18-bit to 24-bit LVDS converter.

You probably will have to convert LVDS to parallel 18 bits, then serialize into 24 bits LVDS again. Just insert 0-s for the lowest two bits of each colour.

Many panels have 18 bits - 6 bits per colour. This is still quite an overkill for TFTs, since the contrast they have is typically around

500 and I have not seen it above 10000 (top class TV-sets/monitors, IIRC), so using just 16 of the 18 bits is a 6 to about 100 times overkill. When doing 16-bits into the 18, the typically used number is 5-6-5 for R-G-B (again, just padd 0-s for the lowest bits you don't have).

Dimiter

------------------------------------------------------ Dimiter Popoff Transgalactic Instruments

------------------------------------------------------

Considering

a) Most Graphics LCD panels on the market as colour displays 640x480 and higher are normally 18bit, I would say it is a surprise to get a REAL 24 bit LCD panel. Many analog LCD panels claimed 24 bit for the same glass and glass drivers.

b) Every LCD manufacturer and nearly every panel they make often have completely different connectors.

c) Consult LVDS specs for panels

has lots of info) and you will see that you need an extra channel to drive 24 bit compared to 18 bit.

So source the right connectors and wire up the 18 bit channels, as that will at least work with LSBs on RGB channels missing, unless you have really GOOD display and image source 90% of people won't notice the difference. For the majority of items displayed it will be sufficient.

--
Paul Carpenter          | paul@pcserviceselectronics.co.uk
    PC Services
              GNU H8 & mailing list info
             For those web sites you hate

The panels are probably using FPD link. You can compare the two data protocols by viewing the OpenLDI spec

and search) which is a superset of the FPD link spec.

You can use a National 24 bit channel link device if you have the data already in it's parallel form.

Cheers

PeteS

You are right on the color resolution point of view but wrong on the interface one. If you look at the data sheets you will see that the fourth pair carries the R6/7, G6/7 and B6/7 signals. If I leave them unconnected the contrast will be poor (assuming the panel works) as I will be using one fourth of the panel's range. I remember having seen a panel that migght be configured for the two formats (or so I understood it) but not the LG-Philips model I use. That is the reason why I am looking for the 18 to 24bit converter. So far the only solution I have found is like Dimiter suggested: using two boards, one

18-bit to TTL and one TTL to 24-bit and route the TTL signals appropriately.

Thank you all for your answers.

Elder.

Yes and NO, in LVDS R6/R7 map to RGB R0 and R1 respectively. The mapping is as follows (same for each colour gun):-

LVDS Display or graphics source 6 0 LSB 7 1 LSB + 1 0 2 LSB + 2 1 3 2 4 3 5 4 6 5 7 MSB

It was designed this way so you COULD leave off the fourth pair for just such circumstances.

If you don't believe see such documents as National Semiconducters AN-1085.pdf, pay particular reference to 24 bit table 1 on page 8, and even more so to the timing diagram on page 9.

On page 9 the fourth pair is labelled as using R6/R7 etc.. with a big BOLD box label to the right stating "LSBs".

You loose nothing of the range that is viewable

Waste time and money if you wish.

I have spent too many years of doing LCD interface designs and video processing before the interface.

--
Paul Carpenter          | paul@pcserviceselectronics.co.uk
    PC Services
              GNU H8 & mailing list info
             For those web sites you hate

Paul, I think I understand your point. I guess both of us are right (I wish I was wrong.) I had a look at NS AP-1127 app note and it does suggest one to connect the least significant RGB bits to the fourth channel. I totally agree that in that case the fourth channel would not be missed. It does make total sense for backwards compatibility.

Unfortunately the LG-Philips folks (and Sharp's/Samsung's for that matter) seeminly did not think this way when they designed the panel: the most significant bits go onto the fourth pair, not the least significant ones. Even NS has an IC that follows this odd (IMO) pattern: in the NS90C387 data sheet, figure 17, the bit stream is compatible with LG-Philips panel's. I saw the same pattern in one panel by Sharp and other by Samsung (both 24-bit.) And the bit stream described on the AMD M690T chipset LVDS output follows the same pattern. :( Thus I presume this pattern has become the de facto standard despite being odd.

FWIW there is one app note by Intel (Document Number: 315975-001) in which they mention two different LVDS bit stream paterns, 24.0 (the rational one) and 24.1 (the "odd" one.)

Maybe I am missing something but I simply cannot see how to connect the LG-Philips panel directly to the SBC as is withouth loosing contrast (BTW it is based on the aforementioned chipset but the manufacturer did not route the fourth data pair.)

Regards.

Elder.

Well I just downloaded to double check I had the latest version and the mapping tables on previous pages (Tables 2 to 4) list the mapping as e.g. LVDS Rx6/Rx7 to graphics and monitor LSBs. This appears to be for all link types (single/dual pixel/clock).

Yet another f*((*&^ing variation in every single model of LCD panel, (connector make/size/pins, panel shape, mounting, number of lamp connectors, connectors positions....) even from same manufacturer.

Got that and the first time I have seen someone show that as an alternative.

If it is based on DS90C387, it would suggest they dropped the fourth pair either because it is set for LSBs on fourth pair, someone cocked up or they expect the panels to twiddle the bits.

Using the Intel parlance the mapping tables of DS90C387 are for 24.0 .

Without knowing the LG-Philips LM150X08-TL01 spec I cannot comment on what the panel is expecting.

--
Paul Carpenter          | paul@pcserviceselectronics.co.uk
    PC Services
              GNU H8 & mailing list info
             For those web sites you hate

I mean the SBC is based on that AMD/ATI chipset which uses the 24.1 pattern. The SBC manufacturer did not route the fourth pair (for both channels) to the LVDS connector. :( I would say it is a design mistake. :(

You may find an equivalent at:

The relevant pages are identical to TLO1 ones and start on page 9.

Thanks for your attention.

My best regards.

Elder.

Or someone laid it out thinking they were using DVI-D i.e. TMDS signalling of 3 channels of 10 bits[1] serialised and clock. Just a thought with these multifunction board layouts and chipsets, not actually expecting someone to actually use the output.

If they have LVDS or TMDS usage, you might be in a worse situation of TMDS to LVDS conversion.

I note they state the DS90C385, which is the 24.1 format compared to DS90C387 which is 24.0, however the same could be achieved in layout changes of shuffling the lines.

Yes it is the weird format and I am trying to see what they are gaining by using it. Unless they have to have all bit positions actually driven.

[1] TMDS uses 3 channels of 8bit data and 2 control signals on each data pair, and a seperate clock pair.

--
Paul Carpenter          | paul@pcserviceselectronics.co.uk
    PC Services
              GNU H8 & mailing list info
             For those web sites you hate