MT6820-B LCD controller, low resolution.

[step_s]

Hi EEV'ers.
This might be more of a computer question, but wanted to ask it anyway.
I bought this http://www.ebay.co.uk/itm/Universal-New-5V-MT6820-B-LVDS-LCD-Monitor-Driver-Controller-System-Board-/381581274868?hash=item58d8042ef4:g:S8oAAOSwLpdW9tVP and wanted to test it out on an old labtop LCD display 1400x1050 max resolution.
After a lot of work and looking up datasheet etc. I finaly got it all working, but the picture only seems to be 600 x 400 or so.
I've set the jumpers on the board to 1400x1050 (2 x 6bit), and this makes the picture sit right on the screen (other resolutions make it half way across the screen, and other funny things). The scaling of the picture also works how it should with setting resolution on the computer, but alas, the picture is still pixelated, no matter how high or low I go.
I know the controller is dirt cheap, and I can't expect too much, but I have seen people on the net make it work well. Is the one i got just a cheap knockoff?

It should be said that I have soldered the connector myself, using the one from the screen, and soldering it on the controller board. Sync and color is spot on, so only the output resolution is too low.

Hope anyone has tried the same, and can help

[bktemp]

When the image fills the whole screen, the resolution going to the TFT is correct. So the problem is most likely at the VGA side.
Did you verify the resolution at the computer? Is there an OSD page showing the detected input resolution?

[step_s]

The image is perfectly calibrated on the screen, and fills it out very nicely, but since the OSD menu is actually affected the same way as the picture from the VGA, it makes me think it's some kind of connection to the screen. . .
The strange thing is, that if any of the connections to the screen is disconnected or connected wrongly, the image gets very distorted, so it also seems to be connected corretly.
This controller only has one set of Clock outputs, but in a double bit configuration there is normaly one for the ODD and one for the EVEN, might this be affecting the screens performance?

I have included a picture zoomed in on the OSD menu, to show how the resolution is messed up.

[bktemp]

You swapped the odd/even channels on the TFT.
It is a dual channel LVDS display: One channel contains the even, the other one the odd pixels.
So you need to change the wiring.

[step_s]

I believe I did try that once, that gave onlystatic, no picture at all, but I'll try again.
If I may ask, how can you see that the channels are swapped?

[bktemp]

If I may ask, how can you see that the channels are swapped?

Look at the characters, for example the displayed resolution 1024x768@75Hz at the bottom: In the "0" the internal diagonal line has clearly the pixels in the wrong order. If you swap odd and even pixels it will be correct. The same applies to any other character. Instead of displaying the horizontal pixels 1, 2, 3, 4, 5, 6 they are shown in the order 2, 1, 4, 3, 6, 5 when both LVDS channels are swapped. That is eaxctly what you see here.

[step_s]

You Sir, are a god
It's all looking good now. Guess you can't really trust the datasheet then :/

Now I just have to figure out how to run it at 1400x1050. This option is skipped in my graphics settings, hmm.
Tried to connect the EDID clock and data directly from the LCD to the VGA, but the computer still only sees it as generic PnP monitor.
Know anything about this?

[bktemp]

You could try generating your own EDID data (Phoenix EDID Designer is a good tool for generating or modifying EDID data). This tool can also extract EDID data from connected monitors, so you can look at the EDID data provided by the LCD.
If you only specify the detailed timing in EDID data it forces a fixed resolution. This works for most video outputs (some graphic cards/drivers ignore the values, but most follow the timing precisely).
You need to write the generated data into an I2C EEPROM (24C02) and connect it to the EDID wires, because the build in EDID is most likely read only.

[step_s]

I de-soldered the connection from the EDID to the controller board, and directly connected the screen. This seems to work after a few reboots of the computer, and a custom resolution setting in NVidia control panel. Since the monitor will be connected to the same PC, it's no issue really.

Maybe I will try out programming an EEPROM. http://www.ebay.co.uk/itm/2PCS-AT24C256-Serial-EEPROM-I2C-Interface-Data-Storage-Module-Arduino-PIC-TOP-CF-/272227376042?hash=item3f62041baa:g:sxgAAOSwuzRXedfi would this module work?

And what can you recommend for connecting the I2C to USB? Have a PICkit 3, that should work?

[bktemp]

Large EEPROMs use an additional adress byte. You need one of the following sizes: 24C01, 24C02, 24C04, 24C08, 24C16.
It looks like PICkit3 can't be used as an I2C EEPROM programmer, but you could use any microcontroller as an EEPROM programmer.

Did you try reading the EDID data from the LCD screen using the EDID tool?

[step_s]

The read of the EDID from the LCD is strange. Windows calls it "Color LCD" and NVidia calls it "Apple Color LCD". The resolutions in the phoenix program seems to only go to 1024x768, on the first page?
Wouldn't the EDID contain some sort of product name?

[bktemp]

EDID supports different ways of specifying supported resolution: Either it gives a list of fixed resolutions (pretty much obsolete for modern resolutions), or it gives a min and max range of frequencies (often used for VGA), or it gives a fixed timing (used for driving digital displays at their native resolution).
In the detailed timing section there are 4 data blocks: One is typically the detailed timining, one is the frequency limits, and the two other two data blocks are monitor name and serial. Those 4 data blocks are often the most important ones.