ET course: low resolution vs high resolution
I wanted to let you know that I appreciated your presentation on June 22 in San Francisco and I also want to offer an alternative viewpoint to one of the ideas you stated about presentations. As a college professor at a primarily teaching institution within the California State system (CSU Monterey Bay) we have given thought to some of the ideas you went over.
Within your presentation and within your book, "Visual Explanations," you talk about the issue of resolution or amount of information that can be conveyed through various means of presentation. On one extreme is the human visual system and on the other the "dreaded" overhead projector and power point. Although your presentation just before lunch was very entertaining (and clearly very well rehearsed) what it seemed to me that you were saying is that low resolution means of conveying information yield "poor" presentations and high resolution means of conveying information yield "better" presentations. While I agree with many of the points you made to improve presentation, such as providing hard copy handouts, etc., I think you were labeling the overall quality of the presentation by the tools used (or mis-used) for the presentation, rather than by what the presenter chooses to do with the tools.
The examples you gave were the stereotypical qualities of a bad presentation, such as reading overhead slides which have several words on them, or leaving an overhead blank for a period of time, but are shortcoming of the presenter, not of the tools. Any tool for presentation has its pluses and minuses and can be misused, including high-resolution tools. A good example of "high bandwidth/resolution misuse" to me is the original "Numerical Storm" video you showed which was very poorly done from the standpoint of producing a meaningful meteorological account of what was going on, yet used a high level of resolution and showed a lot of information.
I think a better way to make your point would be to discuss the appropriateness of a certain presentation tool for the presentation. The Gettysburg address demo you showed is a humorous yet extreme example of how to misuse a tool for a certain type of material, but, it is unfair to say that it is because of power point. Someone without power point could perform in a similarly silly fashion as well. Do you think Lincoln would have done better to produce a high-resolution graphic illustrating his speech? I would guess not, but that the quality of his presence and his spoken word was sufficient to relay his message.
There are times when power point or overheads are appropriate for a certain type of presentation and hopefully the presenter is savvy enough to be able to fill in "between the lines" with the appropriate verbage, gestures, etc. and does better than simply read the slides. I would expect similar performance from the presenter for any type of presentation, whether it be "high resolution" or "low resolution."
As you said in your talk, people have made presentations and taught others for thousands of years before the advent of high technology, which is almost allowing us to do what others have done centuries ago. Let's put less focus on the applications and more on the quality of the presenter's style and the appropriateness of a means of presentation for a given situation.
Thanks for this chance to send some thoughts.
-- Dan Fernandez (email)
It is reasonable to suggest that there are other issues in presenting evidence than high resolution. Any other view would be astonishing.
But the overwhelming truth is that much of the history of progress in science and in communication can be written in terms of improvements in resolution. That is the big idea here, and it should not be compromised away by shortrun situational considerations.
-- Edward Tufte
I had a similar response to Dan regarding the discussion of information resolution, and of PowerPoint as a primary culprit for
many inappropriately chosen low-resolution information presentations.
I think PowerPoint may encourage people to present information badly, *but* it's quite possible to use it to present information well.
Vague bullet lists and useless animations are certainly out.
However, I've found that PowerPoint is an adequate tool for easily producing an information display that follows many of the principles in ET's course -- be it for a presentation, or an illustrative example of some engineering results to send out in an email discussion with a few images next to the text that describes what's going on in them, or even producing a paper handout. It's easy to combine image with word: Just make sure the words are clear, convincing prose, and the images have the right data put together in the right way. ET's example about Galileos "saturns rings look like this (insert hand-drawn image)" is easier to do in Powerpoint than in Word. And the handwriting is more legible if it's done on the computer than drawn by hand! Small multiples are easy, too (again, you still need the images that show the information in small multiples in the first place.)
Certainly you can use a better package to get better formatting that meets the most aesthetically discriminating person's sense of clean, crisp graphical design, but I do think PowerPoint is a simple route
that can be made to work, with proper thought, simply as a convenient
way to format all the information you want on the page/screen/etc.
-- Abbe Cohen (email)
More on low and high resolution presentations
But instead of projecting the images up on the wall, give everyone in the audience a piece of paper. Or many pieces. Paper is high-resolution, exact, portable, efficient, permanent; paper leaves traces, can be copied outside the meeting, says that you are serious and willing to be held accountable. Also people in your audience don't have to take notes; your story and evidence are all there on the handout. Reading aloud from projected bullet lists has none of these features. (There's more on this in an earlier answer on making presentations.)
If you go through all the books on how to use PowerPoint (consider that a sample of presumably skilled PP models and practices), you will find that 80% of the screens are really stupid. I'm making that compilation right now, and the practice, at least in these textbooks, is even worse than I had imagined. So far, there are no really good screens, a few competent screens, some harmless screens, many stupid screens. Fully 10% to 15% of the screens are wonderful bad examples (in the sense of the great book on bad poetry, The Stuffed Owl). Many of the screens seem like PP parodies; these textbook examples make Lincoln's Gettysburg Address in PowerPoint look rather profound!
Of course projected materials can be worthwhile, vital, successful, and very important in presentations. But the PP methodology and Wizard trappings, at least as often practiced in the corporate world and in PP textbooks, do not make for the serious analysis of evidence or for good teaching.
-- Edward Tufte
I provided an answer to an earlier question "How much information on one page," which I think is relevant to this discussion. As Dr. Tufte suggested in one of his reponses, I used a handout for people to actually look at and take notes on, while the PowerPoint slide from which the handout was made (and which was required for the presentation) was on the screen - but not very usable because of the density of the data I was conveying.
Shortly after taking Dr.Tufte's seminar in 1998 in Austin, I had to present a six project research program to the executive committee of an industry association. Prior to that seminar, I would have prepared a vugraph of each project, but instead, I prepared one handout with six small, identically formatted tables in two columns of three tables each. Each individual table had three columns, with the first and second columns having 7 rows. The first column contained the titles (project priority, duration, prior costs, current budget, etc.) and the second the related figures. The third column was one cell containing a short project description and its benefits, hence I was presenting 48 parameters about the research program, 8 for each project. I prepared the tables in MS Word using Times Roman 8 pitch and pasted them into a PowerPoint slide.
While I projected the slide, I used it only to identify to the audience (about 40 people) which project was under discussion, then referenced each small table on the handout of the slide during the discussion of each project.
By displaying the entire program on a one page handout and using a common format for each of the projects in the program, I was able to allow the audience to view the entire program at once and facilitated their ability to make comparisons among all the projects as they were discussed.
After 20 years of making corporate "5 bullet vugraph" presentations, this was the first time that people actually approached me after the meeting to compliment me on my presentation materials.
I have since tried to use handouts instead of slides whenever possible, and I have not hesitated to make "data-dense" vugraphs that required reference to a handout to understand the information being presented, when slides were required.
-- Jim Heimer (email)
New Page 1
web search produced the following limits of resolution:
Best LCD monitor
High quality printing 4800
to photography, professional digital backs commonly have over 25 megapixel
resolution. This is the information capture for the image — it then has to be
displayed. The resulting resolution defaults to the display method.
This brings up
two points. One, if we consider the digital image vs a photographic slide, where
the image is viewed directly from the film with no other intervening process,
the images should be indistinguishable if pixels were the size of the film
grain, which is 3 microns (120 millionths of an inch) for high resolution film,
and adjacent to each other. Current CCD chips have a 9 micron spacing between
pixels, so there's a significant technological development needed here. Note
that this cannot be compensated for by increasing CCD area as this would require
stronger optics and longer focal lengths that would degrade the image before it
even hit the sensor.
The second point is that film itself has a
visual quality, in part from the way it handles colors (combining primaries
serially rather than in parallel) and in part from the fact that film contains
more information, although beyond the resolution capabilities of the eye, not
beyond the discernation of the brain. The rule in digital sound is that, because
of the innate discontinuity of digital recording, sampling must be at twice the
detectable rate in order for the recording to be indistinguishable from natural
(unrecorded) sound. (And there are still aficionados that swear by analog over
192 kHz / 24 bit digital.) This multiplier may be more for visual recording.
This turned out to be much longer than
expected, but in short, although I would agree that high end digital image
quality is quite good and sufficient for most commercial applications, I believe
that the art of film photography will not yet be replaced by the art of digital
photography but will coexist for some time and will continue to have a place in
the recording of critical information.
As for monitors, we need one very large and
very far away.
-- Tom Hartmann (email)
Tom Hartmann has provided a very interesting and helpful answer. Nothing like relevant data.
Can a Kindly Contributor confirm Tom Hartmann's facts and add more elements to his list on resolution limits?
-- Edward Tufte
The issue to consider when looking at high and low resolution displays or printing is the theoretical angular resolution of the human eye. This is given by Dawe's Limit which is applicable to all optical systems. It is given as Angular Resolution in arc seconds = 115.8 / Aperture in millimetres (at visible light wavelengths). So for an eye with an aperture (pupil diameter) of 6mm it is about 20 arcseconds. Therefore if you are looking at a sheet of paper from about 450 mm (18 inches) the closest possible dots that you could perceive as distinct dots is about 0.04mm apart or about 600 dpi. Of course this is an inverse linear relationship so at 9 inches from the page it becomes 1200dpi and so on. This would mean that for a computer monitor 600 dpi would be about as high as you would want to go given normal viewing distances.
-- Andrew Nicholls (email)
Andrew Nicholls suggests that the eye has sufficient acuity to resolve dots to about 600
at normal monitor viewing distances. If so, we ideally would like to use a medium that
*exceeds* the resolution of the eye sufficiently that dots, and hence aliasing, are not
visible. Thus we'll want to be at at least twice that resolution, or 1200 dpi.
There is no question that 1200 dpi output is - at least in high-contrast, monochromatic
applications - easily distinguished from 600 dpi, by my average-good (20/20) eyes. When
I was still in my early 20s my vision was 20/10 20/15, and I expect that I might have been
able to distinguish 2400 dpi output from 1200 dpi.
There is also, of course, the question of pixel depth. Once we are dealing with
and color, resolution issues become far more complicated, both at the input side
(cameras, sensors) and at the output end.
Photographic film, per unit of sensor area, is capable of much, much higher linear
resolutions than ccds are. Yet ccds have all but replaced silver halide emulsions in
scientific imaging because they: (1) provide reliably quiet signals; (2) have the potential for
far, far greater intrascene dynamic range than film; (3) are more reproducible in their
responses (no variability in development and fixation); and (4) provide *extremely* linear
responses instead of the very nonlinear responses of photographic film. This makes
measurements much, much easier.
The bad news is that CCDs will not get a lot better in terms of pure performance.
Tom) They will get cheaper, but pixels cannot be made much smaller without seriously
degrading their dynamic range (larger range requires physically larger CCD "well") and
sensors demand higher lens performance, and high-performance lenses are neither
cheap nor easy to manufacture. So sensors have to get bigger to improve the image.
The newer CMOS sensors are much cheaper to manufacture, and offer performance
approaches CCDs. Canon has exploited this price/performance advantage to produce
larger sensors used in their pro-grade cameras. The best pro-grade cameras that use 35
mm lenses produce images that under normal field conditions equal to 35 mm film. They
are lower in linear resolution but roughly equal in image quality. Sports Illustrated shot the
Superbowl with digitals this year, producing images in up to 2-page size that looked
spectacular. Different from 35 mm but just as good. My guess is that these
cameras will get cheaper, but not drastically better. For that, larger sensors (medium
format-sized) will be required...
-- Alex Merz (email)
Back in the days of Linotronic type output (15 years ago), the type could be ordered at 1200 or 2400 dpi. The difference could be seen easily enough.
This is experienced today in printing images on inkjet printers, with dpi numbers going from 600 to 2400.
Some of the eyes looking at type and image are assisted of course by reading glasses that operate at magnifications up to 2 or 3. After getting my first reading glasses a few years ago, I was so happy at how much more I could see and wished that I had used reading glasses all my life.
-- Edward Tufte
Although Linotype output is indeed ~2400 dpi, this is not usually the "true resolution" of
the final printed output due to color dithering. That is, each CMYK film sheet from the Lino
is 2400 dpi, but they combine in overlay. Colors other than CMYK are simulated through a
halftone process, bringing the effective resolution to ~200-300 dpi. Black text is full
resolution because it is all black (K), and the fine curves in the characters take full
advantage of it.
Obviously, multi-color printing like ET uses in his books is at full resolution for each Lino
film sheet produced.
It is an interesting exercise to take a lupe to a decent hardcopy printout and measure the
drop sizes. I haven't done this in over 15 years, but my memory of earlier research is that
full-color printing via the CMYK process (which is the vast majority of printing) yields less
than 400 dpi in the absolute best case.
Here is a decent intro to color halftoning:
-- Michael Yacavone (email)
Does stochastic screening improve the resolution of CMYK? Or 6-color variants of CMYK?
I'm about to make some decisions about color reproductions in Beautiful Evidence,
probably via some experiments in printing but some good advice would also help.
-- Edward Tufte
A discussion of six colour printing is given towards the end of this article.
There is also a link 'Understanding halftones' on the left hand link bar.
It appears that six colour CMYK (Pantone Hexachrome) increases the range of colours that can be produced in print, not the resolution. This is set by a combination of the halftone process and the resolution of the printing presses.
Stochastic screening also does not increase resolution as such, it overcomes some of the problems with fixed halftone screens, such as moire effects.
-- Andrew Nicholls (email)
As Andrew says, 6-color will help the color resolution (depth) a lot - adding half-strength
C and M to the mix provides a wider gamut. My opinion is that stochastic screening looks
better for most images, though it doesn't change the resolution. I think if you can take the
time and expense of doing actual experiments it would be well worth it, and may even
provide material for a short paper.
(I have a friend who has written imaging software for color printing devices and I'll point
him here to see if he has anything to add.)
-- Michael Yacavone (email)
My first reading of stochastic printing or frequency modulation screening (saw a few
samples today) is that it at least benefits printers, since SP/FM makes it easier to
obtain and maintain registration compared with conventional color-dot screens. And SP may produce better images on average because images stay in register over long printing runs. But I was not impressed by the quality of the images in the particular samples the printer provided today, even though I like the theory behind SP/FM. The samples were high-end commercial work, however, not quite at our exotic exquisitely fussy standard.
A few weeks ago Yoshiki Waterhouse showed me some of his good work, and said that
color fields of light tones maintained a good smoothness over large areas thanks to SP/
Our printer rep did suggest that SP/FM allowed, in effect, 300 dpi screens for CMYK, which
would, if true and relevant, improve resolution. But also some gains in color resolution
have also apparently resulted from digital files direct-to-plate, avoiding the softening
effects (diffraction effects?) of light passing through layers of film in vacuum frames in
burning plates in the past.
Sometimes technological improvements in production are just that, not improvements in
Stochastic is, by the way, just a fancy word for random (at least in statistics)
Good color printing is profoundly multivariate issue, with many important variables
wandering around, not always in control. My strategy is WYSIWYG muddling through: get professional advice, do experiments with the printer printing my images, and see what works locally in the press-room, as the eye (under multiple types of light: sun, light bulb, halogen, 5000 degree Kelvin, florescent) ultimately decides.
In October, we'll be doing some proof press runs experimenting with the actual images to appear in Beautiful Evidence.
I did decide to go to a dull-coated paper (Xantour) for BE on the basis of earlier
comparisons in proof runs between the uncoated paper (Dulcet) used in the first 3 books
and dull-coated stock. That change makes a detectible different in image quality and
particularly in fields of light backgrounds (which have also proved difficult in the past, as
we endlessly fussed over differences between 2% Y and 3% Y). There is an opacity trade-off, however; Dulcet has excellent opacity (better than Superfine and Xantour, for
example). These experiments also indicated that Xantour was better than spot-varnish-prepared Dulcet in terms of overall image quality as well as cost. The spot-varnish did not completely seal the uncoated stock; and spot- and top-varnish made the images too glossy and too different from line-art and type (an important matter in a book that should exemplify image/text/diagram/table/number integration). Xantour is pH neutral, but there is another archival issue with coated stock: if the book pages get wet, the pages glue themselves together because of the clay and clay-glue that form the coating on the paper. Although a few copies might be irretrievably ruined if the coated stock gets very damp, the book will generally survive because so many copies are printed. (Uncoated stock presumably can be dried out with a hair dryer and the pages separated by hand, in what now has become a rather far-fetched scenario). We can't optimize on all relevant variables simultaneously, but at least we can still make certain improvements coupled with some trade-offs.
Finally from an earlier thread on ideas for monitoring business and other processes:
Measurement itself (and the apparent review of the numbers) can govern a process. For
example, in printing my books, I ask that during the press run that the density of the black
ink be measured in 6 or 8 different positions on every 3000th sheet being printed. These
pulled sheets are then inspected shortly after the run and before the next run. The idea is
to try to ensure that the color of the black type is uniform and at the right level of
blackness in 3 ways: (1) across the 8 pages printed up on each sheet of paper, called a
"form", (2) over the 40,000 sheets printed of that form, and (3) over the many forms
making up the entire book. We sometimes review these pulled sheets the next day to
check these density readings and to yell at the printer if there is a problem. But mainly the
mere fact that the printers are making these measurements keeps the process in control.
And the fact that someone might review the measurements.
-- Edward Tufte
| Threads relevant to Edward Tufte's work:|