I set the computer up to run a calculation for every 30 minutes, on thirty day intervals throughout the year, for CIE Cloudy, Partly Cloudy, and Clear skies.   That made for some 600 radiosity calculations in all, so after queuing all that up, I let my desktop run for about two weeks straight.

Untold processor cycles later, I had hundreds of text files that looked like this:

So this was not a terribly useful format for the data to be in what I wanted to do, because I was interested in the behavior of the average and minima of, for example, the “Books Adult Stack 2″ fields as they span all the files that I have, not the value of any one particular field.   Also, the files were named things like sample_3.txt and my computer crashed several times during all this so I had several files with the same name etc. etc. etc. agggghhhh.

Anyway, after trying any number of other things, I wrote some bash scripts to collate and assemble the data for me.  I’m told that future releases of AGI will have more sophisticated data output options, but in hopes they come in handy for someone else, here’s the script I wrote:

#!/bin/bash

echo "Usage: sh extract.sh filename.txt (or *.txt)"

echo ""

echo "Script does two things: creates csv files from a tab delimited input with the pertinent fields in the file names, and collates those parameters along with other data to a central csv file called extract.csv"

echo ""



for filenam in "$@"

do



    #extract the values of the Conditions, Date, and Time fields

    # to a string, with an underbar delimiter

    timestring=$(awk 'BEGIN{ FS="\t"; RS="\r\n"}

        /Conditions/ {print $2"_"}

        /Date/ {print $2"_"}

        /Time/ {print $2}

        END {}' $filenam)



#put the fields you want to match in the /.../, and the fields you want

#returned after the print command

    datastring=$(awk 'BEGIN{ FS="\t"; RS="\r\n"}

        /Main Room RHS/ {printf "_"$1"_"$3"_"$4"_"$5}

        /Floor/ {printf "_"$1"_"$3"_"$4"_"$5}

        /Books Adult Stack 2/ {printf "_"$1"_"$3"_"$4"_"$5}

        /Check in 1_Surface_5/ {printf "_"$1"_"$3"_"$4"_"$5}

        /Computer Area/ {printf "_"$1"_"$3"_"$4"_"$5}

        /Adult Reading Workplane/ {printf "_"$1"_"$3"_"$4"_"$5}

        /Childrens reading area/ {printf "_"$1"_"$3"_"$4"_"$5}

        /Adult Reading Workplane/ {printf "_"$1"_"$3"_"$4"_"$5}

        /Entry Vestibule/ {printf "_"$1"_"$3"_"$4"_"$5}

        /Adult Stack 1/ {printf "_"$1"_"$3"_"$4"_"$5}

        END {}' $filenam)



    #read it out to the terminal, just for confirmation

    echo $timestring



    filenamappend=$(echo $timestring  | tr ":" ".")

    echo $filenamappend



    #extract the base name without extension for concatenation

    dfile=`basename $filenam`

    dfile=${dfile%.*t}



    #create a .txt file with the conditions in the filename

    tr "\\t" "," < $filenam > ${dfile}"_"${filenamappend}.csv



    #send it to one csv file, changing the delimiter to commas

    echo $timestring$datastring | tr "_" "," >> extract.csv



done

Also, so that I could make pretty graphs.  Like this one here:

Graph Porn.

This portrays average footcandle levels on the vertical surface of a section of stacks in adult section.  The levels are very uniform throughout the day (x axis) and year (y axis), except during the winter months, where for a few hours each day there is a large spike in the values.  Since those values are an order of magnitude larger than anything else, they must be caused by direct sunlight.  Since they are occurring in the winter months, I would guess that they are due to sunlight through the windows, rather than the skylights above.  A quick peek at the full-color rendering confirms this.

That behavior would have been discovered if daylighting analysis was done for the ‘worst-case’ summer/winter solstices, but not if only data for the ‘typical’ equinoxes were generated.  It’s important to look at those outliers, is what this graph is telling us.

Rendering of direct sun at low winter angles.

Variability of sky conditions:  Some daylighting programs can take as an input a weather file containing average sky luminances based on weather station measurements.  This is the preferred method for daylighting analysis, but day-by-day sky conditions can vary from cloudless to totally overcast on any day of the year, particularly in Berkeley where this library is to be located.

A question I had was how much numerical variation there is in interior lighting levels between a sunny day and a cloudy day, just as a practical rule-of-thumb.  An average sky condition might be less meaningful if there’s more than an order of magnitude difference between lighting levels on a sunny and cloudy day, since we are concerned about the occupant experience on atypical days as well.

Variation in lighting levels between sunny and overcast sky conditions

This may be somewhat difficult to visualize, but we are interested in the ratio of average lighting levels between clear sky days and cloudy days, i.e. the space over which weather would affect measured lighting levels.  I found a fairly insignificant ratio of up around 2:1 or 3:1 during times when there is no direct sun in the space, and a fairly significant ratio of 5:1 up to 11:1 when direct sun enters the space.  For example, the average lighting levels mid-day in the winter varied from 42fc for overcast days up to 470fc for sunny days (!).  A close-up of that data:

Average daily lighting levels under several sky conditions on the winter solstice.

The take-away would seem to be that unless there is penetration by direct sun into the space, a typical climate can be used to approximate any weather condition.  Notably, the greatest variation in lighting levels due to sky condition was at mid-day points where there was more than enough light in the space in any case.

Cost vs. benefits of dimming:  One question that came up in the design process was the cost/benefit of dimming vs. switched daylight harvesting.  For example, for the computer use area in the main room, we might desire a minimum of 30fc on the task plane 30″ above the floor, as this is sufficient to do the sort of paper tasks that might be performed at this location.   The electric lighting could be controlled by a switched relay, in which case it’s either on or off, or a dimmer, in which case it could provide some of the light but not all of it for cases where there is some daylight but not enough to read.

The thing is, there can be a considerable cost adder for dimming control, particularly for fluorescent lighting.  So the question of whether dimming controls are worthwhile can be stated: for how many hours per year is there some daylight, but less than our target 30fc minimum?

Lighting levels throughout the year, with values > 0fc but < 30fc in orange

In the above graph, time of day is plotted vs. time of year, with contour lines for footcandle levels.  The area where dimming would save energy over switched daylight harvesting is shaded in orange, and it works out to about an hour in the morning and an hour at night.  However, the expected operating hours of the library are expected to be 10am to 6pm most days, so it really only comes into play in the winter.  There are some additional potential benefits to dimming controls as well, but on the basis of this I recommended that we do dimming controls on the LED fixtures where it could be had easily and inexpensively, and forgo them for most of the fluorescent fixtures.

Forecasting total power consumption:  Lastly, I was wanting to get a handle on the real energy usage over the course of the year, as opposed to the total connected load.  AGI unfortunately can’t take real climate data for the area as an input, so I used the partly cloudy data set as a rough-and-ready average value.

For each of the twenty or so calculation grids within the space, I identified a target illuminance level for each and totaled up the number of hours each year where daylight contribution was less than acceptable.  I used the current operating hours of 10am-6pm four days a week, 12am-8pm two days a week, and closed on Sundays (plus an extra half hour after closing for cleanup).

Hours of Electric Lighting per Year

The lighting energy draw of the main room is about 4000W with everything on, so the Real Lighting Power Density is something like 47.2 KWh/yr.  I’m not using a particularly granular or methodical approach here, because in any case the end result is a vanishingly small amount of yearly energy use due to electric lighting.  To give some context, according to DOE figures, the office space averages about 6 KWh per year for lighting, per foot.  The energy savings to be had by good daylighting design are clearly orders of magnitude greater than the energy savings due to efficient electric lighting design.  That’s something to contemplate at a time when code-mandated electric lighting design power densities have reached a point of diminishing returns.

Conclusion:  On a typical project, we don’t have time to run a comprehensive study like this, of course.  I think that looking at this data, the methodology of generating data points for a morning and afternoon time at the equinox, with bounding points at the winter and summer solstices is validated.  I would suggest that for the solstices clear sky conditions be run regardless of the prevailing sky conditions, to generate ‘worst case’ scenarios.  One factor that would make this study less general is that the space is primarily lit from above, with skylights, and the angles are such that there is almost never direct sunlight entering the space.  For spaces with larger windows, I would expect sun angle and sky condition to produce much more variation in lighting levels.  On the other hand, adjustable shading technology is much more available for windows than skylights.

Hope you enjoyed this, it took me months to find the time to generate and assemble and analyze the data.  If you have any questions or comments, please feel free to get in touch or drop me a line below!

I found some planks that had been discarded on O’Farrell street (picture here), and decided to make a cutting board from them.  I’ve only got about 2′ x 3′ of counter area in my comically tiny San Francisco apartment anyway, so I’m essentially replacing all of the counter space.  That suits me because I feel like cutting boards are kind of a sub-optimal solution, in that you’d ideally want the entire workspace to be a cutting area.

My original idea for this was that I was going to stain specific pieces in a semi-random pattern as in this pre-visualization rendering I did:

But that ended up being impractical, first because you would have to stain the pieces before assembling, and after you assemble them there would have to be additional planing and sanding (as things turned out, a *lot* of planing and sanding).  Also important: I couldn’t find a stain or dye that I was sure wouldn’t leach into the food.  I could have achieved the effect by using two different types of wood, but the goal for this project was to use up the free wood I found.

Detail of three of the planks being glued together (in my vise, which I also made :) ). There are internal dowels that run the length of the board for strength and rigidity.

I still wanted that linear semi-random texture, so after some experimentation what I ended up doing was this: I left the top surface untreated except for mineral oil, and used a dye rather than a stain to lay the color all over the back.  Then I went over it with a few layers of water to raise the grain as much as possible.  When I sanded it down flat again, the parts that were raised were completely removed, and so I got a really strong zebra effect from the natural wood color versus the stained dark blue/black parts that weren’t removed.

The finish came out so well that it’s kind of a pity that it’s on the bottom and sides where it’s less visible.

You can see on the end there’s a raised section, so that the end of the board actually overhangs the sink slightly.  That’s great for cleanup because I just wipe everything into the sink– it’s too large and unwieldy to pick up and clean, otherwise.

I’m really happy with the way it all turned out, and it’s such a pleasure to use something every day that you’ve made.  I was a little worried about the durability of pine versus a hardwood for this application, but as long as I treat it with mineral oil it seems to hold up fine.  If I was going to do it all again, I would definitely use a powered industrial planer—it took me a lot of time to get everything flush and true by hand.

I'm a pretty terrible cook, actually.

These are some quick reference guides I made for my dad, for Christmas.  I am posting it now because June is the next month after December, clearly.  So my dad has been hampered in working on his classic cars and rental properties with their myriad electrical problems because he’s never been exposed to the fundamentals of electricity.  I figured that what he needed in lieu of a formal physics course was a quick reference that he could refer to as needed.

Q is for Charge

I is for Current

V is for Voltage

R is for Resistance

P is for Power

I put together some original copy and graphics for each of the major concepts in electricity that were relevant to what he needs to do.  To make the final product, I then laminated each card and included a set of mini Sharpies to make notes on the back or circle key points or whatever.

There’s a lot more that could be done here from a graphic design standpoint, and as I was formatting them for web I found some errors in the material, but for a three-day project I’m pretty happy with how they came out.  PDFs, if you want to print out a set for yourself, are here, and you can download a zip of the photoshop files if you want to make some of your own here.

Also, can I parenthetically remark here that wikipedia is awful?  It’s written to sound like an encyclopedia, not to explain things clearly and accessibly, so it’s completely opaque on any conceptually difficult topic.  Yet when I want to really want to research something in depth, it only has the same shallow gloss of the subject that you’d find in a dead-tree encyclopedia.  Critiques of wikipedia usually revolve around concerns about quality of citations and bias, which misses the meat of the issue, that as a reference guide to anything except pop culture trivia it sucks.  Exhaustive scholarship on that Josh Whedon series though!

South exposure glazing at 3pm, noon, and 9am.

Prologue: I have been wanting for  to better educate myself on daylighting design and analysis, and its coordination with traditional lighting design.  Here in California, we have one of the most aggressive energy efficiency codes in the country, Title 24.  While this and voluntary measures such as LEED have driven impressive technological advancements in smart, lean building, we’re now a point of diminishing returns because, quite simply, most of the low-hanging fruit is gone.  With emerging technologies such as LED lighting still less efficient than good fluorescent lighting (and at three times the cost), there’s no source efficiency cavalry around the corner.

So not to sound like a goddamn hippie here, but it happens we have a free light source that is available all day long.  But before you break out the celebratory granola, consider this: for a typical commercial interior, we might desire 35 footcandles so as to be bright enough to read printed text but not so bright as to wash out our computer monitors.  Direct sun lighting levels might reach 10,000 fc on a sunny day.  Also, the sun moves from east to west, sky conditions vary, and so on.  All of this makes architectural daylighting design a fascinating and bedeviling practice.  I’m still woefully under-qualified to address that topic generally, but I’ve done an extensive daylighting analysis for the West Berkeley Library project I wrote about earlier with an eye to practical application to lighting design, and I present my preliminary findings in this post.

In a daylighting analysis, the software takes parameters such as geographical location, time of day and year, and sky conditions, and simulates the amount and quality of light that would pass through each window and then bounces the light off each surface in the room until an accuracy threshold is reached.  In the example at the top of this post, the light in the room includes contributions from direct sunlight, the ambient light of the sky dome, and the bounce light from secondary reflections.

Knowing how much usable daylight will be present is in some ways less crucial for lighting design than many other disciplines such as glazing selection, since in any case the lighting designer will still have to provide a quality electric lighting solution for when the facility is used after dark.  But daylighting calculations can certainly inform and enhance a lighting design, in particular in controls design.  I would generally like to know:

• Which luminaires in the space might benefit from switching control via daylight harvesting.
• How many hours a year a luminaire would be completely off if controlled by daylight harvesting, for purposes of cost-benefit analysis.
• The value added benefit of dimming or multi-level switching via daylight harvesting control; i.e. how many hours per year daylight would be provide some but not all of the necessary lighting in the space.
• Any design gotchas, things of the form ‘well, I didn’t know mauve was going to look so goddamn pink.’

How it’s usually done: the standard methodology of daylighting analysis consists of:

1. Set up a daylight study for some representative times, e.g. 9am and 3pm on the spring equinox, and perhaps also some outliers such as the summer and winter solstice;

2. Let the computer crunch numbers over the weekend;

3. On Monday, make coffee, look at the results, then make a bunch of assumptions about how the space will behave under myriad other conditions;

4. Adjust the design and repeat, time allowing.

Which is not to say that the above is a bad method.  Any daylight study is necessarily a gross simplification.  As an incremental increase in accuracy can easily result in a doubling in calculation time, it’s clear that the art here is in choosing what shortcuts we can take, what details will not contribute to the accuracy of the calculation and may safely be omitted.

False color rendering of the children's section, 3pm.

September 22nd, 9am: the date used in every daylighting calculation is day of the spring/fall equinox, when the length of day and night are equal.  Also, it’s the day when the angle of the sun is halfway between the steepest angle on the summer equinox, (about 75 degrees above the horizon in San Francisco) to the shallowest midday angle on the winter solstice (about 30 degrees above the horizon).

We think of the sun as being overhead, but for most of us in the norther hemisphere an average angle for the entire year might be 30 degrees above the horizon– more horizontal than vertical.  On the equinoxes, the sun is only above 45 degrees for about four hours.  So as a typical angle of sunlight, a time in the mid-morning and/or mid-afternoon are going to be a better representation than noon.

The data: the above figure shows the location of the key calculation point grids in the library, and the table below it shows the values.

For reference purposes, the lighting levels under totally electric lighting conditions are:

Interior lighting design in 100 words: If you’re not familiar with footcandle levels, 30-50 footcandles at the task plane is about the level you need for comfortable reading, depending on age, size of text, contrast of text to page, etc.  For general ambient lighting, 10fc is sufficient for simple orientation and tasks, e.g. locating the keys in your apartment and not walking into things.   For public or commercial spaces, you want to keep the uniformity to less than 10:1 maximum to minimum (lower is more uniform), to avoid uncomfortably bright spots or very dark shadows. On the other hand, having the entire space lit to one even level is fatiguing to your eyes and makes the space feel sterile and uninviting.  In daylit spaces, 500fc is about where people will take action to reduce lighting levels, e.g. by closing the blinds.

For the vernal equinox, 9 am, clear skies, we have:

And for the same day, clear skies, 3pm:

Conclusions: the daylighting consultant has done a fairly remarkable job of allowing usable amounts of diffuse light into the space, for one.  Most daylighting scenarios include footcandle readings in the thousands in areas of direct sunlight, with a precipitous drop off as you go away from the windows.  For larger spaces, it can require more electric lighting to make the rest of the space feel well lit in contrast to the daylight zones than if there were no daylight at all.  To have this much usable daylight throughout the space takes some real artistry.

In general, lighting levels look pretty good to me, and I think the supplemental electric lighting in this part of the library will be minimal.  The main tasks the space should support are:

• General orientation and navigation through the space: at least 10 fc in the large calculation areas of “Main room floor” and “Main room RHS.”
• Browsing stacks: 30 vertical footcandles to read the spines of books and in particular the call numbers.  This would be the stacks calc areas such as “Adult Stacks 1″ and “Stacks long wall.”
• Children’s reading area: children’s books tend to have larger text, and children’s eyes need much less light– a 65 year old needs twice as much light to accomplish the same task as a 20 year old.  25fc should be plenty.
• Adult reading area: 50fc is about what I’m aiming for here, to support a wide range of text sizes, text to page contrasts, and age of the reader.
• Computer area: lower ambient lighting levels actually help with visibility of the screen, but we also need to support some reading tasks.  35fc is a good number.  The calculated lighting levels of this area are quite a bit lower than this especially in the morning, so I’m expecting the indirect wall wash in this area to be on most of the day.
• Entry Foyer: In an analogous way to its function as a buffer zone for climate control, this area should provide a transition zone from outdoor levels, which might be in the thousands of footcandles, to the interior lighting levels, which are generally less than 100 fc.  From the first table, the electric lighting in that room is good for about 30fc, so I’m expecting to run it during the day to bring the foyer lighting levels up.  At night it will do the same thing in reverse, providing a transition between the high interior lighting levels to the lower outdoor lighting levels, so I’ll need the ability to dim or bi-level switch here.

There are a few areas, such as the the adult reading area in the morning, that show very high contrast ratios.  It might be desirable to turn on the electric lighting in these areas, even though the average lighting levels are quite high, in order to reduce the contrast.  This is where renderings are really helpful, since it can be difficult to tell what’s really happening in the space from a few data points.

9am view of the adult reading area

From the rendering it’s easy to see that the high contrast ratios are due to the occasional patch of direct sun that makes it through the shades on the exterior facade, and is not a cause for concern.  Similarly, for the right side of the space:

The high values are clearly from a patch of direct sun at the end of the hallway.  On the other hand, I’ll probably want to give the librarians at the information counter to the left of the image control over the downlights overhead, as that area might feel a little dark depending on weather conditions.

Since I have the luxury of time for the West Berkeley library project, I thought it would be instructive to generate a much more complete data set, for purposes of comparison.  I’m interested here in what a larger data set might tell me that the ‘key points’ data set above doesn’t, i.e. whether we’re making the right simplifications when we do daylighting calculations.  To that end, I’ve generated about 20,000 data points for this project, and I’m sorting and parsing and analyzing that mound of data now.

Please feel free to drop me a line or leave a comment if you have questions or feedback!

Some months ago, I went with my family to the Alexander Calder exhibit at MOCA Chicago.  The collection was impressive, dozens of his collages and a few sculptures for good measure.

Afterwords, as we were eating our ice cream, my father wondered aloud whether how much calculation or adjustment Calder would have had to do, to get things to balance out.  My off-the-cuff answer at the time was that only the last, smallest leaf would have to be correct, but I’ve since started to doubt whether that was correct.

Here is some cryptic math business for you to look at, so you can see how very, very serious this is.

You know what Newton would be doing right now if he were alive?  Clawing at the lid of his coffin. Recall that torque is equal to the vector cross product of force and radius vectors:

If your physics is a little rusty, intuitively you know that to double the torque you could use twice as much force or double the length of the lever, right?  In our case, the force is due to gravity, and thus equal to the mass times the acceleration due to gravity:

So for a very simple mobile with one arm and two weights such as the following:

This mobile will be in balance when:

Some simplifying assumptions: We’re really only interested in the magnitude of the torque here, and let’s assume that each bar is hanging roughly horizontally and so the force and radius vectors are at a 90 degree angle.  In which case, we can just multiply the scalar magnitudes of the terms together, and forgo the vector cross product.  I’m not even sure I remember how to do a cross product, so this is a happy accident.  Also, notice that each term in the above equation is multiplied exactly once by gravity, so we can just pull it out of both sides of the equation and forget about it.  Then we have:

m₁·r₁ = m₂·r₂

e.g. a 3 kg weight on a 2 meter arm would be in balance with a 6 kg weight on a 1 meter arm, right?  Both sides of the equation would multiply out to 6 kg·m.  The general solution for a 4-weight system such as:

would be:

m₄·r₄ = m₃·r₃ + r’₃(m₁ + m₂)

Notice that the length of the lower arms become unimportant in balancing the upper arms– only the magnitude of the weights and the distance r’₃ contribute to the system torque.  More generally, for an arbitrarily large number of arms:

Perfect vacuums and frictionless pulleys: Of course, this isn’t Physics 101 and we can’t neglect the weight of the arms.  Say that the arm material has a linear density of λ (kg per m would be the units).  Then the weight of each arm is:

λ·r_i

My sense is that this figure would only matter in aggregate, as the sum of all the arm mass vs. the largest weight on the top.  Then the last equation becomes:

Elegant special case: If the arms are all of equal length, and each mass is twice as large as the one below it, e.g. m_i = 2·m_i-1, equation (1) above becomes one of my favorite geometric series (because I play favorites):

Or in diagram form:

I think that a lot of what makes Calder’s mobiles so immediately appealing is in the orderly way that the proportions step down from one branch to the next.

We now return to our original hypothesis, that only the last weight must be correct to balance out to make the mobile come out in balance: sort of true, but not really.  Pulling out the first mass from equation (1):

m₁ can be chosen so as to balance the equation, as can any other term for that matter.  r₁ can then be chosen so as to balance out m₂·r₂, but that’s as far as it goes– the second level consisting of m₃·r₃ and m₄·r₄ will need some entirely different mass to level out.  It would be very difficult to build from the top and leave the final balancing until the end– every mass added would throw the balance off for the system above it.  In fact, I think we can say with some confidence that Calder would have had to build his mobiles from the bottom up rather than the top down.  So that’s kind of an interesting result.

Aerial shot from Adult reading section, looking towards Children's.

If you were wondering what I do all day: Here’s a project that is still very much a work in progress, but the design process is kind of interesting so I though I’d post up where we’re at with it so far.  West Berkeley Library is a planned Zero-Net branch library for which I am consulting on the lighting.  The architects are Edward Dean and Michael Bulander of Harley Ellis Devereaux.

Zero-Net means that the building has sufficient power generation that it generates at least as much energy as it consumes.  In this case, the building has photo-voltaic panels mounted on the roof that feed excess power back into the grid during the day, offsetting the amount of power consumed at night.  This is environmentally preferable to a fully off-grid solution, as peak electric usage is typically during the hottest part of the day, and so that’s when power companies run their dirtiest and most inefficient plants.

The numbers are predicted footcandle readings at that location.

An ounce of reduction is worth a pound of generation: In addition to on-site generation, Zero-Net facilities must implement cutting-edge strategies to minimize consumption.  This is a practical necessity, as there is typically a limited amount of real estate available to give over to generation, and a cost cutting strategy, as it is much less expensive to reduce usage by a watt than to add a watt of generation.  Efficiency strategies for lighting include:

• Extensive use of daylighting
• Dimming fixtures, which may be manually or automatically adjusted for less than peak load
• Efficient lamps and luminaires
• Occupancy sensors to turn off the lights when a space is unoccupied

Calculations showing the predicted light levels.

Overall lighting strategy: Often lighting design for interior spaces involves lighting the entire space up to the highest level necessary for the task typically performed there.  This approach, while having many virtues, is not particularly great from the standpoint of energy consumption.

After looking at a number of options, we settled on an approach known as ‘low ambient/task.’  In low ambient/task lighting, you instead light the space up to the minimum level necessary to get around without walking into walls (low ambient), and then put concentrated task light in the areas where you need the higher levels.  As a designer, it’s kind of a pain to implement, because you have to think about all the tasks that will be performed in the space and make sure that you’re providing good lighting for all of them.  However, done right you can achieve very low energy densities without sacrificing the quality of the lighting.

Task lighting is provided by indirect/direct floor lamps at the reading areas, which also serves the aesthetic goal of making these areas feel as like a personal reading area in your home, rather than an institutional study space.  Task lighting on the stacks is provided by linear LED fixtures, about which more will be said later.  The ambient lighting is provided by the uplight component of the floor lamps, each of which has two 55W biax lamps with about a 70% up/30% down distribution.  The uplight component does not do much in the way of footcandles on the ground, but does make the space feel airy and open instead of cavelike, contributing to the perceived brightness.

Section showing arrangement of skylight wells and PV panels

A unified design aesthetic: The main room is divided into adult stacks and a reading area to the south, and children’s to the north.  The approach to lighting in the two spaces was necessarily somewhat different, as the linear LED fixtures cannot be used on the lower stacks in the children’s areas, nor can floor lamps.  However, an important requirement was that the two halves of the space not feel distinct from each other, but rather that the whole space feel unified.  This is accomplished by placing an appropriate amount of direct/indirect fixtures in each area of the space so as to keep the lighting levels on the walls and ceiling consistent.

The skylight wells created a repeating dark-light-dark pattern, and that theme was used throughout to tie the space together.

A stack lighting study, one of several I did to compare different approaches to this issue.

Stack lighting is a problem. Best practices for stack lighting, that is the lighting on the spines of the shelved books, is 30 fc average with a 5:1 max:min ratio between the brightest point and darkest point.  That’s a very high average lighting level, and in addition a *very* difficult uniformity to achieve.  Common approaches are to run a narrow beam fluorescent fixture down the middle between each row of stacks, or to put a small linear task light along each shelf.

We looked at these and other approaches, but the one that had the lowest lighting power density (LPD) and best performance was to put a linear LED fixture along the top of each stack and graze down the front of the shelves.  This puts us at 23 fc average with an 8:1 max:min in this study, so with contributions from ambient light we’re right at the recommended levels.  Power consumption is a fairly remarkable 5W per lineal foot of stack (10 for double sided stacks as shown here).  Additionally, LEDS are easier to dim than fluorescent, so in our dawn-dusk scenario they could be run at less than full power.

Exterior Lighting: Aesthetically, the goal is to create a lantern with the appearance of light streaming out from the interior.  The desired feeling is one of approachability, that would welcome pedestrians inside, rather than the grandiose tradition of civic architecture that creates a space between the viewer and the architecture.

Near the entrance, ingrade uplights wash the wall, while LED pinspots pick up repeat the dark-light-dark pattern of the interior on the overhead beams.  The brightly lit foyer creates a focal point to attract the user into the space.

I’m working on daylight studies to analyze the actual projected energy usage now, hope to have that post up in a week or two.  A big shout out to Edward Dean and Michael Bulander for for having both the vision to conceive of this project and the skills to realize it.  Also, the inimitable Dal McGinnis, my colleague at Lighting Systems.

It’s finally done, jesus. A few years ago, I bought some electronics components from a former Eastern Bloc country.  Nixie tubes are apparently still widely available over there, so since the largest line item in the order was shipping costs, I had them throw some Nixies in with the order.  They sat in my parts bin until, a few months ago, I decided to make a clock out of them.

Nixie clocks aren’t exactly novel if you read the DIY blogs, and you can buy yourself any number of commercially available kits or even mostly finished clocks to simplify the build.  I’m a big fan of using kits to allow you to focus on the parts of the project that interest you, and honestly designing the high voltage supply and multiplexing from scratch would probably be more trouble than I would want to go to for this project, if not beyond my abilities with electronics altogether.

But, the kludge. My issue with the available kits for this type of project are that you end up with a large box to hold the control electronics and drivers that’s all out of scale to the tubes themselves.  It’s one of the giveaways that something is DIY, even if the fit and finish is quite good you end up with a finished product that’s several times as large as one would like.  I wanted the final project to look like a  industrial mass-produced product, one that had been found in daily service for many years.  So the circuit layout is custom to fit in my enclosure, although I borrowed code and the schematic from the Arduinix project, many thanks to them for open sourcing their hardware and software.

This isn't even the clever bit.

Microprocessor control is done with the RBBB Arduino, which is a fantastic solution to embedded Arduino projects.  They’re quite small, and if you buy five of them at a time they’ll cost you a mere $10 apiece.  My only complaint with the RBBB design is that while there’s any number of places to tap into the 9-12V supply voltage, there’s only one pin that puts out regulated 5V supply.

I also added a real time clock to keep time when power is lost using the newly released Adafruit DS1307 kit ($9!).  The number of ready-made solutions now available within the Arduino ecosystem is very impressive.  Sure, you could engineer $20-30 of cost out of all of the above by building it yourself from scratch, but life is short, and I personally like to finish more than one project a year.

Even with all these building blocks, this build took me a few months to complete, mostly because of the difficulty of fitting all the parts within my chosen enclosure, and, you know, life.  Since I’m using mostly other people’s design material, the most interesting part is the build process.  Without further ado:

For the record, this is my preferred brand of PCB blank.  I usually buy the 1/2 oz, it works just as well and it’s easier to etch.  The brand they sell at Radio shack is shit, it’s this ugly brown color and it has a texture that makes it difficult to get all the copper off your etched areas.  This stuff is also milspec, so that’s cool, I guess.

I used this commercial toner transfer product for the first time, and it works much, much better than any paper I’ve ever used.  Well worth the $1 a sheet.  I’m still figuring out the heat setting for the iron, however– it has a much lower fusing point than standard inkjet toner, and I’m having trouble with the traces smudging.  The same company also makes also a foil that seals the copper, and a stencil product.  I haven’t tried those, but I’m curious about them.  Shout out in the comments with your experience.

Another protip I recently figured out– brushing the etchant onto the board with a sponge works a lot better than the dunk and slosh method.  Also uses less etchant.  The part on the left covered by the tape is where the time set controls ended up, I got impatient and etched the board before I had them fully designed.

Saturday night at the Pierson residence.

Testing the high voltage supply. Yes, that’s 180V DC.  I kept shocking myself on the high voltage parts throughout the project, but by luck more than skill didn’t kill the microprocessor due to short circuiting.  You can see my Arduino Duomilanove in the background, which I use all the time to produce 9V and 5V DC for testing purposes.  Final voltage was 165V on my IN-4 Nixies.

About this time the sockets arrived from the Ukraine, with love.  I usually work by designing one part while building another, and trying to order parts just before I need them so that I know all the parts I need and I don’t have to place two orders.  Sometimes that means painting yourself into a corner, but it keeps things lively and if I waited until the project was completely designed and thought out to begin building, I’d never start anything.

A first attempt at wiring up the sockets.  I realized pretty quickly that this would be very bulky in terms of cabling, and all that identical gray cable would be a nightmare to troubleshoot.

Assembling the sockets. Each socket has 11 (tiny) wires, times two ends, times six sockets total.  Then I had to redo them with shorter cables, because all that cable wouldn’t fit in the enclosure.  You know how Sartre imagines hell as this tiny room full of tiresome people, forever?  I can do him one better.

So goddamn much soldering.

First test of the socket. One thing to know about if you’re using the Arduinix as your reference design:  it does not include the necessary 10KΩ current limiting resistors on the anode pins.  I can see why they did it this way, but I think it needs a big stupid warning on their literature.   I came very close to burning up my supply and/or a tube because I did not realize that crucial point until it was almost too late.  So that’s why I have a resistor soldered on to the socket, instead of the main PCB where it belongs.

Figuring out what pins on the Arduino connect to what pins on the K155ID1 Nixie drivers.  I ended up using almost all of the pins on the Arduino.  More soldering.

About this time the enclosure for the project finally arrived, from Really_Lethargic_Shipper_87 on ebay.  It’s some kind of toolbox from the early 20th century.  The right enclosure can make a project and the finish and size were perfect, exactly what I had in mind.  I carefully cleaned off some but not all of the ingrained dirt and put a coat of polyurethane on it to fix the patina.

Uh Oh. It was at this point that I realized that I had been optimistic about how much space was needed.  The seller had been nice enough to send me the interior dimensions of the box.  But, while on paper a 8 7/8 x 1 7/8″ x 7/8″ deep interior dimension was large enough to contain everything, in practice I found that I desperately needed another 1/8″ of depth.  Also, bad day story: The box is old and very brittle, and so the very first hole that I tried to drill in the lid split the entire lid in half.  I regret the intemperate nature of my ensuing remarks.  Fortunately, it glued together really cleanly, and you can’t even see the seam.  I added some wood strips on the back of the lid to reinforce it, though.

I removed 1/4″ wood backing on the box with a dremel and wood chisels, and replaced it with 1/8″ plex.  The fasteners are dollhouse hinges that I drilled the pin out of and replaced with stainless steel clevis pins.  I wanted the sliding lid type of box for aesthetic reasons, but I knew ahead of time that I might have to figure out another way to get at the innards for servicing.  As it is, the lid slides open just enough to reach the time set controls.

The power jack. The mounting bracket was milled down from a spare piece of angle iron.  I think if my left hand was replaced by a dremel, that would really be just fine.  Also, a tip:  Goodwill is great for power supplies.  They have a bin of them, $2 a pop.   It’s getting harder and harder to find ones that aren’t 5V, though.

Final assembly begins. You can see the mass of ribbon cable in the back, which I had to cut down and re-solder, twice, in order to fit in the enclosure.  More soldering.

Here I took a break. To do two shows and some other projects, which I’ll post up eventually.Serendipity! In the meantime, the DS1307 breakout board was released by Adafruit.  It connects to the Arduino via I2C, so only two wires.  It will keep the time for up to five years if power is lost, and is somewhat more accurate than the crystal found on the standard Arduino board, and much more accurate than the oscillator used on the RBBB.  And it just fit within my enclosure.  The board was initially unresponsive, and I spent at least a week troubleshooting it until I finally figured out that I2C on the Arduino is done on *analog* pins 4 and 5, not digital pins 4 and 5.  Facesmack, but I think the support pages on arduino.cc and adafruit could be a bit more forceful on this point.

Controls: The trim pot is to set the high voltage, which runs from about 140V up to 190V.  I’ve got mine trimmed out to about 165V.  The switches are:

• Time set: Shows hours only, minutes only, seconds only, and then restores to regular time display.  I tried using the external interrupts for this because I thought it would simplify coding, but with all the switching high voltage it ended up being very unreliable.  Hey Internet, how do you fix that?  I’m thinking it’s something clever with a resistor and capacitor, but I’m not sure exactly what.
• Time Increment: increments hours, minutes, or seconds, whichever is displayed currently.
• Time Decrement: as above, but –.
• Spare switch: something I’ve learned to put into my designs.  I was thinking that it would be a display blank, because I was worried about the clock being really bright, but it’s not an issue in my current apartment.  I’m planning to re-program it as a cathode exerciser, i.e. cycle through all ten numbers in turn, to prevent cathode poisoning.

The final assembly. I used dental floss to do the cable lacing, because cable ties were too bulky.  It had the added benefit of making the enclosure smell minty fresh, instead of something unspecific but foul.

Here’s a detail I’m particularly happy with.  The power cable reveal is a bronze bearing, which hides the edges of the wooden hole.  I do some of my best thinking in the hardware aisle of Cole Fox.

In situ.

So that’s it.  Took a bit longer than I would have like, but I’m really happy with the results.  It looks fantastic up on my wall, and really came out exactly as I hoped.

]]> http://blog.wingedvictorydesign.com/2010/10/13/my-nixie-clock-build/feed/ 4 http://blog.wingedvictorydesign.com/2010/10/09/rev15-of-the-arduino-dmx-reception-software-released/ http://blog.wingedvictorydesign.com/2010/10/09/rev15-of-the-arduino-dmx-reception-software-released/#comments Sun, 10 Oct 2010 04:10:16 +0000 Max http://blog.wingedvictorydesign.com/?p=1272

New in this version:

• Tested and working with IDE version 0021.  I have not tested it with the new Arduino Uno hardware, but can’t think of any reason why it wouldn’t work.  If you have any success or otherwise with the new hardware, drop a line in the comments.
• A bug has been fixed in the addressing routine caused by button debounce, and the addressing routine logic has been simplified and made clearer.  Big thanks to Ron Barber for pointing out the source of the bug and contributing new code.
• There is now a check when the previously stored address is read from EEPROM to ensure that it is in the valid range of 1-511, to prevent a bad value being read in from uninitialized EEPROM.
• Fixed a potential bug in the break detection routine that could cause the read values to be off by one in cases where the break from the microcontroller was exactly 88uS.

You can grab it here or mosey on over to the original post for the instructions.

I went to Chicago last weekend for the premiere performance of my sister’s latest musical project, Nylon Radio (My sister is doing a doctorate in Music, um, Theory? at the University of Chicago).

All of the pictures were taken with my Panasonic G1 with 20mm/f1.7 lens.  It’s a digital camera, and a good one, but I feel like something has been lost with the switch to digital photography.  With film media, you have this really fascinating and unpredictable interaction between the type of film stock/processing and the environment that you were trying to capture when you opened the shutter.  When I shot on film, this was usually very, very frustrating, as you could do everything right and still end up with a bad shot, due to some quirk of the processing or stock or spectral distribution of the ambient light or measurement error in the metering or whatever the hell else.  But I find that digital cameras, especially high end ones, can render the scene so neutrally that the results are nice enough but boring.

So I was presented with something of an opportunity when I loaded the SD card into my laptop on the plane back to San Francisco.  I had to shoot nearly all of the shots wide open at f1.7 at 800-3200 ISO– there was no light, except for some ambient bounce that didn’t give any modeling and some LED music stand lights that look like death warmed over on skin tones.  If I was going to have anything usable, it was all going to be in post-process.  I decided that rather than trying to make a purse out of a sows ear, I’d instead play around with all of the fun ways you can abuse the Camera Raw filters, and let taste be no object.  Here are some of my favorites, with a quick description of the techniques used in each.

Cheap Photoshop trick #1: the automated panorama stitcher. I really love the look of these.  While the automated stitcher routine does a pretty good job of adjusting for different exposures, I set the exposure lock before I took the set, in order to have distinct light areas and dark areas in the picture.  I pulled the color temperature slider way down, but subjectively this was the feel of the lighting at the event.

You damn kids won't remember assembling these by hand prior to CS

Most of the really objectionable noise was in the color rather than luminance, so a lot of the remedies involve removing the color before applying some sort of false color process.  In the above picture, I also grabbed the midranges and pushed them into the highlight range, which had the effect of popping the band out from the brick wall behind them.

Here’s the warm-up act.  There really wasn’t any light on them at all, this was shot at 3200 ISO with the aperture wide open.  I compensated by shifting to the left until the guy was standing in front of the lamp, which at least popped him out from the background.

The original was a solid mess of color noise, so I desaturated the image completely and replaced the highlights with an orange tone, and the shadows with a lavender tone.

Here, I nibbled away at the saturation by color and value of everything but the central figure (my sister, Marcy).  You can do the same thing in half the time by using a mask and a saturation layer, but I find that this gives a more organic feel.

I’m still working at getting better at portraiture.  I got into photography as a matter of necessity, because I needed good photos of my shows for my portfolio, and the production photographer tends to shoot close-ups without any regard for preserving the look of the lighting (don’t shoot with a flash, please), if there’s a photo call at all.  I didn’t want to get right up in people’s grill while a performance was going on, so I don’t have a lot of close-ups for this shoot, but here’s one I quite like.