DVI monitor switch help

[qa9b]

Hello all!

I've recently embarked on a little project... I intend to build my own KVM switch from scratch. Part of this device's functionality is to switch a single monitor between two inputs (i.e. two computers.), using single-link dvi-d cables.

Now, I've run into a little snag. Obviously, the first step in building a DVI switch is to successfully interface the DVI cables with my project. I purchased these i/o connectors from mouser and soldered leads to them.

Before I connected them to any kind of circuitry, I wanted to first see if a monitor signal survived what I had done so far. Well, it couldn't. Pictures seem to tell the story better:


This was my basic setup. I just have two i/o connectors with all the corresponding pins connected. On the left is the dvi cable from the computer, and on the right is the dvi cable going to the monitor.

**broken link removed**

Just let me say that I've spent at least an hour and a half double and triple checking for continuity and shorts. Everything looks perfect on the volt meter, but something isn't working with high amounts of dvi data.

I'll clarify a little. If I have my computer set to my usual 1920x1200 resolution, my little setup produces a "signal out of range" message on my monitor. So I decided to try a lower resolution, and hence, it works. Here's a little illustration of what happens as I try to increase the resolution:

1152x864:
**broken link removed**

1280x1024:
**broken link removed**

1600x900:
**broken link removed**

1600x1200:
**broken link removed**


With that last resolution (1600x1200), the monitor displayed that for about a second, then went to black. When I try to get to 1920x1200, my monitor displays a "signal out of range" message.

My dillema is obvious. I don't intend to admit defeat, though.

Somehow my setup can't handle the data throughput of higher resolutions, which also makes me doubt the integrity of even a 1024x768 signal. My guess is that there's maybe some type of interference going on or degradation of the signal as it travels through my expert soldering joints.

One of the tests I did do was check to see that if I stuck 6.3v in one end, that I got exactly that voltage out the other end. That all went fine. Like I said, whatever this problem is, it isn't going to be found with a voltmeter.


I'd greatly appreciate any help you guys can give me. Thanks for reading...
qa9b

 

[Nigel Goodwin]

For a start don't try and use a breadboard - at higher resolutions layout becomes more and more critical, and breadboards are only suitable for low frequency use.

 

[3v0]

If I were building it I would use DVI connectors that soldered to the PCB. Put a gnd trace between each signal if there is room. That sort of thing. With that many signals coming off the convector it will not be easy.

If you can not do a PCB start by shielding the various clock and sync signals. Shorten all the wires as much as possible.

This project is not an easy one for a couple of reasons. You have already found the first one.

A KVM need to fool the computer without a monitor into thinking that it still has a monitor attached. When a computer boots up without the monitor the KVM has to make it think the existing (but not connected) monitor is attached.

It requires a microcontroller and lots of switching both digital and analog.

The only good reason to build this project is learning.

3v0

 

[qa9b]

Thanks for the reply, guys.

I will eliminate the breadboard and see if that helps at all.

For me to connect these dvi connectors to a pcb meant either constructing one myself or having a special one made, and I wasn't ready for that kind of investment. I am fairly confident in saying that there really was no other option on my part in terms of interfacing my project with dvi cables.

3v0-
I did consider using a micro-controller in the beginning, but eventually settled for something called the 4066 quad bi-lateral switch. I tried running a few TDMS wires through one of these switches, and it didn't harm the signal any further. About 10 of these switches will together form something like an 18PDT switch.

I'll admit that I hadn't foreseen this problem. I'm having trouble sending this signal through a few solder joints, but what about the maze of switching circuitry that it will eventually go through?

About shielding... there are four wires inside the dvi cable that are devoted to shielding. What should I do with these to preserve the shielding throughout my project (i.e. not just treating it as another signal that needs to get from point a to point b). Are "ground" wires typically used for shielding?

The 4066 is extremely cheap, and being the center of my project, other costs such as wire, connectors, and pcbs don't add up to much more. If all of this pans out, my end cost will be significantly less than a stand-alone dvi switch.

At the moment, I'm not too concerned about "tricking" the computer to see a monitor connected. I'm aware that it's rather crude in this type of scenario, but hot-swapping works and bios warnings can always be disabled. I might, however, put some more thought into this when it comes time to make the usb switch, to which a mouse and kb will be connected.

 

[3v0]

You could start by converting the shielded wires to twisted pairs. If that is not enough they make shielded wire. I do not know off hand where to get it in small quantities.

Making your own PCBs is not as big a deal as it used to be. Once you are setup with a laminator and a laser printer the actual cost is under 10 cents a square inch depending on the PCB stock you use.

Fill in your locaton.

3v0

 

[qa9b]

Ok. I have an idea, albeit a little expensive.

What if I were to get a hold of a few dvi cables with female connectors? Couldn't I just hack off the other end and be left with a bunch of wires that I know are soundly interfaced with the connector? Of course, female to female dvi cables pretty much are non-existant, so I'm looking at another $30+ to add to this project's cost. (But if it works, hey, what the hell).

Also, instead of inputting said bunch of wires through my not-sure-if-it-really works dvi switching circuit, I put them into something like this:

https://www.electro-tech-online.com/custompdfs/2008/10/MAX4886.pdf

If I'm looking at that correctly, all I would have to do it connect my cables to the right pins and I'd be in dvi switching circuit heaven.

Comments?

 

[3v0]

Observations.

The chip looks nice but it is only the DVI section.

Most often these projects cost more to build then buy.
If you flipped burgers for a week you could buy one.

3v0

 

[qa9b]

Would this take care of the rest, then?

https://www.electro-tech-online.com/custompdfs/2008/10/MAX4929E.pdf

I'm a bit of a noob at interpreting what all the pins on that one do, but it looks to me as if it could switch the rest of the signals?