2019-08-10 14_ChrisPineSpideyBellsYouTube

Captain Kirk is filled with deep regret. Still from Chris Pine on YouTube

by George Taniwaki

There are not very many popular culture references about getting a degree in chemical engineering. But there should be. I’m only aware of three. That’s not enough to fill a blog post so I added a few extra.

Spidey Bells–2018

If you watch a Marvel movie all the way to the end, hidden in the credits are wonderful Easter eggs that tend to be somewhat unrelated to the movie.

In the opening of the 2018 animated movie Spider-Man:Into the Spider-Verse, the first Peter Parker, voiced by Chris Pine, says that he has recorded an album and is selling some related merchandise. At the end of the movie, as the credits roll, one of the songs from that supposed album plays. It is a Christmas carol called Spidey Bells (A Hero’s Lament). In the song’s middle bridge, he reveals a dark secret:

“Why did I agree to do this stupid song?
I have a degree in chemical engineering”

Who knew? We know that Peter Parker dropped out of college to become a photographer for the Daily Bugle. But until now, we didn’t know what his major was.


Long before Mark Wahlberg became famous as a restaurant franchisor, he was known as Marky Mark, the lead singer of the Funky Bunch. They had a hit in 1991 with a cover version of Lou Reed’s Walk on the Wild Side. Their version includes samples from the original but had different lyrics and was retitled Wildside. The song include these great lines:

“Cause deep inside Annie had aspirations
Besides that she had expectations
Wanted to be a chemical engineer”

Unfortunately, it does not end well for Annie, who is played by an uncredited actress in the music video.


A ChemE major takes a hit on the wild side. Still from Marky Mark on YouTube

The Future’s So Bright, I Gotta Wear Shades–1986

In 1986, the U.S. band Timbuk 3 released their debut album which contained the hit song  The Future’s So Bright, I Gotta Wear Shades. The music video got extended play on MTV, back when MTV consisted almost entirely of music videos in rotation. The video features a burro, the mascot for my alma mater, Colorado School of Mines. The song refers to nuclear science rather than chemical engineering, but really, that burro and the harmonica solo.

Incidentally, The Future’s So Bright, I Gotta Wear Shades is the title of a 1990 episode of ALF, a television sit-com.


The future’s so bright I put a TV on a burro. Still from Timbuk3 on YouTube


Just a few weeks before I graduated from college in 1981 there was a Doonesbury cartoon that became wildly popular among my classmates. It became the buzz at school because it pointed out how crazy high the demand for chemical engineers had become.

Earlier in the year, all of us seniors were flying around the country interviewing with the major oil companies and receiving job offers from every company we talked to. It was madness. I interviewed with eight firms from California to Louisiana and received offers from all of them at starting salaries that exceeded what my parents made combined.

Little did we know, the oil industry would collapse a year later, throwing many of us, including me, into unemployment.


BD is tired of coddling lazy chemical engineers. Image from GoComics.com

The Graduate–1967

The granddaddy of all chemical engineering pop culture references is a famous scene in The Graduate, the 1967 movie directed by Mike Nichols (who won the Academy Award for his work) and co-written by Calder Willingham and Buck Henry. In the scene Mr McGuire (Walter Brooke) advises young Ben (Dustin Hoffman) that that there is a great future in plastics. However, he doesn’t say you have to know anything about chemistry to reap the rewards of that future.


I just want to say one word to you. Still from Movieclips on YouTube

Space Oddity–2013

And finally, I want to point to a 2013 cover version of David Bowie’s Space Oddity performed by Canadian astronaut Chris Hadfield while onboard the International Space Station (ISS). This song has nothing to do with chemical engineering and Mr Hadfield is a mechanical engineer, not a chemical engineer. I just like the song and the visuals of the earth zipping by in the background of this music video.

I’m glad David Bowie had an opportunity to see this before his death just over two years later. In fact, without his intervention, the video may not have been made available (Independent Jan 2016).

According to NASA, Mr Hadfield is the first and so far only astronaut to record music videos in space.


Can you hear me major Tom? Still from Canadian Space Agency on YouTube.

Update1: To complete the circle, check out this video of the “real” Captain Kirk, William Shatner, and a fellow Canadian, singing RocketMan at the Science Fiction Film Awards.

Update2: Inquiring minds want to know. If Peter Parker dropped out of college, when did he earn a degree in chemical engineering.



by George Taniwaki

This post describes the final six classes I took to obtain the Microsoft Artificial Intelligence Certificate. Four were required and two were optional. For a description of the first six classes I took, see this May 2019 blog post.

DAT236x – Deep Learning Explained

Deep learning is the use of machine learning on large datasets, often using neural networks. It is used in fields such as computer vision, speech recognition, and language processing (topics covered in more detail in later classes). Techniques include logistic regression, multilayer perceptron, convolutional neural networks, recurrent neural networks, and long short-term memory.

Time: 10 hours on 6 modules

Score: Missed 4 homework questions and 2 knowledge check questions for a score of 92%

DAT236x Score  DAT236x Certificate

DAT257x – Reinforcement Learning Explained

Reinforcement learning assumes a problem can be modeled as a Markov decision process. There is a set of discrete states (S), an agent that can perform a set of actions (A) selected from a set of decision policies (Π). Each possible action will result in a reward (R) and a new state (S’). The goal is to find the optimum policy π(s) for all s in S or to determine if a given policy is optimal.

Solutions to the reinforcement learning problem include use of multi-arm bandits, regret minimization, dynamic programming (Bellman equation), policy evaluation and optimization, linear function approximation, deep neural networks, and deep Q-learning. Advanced topics include likelihood ratio methods, variance reduction, and actor-critic.

I have an undergraduate degree in chemical engineering where I learned about control theory and Markov chains. However, that coursework only covered analog PID controllers. The topics in this class were new to me and so it was slow going.

Time: 16 hours for 10 modules

Score: Missed 7 knowledge check questions early on, then slowed down and got the rest right. Missed 3 lab questions. Final score of 91%

DAT257x Score  DAT257x Certificate

DEV287x – Speech Recognition Systems

Speech recognition is an interdisciplinary activity that combines signal processing, acoustics, linguistics, and domain knowledge with computer science. The topics covered in this course include:

  1. Fundamental theory – Phonetics, words and syntax, performance metrics
  2. Speech signal processing – Feature extraction, mel filtering, log compression, Feature normalization
  3. Acoustic modeling – Markov chains, feedforward deep neural networks, sequence based objective function
  4. Language modelingN-gram models, language model evaluation (likelihood, entropy, perplexity), LM operations (n-gram pruning, interpolating probabilities, merging), class-based LMs, neural network LMs
  5. Speech decoding – Weighted finite state transducers, WFST and acceptors, graph composition
  6. Advanced topics – Improved objective functions, sequential objective function, sequence discriminative objective functions

This class was pretty awful and I’m glad I didn’t pay for it. It consists mostly of text displayed in the edX courseware. It would have been helpful to if video or audio lectures were included to show voice recognition in action. The text itself was a split into multiple web pages containing embedded MathML equations, making is unsearchable. I ended up copying and pasting all the text into a Word document.

The lab assignments in this class are provided as Python files designed for Linux. Some labs require a Linux shell and would not run in Visual Studio on Windows. I would expect instructors in a Microsoft sponsored course to design lessons that could run on Windows. Simply putting the code in Jupyter notebooks would have made it easier to read and to work with.

Some labs require a Python package called OpenFST that does not compile with the latest build tools available from Microsoft. Again, I would expect instructors to design lessons that could run on Windows.

Time: 6 hours for 6 modules

Score: None, I did not take this class for credit

DEV287x Score

DEV288x – Natural Language Processing (NLP)

Natural language processing consists of many separate but related tasks. These include transcription, translation, conversation, and image captioning.

Machine translation has evolved from conventional statistical machine translation (STM) that uses hand-coded phrase pairs, to neural machine translation that use deep neural networks to create end-to-end sequence probabilities to translate entire sentences at a time.

The deep semantic similarity model (DSSM) is a DNN model for representing text strings as vectors. It can be used for information retrieval and entity ranking tasks.

Natural language understanding requires spoken language processing, continuous word representations, neural knowledgebase embedding, and KB-based question answering. NLP can be enhanced using deep reinforcement learning.

Finally, image captioning requires multimodal intelligence, combining image recognition and assigning labels to images in a natural language format.

Time: 10 hours for 6 modules

Score: None, I did not take this class for credit

DEV288x Score

DEV290x – Computer Vision and Image Analysis

This course is an excellent overview of the state-of-the-art in computer vision. It starts with a description of classical methods including thresholding, clustering, region growing, template matching, and feature detection (Sobel edges and Harris corners).

Next it covers object classification and detection algorithms such as Viola-Jones, histogram of oriented gradients (HOG), deep learning, extending classifiers into detectors, object proposals, and introduces convolutional neural networks (CNN).

Finally, the course introduces advanced topics such as super-pixels and conditional random fields, deep segmentation, and transfer learning.

Time: 10 hours for 20 modules

Score: Missed 4 quiz questions and 1 final exam question for a score of 94%

DEV290x Score  DEV290x Certificate

DAT264x – Microsoft Professional Capstone : Artificial Intelligence

This is the last class for the certificate. Similar to the capstone for the Microsoft Data Science certificate, it is a month-long project designed as a contest hosted by datasciencecapstone.org. Unlike the capstone class for the Data Science certificate, there is no report, the grade is based solely on the contest score. For a description of the April 2019 contest, see this [future date] blog post.

I used Microsoft’s cognitive neural toolkit (CNTK) package for Python for my solution. I had a hard time debugging my code. CNTK is not as popular as Google’s Tensorflow, so searching error messages on the web gives few results.

Time: 30 hours for single assignment

Score: Log-loss error of 0.22 for a final score of 95%

DAT264x Score  DAT264x Certificate

Final Certificate

Below is my certificate of completion for the Microsoft Professional Program, Artificial Intelligence Certificate.


* * * *

As an aside, starting in January 2019, edX has changed the way it handles students who audit courses. To encourage more students to pay for its courses, edX now limits access to course content to 30 days after enrollment. After 30 days, you lose access, even if you have posted items on the discussion board. Further, it eliminated access to the assessment content (quizzes, labs, and exams) entirely unless you pay. This sucks.

I hope Microsoft will provide funding to edX to allow audit students to participate. Or Microsoft should stop working with edX and move its content to another MOOC platform that supports audit students. I’ve paid over $2000 to participate in the edX courses. But I always audit a course before paying for the content. I think the try-before-you-buy model is essential to get students to trust they will get their money’s worth. Preventing audit students from seeing the assessment content will make it difficult to gain their trust in the value of that content.


Digital billboard by LKDN and Blip

by George Taniwaki

If you want to gain attention of people outside your social circle, you need to get your message in front of strangers. One way to do this is to use billboard advertising. Billboard advertising, also called outdoor advertising or out-of-home advertising, can be expensive. It is also impersonal. However, it can generate a lot of impressions and can generate additional word-of-mouth or news publicity.

Every form of publicity you create should have a call for action. It should include a URL, phone number, or email address. In the example billboard at the top of this blog post, note the short URL that points to the Living Kidney Donor Network and includes the patient’s first name. It’s easy to remember and the LKDN web page can contain additional information to help potential donors learn more about you and the donation process.

Do-it-yourself billboard

If you own a billboard or know someone who does, you can paint your own billboard. A do-it-yourself billboard is pretty uncommon, but I found one case (Kidney Dialysis News, Dec 2013).


Example of hand-painted billboard

Rent a static billboard

More likely, you don’t own your own billboard and so will have to rent one. The most common type of billboard is static. The image is printed on paper or canvas and will remain in the billboard until it is manually covered by a new sign.

There are two costs to consider when renting a billboard. First is the up-front costs for design and production. This can range from free to a few hundred dollars. The second is the rental fee. The rental fee will be based on the amount of traffic that passes by the billboard and the scarcity of billboard space in the area where you want to place your sign. You will want to compare the total cost per impression of a billboard versus the other activities you are employing to find a donor. For a good primer on measuring advertising and publicity effectiveness, see Real Numeracy Sep 2019.

All billboard companies provide design services. If you want to do your own design work or hire your own designer, ask the billboard company for the dimensions of their billboards. Also ask for the required resolution for artwork and what file format the artwork should be delivered in. Check if they accept RGB color PDF files or require CMYK color.

For smaller billboards, the images are printed on a 27” x 40” sheet of paper. Multiple sheets can be tiled to produce larger images. Larger billboards are usually printed on 10-foot wide sheets of canvas using a special ink jet printer. Again, larger images are created by tiling sheets.

Additional benefit of a billboard

There is another impact that a billboard can generate. Your effort could get picked up by a local TV station or newspaper where an even larger audience will see it. I found several stories about billboards used to find kidney donors on the web. A few are reproduced below along with links to the news stories they were contained in.


PaulBillboard  RandyBillboardJoshuaBillboardMandieBillboardMirandaBillboard

Eight examples of static billboards that got in the news

First row: Tracy (from WBIR-TV, Apr 2016), Jacob (from WKYC-TV, Apr 2019), Jessica (from WDJT-TV, Apr 2018)

Second row:  Paul (from NY Daily News, Jul 2015), Randy (from Atlanta J. Const., Oct 2017)

Third row: Joshua (from USA Today, May 2018), Mandie (from ABC News, Apr 2018), Miranda (from Daily Item, Apr 2016)

Mobile billboards

In urban areas, billboard space is limited. One option to overcome that is to put signage on the side of a truck or attach it to a trailer being towed by a truck or car or even a bicycle. This is different than putting a sign on the side of your own car or truck, which is discussed in a separate blog post (Real Numeracy, Aug 2019).


Example mobile billboard for Kidney Car, from Kidney Foundation of Canada

Rent time on a digital billboards

The latest advance in billboard advertising is digital signage. A digital billboard, like a web display ad, does not require the expense of a physically printed image. It can be placed on any billboard in the world that participates in the digital billboard network. Your ad can be displayed for any length of time on any date or day-of-week and time-of-day you desire.

Your ad will run for 10 second increments that cost between 1 cent to a dollar, depending on the amount of vehicle and pedestrian traffic expected. Most traditional billboard companies like Lamar and Clear Channel have digital options. There are also digital-only companies like Blip. (In advertising parlance each 10 second increment is called a blip.)

Donor billboard

Finally, below is an example of a billboard promoting living donation without a specific patient in mind. The billboard is sponsored by TransplantFirst Academy, an organization started by Risa Simon, the author of the book Shift your Fate.


Donor billboard from TransplantFirst Academy


Westbound on Evan at Santa Fe. Photo by James Taniwaki

by George Taniwaki

Take a look at the intersection above of two busy streets in Denver. The photo was taken from the left-hand turn lane of westbound W. Evan Av. turning to southbound S. Santa Fe Dr. From this angle you cannot see the on-ramp you are turning onto. It is obscured by a concrete jersey barrier.

In fact, the end of the barrier is separate from the rest of it. One gets the impression that it was originally connected to the rest of barrier but was accidentally hit, forcing it to be perpendicular to the remaining section. This creates the illusion that there is no road behind it, and you should turn left before the barrier, not after.

If you do turn left into the lane on the east side of the barrier, then you will be headed the wrong direction into oncoming traffic on Santa Fe Dr. This is really dangerous.


Synthetic aerial view of same intersection. Image from Google Maps

A better view of the intersection can be seen using the 3D tilt view feature of Google Maps. Looking at the synthetic aerial view, one can see that the intersection is directly over Santa Fe Dr. This allows large volumes of traffic to flow without taking up a lot of real estate. This configuration is called a single-point diamond interchange.

Two changes would probably make this intersection safer. First, the end of the jersey barrier should be a continuous piece of rounded concrete, not jagged, so the user can imagine there is a traffic lane behind the barrier. Second, the barrier should be set back about 3m (10’) so that a driver in the left-hand turn lane can clearly see there is a paved surface beyond the barrier before entering the intersection.

I don’t know how many accidents happen at this intersection and I don’t know how the state of  Colorado measures the social costs of traffic accidents. But if there are ten accidents (some with injuries) per year here, and the lifespan of the interchange is twenty years, then it is probably worth spending the one million dollars I estimate it would cost to make the suggested modifications to this intersection.

[Update: There was a paragraph here describing a fatal accident on Santa Fe and Dartmouth involving a driver going the wrong direction. However, that accident is unlikely to have been caused by a driver entering the highway at Santa Fe and Evans. I have deleted the paragraph.]


Righty tighty, lefty loosy, when viewed using a mirror

by George Taniwaki

On a visit to the University of Washington Medical Center laboratory, I was asked to provide a urine specimen. As I approach the restroom, I notice the door is about 1m (40") wide to accommodate a wheelchair, which is good.

Both the outside and inside of the restroom door have solid metal lever handles. You push down to open the door. Levers are easier to grip than knobs and are now the preferred method to open and shut doors. Also good.

Once inside the restroom, there is an easy to grip lever above the handle that controls a lock for privacy. Good again.

As shown in the image above, the hinge for the lever is on top of a circular escutcheon and it flips left or right. But which way locks the door?

The rotation direction to lock the door is ambiguous. Apparently, there have been complaints, so someone printed a sign and taped it above the lever. But the sign is somewhat ambiguous as well since it is posted above the hinge, but the lever is below it.

Finally, it seems one of the lab techs used a grease pencil to indicate the direction to turn the lever to lock the door. But the grease is now smeared and illegible. The hand-drawn arrow points in the direction to lock the door, meaning the bottom toward the door edge or a counterclockwise turn. This is not standard in the U.S. and the likely source of confusion. Oh well, good thing I don’t have a shy bladder and don’t care if someone accidentally walks in on me.

Oddly, the convention to rotate a lock lever so that when the top points toward the door edge to mean the door is locked is not universal. In Japan, most locks are installed so that when the bottom points toward the door edge it is locked. You may notice this on some Japanese car doors.

* * * *

P.S. Ever have trouble knowing if a door opens toward you or away from you? To learn more about this design problem, read this blog post from 99% Invisible. Also watch the video. And read Don Norman’s book, The Design of Everyday Things.

[Update: Clarified the description of locks in Japan.]


Google rewards reputable reporting, not left-wing politics, from The Economist

by George Taniwaki

A few months ago The Economist added a new feature to its back section called Graphic detail. It’s a pleasure to read because it nearly always contains bivariate plots where the x-axis is something more interesting than the date.

This week’s entry does not disappoint. It is entitled, Seek and you shall find and contains two charts (see above) with interesting x-axes. The charts analyze the impact of Google news search on the traffic a news source receives. It uses two independent measures, Accuracy score and Ideology score to rate different news sources. Accuracy and bias were determined using data from Adfontesmedia.com and Mediabiasfactcheck.com.

Many people claim that Google favors liberal news sources to the detriment of conservative views. Google claims it has a set of outside reviewers who check news sources for accuracy and reach. Point of view is not considered. However, one could imagine that a news source that has a strong point of view may report facts to match a point of view and that would reduce accuracy. As can be seen on the chart on the left, news sources with a strong ideological bias (darker red and blue dots) tend to have lower accuracy scores than less biased sources. I encourage you to go to the website because the data is interactive.

The dependent y-axis is the share of web traffic that comes from search engines. This is a bit problematic since if users believe that Google’s results are biased against their favorite news sources, they will visit it directly without using a search engine. Nonetheless, the data shows that search engine (mostly Google) share of web traffic increases with accuracy, not with ideology. That is, the plot on the left shows a linear relationship while the right plot does not.

Expected v. Actual

A separate experiment confirms the results. The Economist built a model to predict the number of news results appearing in 37 publications should receive from Google’s search engine based on their accuracy and their reach. It then compared the model results to actual search results on a “clean” computer using “a browser with no history, in a politically centrist part of Kansas.” (Why Kansas, you wonder? I’m guessing that is where the author lives.)


No bias detected, from The Economist

Again, no bias was detected. The difference between left and right are small and could be due to how they are defined, time of study, keywords searched, or other factors. The story is an excellent example of combining data from multiple sources, programming a bot to collect data, and visual display of statistical analysis.


The versatile design of this shoe rack can fit many needs

by George Taniwaki

I’m a member of the Washington Karate Association, which has a dojo in Bellevue. Traditionally, one removes shoes before entering a Japanese home or a sacred area. The genkan or foyer of the dojo has a shoe rack for the students. The current one needed to be replaced and I offered to make one. An exploded view of my design is shown in Figure 1 below. The design ideas and the techniques I used to make this shoe rack can also be scaled up or down for your needs.

Shoe rack design

The shoe rack will get heavy use as there are up to 4 classes a day at the dojo. On rainy and snowy days, the shoes can be wet. For durability and water resistance, I decided to make the shoe rack out of 19mm (3/4") particle board with melamine surface on both sides. I will cover any exposed edges with melamine edge banding.

Children tend to lean on or step on the shoe rack. To help resist damage, I will reinforce the shelves by adding a lip to the front.

To keep water from wet shoes from dripping on the floor, I will tip the shelf toward the front by 50mm (2"), or about 12 deg. The lip will then catch any water.

The melamine boards I used come in 2.4m (8′) lengths. Even with the reinforcement provided by the lip, I decide the shelves should be 1/3 the length of a full board or 0.8m (2′-8") long. (I felt that a shelf 1/2 the length of a full board or 1.2m (4′) could sag or break.)

To ensure tall boots can fit on the shoe rack, I space the shelves 250mm (10") apart.

I want every child to be able to reach every shelf, so the rack will only have 4 shelves, making it 1m (40") tall. This also  leaves room under the shoe rack for any overflow shoes.

Finally, the rack will be heavy, so I  will include two holes, each 100mm (4") in diameter on each side near the top to act as carrying handles.

There are many ways to attach the shelves to the sides of the shoe rack. I decided to rout out a mortise into the sides and use glue and screws to attach the shelves. The mortise should support the weight of a child without a bracket and also help resist racking.


Figure 1. Exploded view of the shoe rack

The basic design of this shoe rack is very versatile. More shelves can be added to make a taller rack for a walk-in closet. Or it can be made with fewer shelves and using furniture grade wood to fit the entryway of a Japanese-style home.

Cut the stock and prepare the shelves

The first step is to cut the sides and shelves to length using a panel saw, table saw, sliding miter saw, or handheld circular saw (Fig 2a). Use a sharp blade and push the stock (or the saw) slowly to get a clean cut.

I used a circular saw to make my cuts. Notice that I use a 50mm (2") thick sheet of polystyrene foam insulation as a sacrificial backing board under the work piece. It’s cheap, flat, smooth, light, easy to keep clean, sturdy, and protects your blade (or router bit) from damage.

Each shelf will consist of a bottom and a front cut from a single piece (Fig 2b). The front will have a dado that the bottom will fit into. Use a table saw with a dado bit to plough a 19mm wide by 5mm deep (3/4" x 0.2") dado (Fig 2c) on each shelf. The dado will be on the front of the shelf and will receive the bottom. Rip the shelves to separate the bottoms from the fronts (Fig 2d). Before cutting the actual stock, make a prototype and check dry fit (Fig 2e).

CutParts  ShelfDetail2

PlowDado  RipShelf


Figures 2a to 2e. Cut parts to length; Design detail for the shelves; Plow dado for shelf front; Rip the shelf to separate the bottom from the front; Test dry fit of prototype

Note: Table saw riving knife, blade guard, and dust collector removed for demonstration only. Do not operate power tools without safety features in place

Assemble the shelves

Dry fit the shelf bottoms and fronts, selecting the better surface to be the top. Then glue them up, clamp, and let dry overnight (Fig 3a). Notice how I clamped all four shelves at once. I laid heavy tiles on top to keep the shelf bottoms from popping up. I also put shims underneath the back of the shelf bottoms to keep the shelf bottoms parallel to the bar clamps and the shelf fronts perpendicular to the bottoms.

Unclamp the shelves and stand them front side down. The backs of the shelf bottoms will have a raw edge. Cover them with melamine edge banding (Fig 3b) and trim off the excess (fig 3c). For more details on using hot-melt adhesive banding, see this Oct 2012 blog post.

I did not bother to add banding to the raw edge on the bottom of the shelf front.

GlueUpShelf  ApplyBandingShelf


Figures 3a to 3c. Glue up the shelves; Apply melamine edge banding to the back of the shelf bottoms; Trim the banding

Mortise the sides

To cut a clean mortise, you must use a template. See this Jun 2019 blog post on how to make a perfect mortising template for a shape that consists of right angle corners.

Using a pencil and straight edge, lay out the position of the shelves every 250mm (10") on the left and right sides with the 50mm (2") slope. Lay the left and right mortising templates onto the sides and ensure they align to the pencil marks.

Remove the templates and drill two 4mm (5/32") holes in the sides for each shelf (Fig 4a). There may be a small amount of tear out. Flip the side over and countersink the screw holes by 3mm (1/8")  (Fig 4b).

Flip the sides over again, align the templates to the pencil marks, and rout out 5mm (0.2") deep mortises using a plunge router (Fig 4c). If you do not own a plunge router, first use a Forstner bit to create 5mm (0.2") deep holes for the router bit. This will prevent kickback that could gouge the sides or the templates.

DrillPilot1  DrillCountersink


Figure 4a to 4c. Drill the pilot holes; Countersink the pilot holes; Use the jig and cut the mortises

Add handholds and banding

The shoe rack will weight about 25kg (55 lb). To make it easier to move, cut handholds at the top of each side. The most important consideration is to avoid tear out. Mark the location of the center of the hand hold and drill a pilot hole through the side (Fig 5a). Using a drill with a 100mm (4") hole cutting bit to cut through one surface of the melamine (Fig 5b). Make sure the drill is up to speed before making contact with the melamine surface to reduce tear out. Press down on the surface lightly so that the speed stays high. Cut about half way through the side.

Flip the side over and repeat on the other side (Fig 5c) until the hand hold is complete. Cutting from both sides rather than one side reduces chance of tear out.

Cut the melamine edge banding to length (Fig 5d) and apply it (Fig 5e). Note that you probably cannot get a hot iron to follow the concave surface of the handhold and get hot-melt banding to adhere to the surface. Instead I used self-adhesive banding. The adhesive is not as strong as hot-melt adhesive.

CutHandhold  CutHandhold2

CutHandhold3  ApplyBanding


Figure 5a to 5e. Drill a pilot hole for handholds through the side; Use hole cutting bit and drill through one surface of the melamine; Turn over a drill through the other surface; Cut and lay out the self-adhesive melamine banding; Apply the banding

Final assembly

Before final assembly, remove any adhesive residue from the shelves and sides using mineral oil (dissolves most sticky adhesives) followed by warm soapy water (Fig 6a).

Label the four shelves to indicate which mortise it will fit into, selecting the best shelf for the top, which is the most visible. Use a chisel to clean the corners of the mortises and to make any adjustments to ensure the shelf fits (Fig 6b).

Dry fit the shelves into the sides and drill two 2mm (5/64") pilot holes in each shelf (Fig 6c). Glue up the shelves to one of the sides and drive two 4mmx40mm (#7 x 1-5/8") screws into each shelf (Fig 6d). Flip over and repeat on the other side. The screws will hold the shelves in place so there is no need to use clamps.

After the glue has dried, fill the screw holes and any gaps in the joints with white caulk.

RemoveGlue  ChiselCorners

DrillPilot2  ReinforcementScrews

Figure 6a to 6d. Clean the finished shelves and sides; Use a chisel to clean the corners of the mortises; Drill pilot holes; Glue up, add screws, flip over a repeat

[Update: There was a flaw in my design of the shoe rack. The 10" depth for the shelves works fine for children’s shoes, but not for adults. The solution was to turn the rack around. A picture is shown below. I think it looks fine this way.]


Figure 7. The shoe rack flipped around so that adult shoes will fit