[DIY] Modify XPAND's to reduce crosstalk.

[Leonos]

Lot's of us have problems with Ghosting/Crosstalk and use XPAND's.
This problem can be resolved if we could tune our XPAND's to be dark a tiny bit longer.
Unfortunatly it's not possible to achief this by modifying the IR, because the XPANDS only have two states: right open/left closed and left open/right closed.
So we have to do something inside the glasses.

In another post Petrus has sorted out how the XPAND's lcd's are driven.
Open=2volts/-2volts and Closed=14volts/-14volts (these are squarewaves).

What I need is a (simple) diagram or example on how to let the 14volts pulse to last longer, but the start of the pulse shouldn't be affected.

Are there some smart guys out there how know how to achieve this ?

(would be nice if it could fit inside the casing)

[Likay]

I could give it a shot but need to know a few things. Where's the thread about the xpand lcd-panels? I need to know which frequency the drivercircuit for the panel operate with, available voltages and if it's (most probably is) ok with a dual symmetrical driverstage because of the high voltage (high compared to logic levels...).

[Leonos]

Hi, Thanks that you wanna give it a try.

Here is the forums link: http://www.mtbs3d.com/phpBB/viewtopic.php?f=26&t=10169

Maybe you could use the others side pulse and some sort of HighPass filter from that to generate that extra pulse.

What I mean is something like this:
Right is dark (14volts) and left is open (2 volts).
At one time the switching is made where right should be open and left closed.
At that time left will go to 14volts and right will drop to 2volts.
Maybe we could at a highpass filter between the left side and right side, resulting that the left upgoing pulse will generate an extra spike to the right side (and otherway round).

I don't know if there will be any leakage or something or enough 'power' to get the spike to reach the other side.

[Leonos]

From Petrus I've got an intresting image containing the wiring diagram of the xpand 101, which is very close to the 103 (I guess).

See attchment

x101_full.png

[Likay]

The schematic helps more than anyhing else and i'll see when i got the time to solve it. The frequencies are not given but i have a slightly different idea that i think it would work anyway without any extra delays.
I also have to ask: Are you sure it's a timing issue? In this case only a part of the screen shows crosstalk. If the crosstalk is somewhat equal over the entire display there's nothing to do.

[Leonos]

The timing of the xpand with the television is very accurate I think. The lag of the TV's lcd's seems to be the cause. You'll see crosstalk there where the difference between left and right is a lot AND if the contrast difference between object and background is a lot.
So it's not always and not everywhere at the same time.
Example, white letters going towards you on a black background.

How do I know its the lag of the screen ?
I have read that 'heating' up LCD screens, shortens the switching time of the lcd's.
So that's what I did, I heated up the entire panel and the crosstalk dissapeared while heating it.
In the crosstalk test with the numbers, the highest visible number on the 'closed' side went from 8 to 78 (that's a good thing).
(also it is completly gone at the white letters on black background).

So I figured that I have two options:
1. Heat up the TV's panel with some kind of warming-mat (scary).
2. Give the TV's lcd's just a little more time to switch.

Obviously option 2 is the logical one and can be achieved by modifying the xpand.

Now the xpand has two states: 1. left open, right closed and 2. right open, left closed.
With the modification you'll be creating an extra state: both eyes closed for a little time. This is where both eyes are 14volts. This state is not controllable, but that is not needed where it's just to give the TV some extra time.
Probably the picture will lose some brightness, but that's no problem.

I think the panasonic glasses already do this, they have a gap between switching where both eyes are closed for some time.

[Leonos]

Some studying on the diagram:

There are 4 states: dark1=+14volts, dark2=-14volts, open1=-2volts, open2=+2volts. These states go from plus to minus and back etc. because LCD's do not like DC but prefer AC.
The 4052 is driven as follows through pin 10 and 9 (A and B).
B wiring on the left and right is connected to eachother.

RIGHT EYE----------------------------- LEFT EYE
Dark1: A=1, B=0 (+14 volts) ---> Open1: A=1,B=0 (-2 volts)
Open1: A=0,B=1 (-2 volts) ------> Dark2: A=0,B=1 (-14 volts)
Dark2: A=1,B=1 (-14 volts) ----> Open2: A=1,B=1 (+2 volts)
Open2: A=0,B=0 (+2 volts) ------> Dark1: A=0,B=0 (+14 volts)

Frequency on my TV seems always 60Hz per eye.

[Leonos]

Oke, I've done some thinking again.
See the next two schematics, I think they should work. I've tested it at an online tool and in both sides the dark state can be delayed.

Some explaination:
RIGHT SIDE.
The right side is closed and wants to open, pulse A is going LOW.
The drop of this pulse is inverted through the first transistor and a spike is generated with the RC filter.
The second transistor is inverting the spike again.
The LM555 is triggered by this spike and generates a pulse with x-length which can be modified. This pulse is added to the original through a diode.
This makes this 4052 input HIGH throughout the length of the 555-pulse and overriding the original status (LOW) for that time.
The two diodes are there so no current can flow back to the controllers outputs. A pull-down resistor should be added from the new output to ground (100K ohm).

LEFT SIDE.
The left side is closed and wants to open, pulse A is going HIGH.
No inverter needed and the spike is generated from A through the RC filter.
Again this spike is inverted and the spike is triggering the 555.
This generates a pulse with x-length (can be modified).
Other than the right side, the dark-state can be achieved by setting the 4052 input A to LOW. To achieve that, the original pulse is connected through a diode (D1) and a high resistor to the 4052. The 555's pulse opens a transistor which results in the original pulse (HIGH) to flow through D2 through the 1k resistor through the transistor to ground, pulling the output to LOW for the time of the 555 pulse duration. After that the transistor closes and the output will be HIGH.
Note that a pulldown resistor should be added as well, 1M ohms between new output and ground.

xpan_mod_right.png

xpan_mod_left.png

[Petrus]

Your design can be simplified to this :

Code: Select all

Right:
     D
  .-->|--.
  | ___  |
A-.-___--.-- OUTA
     R   |
        ___
        ___ C
         |
        GND

Left:
     D
  .--|<--.
  | ___  |
A-.-___--.-- OUTA
     R   |
        ___
        ___ C
         |
        GND

The capacitor hold the signal A during the Dark to open transitions and the diode bypasses the resistor for Open to close transitions.
C must have a low value (< 1nF) to keep a fast Open to close transition and limit power consumption.

But the problem is the drive signal will go from +14V to -14V. Even if PI-cells are fast, I think it will take some ms for the crystals inside the LCD to rotate 180° and the LCD will not be completely closed during this time.

I think it's more reliable to add a small microcontroller between the original microcontroller and the two 4052 multiplexers. And with a microcontroller, you will be able to adjust the phase too.

[Leonos]

Wow, what easy life can be.

For the left side, the issue with the voltage going from +14 to -14 isn't an issue, because the states of B stays the same when it goes from dark to open.
So this voltage will stay the same for the extended time.

The right is indeed an issue BUT can be 'easily' solved by adding triple XOR gate with the output connected to the rightside B-input on the 4052.
If we connect the rightsides originals A, new (extended) A and B to the input we're there.

States will be:
dark1: A=1, new A=1, B=0, XOR output B=0 (same as original)
extended darktime1: A=0, new A=1, B=1, XOR output B=0 (=same as dark1)
open1: A=0, new A=0, B=1, XOR output B=1 (same as original)
dark2: A=1, new A=1, B=1, XOR output B=1 (same as original)
extended darktime2: A=0, new A=1, B=0, XOR output B=1 (same as dark2)
open2: A=0, new A=0, B=0, XOR output B=0 (same as orinal)

(looked it up at wiki: http://nl.wikipedia.org/wiki/XOR-poort

[Leonos]

Well I made the test setup for the left eye.
Added a variable resistor to it and watched a 3D movie holding the screwdriver.

The setup does work and I'm able to add delay time before the left glass opens. I could make the delay zero with result no changes to complete darkness of the eye (and off-course all between).
But....in all cases the ghosting did not disappear.

My conclusion is that the timing between TV and glasses is inaccurate.
Could be the glasses that take some time to switch or it's the signalling from the TV that inaccurate.
Strangly enough now i have no idea why the crosstalk is so much reduced when I heat the TV's screen.

So my project now continues from extending the dark time to shorten the open time. So the pulse to dark should occur sooner.

[Leonos]

Left side is pretty easy

Code: Select all

       C
A---||-------OUTA (C=100n, R=100k)
           |
          ---
         |   |
      R #   ^ Diode 
         |   |
        ground

The diode is discharge the capacitor when A turns low.
So I tried this (disadvantage is that now the voltage goes from +14v to -14v to -2v and then from -14v to +14v to +2v).
But I took this for granted and now I had more succes.
While turning the resistor, the numbers for the other eye became less visible, while the left side brightness stayed almost the same. The other side less readable number went from 18 to 26, while the left side stayed at 4.
I couldn't get more succes, because then the left became darker.
But that's probably caused by the lag of the LCD's rotating 180 degrees in the glass because it goes from +14v to -14volts.

Next thing is to fix this voltage issue.

For the right I haven't got a solution yet.

So Petrus, could you help me out again ?

[Petrus]

I don't understand, You put this circuit on the left side and notice an improvement on the right side ?

Your circuit create a spike when A go from 0 to 1, you want to shorten the clear sequences instead of extend the dark sequences, right ?

So you can try this on the right side :

Code: Select all

           VCC
            |
         .-----.
         |     |
         _     |
        | |   __
       R| |   /\ D
        |_|    |
         |     |
A --||---.-----.-- OUTA
    C

You will not have the -14V to + 14V problem as B does not change during clear to dark sequences.

To avoid the -14V to +14V issue on the left side, your idea of using XOR gates is good.

[Leonos]

Ok, I'll try that asap.

The improvement is noticable for the left eye.
The test is a SBS file with numbers, each side has numbers.
The intensity of the white is increased with the numbers, 1 has the lowest itensity and 80 the brightest.

Before with the left eye I could see the lowest number '19' ment for the right eye. With the modification it went up to 26. So this is an improvment. Best score is that no numbers can be seen from the other eye.