Dear Hoover,
I would appreciate it if you could post the following summary of
replies to my question about DLP projectors, which was sent to CV-NET a
several weeks ago.
Thanks very much,
Curtis Baker
-------------------------------
original query about DLPs:
Does anyone know of an available DLP projector which provides the
option of linear gamma ?
Projectors based on "Digital Light Processing" (see
http://www.ti.com/dlp/main.html) exhibit several advantages over
conventional LCD projectors or CRT displays. An important potential
advantage, the z-linearity of the DLP chip itself, is lost in the commercial
implementation of most or perhaps all available DLP projectors - evidently a
gamma-nonlinearity ("degamma") has been added to provide compatibility
with other devices.
A collection of replies to this question will be made generally available.
Thanks,
Curtis Baker
send replies to:
curtis@astra.vision.mcgill.ca
------------------------------
The following is a compilation from a series of correspondences
with Yi-Xiong Zhou, The Salk Institute, California:
The advantage of the DLP technology to vision research was first
discovered by Lisa Croner in our lab. She did some initial tests to
show that 1) the luminance addition of the three guns was linear
and 2) the luminance of a stimulus was independent of the stimuli
presented elsewhere on the screen. She also did an initial survey on
the DLP technology, which convinced me of its potential
advantage over all the other existing ways of presenting visual
stimuli.
I recommend the following stimuli when you test a projector:
1) Put up a checkerboard with check size equal to 1 pixel. If the
dot clock frequency or the synchronization frequency of the
DLP is not correctly matched to the VGA signal, the checker-
board will not be displayed appropriately.
2) Put up a horizontal grating and a vertical grating side by side,
both gratings having a line width equal to 1 pixel. Both
gratings should be perceived as having the same mean
luminance and contrast. On our projector, the two contrasts
need to be balanced by adjusting the focus. By the design of the
DLP chips, the mean luminance of the two patches should be
matched exactly.
3) A continuous moving field of dots (immortal). When you track
the dots, you should not see the dots being doubled. This is the
problem that may be introduced by an asynchrony between the
spinning color wheel and the frame refresh.
4) A flickering grating; one frame on and one frame off. I have
seen failure producing this stimulus due to the wrong
connections between a Mac and the Proxima 4200. The problem
was corrected with a Mac to VGA adapter.
5) A black screen in a dark room. If you see a field of dynamic
noise, that is an indication of a problem. I have seen that on
a Number Nine SGT plus card with NNIOS driver.
I have resolved two problems with using the Proxima DLP 4200:
1) Setting the internal lookup table of the projector to be linear. The
procedure is described below.
2) Finding the video frequency range, within which the color
wheel inside of the projector is synchronized to the video frame
rate. The way to check the color wheel synchronization is to
use a photo diode and an oscilloscope to see exactly how the
micro-mirrors turn on/off with the frame rate and the color wheel.
For the Proxima DLP 4200, the synchronization range is 64Hz to
75 Hz video rate.
Unfortunately, Proxima 4200 is out of production. I am trying to
persuade Proxima to add a switch for setting the linear lookup
table on their projectors. Hope they will listen to me. The trick I
found to reset the lookup table may not be applicable to other
models of projector, and the lookup table in some models may not
even be reset, as the engineer in Proxima told me.
To linearize the 4200, you need to have a pc running kermit (from Columbia
U., free ware), or use hyper terminal from windows (actually, any
ascii terminal will be fine). Use a pc Null-modem cable (DB9 to DB9) to
connect the pc com1 port with the projector's auxilary port. Set kermit
speed to 19200 and start "connect'. Then turn on the power of the
projector. If there is only one button illuminated with red (orange) light,
press that button to start the projector. You should see some stuff showing
up on your kermit terminal. If not, something is wrong.
If you do get something, press the VOL+ and Menu buttons together. Your
kermit terminal should have:
DEBUG>
DEBUG>w dde_lg
dde_lg: 000=1
dde_lg: 001=
After you typed 1 at 000= and hit the return, you should see the
effect of changing the LUT on the projector screen.
-------------------------------
from: "Ian Holliday" <hollidie@pcmail.aston.ac.uk>
Organization: Aston University
Hi Curtis,
I don't know about the linearity question, but I looked at some of
these projectors last year with a view to their suitability for
vision experiments and found that there is a big problem with eye
movements, which are very apparent and distracting. I think it
arises because luminance is controlled by the pulse width of the
light from each pixel, and when the eye moves, for example in a
saccade, you see high contrast edges around the stimuli. Even worse
if the stimuli are also moving.
regards,
Ian Holliday
Dr. Ian E. Holliday
The Psychology Institute
Life and Health Science
Aston University
Aston St. Birmingham B4 7ET UK
tel. +44 121 359 3611 ext.4930
fax. +44 121 359 3257
--------------------------------------
> Hi Ian,
> Have not heard of the saccade-artifact. Which make/model of
> DLP was it ?
I have a brochure for LitePro 620 by InFocus systems, www.infocus.com
is your US contact.
>And, what kind of image gave the effect ?
Everything. The greatest disappointment was that it occurs even for
static images.
> I wonder if
> it is specific to sharp-edged and/or high-contrast stimuli, and might
> not be a problem for the typically moderate-contrast sinewave gratings
> or Gabors, etc, that we normally use.
Well maybe that's right. I think the demo I used was the rotating
vane pattern from the VSG demo set. We liked that as it was easy to
get going with different devices, and you can easiyou are
getting decent temporal resolution, and if the contrast etc. can be
fiddled with to give a reasonable impression of linearisation. With
the DLP it looked bad whatever you did. In the end we bought an LCD
projector at half the cost. One thing about these gadgets I didn't
realise ( well, the one we bought) is that although they are
'multisync' it seems they resample the incoming video, and stick it
out the front at a fixed normal video rate. Bummer.
If the DLP projectors were to run very fast they would be great I
think.
Nice and bright though, arn't they?
Ian
--------------------------------------
From: Steve Panish <scp@vrg.com>
To: curtis@astra.vision.mcgill.ca
Subject: DLP Z
Curtis,
As I understand it, TI supplies the "engine" of the DLP as a unit to
mfrs of the projectors. So if the non-linearity has been added by TI it
will be in all of them. That doesn't bother me too much, as it can be
undone by calibration. The 60 Hz effective refresh is the bummer for
me. If anyone knows if that has been changed, I'd be interested in
hearing about it.
-- ============================================= Steven C. Panish, Ph.D. Mailto:scp@vrg.com CEO, Vision Research Graphics, Inc. http://www.vrg.com (603) 868-2090 voice (603) 868-1352 fax----------------------------------------- To: scp@vrg.com Subject: Re: DLP Z
Hi Steve, As I understand it, the nonlinearity is not added by TI, but by the companies who design the projectors using the TI chip. I will try to verify this, though. I think some of the projectors can lock to a limited range of vertical rates which may extend somewhat above 60 Hz - probably quite model-dependent. Thank you for your interest.
Curtis --------------------------------
From: "James A. Ferwerda" <jaf@graphics.cornell.edu>
Texas Instruments used to offer a developer's kit for their DLP engine that might let you set/bypass the gamma correction for the device. Check out their DLP web page for more info.
-Jim
James A. Ferwerda Program of Computer Graphics 580 Rhodes Hall Cornell University Ithaca, NY 14853 607-255-7365 607-255-0806 (fax) jaf@graphics.cornell.edu http://www.graphics.cornell.edu/~jaf
------------------------------------------------------------
From: Rob Jacobs <r.jacobs@auckland.ac.nz>
Dear Curtis
I am not replying to your original question about gamma functions but as you have not specified what you will be using the displays for I thought the following information may be usefu,l even if only peripheral to the query.
I looked at DLP projectors as a potential source of bright stimuli for VERIS (topographical ERG) measurements.
Yes DLP projectors are available that will reproduce a 67 - 72 Hz frame rate faithfully (not all will)
Unfortunately for what I wanted to do, the luminance wave form of DLPs proves to be totally different from that of a CRT based display. The luminance profile for each frame of the DLP is a square wave (the better ones have a sharper onset and offset - examine with a photodiode and an oscilloscope) while the luminance profile for each frame of a conventional CRT is an intitial superbright spike with a subsequent decay. This means that the retinal stimulus from the two display types has a totally different time course per frame.
kind regards
Rob Jacobs Optometry and Vision Science The University of Auckland Auckland, New Zealand
-----------------------------------
PS, from Curtis Baker:
Following up on the information from Yi-Xiong Zhou, my colleagues and I at McGill arranged an opportunity to test a Proxima 4200 DLP projector (while no longer manufactured, demo and used models can still be obtained). We successfully used a PC running Kermit, connected to the projector's serial port, to implement the procedure provided by Yi-Xiong to disable the degamma. All three color video signals were then truly, impressively linear.
However, we decided not to purchase the unit, primarily due to distortions it caused to smooth-edged stimuli (such as drifting sinewave gratings) when they move at higher velocities. An engineer at Proxima indicated this kind of problem would probably be generic to nearly all DLP projectors, due to the temporal sampling introduced in conjunction with syncing to the sectored color wheel (a possible exception might be those which devote separate DMDs to each of the colors, but those are very much more expensive).
At the present time, DLP projectors appear to be very attractive for spatial vision experiments, in which there will be no eye movements (we also noted the saccade-related artifacts mentioned by Ian Holliday). We have not had any indication of the existence of DLP projectors other than the Proxima 4200, which provide disabling the degamma. Also, we have not confirmed that the artifact for rapidly moving, soft-edge stimuli is generic to all DLPs, or a peculiarity of this model.
For further information about DLP technology, a good place to start is:
ead="start" -->