Prev: SD card: 1G vs 2G
Next: Thanks Paul
From: yaugin on 20 Jun 2010 17:07
How do you translate resolution and refresh rate to bandwidth? I
usually see something like 150-200 MHz on consumer grade switches and
have no idea if that's enough for my setup. Also wondering if color
depth has any effect.
From: Paul on 21 Jun 2010 04:21
yaugin wrote:
> How do you translate resolution and refresh rate to bandwidth? I
> usually see something like 150-200 MHz on consumer grade switches and
> have no idea if that's enough for my setup. Also wondering if color
> depth has any effect.

I can't give you a very good answer on this, because I can't get
any two products to line up in quite the same way :-( So if there is
some "accepted" method generating these numbers, I'm not seeing it.

This VGA splitter, is an example of a device that buffers a VGA
signal (it uses a power adapter, so there is an active device inside).
It produces two copies of the signal. This is the specification.

http://homeavcables.com/vga1x2.html

"the distribution amplifier has a high bandwidth of 450MHz,
up to 2048x1536 maximum resolution and a two level cascade
capability that provides for future expansion."

Now, even if you compensated for putting two of the devices in a
row, by dividing the bandwidth by root of two (1.414) and getting
a cascaded bandwidth of 318MHz, the results still can't be made
to line up with this as a second example.

http://www.zonet.com.tw/product-61.aspx

16 port rackmount switch
video resolution 1920x1440
bandwidth 200MHz

If I take 318 * (1920x1440)/(2048x1536) I get 279MHz instead
of 200MHz.

I would have thought resolution and bandwidth, would go hand in
hand for analog VGA.

*******

This is how I started looking at it. First, imagine a specific video test
pattern. Alternating white and black pixels. This is a square wave
at half the pixel clock rate.

To work out the pixel clock rate, and incorporate time for blanking
interval (hides CRT retrace), I can use an approximate formula or use a
GTF modeline program, to do the same thing. These give reasonably
good agreement with one another. The pclk or pixel clock, defines the
frequency, and the inverse of that is the time period per pixel.

http://www.monitorworld.com/faq_pages/q29_page.html

http://www.arachnoid.com/modelines/

This article says the video bandwidth should be about 3 times the
half pixel rate, for the picture to be non-blurry.

http://www.kdsusa.com/FAQCRT.asp

"RGB bandwidth determines how fast the electron beam can change
state as it sweeps across the screen, setting an upper limit on
horizontal resolution. The most torturous image for a CRT is an
alternating series of one-pixel-wide vertical B&W lines. The input
signal for such a pattern is a square wave, but unfortunately, bandwidth
measures only sinusoidal frequency components. The sinusoidal bandwidth
required to adequately approximate a square wave is roughly three times
the binary frequency of the square wave. The additional bandwidth is
required to capture the third harmonic, which is essential if a decent
approximation of a square wave is to be obtained (most of the signal
energy for a square wave is in the first and third harmonics of its
Fourier series)."

For fun, to check the results of that little blurb, I tried my old CRT,
now sitting in the basement.

http://www.dooyoo.co.uk/crt-monitor/sony-gdm-2038/details/

Max Resolution 1600 x 1200 (I doubt it would do this)
Factory Preset Resolution Modes 1280x1024 @ 76 Hz (It can do this)
Video Bandwidth 120 MHz

The 120MHz number is too low to meet the criterion suggested in the KDS article.
pclk 141.82 MHz times 1.5 = 213Mhz, not even close.

A further factor, but perhaps less important, is the overall system bandwidth,
is a mathematical function of the bandwidth limited components connected in
series. Equation 1 on PDF page 5 here, shows how to take the square root of
the inverse sum of squares, to get the overall system bandwidth. To
give an example, if the monitor was 400MHz, the KVM 200MHz, the overall
end to end response 3dB down, is 178.9 MHz. The total system response must
be lower than any one element in the system. In this case, the KVM dominates
the monitor, so we'd use a little less than the KVM result to estimate
how well the whole thing works. You can see for my old CRT monitor, in fact
the CRT bandwidth is less than the KVM. I'd be seeing a number below 120MHz.
That KVM wouldn't be upsetting my CRT very much, in terms of making the
image blurry.

http://physics.bu.edu/mcen/MCCM_Parts/Hotlink%20design%20considerations.pdf

As I cannot get the numbers to line up, I'd suggest just going with the
1920x1440 number, as being the equivalent of "200MHz" and work with that.
I would think they mean 1920 x 1440 @ 60Hz, as assuming a higher frame
refresh rate works against them. They'd *probably* assume 60Hz for
refresh rate, but it really should be stated.

I don't think the color depth is a factor. The above analysis focuses on the
analog response to the RGB signals, which come after the DAC has done
the conversion from some color depth. For example, you could think
of two adjacent four bit per pixel colors on the DAC, as being 10100000 and 10110000.
The four lower bits are still there, they're just always zero.

Good luck,
Paul
 | 
Pages: 1
Prev: SD card: 1G vs 2G
Next: Thanks Paul