the same strip.

I have built my strips and commenced more careful measurement using my spectrometer rather than simply trying to use my DSLR and making educated guesses
The binning process was a success-- my strips have nice even color and brightness. Unfortunately, the final specs are not as perfect as hoped. The color balance is good, but a bit warm (~6000k instead of 6500k) and the brightness is slightly below that one perfect strip I got as a sample using the same manufacturer's LEDs. I am being very picky of course-- they're fully twice as bright at the same power as CCFL. But I wanted so badly to recreate that one perfect sample that came out of nowhere and I've never found again. It's the best backlight strip I've ever measured.

I'm also retesting strips I got from other suppliers. The Simpleboost LEDs are better than I remembered-- not quite as bright as my strips, but nearly perfect whitepoint. I have an iCCFL strip that is slightly greenish, but hits nearly 300% the original CCFL brightness without seeming too far off on color. I'll have some actual numbers in a bit.

[wileE]

Yes, there's something to do, you'll need to replace a resistor with a different value. Let me try it here today on one of my scratch boards to get the right ballpark. You said you're unlikely to ever want less than 60 nits under any circumstances, right?

Just one resistor? That would be nice. And no, I do not care about the steps below 60 nits.
I was a little surprised that the brightness steps are so much different in the T500. I did not much like how the steps worked in the X61, but it was still much better.

The 'best' way to fix that problem is to slip a stronger diffuser layer between the waveguide and prism layer, but it requires popping the screen frame open.

Yes. I have done that with enough X6x screens, and I absolutely hate opening screens that far.
Not going to try anything like that for this LED strip. Maybe I will try soldering different LEDs on that strip too.

The binning process was a success-- my strips have nice even color and brightness.

Well deserved success. Measuring each single LED is a crazy idea. Handcrafted workmanship applied to modern electronics.

[wileE]

My screen in the T500 is back to CCFL. No need to figure out the resistor for different brightness steps anymore.
The LEDs used by creatall on the strips were too ugly in that screen, even worse when I tried them in a BoeHydis AFFS.

The inside of the bracket for the CCFL was interesting. The white plastic bracket made it very easy to attach it to the waveguide in the original position.
pic: http://i.imgur.com/wKBqK4m.jpg screen type: LP154W02-TL-10

[xiphmont]

My screen in the T500 is back to CCFL. No need to figure out the resistor for different brightness steps anymore.

Sorry, I should have given an update (and I need to figure out a better scale selection anyway. I got very bogged down in writing some testing software because I too just can't put up with the poor color LEDs.

The LEDs used by creatall on the strips were too ugly in that screen, even worse when I tried them in a BoeHydis AFFS.

Unfortunately, I agree completely.

And thanks for the pic--- not all of the brackets clamp tightly around the waveguide. Probably not even most. The AFFS screens make some things decidedly trickier.

[xiphmont]

Here's what I've been working on-- designing add-on optics for my spectrometer to make it appropriate for screen calibration, and all the software to collect raw data and plot it. I'm figuring out how to put the pictures together on a webpage that's comprehensible (easy to read and compare), but here's a teaser....

To compare against the later LED conversion measurements, a gamut plot of a factory new HV121P01-100 with original CCFL: https://people.xiph.org/~xiphmont/think ... 0-CCFL.png

A couple notes about that screen: I believe it was factory new, but it's always hard to tell for sure. All the warranty seals and shipping wrap were in place, and it didn't look tampered with. It tested a bit brighter (156nits vs 128ish nits) than any other 'new' HV121P01-100 I've tested in the past, but the gamut looked spot-on. So it may just have been a brightness freak. It was tested with the stock front 'glass' cover in place and allowed to warm up for 10 minutes.

These screens are obviously not saturation monsters That test makes it look kinda bad, but it's a typical measurement for a good quality low-power IPS panel of the era. The first few iPads were pretty much identical.

[xiphmont]

Yes, there's something to do, you'll need to replace a resistor with a different value. Let me try it here today on one of my scratch boards to get the right ballpark. You said you're unlikely to ever want less than 60 nits under any circumstances, right?

Just one resistor? That would be nice. And no, I do not care about the steps below 60 nits.
I was a little surprised that the brightness steps are so much different in the T500. I did not much like how the steps worked in the X61, but it was still much better.

I *finally* got to working out some resistor values, though I only have X61 machines to work and test with. This will produce exactly what you want for an X61, probably still not quite what you want on the T500... but I have more of a recipe now.

On the X61 at least, replace RA with 120k, leave RB blank, and R3 with 120k. That will give a roughly 12:1 total brightness range, but more importantly it imitates the exponential of the original CCFL so that the high-brightness steps are much closer together. Most users I think will also want an RB trim resistor to lower the minimum brightness (1.5M gives approximately exact behavior to the original CCFL) but you wanted fewer steps under 60nits. Leaving RB off will do it.

The recipe, if you want to tweak it, is that RA||RB + R3 == 240k gives ~ a 12:1 range. Decreasing the total increases the range, increasing the total decreases the range. It is very sensitive to adjustment as range increases, which is to say as minimum brightness decreases. You can't get past 40:1, but I don't think anyone would really want to. BTW, I've taken more careful measurements of new, sealed screens and the original CCFL is actually closer to 30:1 total range than the 25:1 I originally thought...

When RA||RB == R3, the exponential (brightness steps) will be almost exactly like the original CCFL. As you increase RA||RB relative to R3, the exponential flattens out, and decreasing the RA||RB to R3 ratio makes the exponential sharper (high brightness steps get farther and farther apart relative to low brightness).

I think the correct thing for me to do with my boards is to use the new resistor values to imitate the original behavior closely (and maybe decreasing the range slightly to bring min brightness up just a bit). Principle of least surprise, and more likely to work better with other models. I'll figure out new values for the v2 board as well.

[wileE]

Thank you. I would like to try that.
Do you have a picture of the version 1 board with the labels? You had a nice one for the version2 board. The labels on the board itself are mostly impossible to read.

A couple notes about that screen: I believe it was factory new, but it's always hard to tell for sure. All the warranty seals and shipping wrap were in place, and it didn't look tampered with. It tested a bit brighter (156nits vs 128ish nits) than any other 'new' HV121P01-100 I've tested in the past, but the gamut looked spot-on.

That is very unusual. Never had one above 130 nits. Contrast value is therefore also unusually high. Most are around 500:1.

Looking forward to seeing what you did with those "pounds of glass".

[xiphmont]

Thank you. I would like to try that.
Do you have a picture of the version 1 board with the labels? You had a nice one for the version2 board. The labels on the board itself are mostly impossible to read.

Oh! yes: https://people.xiph.org/~xiphmont/thinkpad/tld1.png <-- large pic warning

Advanced Circuits makes very reliable boards, but their silkscreening is definitely behind the tech curve

A couple notes about that screen: I believe it was factory new, but it's always hard to tell for sure. All the warranty seals and shipping wrap were in place, and it didn't look tampered with. It tested a bit brighter (156nits vs 128ish nits) than any other 'new' HV121P01-100 I've tested in the past, but the gamut looked spot-on.

That is very unusual. Never had one above 130 nits. Contrast value is therefore also unusually high. Most are around 500:1.

I have six HV121P01-100 here right now-- all measure better than 800:1 contrast. I think that comes down to equipment difference; I'm using a very narrow field for my sensor now (<10 degrees), eliminating off-angle glow and bleed, so it gives a near-optimal 'dead on center' contrast score.

The IPS screens only measure 150:1 for averaged [cosine corrected] contrast over a full 180 degrees wide-angle view. If you use a sensor with a field of view in-between, you get a figure in-between.

Looking forward to seeing what you did with those "pounds of glass".

I made a large-aperture narrow field-of-view adapter for the spectrometer

https://people.xiph.org/~xiphmont/pics/glass.jpg <- large pic warning

[wileE]

I made a large-aperture narrow field-of-view adapter for the spectrometer

Yes, I saw that on your blog. Much more interesting is what you might have done with that thing.

[xiphmont]

I made a large-aperture narrow field-of-view adapter for the spectrometer

Yes, I saw that on your blog. Much more interesting is what you might have done with that thing.

Here's a preview: https://people.xiph.org/~xiphmont/think ... amut.shtml <-lots of big pics + javascript warning

Not live quite yet, will be soon. I need to comprehensively update the pages in general.

[wileE]

Incredible work. Very very nice.
I am somewhat surprised by how different the strips from creatall turn out. And I am going to forget all about experimenting with the 6500k LEDs commonly available. Those readings look awful.

Your maximum brightness measurements are mostly higher than I would have expected. Certainly higher than my Spyder ever showed on the same screen type. Equipment differences due to your narrow angle of measurement again?

iCCFL 'unboostyB' 9.6V LED backlight

That would work with one of your boards?
Is it this? http://www.iccfl.com/product_info.php?c ... bd7i7oevj0

That reclaimed HV121P01-101 screen is strange. I cannot understand why they would swap the diffusers.

[khseal]

Xiphmont maniac!

Your maximum brightness measurements are mostly higher than I would have expected. Certainly higher than my Spyder ever showed on the same screen type. Equipment differences due to your narrow angle of measurement again?

What type of screen you have in mind?
The measurements are very similar to what I did Spyder.
Only the brightness values in some of your measurements are very high.
431 nits for HV121P01-100(Sample C) it is crazy result.

[xiphmont]

Incredible work. Very very nice.
Certainly higher than my Spyder ever showed on the same screen type. Equipment differences due to your narrow angle of measurement again?

Maybe? I have to admit I'm not sure. The narrow angle would affect it at least a bit because I'm measuring only the direction of the strongest emission...

How far off are the values from what you get? Boe/Hydis and Lenovo always claimed the HV121P01-100 was 190 nit, and I measured just over 160.... (they use a 2-degree FOV, even narrower than mine)

iCCFL 'unboostyB' 9.6V LED backlight

That would work with one of your boards?
Is it this? http://www.iccfl.com/product_info.php?c ... bd7i7oevj0

Yes, any of the buck-down boards should be able to drive any of the 9.6V-nominal strips. The measurement tests were all run using one of my boards (except for the boost strips, I used a fancyboost for those).

[edit: not sure about the iCCFL listing--- the board in that ad is a cheapybuck and I'd be really suprised if any of their kits still came with the same LEDs I got in an order last year.]

That reclaimed HV121P01-101 screen is strange. I cannot understand why they would swap the diffusers.

I don't have any idea why, I'm just sure it's not stock. The manufacturing retainer clips are missing on those screens, and the diffusers aren't keyed. Previous versions of those panels I've disassembled had a single bottom diffuser with full keying and retainers.

I'm looking through my stock and notes more carefully, the only other thing I can think is I got my labels mixed up and actually measured a screen *I* had previously been messing with... I don't think so, but once I'm back from my current trip I can check it more carefully for my own handiwork (that would be embarrassing, but oh well In any case, color was the important parameter, and the prisms/diffusers shouldn't affect that.

(I would expect the HV121P01-101 to be dimmer than the HV121P01-100 or the XGA screens because it uses a different, wider-spread emission pattern. But I wouldn't expect it to be *that much* dimmer.)

[xiphmont]

Xiphmont maniac!

Your maximum brightness measurements are mostly higher than I would have expected. Certainly higher than my Spyder ever showed on the same screen type. Equipment differences due to your narrow angle of measurement again?

What type of screen you have in mind?
The measurements are very similar to what I did Spyder.
Only the brightness values in some of your measurements are very high.
431 nits for HV121P01-100(Sample C) it is crazy result.

It's easy to get 431 nits when you're willing to be that yellow. Nits is practically a measurement _of_ the yellow component (well, yellow-green). It's a weighted measurement.

[bwaldow]

The good news is that I just sent in two prototypes for a short fabrication run. We shall see soon if they pan out. They worked well when assembled point to point

I would like to hear the current state of this.

I have an X61T with the CCFL panel from an X60T (no bubbling). Having a bright backlight would certainly extend it's usefulness to me.

Am I better selecting one of the existing solutions, or do you have devices you can provide?

Thanks in advance,
Bret

[xiphmont]

Am I better selecting one of the existing solutions, or do you have devices you can provide?

I have several of the original prototype run I can send to people, and I'd rather give away the prototypes for now so that there's more testing before I actually sell anything. The prototypes have been reworked a few times in testing (so the soldering isn't as beautiful as a factory-made board) but they're all in perfect working condition. SO, if you're OK with being a guinea-pig, I can send you a board for free immediately. PM me.

[xiphmont]

Last year I bought a couple of 'new-other' HV121P01-101 screens from ebay that were sold without glass. These are the X61T screens that normally have the glass fully bonded on with the nasty adhesive that eventually bubbles.

Upon arrival it was clear they were reclaims that had been de-glassed as there was still some adhesive residue inside the frame. They also had some off-angle color shift which struck me as odd. I hadn't remembered these screens having any shift at all.

I noticed at the time that the front polarizers had square corners; I thought BoeHydis screens always have front polarizers with beveled corners. I speculated at the time that the process of de-glassing the screens damaged the polarizers, and the Chinese places doing the work slapped a cheaper polarizer on front to replace the original damaged one. (I seem to recall the most expensive pieces of an LCD screen are the polarizers, not the TFT matrix itself) But I couldn't tell for sure, since the edge of the polarizer isn't visible on a fully assembled HV121P01-101 with the glass in place.

Last night I pulled the glass off a bubbled -101 to finally check--- same polarizer as used on the -100, and it has a beveled edge. That seems to confirm the theory

(Sadly, after successfully getting the glass off I found I'd cracked a corner off the matrix during initial disassembly. D'oh! I did get spectrometer testing data out of it before trying the debonding as it otherwise worked fine. And yes, the previous de-glassed -101 I tested was in fact weird. This one gave numbers much more like the -100)