The Stratograph

A kinetic sculpture, and component of the Science Museum of Minnesota's "Playing With Time" traveling exhibit.

-- Concept:  at each venue, the piece creates a "core," a tube filled with colored sand which is slowly deposited over a 3 month period.  The sand patterns depict both text and graphics relevant to the venue locale, as well as an actual record in time of a physical phenomenon occurring within the Stratograph's immediate environment.

May, 2002
Everything was going well until 5/18, when apparently a fairly tall visitor decided to vandalize the sand valve assembly by pulling the spring mount off the black sand valve.  This resulted in the dumping of black sand  until the tube was full.  I was informed two days later that the funnel was "buried under sand , and the machine was not functioning."  Although rather annoying, the remedy was simple:  I moved the top assembly, vacuumed the sand above the funnel, thereby freeing it.  After bringing the funnel up and removing it, I could then vacuum down to the undisturbed level (in Hubert's forehead).  This done, I sent the funnel back down, replaced the head, and loaded up the remaining portion of the plot (the Minnesota state map).
April, 2002
_23 days since the opening, and all is well.  Visitors appear to be genuinely engaged, and despite the lack of finished signage, many seem to be "getting" it.  The only unexpected finding has been that the archived tube has settled an additional ~1/2"-- undoubtedly due to visitors moving it back and forth.  This raises the need for a threaded plug on the top wood block, which will allow periodic "topping off" with black sand.
 

March, 2002:
Actualization time:
Steel bases and connecting structure cut.
Calibrations.
Opening day.
 

February, 2002:
2/1
- The final design of the Stratograph is taking shape, as fabrication begins with a trip to Youngblood lumber, and the choosing of ~100 board-feet of Honduras mahogany.  After gluing up the active chamber's side panels, it was time to refine some tool paths, and fire up the CNC router-- crucial for the tight fit needed at the core tubes' ends, as well as the structural integrity of the sand delivery assembly.  Mark Balboa, of Balboa Plastics, in Richfield, was extremely helpful with the acrylic fabrication-- selling me the small quantity of 1/4" sheet I needed, and lending me a table saw blade optimized for acrylic.

1/11
Mona's almost complete (laptop screen shot).  She's a bit fuzzy, but still recognizable-- especially from afar.  This suggests that faces will need to fill at least 150x 150 or so pixels to be recognizable.  The motion sensor recording continues to function well.

1/9
Front:                                                           Back:
CAD drawing

The final form of the piece is taking shape.

Features of this design:

The CAD drawing lists pertinent dimensions.  The renderings are intended to give a feel for the overall appearance of the completed work, but to save time I did not create different maps for the archived tubes, or try to finesse the appearance of the wood grain.  Despite the light wood materials used throughout the exhibit, I have chosen a dark hardwood (Honduras mahogany) for a variety of reasons-- both practical and aesthetic.

1/8

Since changing back to the 3M aggregate on 12/29, the Stratograph has been steadily dropping sand at a rate of every ~7 seconds-- the actual rate which will fill a core tube in 3 months time.  For the most part, its behavior has been what I expected-- nightly black bands, interlaced with daily activity records.  Close inspection of the record reveals some interesting findings-- one of which is that the occasional blue bands seen during the middle of the night are likely due to a late night cleaning crew.  One glitch which I believe is related to a timing routine error on my part (since corrected) resulted in the complete absence of activity recording during New Year's Day.  No problems since the fix.  Today, the Stratograph is about midway through a plot of the Mona Lisa (on the flip side of the activity record) -- expected to be complete on 1/12.

Dec. 28
The prototype is up and running at the Museum, as of this morning.  Using the protective display case that Dan was able to dig up, it is nestled under the staircase on level 3, facing the main thoroughfare to "the bones" (paleo).  I met with Dan, Randy, Russ and Richard and explained the key features of operation.  To begin with, we will use a small test image (testpat3.r8, which contains the Museum logo), and spread it over 5 days.  This is approximately the same rate that would fill the entire tube in 3 months.  The Museum is packed with holiday visitors today, and I anticipate a fair amount of red sand.  The PIR sensor is currently located about 8' up, and looks down and in front of the case.

Dec.  26-27
Continued testing of the PIR sensor and new software changes.  Also, samples of colored sand from Dick Blick's web store have arrived.  They carry a wide variety of brightly colored sands, and the price is not dramatically more than the 3M aggregate.  It is, however, much finer-- which may add to the resolution capabilities, but definitely makes it more difficult to actually see the sand come out of the spout.
In order for me to transport the Stratograph to the Museum, without spilling the first tube's sand, I have fashioned a plunger.  It appears to work just fine.

Dec. 19-21
_ Focus has turned toward a major overhaul of the Stratograph software.  I've run into several annoying bugs-- from simple oversights, to subtle problems dealing with numerous timing routines.  As always, the sand records my progress (or lack thereof).  My first attempt at mapping motion detector data to colors turned out to be too insensitive (mostly black).  The scheme I am now using (strato4.exe) produces more color deposition.  It is difficult to mimic random visitor motion.  I tried locking my dogs in the room when I was out, but it appears that they both took  naps (lots of blue and black). Close inspection does reveal a correlation between inactivity and night-- in this case, the nocturnal black band is about 4x thicker than would be expected in an actual 3 month plot, since I am running at about 4x speed).  As expected, text and graphic quality are far better when reproducing horizontal lines than vertical.
Tentative plan is to set up the prototype at the Museum next week.

  Dec. 13-14
After destroying two $7 security lights from Home Depot without much success, I was able to hack into the circuitry of a PIR security module being sold at AxMan for $10.  Two of the three PIR devices I ordered from MPJA arrived on 12/12-- far easier to deal with!  I spent most of yesterday writing a program that will allow testing and logging of the "hit rates" form three sensors simultaneously.  I had two sensors hooked up (AxMan's, and MPJA-1) by 10pm and started recording.  The "triple chart" shows the motion detection hit rates for these two ( the top chart is me messing with the unconnected input wire for sensor #3).  As expected, both sensors remained unactivated over night, until this morning.  The vertical scaling numbers refer to the number of times per minute that the sensor was activated (the software checks the status every second).  Each PIR module has different latency times-- hence, although 60 hits / minute is the absolute max-- far lower numbers are expected, even when there is relatively constant motion in the sensing area.  Each vertical bar (one pixel wide) represents 1 minute.

Dec. 10
Finalizing design of top assembly.  CAD layout.  Also, finalizing electronics:  schematic, PCB layout, circuit board under construction.
Awaiting PIR sensor from MPJA-- should be here tomorrow.

Dec. 5
Conference call with Robert, Katrina, Bob and myself-- "Plan A" (multiple buttons corresponding to sand colors as input) vs. "plan B" (single button or sensor, sand color related to "hit rate") discussed.  Plan B favored.  Sensor (passive detection) vs. active input-- passive favored.  Will proceed with current plans to use 24 hr/day, metronomic sand drops.

Have begun to experiment with PIR sensors.  Bought a  $7 security light at Home Depot, and hacked the sensor board.  Have ordered a variety of PIR modules from Marlin P. Jones, but the bare sensor is on back order.  Also, have found a couple articles on hacking this sensor -- Jack Dollhausen's; Seattle Robotics'

Dec. 3
  Rendered tube front, back, animation (682K mov)

After much thought, have decided to try going back to an interactive approach for the "data driven" aspect.  Key reasons:

• Use of archived temperature data is not as "natural" as recording phenomena from local environment.
• Aesthetic promise of temperature data questionable.
• lack of interaction at night will create recognizable, and visually pleasing, bands representing a 24 hour period.

 

Visualization images above, while crude, give at least some idea of what the filled tube will look like.  In this case, interactive sand pixel colors were chosen randomly, with random repeats.  The palette is my best approximation of the five non-black/white 3M sands that I have: .

The visualizations do not yet show the "queue," the sand valves and storage on top, or the interactive input buttons.  Also, to be added are the archived tubes from prior venues on display.

Assuming: Tube length of 72", 100 day duration, 24 hr/day operation, similar sandpixel size to what I've been doing (12ms open time)--
Time between pixels = 6.7 sec.
3/4" / day
13000 pixels / day
400 pixels per rotation (row)
32 rows / day
1.3 megapixels / venue -- (entire record would fit nicely on a floppy)

Nov. 20
An example of Bob's early experiments with visualizing data (in this case, temp and daylight hours).  Also, a rendering of the pattern mapped upon a cylinder.

Nov. 19
Met with Robert, Dan and Katrina at my place, while stratograph continued to drop sand.  Salient points:

• Robert raised the importance of a "data-driven" aspect to the piece as a direct link to the idea that there is a direct connection to time.  We all agreed. One potential strategy would be to use half the tube for this, and half for text or bitmap information.
• One potential source of data could be weather-- temp and precip.  We could use almanac data (Katrina), to avoid real time data entry compliance issues.
• We'll try to test various strategies using the visualizing tools Bob is developing.
• The favored barrier system for the final version has shifted from a railing system to acrylic / glass plates in front and behind that will allow viewers a closer look.
• Dan would like to get the prototype out on the floor as soon as possible.  I agree-- but want to make sure we can test pattern ideas quickly.  I am considering the  possibility of building a second machine.

Nov. 13
Spurred on by the successes with simple text, I turned to the next level of complexity.  Choosing simple images of my daughter's face and my own, I attempted to do them in sand.  Unfortunately, I had made a small (but dumb) error in my program, which caused a slight shift in alignment with each row.  I discovered the error upon awaking (11/11) and finding what I hoped would be faithful renditions, smeared to the left.  After correcting this bug, I began again.  Now 28 hours into what will be a 41 hour plot, the results are telling.
Nov. 11
A little more ambitious-- tried doubling to 192 pixels/row, and more difficult text.  Not bad (3 hours, 25 min. to complete, settings: valve open 30 msec, closed 1500 msec). To facilitate this evaluation, have moved programming platform to C (stratro1.exe). Can now load bitmaps, adjust solenoid timings, and pause plot to adjust parameters.
Nov. 10
Having a reliable sandlevel-following apparatus, I returned focus upon feasibility of reproducing bitmap images.  Starting with 96 sanddump pixels per revolution, I tried my first test image-- my illustrious initials, which turned out upside down and backwards.  Naturally.

Nov. 9
While the initial sand sensor circuitry appeared to work well, as night fell, it became clear that ambient light conditions had a marked effect.  This appears due to the relatively weak IR signal coming from the LED.  To increase the "signal to noise ratio," I decided to abandon the LED in favor of a visible laser diode (using a hacked keychain laser pointer).  Substituting the pointer's three watch batteries with my 5 volt supply, knocked down to ~3.2V via diodes, it works just fine.  The tight light spot raises the potential for misalignment, so I spread the beam using a hemispherical plastic lens.  Initial tests (over night) worked perfectly. New schematic.

Nov. 8
Focused upon instituting sand sensing.  Used IR LED and phototransisitor similar to Pipedream II.  Had to add amplification in order to ensure enough "distance" between light and dark.  Used empiric method to discern optimal resistor values.  Schematic.

Using new feedback capability, returned to pursuit of max res.  Soon observed the first clog. Closer inspection reveals a tiny piece of paper (from sand bag) lodged in orifice of nozzle.  This raises the obvious need for a filter (screen / mesh) in funnel.

Nov. 7
Initial trials to ascertain max resolution.  Early results are encouraging-- able to discern 96 distinct "pixels" around circumference.

Nov. 6

The first attempt at creating a decent seal between the sand director funnel and the inside wall of the tube using felt was clearly unacceptable.  This new design is far better.  Also, have started narrowing the director's output spout, in preparation to evaluate maximum resolution.

Oct. 29

Demonstrated the full-scale prototype for Dan, Robert, Katrina, Eugene and met Robert Hone of Red Hill Studios.  Discussed a multitude of ideas, including possible strategies for interactivity.  Notable ideas:  "sand pixel" color as indicator of sensor activation rate / splitting column of sand into front and back, with different pattern strategies.

Plan:  I will investigate the resolution limit of the machine, in hopes of being able to faithfully recreated bitmap images in the sand. Robert H. to begin experiments with visualizing color patterns in a virtual environment.  Tentative meeting to show results in 3 weeks (prior to Thanksgiving).

The acrylic tube was purchased  from Crown Plastics, and came in at $114.  This is cast (vs. extruded), and although significantly more expensive, is gorgeous.  The lead screw is overkill-- but it's the right length, and looks good.  The Z-axis stepper and the Theta-axis steppers took a while to mount-- theta has an 18:1 gearhead which is also overkill, but since the tube will weigh in ~100lbs when full, why not. (view of the motor and drive layout on the underside)
 
 
 
 
 
 
 
 
 
 

To accomplish the capping off of the tube's ends, I enlisted the use of my CNC router.  Using a 1/8" end mill, and a 3-15/16" diameter circular path, resulted in a perfect fit.
 
 
 
 
 

After testing the two axes, and mounting the top tube holder / lazy susan, turned attention to the magnetically coupled sand director.  After toying with a number of strategies, I decided to follow the simplest:  a funnel, a 45 degree copper coupling, and a handful of nuts and bolts.  Plus two rare-earth mags.   While a number of improvements will be made, this works reasonably well!
 
 
 
 
 
 
 
 
 
 

Because of time constraints (I leave for NYC on 10/24),  I have stolen the maquette's sand delivery system for testing of the full size proto.  There are some sand leaks around the funnel, but the initial tests look great.
 
 
 
 
 
 

September 12,  2001
Met with Dan, Katrina, Robert and Bart.  Prelim design presented, along with weights and estimated costs of various tube lengths and diameters.  6' length and 4" diameter agreed upon for first full scale testing.  Full scale, functional prototype to be accomplished by end of October.