[KennyG]

OK, you guys have me going now.

I started a new thread because I don't want to move Jon Carpenter's thread in the wrong direction.


If I'm understanding correctly, the correct formula to calculate bandwidth is:


(H-res X V-res X Refresh Rate) divided by 2


I have been staying away from the res I really want to use on this Ampro because it's bandwidth is only 100mhz.

What I want to use is [email protected] I thought was 99.5mhz...however from what I understand now, it's really more like 50mhz???

 

[Dizzman]

one of the reasons that 48 hz is so popular with those who try it is this...


Bandwidth is THE slipperiest (i am pretty sure that is not a real word) number in video. all calculations, and most specs are somewhat interpretive.


If a projector says that it has a given BW then that means it can do a reasonable reproduction of the signal. As you come closer to the max bandwidth, you DO NOT get a better picture. as you push harder and harder then image starts to suffer.


My desktop computer monitor has the capability (and the specs) to display 1600 x 1200 @ 72 Hz. however it looks like crap. run it at 1280 x 1024 @ 70 and it looks much better. (go down to 1024 x 768 and it looks even better, but i hate that res) the reason for this is easy and complex at the same time.


As the display gets a higher and higher freq, it has to "draw" pixels faster and faster. there is a point you get to where it starts to blur them together. In essence, the rise time starts to fall off, and you get one pixel blending into the next. when that happens, the image loses "crispness" it is a simple matter of physics.


THe other (albeit smaller) part of the equation is the system as setup... Cables, Connectors (VGA BAD!) switchers DA's. they all contribute, and again, as the frequency gets higher... they all start to noticeably effect the image.


THink of this analogy. remember the game where you whisper something to somebody and then it gets passed on down the line... Video systems kinda work the same way. (in this example we are using the quick brown fox sentence) If the sentence is two slow, we get frustrated at waiting. (this is like too low a signal freq, we can see pixels and we do not like it) if the sentence is too fast, then everything blends together. However, at some point there is a speed that is perfect for intelligibility, and everything is good. however, we must not forget what happens when there is inconsistency in our line. (lets say one guy is a little slow) then there is a bottleneck at that point, an error is introduced and it continues down the line.


FOr bandwidth calculations go here...
http://www.extron.com/technology/archive.asp?id=vidband
http://www.extron.com/technology/arc...p?id=bandwidth


So if you were to take a signal doing resolution...

X @ 72 Hz. BW= 100 MHz

X @ 48 Hz. BW= 66 MHz


So in essence, the device is drawing individual pixels with an extra 33% of time. they are cleaner, MUCH sharper, and basically Mo' betta.


Of course, the final point on this is that you never have more resolution than you started with. SO the real frequency is not really higher than about 7 MHz. (in the case of NTSC/PAL)


So to test a system, do two things... Get a test pattern program and get a computer (HTPC) and feed an alternating pixel pattern to the display at your chosen frequency. If you CANNOT discern individual pixels, then you have passed the practical bandwidth of the display. Secondly, get VE or AVIA, and put up a multiburst pattern. and see if the resolution is clear all the way out.



Just trying to go as high a res as possible will not always translate to the best possible picture. sometimes it is best to back off a bit.


Of course, this is just my $.02 worth.

 

[kal]

Good lucid post Dizzman! I'm going to use some of your links on another thread in the HTPC forum where people don't seem to understand the importance of using good switchers...


Thanks!


Kal

 

[KennyG]

Thank you Dizzman...this is one reason I never use extra switching in my signal path...straight runs to the projector...I don't even go through my pre/pro's video switching.

Thanks for the links also.

 

[Dizzman]

i would not go so far as to say that switching gear is problematic. I would however agree with your decision in regards to the consumer AV gear.


pro gear is usually pretty darn good, and if the right decisions are made it can be seamless to the resulting image. But the right decisions have to be made and more importantly, you have to understand what a piece of gear does to the system, and why you are using it. Anything you do not need, remove it. but do not be afraid to add if it is going to add to the system.

 

[Dizzman]

i thought i would bump this as it is a discussion that is always kind of "beaten around" but never directly discussed.


I am sure that others have thoughts...

 

[KennyG]

OK...between my projector gutting and building I've finally taken some time to read the above Extron links...and it looks to me like all our calculations are wrong!?!?


Extron says bandwidth calculation is as follows:
SF=[(TP X Vt)/2]3


Where SF =Signal Frequency...TP =Total number of pixels...Vt =Vertical scan freq. (refresh rate)


So a [email protected] signal would actually use 99.53mhz of bandwidth.
TP = 1280X720, or 921,600 total pixels...with a Vt of 72

So: [(921,600 X 72)/2]3 = 99.53mhz


Perhaps this is why everyone sees such good detail when slowing the refresh rate down to 48hz?...A res of [email protected] uses the exact same amount of bandwidth.

 

[Iceman]

For a parallel distribution system such as RGB, you can forget about the factor 3. Unfortunately, the Extron links are somewhat lacking and oversimplified. As I noted previously ( http://www.avsforum.com/avs-vb/showt...hreadid=136254 ), the limiting bandwidth fB can be calculated as:


fB = pixel count / 2 x refresh rate x scan factor


The scan factor takes the retrace and sync parts of the signal into consideration and varies depending on resolution, but somewhere around 1.2 is a good starting point.


In practice, this bandwidth may be somewhat reduced in some cases (due to anti-aliasing filtering), just as VideoGrabber so rightly noted in the linked thread above.

 

[Dizzman]

In actual fact, the extron formula is very accurate. I had it explained to me as such;


once you have figured out the total number of pixels per second, you divide by two to simulate alternating pixels.


THe highest frequency pattern one could display on screen is an alternating pixel pattern (one on, one off and so on)


At this point a sine wave is what we have theoretically calculated. So in order to simulate a square wave, we multiply by three to take it out to the third harmonic (the point where 98% of the energy is transferred)


However, when we do these calculations based on a "video signal" they are too high. the reason for this is because video signals are not based on a square wave calculation. THere is a certain amount of bandwidth limiting so as to fit within the frequency allocations determined by the FCC and built into the various standards.


But if you want to determine what you need to get the signal CLEANLY from your computer to an image on screen, this gives you a very close theoretical starting point.


HOWEVER, there are many other factors included in all of this, like the fact that if you are using a VGA connector then it severely limits the high frequency response. Also, most graphic cards out there do not have the capability to deliver smooth response all the way to the high end we are talking about (as far as HTPC's go at least)


Most discussions about bandwidth cannot be truly entered until people are at a very high level of knowledge (i am kinda spewing back things i learned in classes previous to this) and allot of it is strictly theoretical. Of the three major "interfacing companies" all three of them have different calculations. Based on different resoning, logic, and engineering knowledge. Some do a calculation where the blanking and retrace timing is factored in, but those portions have no detail to them, so their importance in a theoretical calculation could be questioned. Either way, most of the calculations bring you somewhere around the same spot.


If you want to know how much bandwidth a component has, you need a scope and a GOOD pattern generator with a very clean rise time (under 2 ns) then you put the signal from the generator into the component and measure the rise time on the output side. then you just divide 360 into the rise time.


Rise time of 2.4 ns... 360/2.4= 150Mhz @-3dB


Of course, if you have a network analyzer, then you can get a true reading, but not everybody has one of those around.


So no matter how you calculate it, there will always be slightly different interpretations... It becomes a matter of which one sounds the best to a certain extent.

 

[BenjaminWoo]

Hi,


Check out Greg Roger's answer to my question in the Special Guest forum. The scan factor does indeed vary as Iceman explains.

 

[Dizzman]

Quote:

Originally posted by KennyG Perhaps this is why everyone sees such good detail when slowing the refresh rate down to 48hz?...A res of [email protected] uses the exact same amount of bandwidth.

BINGO... slow down the signal, give the display more time to resolve individual pixels, get a sharper image.


Quote:

Check out Greg Roger's answer to my question in the Special Guest forum. The scan factor does indeed vary as Iceman explains.

I never said it did not.


THe point of what Greg said is that pixel clock varies. Nowhere in there will you find a "fudge factor" In his calculations he uses EXACT timings from either SMPTE or VESA or whoever put the standard together.


so in his formula... (which is correct also for pixel clock)


XGA= 1344 x 806 (VESA total pixels and lines) 1024 x 768 (visible)


1344 x 806 x 60= 64995840

1024 x 768 x 60= 70778880


So in one formula we end up with a slightly higher number, that gets bigger as we get higher (around 7-10 %)


If somebody is debating the 5 MHz then they need another hobby. If somebody is putting a system together and they are using either of these numbers then they are more than fine. I just dislike the fudging. The numbers are clearly laid out.


THe difference is that one is a "digital" pixel clock.

One is an analog Bandwidth with a third harmonic factored in.


An important point that greg brings up is to put the alternating pattern up, if you see all the lines... then you are fine. That is what needs to be shot for. If you cannot see them all, the resolution might be fine... just lower the refresh and suddenly it might be OK.


When people ask things about "what res can ___ projector display" this is a big variable.


And for the record... this is my 520th (or so) post, and the first %*&$ time i ever figured out how to use the quote thingie.

 

[BenjaminWoo]

Hi Dizzman,


Thanks for clarifying that the difference is between the digital pixel clock vs analog bandwidth with third harmonic factored in.


Ben