Connecting a 4 channels LVDS LCD (without EDID) to a 3 channels motherboard

[PaddleStroke]

Hey guys!
I started a video project of upgrading a laptop screen which includes heavy modding. Though I stumble on issues for connecting the LVDS interfaces :

My plan is to fit a 20" 1600x1200 4:3 LCD (LM201U05-SLA1) from a Dell Ultrasharp monitor (2007fPB) inside my 17" 1600x900 Asus (N71JQ) laptop (LP173WD1). From the 17" datasheet I found what I was expecting : 3 LVDS channel + 3.3V power + EDID + backlight power.

For the 20" it's troublesome :
- First it's a 4 channel LVDS : Can I just leave the 4th channel floating or grounded and only connect the 3 other channels?

- Also the 20" does not have EDID storing LCD timing/resolution settings. I am unsure how to proceed. I was considering keeping the 17" controller board with only the wires for the EDID (while the other LVDS wires would go to the 20" controller board), and overwrite the EDID set to the new LCD timing controls. Though the issue is that I have not clue how to write the EDID (both how to flash it and how to write it's code).
Anyone one with some clues is most welcome !

Thanks!

[c64]

- First it's a 4 channel LVDS : Can I just leave the 4th channel floating or grounded and only connect the 3 other channels?

Usually 4th channel is the MSB for all colors. What is in the datasheet for the screen?

[PaddleStroke]

It seems to me from the LM201U05-SLA1 datasheet (which I can't attach here but is easily found online) that you are correct. The 1st and 2nd (R0 R1 G0 G1 B0 B1) bits are the less significant indeed. Would that mean that the channel 0 is to be left unconnected instead of the channel 3?

Though I am unsure if the 0 and 1 bits are the ones carried by the channel 0.

[PaddleStroke]

I have been having a closer look at the datasheets. In the 17" datasheet I found that

Code: [Select]

Odd_Rin0- - LVDS differential data input (R0-R5,G0)
Odd_Rin1- - LVDS differential data input (G1-G5,B0-B1)
Odd_Rin2- - LVDS differential data input (B2-B5,HS,VS,DE)

So it seems that channel 0 is carrying R0 to R5 + G0 and so on.

Looking at that it feels really really difficult to match the LVDS interfaces...

[c64]

Looks like your only option is to de-serialize lvds -> rgb, convert 6 bit color into 8 bit, when serialize it into lvds

[PaddleStroke]

Yes you are right. Sounds difficult. Well not for someone who is already familiar with such I guess.
I think what is needed is a board like this : https://fr.aliexpress.com/item/32809909192.html
But as it's custom to both panels there's not hope of finding such a board. I guess it needs to be custom made for both PCB and software.

[c64]

Come on, this is electronics forum, you need to do it yourself 
All you need just pair of chips like SN75LVDS82 (lvds receiver) + SN75LVDS83 (lvds transmitter) - easy to use, I used them on 2 layer board to problem.

[c64]

You have 8 lvds pairs so you need 2 LVDS82 and 2 LVDS83

Or better use DS90C387A/DS90CF388A

The DS90C387A/DS90CF388A transmitter/receiver pair is designed to support dual pixel data transmission between Host and Flat Panel Display up to QXGA resolutions.

[PaddleStroke]

Thanks for your feedback!

I have actually kept digging the datasheets this afternoon and found out that as you say it just needs a LVDS receiver and transmitter! I should have checked your replies before !

In the 20" panel datasheet, they say "Interface chip must be used LVDS, part No. DS90CF383MTD(Transmitter) made by National Semiconductor.
Or used the compatible interface chips(TI:SN75LVDS83)."

It's not exactly the same P/N as you said, but it's probably very similar chips I guess? Though you also advised the SN75LVDS83 too which is said to be compatible in the datasheet.

So if I'm guessing correctly, it just needs to connect as follow :
Host LVDS connector -> receiver -> RGB18 bits -> transmitter -> panel LVDS connector.

It only needs to connect correctly the RGB output from the receiver to the input of transmitter while leaving LSB to GND, right? Following the mappings given in the panels datasheets ? Now the 20" panel datasheet does not give the mapping of the bits inside the channels. But looking at DS90CF383MTD datasheet at figure 2, I think that's the mapping to follow isn't it?

This sounds indeed pretty easy actually. I was kind of afraid that it needs to use a MCU and write firmware.

Now there is still the missing EDID for the 20" panel. As it needs to make a PCB anyway, I can put a EDID on the same board. Though I need to find a reference schematic for EDID.

[coromonadalix]

EDID is just an eeprom, on the data lines, storing all the resolutions infos needed for the screen to be driven,  how's to are on google and examples on how to add it, connect it or reprogram it ...

[NiHaoMike]

Might it be simpler to just use the original driver board to connect it to DVI/HDMI/Displayport?

[hlavac]

Maybe try to duplicate the least significant bit into the extra lsbs or you may lose some dynamic range...

[c64]

In the 20" panel datasheet, they say "Interface chip must be used LVDS, part No. DS90CF383MTD(Transmitter) made by National Semiconductor.
Or used the compatible interface chips(TI:SN75LVDS83)."

It's not exactly the same P/N as you said, but it's probably very similar chips I guess? Though you also advised the SN75LVDS83 too which is said to be compatible in the datasheet.

Should work, the interface is standard.

So if I'm guessing correctly, it just needs to connect as follow :
Host LVDS connector -> receiver -> RGB18 bits -> transmitter -> panel LVDS connector.

It only needs to connect correctly the RGB output from the receiver to the input of transmitter while leaving LSB to GND, right? Following the mappings given in the panels datasheets ? Now the 20" panel datasheet does not give the mapping of the bits inside the channels. But looking at DS90CF383MTD datasheet at figure 2, I think that's the mapping to follow isn't it?

There is no "standard mapping", lvds83 datasheet describes some most commonly used ones. If the panel datasheet doesn't provide mapping, you can only guess which one they used. Or if you have fpga development board, you can try to drive your monitor, to find out the mapping.

[Berni]

Often the 4 lane LVDS mapping is made in such a way that the first 3 lanes carry the upper 6bits of each color + sync signals. Then the 4th lane just carries the extra 2 bits of each color. In those cases the display will run just fine with the 4th pair not used, it will just only have 18bit colors.

Last time i had to identify LVDS lines (It was a scope motherboard with a LVDS display and i needed the pinout) i simply poked the LVDS lines with a scope and looked at what they are doing. Various colors can be put on the screen to find where the color bits are. Tho you do need a fairly fast scope since the bitrates are very high.

The EDID is indeed just a I2C EEPROM on the lines, you can find the data structure that needs to be written into them on the internet, since you have a different resolution on the new display means you have to modify the data to report the correct thing. However there is no grantee that the motherboards BIOS actually reads the EDID. On boot the BIOS will set up the graphics card to output the correct LVDS format and in motherboards that are only ever supposed to have one display connected this is sometimes hardcoded in. Or perhaps it might use the EDID to determine from say a list of 4 different displays that this laptop model comes with and choose the right hardcoded initialization routine for it. So check that you can actually convince your motherboard into outputting the correct resolution first.

If not then you can still tap into something like a HDMI port where EDID is always used so you can make it give you any resolution you want. But that involves HDMI to LVDS conversion too.

[PaddleStroke]

Thank you all for your replies and insights!

However there is no grantee that the motherboards BIOS actually reads the EDID.

You raise a very interesting question concerning the bios. I kind of fear going all the way then realizing that the bios only supports 1600*900. As 1600*1200 is not the same aspect ratio there is no reason the manufacturer would include it. Though it sounds like the bios could support any resolution without trouble and it would be rather the bios excluding voluntarily other resolutions.

To test that I could try to write a EDID for another resolution and load this EDID to the current panel. Then see if the display shows garbage or a somewhat truncated image. As I can use an external monitor to setup back the original EDID I guess.

Often the 4 lane LVDS mapping is made in such a way that the first 3 lanes carry the upper 6bits of each color + sync signals. Then the 4th lane just carries the extra 2 bits of each color. In those cases the display will run just fine with the 4th pair not used, it will just only have 18bit colors.

What you say makes sense, it's what I was expecting. But the thing is the 20" datasheet does not provide any mapping for the LVDS signal. It only says "Interface chip must be used LVDS, part No. DS90CF383MTD ..." Then looking at this DS90CF383MTD datasheet there seem to be an offered mapping in figure 2, which is not 4th lane with extra 2 bits of each color, and I don't know if that is what they used anyway.

[Berni]

Well there are two standard LVDS signaling formats called VESA or JEIDA. Both have the extra color LSBs on the 4th channel, but have the other bits jumbled up, but its perfectly possible there are displays out there that use some proprietary LVDS bit format. This format is one of the things that the BIOS needs to tell the graphics card, and i don't think this is part of the EDID since the interfaces that is was designed for such as DVI HDMI..etc have the bit mapping format defined by the standard so it is always the same.

The last industrial motherboard with LVDS video output that i have used had the LVDS output formats and resolutions configurable inside the BIOS settings (And also could not support any arbitrary weird format and resolution combination, it had to be something standard from the list).