Prisjakt
Sökningen pågår
Avancerad sökning
6 bilder
En artefakt med alla DLP-projektorer som inte nämns så ofta är en typ av brus som syns vid panoreringar på skärmen. Jag hittade lite intressanta trådar på avsforum som diskuterar detta:
"Has Single Chip DLP Reached Its Limit?"
http://www.avsforum....threadid=202617
-----
In a review at Projector Central, I found that the Sharp 9000 (I'm not picking on Sharp here - this is what I found in a quick search) has a 5x, 6 segment color wheel. I'm assuming the 1x base case is a 3 segment wheel at the 60 Hz video rate - please let me know if I'm wrong. So the Sharp is actually 10x. To be conservative, I'll use 5x. For 24 bit video, R, G and B must each have 8 bit resolution or 256 discrete steps. Since a DMD is a digital device, it uses pulse width and/or duty cycle modulation to get these 256 steps. 0 is black and 1 is the next step above black or 1/256 of the 256 wide period required for each color. In a single chip PJ, the colors are presented sequentially as the color wheel turns.
Some quick math using 5x and 60 Hz: 1/(256*3*60*5) = 4.3 microseconds (us). Ideally, using the stated assumptions, a mirror should be able to instantly flip from off to on, stay on for 4.3 us and then instantly flip back to off to accurately present the first step up from black in the gray scale. BTW, for a 3 chip DLP PJ at 60 Hz and 8 bits per color, 1/(256*60) = 65 us and for 3 chip digital cinema at 24 fps and 10 bits per color, 1/(1024*24) = 41us.
On TI's dlp.com website, I found a white paper by Larry Hornbeck called "Current Status and Future Applications for DMD-Based Projection Displays" dated 9/30/1998. This paper discusses the older +/- 10 degree swing mirrors and gives a "…mechanical switching time of ~15 us and an optical switching time of ~2 us." These times are not further defined - I assume mechanical switching is accelerating the mirror from rest at -10 degrees, swinging from -10 to +10 degrees and decelerating and settling the mirror at +10 degrees and optical switching time is the actual swing time. If my assumed definitions are right, there is a problem with single chip - the mirrors cannot flip fast enough and the problem will be most obvious just above black where the required pulse widths are the narrowest. Even if I’m wrong, the calculations above give a relative idea of the heavy burden on a DMD in a single chip PJ. I did not find switching times for the latest devices.
-----
Som beskrivs ovan så hinner helt enkelt inte varje spegel med att byta färg varje frame. Att 3-chips DLP ser så bra ut på biografer är just att det inte är samma krav på uppdateringsfrekvens. I samma tråd så beskrivs även hur man löst problemet med tiden det tar för varje spegel att gå till rätt position varje frame:
-----
The full colour depth is not presented in each segment, but instead it's spread out (dithered) over time. The result is mosquito noise in the image and a drop in colour depth (banding) during pans and for moving objects. I've noticed these artifacts in all single chip DLP's I've demoed, so I think the technology is already being pushed past its limits.
-----
Mer information om hur man fördelar färgdjupet över bilden finns beskrivet här:
"How DLP works: PR vs Reality"
http://www.avsforum....threadid=147363
Ni som har DLP hemma idag (även ni med HT1000 och Infocus7200) kan lätt se de här problemen genom att sätta på en mörk bild på er DVD-spelare och pausa bilden. Gå nära duken och titta så ser ni konstant "mosquito-noise" i bilden som flimrar. Orsak till detta?
-----
Actually, I think the low IRE DLP resolution problem is much worse than was calculated above. The 256 counts are pretty nearly in uniform perceptual space. To get into energy space, you need to transform it by gamma. For the sake of brevity, let's assume gamma is 2 (close enough) then energy = IRE^2 (close enough once you get past the initial linear relationship). The 10 IRE is 1% of the energy of 100 IRE. 1 IRE is less than 1/10 that, making it over 4x worse than previously calculated. That's why dithering is required.
Speaking of dithering, put a grey scale ramp on a properly tuned DLP and take a real close look. I've done that with units up to the HD1 generation. As Dave pointed out, you will see some pretty glaring artifacts. From regular viewing distance, most people probably wouldn't notice, but it does say something about the technology.
-----
Så... Även om TI nu lyckas få till sitt Arkimedes-färghjul för att motverka regnbågarna så finns fortfarande problemet med den fysiska tiden det tar för speglarna att ställa in rätt färg.
"Has Single Chip DLP Reached Its Limit?"
http://www.avsforum....threadid=202617
-----
In a review at Projector Central, I found that the Sharp 9000 (I'm not picking on Sharp here - this is what I found in a quick search) has a 5x, 6 segment color wheel. I'm assuming the 1x base case is a 3 segment wheel at the 60 Hz video rate - please let me know if I'm wrong. So the Sharp is actually 10x. To be conservative, I'll use 5x. For 24 bit video, R, G and B must each have 8 bit resolution or 256 discrete steps. Since a DMD is a digital device, it uses pulse width and/or duty cycle modulation to get these 256 steps. 0 is black and 1 is the next step above black or 1/256 of the 256 wide period required for each color. In a single chip PJ, the colors are presented sequentially as the color wheel turns.
Some quick math using 5x and 60 Hz: 1/(256*3*60*5) = 4.3 microseconds (us). Ideally, using the stated assumptions, a mirror should be able to instantly flip from off to on, stay on for 4.3 us and then instantly flip back to off to accurately present the first step up from black in the gray scale. BTW, for a 3 chip DLP PJ at 60 Hz and 8 bits per color, 1/(256*60) = 65 us and for 3 chip digital cinema at 24 fps and 10 bits per color, 1/(1024*24) = 41us.
On TI's dlp.com website, I found a white paper by Larry Hornbeck called "Current Status and Future Applications for DMD-Based Projection Displays" dated 9/30/1998. This paper discusses the older +/- 10 degree swing mirrors and gives a "…mechanical switching time of ~15 us and an optical switching time of ~2 us." These times are not further defined - I assume mechanical switching is accelerating the mirror from rest at -10 degrees, swinging from -10 to +10 degrees and decelerating and settling the mirror at +10 degrees and optical switching time is the actual swing time. If my assumed definitions are right, there is a problem with single chip - the mirrors cannot flip fast enough and the problem will be most obvious just above black where the required pulse widths are the narrowest. Even if I’m wrong, the calculations above give a relative idea of the heavy burden on a DMD in a single chip PJ. I did not find switching times for the latest devices.
-----
Som beskrivs ovan så hinner helt enkelt inte varje spegel med att byta färg varje frame. Att 3-chips DLP ser så bra ut på biografer är just att det inte är samma krav på uppdateringsfrekvens. I samma tråd så beskrivs även hur man löst problemet med tiden det tar för varje spegel att gå till rätt position varje frame:
-----
The full colour depth is not presented in each segment, but instead it's spread out (dithered) over time. The result is mosquito noise in the image and a drop in colour depth (banding) during pans and for moving objects. I've noticed these artifacts in all single chip DLP's I've demoed, so I think the technology is already being pushed past its limits.
-----
Mer information om hur man fördelar färgdjupet över bilden finns beskrivet här:
"How DLP works: PR vs Reality"
http://www.avsforum....threadid=147363
Ni som har DLP hemma idag (även ni med HT1000 och Infocus7200) kan lätt se de här problemen genom att sätta på en mörk bild på er DVD-spelare och pausa bilden. Gå nära duken och titta så ser ni konstant "mosquito-noise" i bilden som flimrar. Orsak till detta?
-----
Actually, I think the low IRE DLP resolution problem is much worse than was calculated above. The 256 counts are pretty nearly in uniform perceptual space. To get into energy space, you need to transform it by gamma. For the sake of brevity, let's assume gamma is 2 (close enough) then energy = IRE^2 (close enough once you get past the initial linear relationship). The 10 IRE is 1% of the energy of 100 IRE. 1 IRE is less than 1/10 that, making it over 4x worse than previously calculated. That's why dithering is required.
Speaking of dithering, put a grey scale ramp on a properly tuned DLP and take a real close look. I've done that with units up to the HD1 generation. As Dave pointed out, you will see some pretty glaring artifacts. From regular viewing distance, most people probably wouldn't notice, but it does say something about the technology.
-----
Så... Även om TI nu lyckas få till sitt Arkimedes-färghjul för att motverka regnbågarna så finns fortfarande problemet med den fysiska tiden det tar för speglarna att ställa in rätt färg.
Minhembio - Senaste nyheterna
nimman 
Fler
Vilka bilder visas här?
