DVI Switching

Jason,

That is a very interesting setup, but I am not sure that it will work.  TMDS442 is a true CML driver that requires a 50 ohm pullup on each signal to ~ 3V.  The DP128 is basically a voltage follower that does not have the input structure to support the 442's termination.  You could cascade two of the 442's with some provisions for loss in between the cascades, but that only gets 4 outputs, and you will be also concerned with "blocking" if you plan on switching any input to any output (even with the 128 this will be an issue)..  Do you know the maximum data rate for this application?  I am guessing that this is not dual link DVI either.  I think that a block diagram and the connection options really need to be identified before much more effort is involved.  If any input needs to go to any output at any time, this will not work and we would probably want to look at some of the cross point switches.

Ken,

This is an application for testsing DVI functionally.  I do not require dual link and I only need 1 output at a time.  I would like to control with discretes as oppossed to the serial bus, since the units under test require the usage of my serial bus.  Currently, I need to switch WXGA 1440 x 900.  I believe the data rate is around 400 MHz.  Using 2 TMDS442 switches, my application looks as follows.

That probably has a good chance of working with the lower data rate.  Are you sure it is around 400 MHz?  Reduced blanking 1440x900 is around 89MHz pixel, so data would be close to 890MHz, which might still work as the eye is still fairly large, so the fact that our device doesn't retime the signal should be ok.  What is the final UUT?  Is this a go/no go test, or will you be adding cable and trying to close the eye for limit testing?

Ken,

I believe it is a lower data rate from the UUT, but for testing this design 890MHz may apply.  The final UUT accepts DVI from multiple sources, merges them, and retransmits HDVI and A818 video.  This will be a go/no go test. 

To test this design, the test connections will be connected to a monitor via 16 feet of DVI-D cable.  I will connect a video generator with different data patterns and color to the first input and switch to either output connector to verify design and functionality before interfacing the UUTs.

Thanks,

Jason

Jason, I think this has a good chance at working.  16ft DVI cable may have a fair amount of loss, is the UUT cable of recovering that much loss?

Ken,

My video generator has overcome the cabling and the UUT is required to.

I have another question.  I have connected my video generator to 50 Ohm pull-ups and verified good clock and data with an oscilloscope.  When I disconnect from the 50 Ohm pull-ups and connect my video generator to 1 of the TMDS442 inputs, I can only verify the pull-up voltage on the input.  Any ideas why resistor terminations work better than the TMDS442 in my experiment?

Jason

I am not really sure how your setup looks.  Are you saying that the video source stops putting out a signal when the 442 is connected?  Are you using a high impedance probe with appropriate bandwith for the signal of interest?  If your 50 ohm pullup setup is the same as you have connected to the input of the 442, I don't understand why you would not get a similar signal as the 442 has the internal 50 ohm to Vcm (~3V) termination

Dear TI and All,

Can I get SN75DP128ARTQT reference schematic?

I could 't find it in TI web...

thanks, 

eric

Hello Eric,

Please send me a brief explanation of the project to diego.cortes@ti.com, in order to provide you the reference design.

Best regards,

Diego.