IOGEAR DVI KVM Problem?

  • Thread starter Thread starter Frederick
  • Start date Start date
F

Frederick

I have been using a IOGEAR DVI 4-port 'KVMP switch' for a while now.
It has an irritating problem, and I am wondering if it is defective,
if anyone has similar trouble, or is it just the nature of the beast.

Three desktops (All XP SP3), use the following DVI video cards:
1) RADEON X300 SE 128MB set at 1024x768 pixels.
2) NVIDIA GeForce 9300 GE 512MB set at 1360X765 pixels.
3) VISIONTEK RADEON 9250 AGP 128 MB 1024x768 pixels.

All boot up correctly at the above resolutions.

When I use the KVM to switch back and forth, the second and third
machines quite often forget the display setting in favor a totally
useless 640X480. On top of that, the boot setting is no longer
selectable. Which means I guess that the system is no longer using my
video card (yes I tried reloading its driver). Re-boot of that
desktop is required to fix things which it always does.

Anyone?

Frederick
 
Frederick said:
I have been using a IOGEAR DVI 4-port 'KVMP switch' for a while now.
It has an irritating problem, and I am wondering if it is defective,
if anyone has similar trouble, or is it just the nature of the beast.

Three desktops (All XP SP3), use the following DVI video cards:
1) RADEON X300 SE 128MB set at 1024x768 pixels.
2) NVIDIA GeForce 9300 GE 512MB set at 1360X765 pixels.
3) VISIONTEK RADEON 9250 AGP 128 MB 1024x768 pixels.

All boot up correctly at the above resolutions.

When I use the KVM to switch back and forth, the second and third
machines quite often forget the display setting in favor a totally
useless 640X480. On top of that, the boot setting is no longer
selectable. Which means I guess that the system is no longer using my
video card (yes I tried reloading its driver). Re-boot of that
desktop is required to fix things which it always does.

Anyone?

Frederick
Click to expand...

This is related to auto-detection of connected monitors, on
the video cards.

The DVI connector has high speed serial signals for R, G, B.
They're differential pairs, with a termination resistor across
them.

It's possible, at the video card, to detect the termination
resistor is missing. That is how the video card knows the
monitor is not connected. A KVM needs to prevent the computer
from seeing an "open circuit", when the KVM is switched
away from that computer.

If the KVM design is a good one, it should maintain termination
impedance on all the four computer DVI connectors, whether
the monitor is switched to that port or not. To do that,
the designers would use a digital multiplexing chip, rather
than some crude scheme involving relays for example. If relays
were used, the "glitch" on the diff pairs, would be enough
to trigger an attempt to auto detect the monitor.

The other part of the puzzle, is the DDC serial interface.
The monitor has clock and data signals. The computer queries
the monitor, via that interface. The interface, as far as I
know, was not designed for multi-master operation.

Computer #1 ---- KVM ----- Monitor DDC clock and data
Computer #2 ----
Computer #3 ----
Computer #4 ----

When you switch computers, the computer should read the
DDC, if it wants to know about the monitor. I doubt all
the computers are connected to those wires at the same
time. Each one of them is likely a separate bus master.

If Computer #2 attempts to read EDID via DDC in the picture,
it's currently got no connection. It can't get any data,
so it switches to a safe 640x480.

What the KVM should be doing, is "faking" the EDID, like
this box does. This box will "play back" the EDID, as it
is stored inside the box. The copy in the monitor now
doesn't need to be consulted. Of course, buying four
of those, at $79 each, completely defeats the low cost
of the KVM in the first place. I'm not suggesting
you buy these, merely illustrating there is a device
that can solve the problem of EDID reading. The chip
inside this thing, would cost less than a couple dollars,
and the KVM should have integrated that chip for you.
Then, the KVM could have "faked" the EDID readback, itself.

http://www.gefen.com/kvm/dproduct.jsp?prod_id=4714

Computer #1 ---- Gefen --- KVM ----- Monitor DDC clock and data
Computer #2 ---- Gefen ---
Computer #3 ---- Gefen ---
Computer #4 ---- Gefen ---

Such a box, solves the "reading EDID" problem. Depending on
whether the little Gefen box also reclocks DVI data, it might even
stop impedance discontinuities or glitches, from triggering
auto detection.

It's possible to design the box, such that this doesn't
happen. Who knows why they don't bother... It could be
done with one custom chip, and done properly. And I would
expect, with a market demand for such a chip, there is
likely one already designed to do just that.

This isn't exactly the right chip, but what it does have,
is a block labeled "EDID memory". If any of the four
devices on the left, want to query the EDID via DDC, they
can get a copy.

http://www.nxp.com/documents/leaflet/75016975.pdf

The only thing wrong with that chip, is the interface on the
right hand side isn't right. Since the chip is for usage inside
an LCD TV or the like, it's a front end chip, rather than
ready to be used in a KVM.

I suppose you could contact Iogear tech support and see if
your problem is "normal" or not. I'm sure they'll just
blame it on your setup, etc.

Paul
 
This is related to auto-detection of connected monitors, on
the video cards.

The DVI connector has high speed serial signals for R, G, B.
They're differential pairs, with a termination resistor across
them.

It's possible, at the video card, to detect the termination
resistor is missing. That is how the video card knows the
monitor is not connected. A KVM needs to prevent the computer
from seeing an "open circuit", when the KVM is switched
away from that computer.

If the KVM design is a good one, it should maintain termination
impedance on all the four computer DVI connectors, whether
the monitor is switched to that port or not. To do that,
the designers would use a digital multiplexing chip, rather
than some crude scheme involving relays for example. If relays
were used, the "glitch" on the diff pairs, would be enough
to trigger an attempt to auto detect the monitor.

The other part of the puzzle, is the DDC serial interface.
The monitor has clock and data signals. The computer queries
the monitor, via that interface. The interface, as far as I
know, was not designed for multi-master operation.

Computer #1 ---- KVM ----- Monitor DDC clock and data
Computer #2 ----
Computer #3 ----
Computer #4 ----

When you switch computers, the computer should read the
DDC, if it wants to know about the monitor. I doubt all
the computers are connected to those wires at the same
time. Each one of them is likely a separate bus master.

If Computer #2 attempts to read EDID via DDC in the picture,
it's currently got no connection. It can't get any data,
so it switches to a safe 640x480.

What the KVM should be doing, is "faking" the EDID, like
this box does. This box will "play back" the EDID, as it
is stored inside the box. The copy in the monitor now
doesn't need to be consulted. Of course, buying four
of those, at $79 each, completely defeats the low cost
of the KVM in the first place. I'm not suggesting
you buy these, merely illustrating there is a device
that can solve the problem of EDID reading. The chip
inside this thing, would cost less than a couple dollars,
and the KVM should have integrated that chip for you.
Then, the KVM could have "faked" the EDID readback, itself.

http://www.gefen.com/kvm/dproduct.jsp?prod_id=4714

Computer #1 ---- Gefen --- KVM ----- Monitor DDC clock and data
Computer #2 ---- Gefen ---
Computer #3 ---- Gefen ---
Computer #4 ---- Gefen ---

Such a box, solves the "reading EDID" problem. Depending on
whether the little Gefen box also reclocks DVI data, it might even
stop impedance discontinuities or glitches, from triggering
auto detection.

It's possible to design the box, such that this doesn't
happen. Who knows why they don't bother... It could be
done with one custom chip, and done properly. And I would
expect, with a market demand for such a chip, there is
likely one already designed to do just that.

This isn't exactly the right chip, but what it does have,
is a block labeled "EDID memory". If any of the four
devices on the left, want to query the EDID via DDC, they
can get a copy.

http://www.nxp.com/documents/leaflet/75016975.pdf

The only thing wrong with that chip, is the interface on the
right hand side isn't right. Since the chip is for usage inside
an LCD TV or the like, it's a front end chip, rather than
ready to be used in a KVM.

I suppose you could contact Iogear tech support and see if
your problem is "normal" or not. I'm sure they'll just
blame it on your setup, etc.

Paul
Click to expand...


Thanks again Paul.

I guess I will just have to live with this problem. I didn't tell
you, but this KVM replaced its twin, that died on me about a year ago.
I don't recall ever suffer my current problem with that one. So, I am
going to assume I have a faulty KVM now. Sobeit.

Frederick
 
Frederick said:
Thanks again Paul.

I guess I will just have to live with this problem. I didn't tell
you, but this KVM replaced its twin, that died on me about a year ago.
I don't recall ever suffer my current problem with that one. So, I am
going to assume I have a faulty KVM now. Sobeit.

Frederick
Click to expand...

It might just be the way it is designed.

To prevent the computer from seeing a "monitor change", you must

1) Maintain termination resistors on the end of the monitor cable.
2) Support DDC EDID reading at any time. If you can't maintain
termination resistance, at least eh EDID should always be readable.
The computer may make a video output switch, but eventually
return to the normal resolution.

If the design doesn't support these features properly, then
switching away from the monitor, may cause the computer to
revert to 640x480. That's my guess as to what is happening.

You'd need to find a product review for the KVM, to see if
other users also see the problem. Check Newegg and see
if they carry your product, and see if the reviews there
detected a problem.

Paul
 
Back
Top