In a perfect world, every dimension on a part would match the drawing exactly and there would be no variation between parts. In the real world every manufacturing process has some amount of uncertainty, a statistical account of the variation from part to part.  To ensure that all of the parts will fit together nicely even with the part-to-part variation, we create a calculation often called a “tolerance stack-up”.

Input

The first input is specified by the manufacturers of off-the-shelf parts. These are fixed inputs to the tolerance calc. “As built” parts may or may not be tighter than the specified tolerance, but it probably wouldn’t be prudent to depend on that. Unfortunately, most manufacturers will not adjust the given tolerances unless you’re buying big, big quantities, so we treat these values as fixed.

The second input is based on the process – these are slightly adjustable input tolerances.  For example, most machinists should be able to meet a standard tolerance for machining steel or aluminum of 0.005”. However, you can specify better tolerances if necessary, for the right price. Plastic part tolerances can vary widely based on the molding process, part geometry, and material characteristics, but a molder should be able to give guidance.

Stack it up

Once we have all of the inputs, we create a table to determine the overall tolerance. For medical devices, we often consider absolute tolerances, adding everything up as stated to account for 100% of the variation. As a less-conservative alternative, the Root Sum of the Squares method accounts for a large portion of the variation by accounting for independent variables differently than dependent variables. As the number of variables increases, the probability that all of those variables measure to their extreme tolerance values independently gets smaller. Thus, the population that is accounted for gets larger, but it never reaches 100%.

Summing tolerances together is as simple as adding the values and adding the tolerances to determine the total value. However, subtraction, multiplication, and division are a little more complicated. This document explains the methods (and a few examples) for accumulating absolute tolerances for some of these cases.

Output

The output of the tolerance calculation should ideally be a dimension, such as the width of the display window for the enclosure. For example, if you’ve added up all of the tolerances to be 0.017” and the width of the display is 3.886” ± 0.0197”, the display window should be 3.886” + 0.0197″ + 0.017″ = 3.923”. These are the inspection dimensions the quality control department will be checking against before a new fabricated or off-the-shelf part makes it into the assembly.

If Erik “with a K” Askin sounds familiar, it’s probably because his work has recently been featured on Core77, where he designed a hard to use packaging to discourage the use of cigarette smoking. We are excited to have another talented designer at Key Tech. Watch for more of Erik’s work soon and keep another eye out for him cruising on his fixie through the streets of Baltimore!

Photo credits: Erik Askin

NJT28KF249CD

Our staff of cross-disciplinary engineers, an industrial designer, admin support, and interns all work together to create remarkable solutions every day. We usually have 5-10 projects going on simultaneously, so it’s tough to stay current on everything that’s going on outside of my own projects. Just recently, I was getting the 10-second tour of a prototype when I started asking about a robotic end-effector densely packed with numerous individual pneumatic actuators and complex linkages.

“Oh, that? Yeah, we had to make almost all of those parts ourselves – just the pneumatic cylinders and screws are stock. Assembled, the density of the tips is probably twice anything we could find off the shelf.”

Seth Godin writes about having one or two Linchpins that can make a company. These are the indispensible go-to people that are creative, willing to be unconventional, and passionate about their work. They make their work personal and are not satisfied until it’s done right. I don’t know anyone at Key Tech that doesn’t act this way.
Most of our work is highly confidential, so remarkable work rarely gets remarked upon outside of these walls. But, internally, annual reviews often include feedback from peers and project managers that looks like this:

“I’ve worked with him on the ************* project and it has been awesome. He has been handling the many headaches of designing, re-designing, and re-re-designing the critical [custom hardware] for the various systems. Additionally, it has been a pleasure working with him the whole way. He is always in good spirits, does his work in a very timely fashion, and still finds time to host things like ASME meetings. Very impressive. No criticisms what-so-ever.”

So, thanks. You rock! It’s pretty cool to work with all of you.

If you’re not the one being wheeled in, when you enter an operating room you’ll see a lot of duplicate hardware. Multiple displays and computers, various sensors and diagnostic devices to monitor the patient. And, none of it knows what the other devices are doing. If every device could communicate over common protocols, at least some of that hardware could be eliminated. This could cut the cost of the devices and reduces the clutter in an already cramped space. Communication could also improve the techniques by which health care is administered. But, is it possible to verify and validate the safety and efficacy of medical devices if they could be connected to unknown current and future products?

In the current system, the answer is probably “No”. But, if inputs and outputs could be strictly enforced and regulated, as is the case with FDA-regulated medical devices compared to the “wild west” of computer and home theater peripherals, perhaps it wouldn’t be so difficult to imagine a future with compatible tools.

The medical device industry has not ignored the potential of device interoperability. In fact, it is a recognized issue. There have been many conference presentations and on the subject, and an FDA held a workshop this past January focused on working out these exact issues. While I don’t expect the solutions to arrive overnight, I do look forward to seeing developments coming down the pipeline.

Frank Regan has just made the long trek to Jinja, Uganda with The Giving Circle to help bring needed resources to a Wanyange village orphanage, the Koi Koi House. Since March of this year, the nonprofit has been working to build a permanent home for orphans afflicted with AIDS. So far, the group has purchased land and already helped dig a well. Among other things, Frank will be working to bring electricity to the village by designing and building a small windmill. We’re following their progress on their blog.

This is a very worthy cause. Of course, back at the office, that gives us two weeks to setup a few practical jokes.

Good luck, Frank.

Between performance, theatre, film, paintings and sculptures, the average attendee is hard pressed to not find some type of art that connects with them. This year, for the first time, Artscape included an area called Betascape, a geek/nerd/DIY-enthusiastic gathering full of home built robots, hacked circuitry to compose music and power tool races. You could even listen to the activity of ants in an ant farm by way of a few infrared LEDs and photodetectors.

One of the really interesting exhibits was an electronic marionette of a wizard. Click the first video to see the wizard’s song and dance, and then click on the second video to see the set of servo motors and circuitry that make it all work. It goes to prove that there’s no reason engineering can’t be a form of art when enough creativity is applied.

A 10 foot Zorb ball sent Key Tech employees and their families rolling down the terrain at Hilltop Farm where Scott and Ben live.  We usually have a stunt or two that involves a lot of screaming, such as a 1/4-mile zip line or 100′ tall super-swing, but this year, it was mostly hysterical laughing all the way down. And, many of the riders went in pairs!

The double sectioned ball is comprised of an outer 10 foot ball and an inner 6-7 foot ball which acts as a shock absorber.  The rider or riders can be strapped into the inner cavity of the ball or ride down inside freestyle.

Many thanks to the ball stoppers at the bottom of the hill who were knocked down, run over, and generally thrown about the yard.   Without you we would have certainly smacked into a few trees.

Let’s go to the videotape! Tim recorded several videos of the event and posted them on YouTube for all to enjoy. Below is one of them, but you can check the rest out here.

Are you a tinkerer? Even among engineers, I find that very few of us are willing to delve into the guts of a device with complexity outside our comfort range. I think kids have the advantage of not fearing whether they will be able to fix it later. They don’t think about how much it cost or whether it should be taken apart, and they’re confident an adult can provide a safety net. But, for those of us that are concerned about how best to crack open a $1,000 MacBook with a critical soda problem yet have the compelling desire to take it apart anyway, there is help.

Online video isn’t just for watching other people’s attempts at stardom. There are a host of instructional videos put together by generous people willing to share, learn from their mistakes, or just help you recover from what could be a $1,000 mistake. A quick search of YouTube or iFixIt can provide step-by-step examples of how to take apart a MacBook, replace a cell-phone screen, or repair a chainsaw carburetor.

As a tinkerer, perhaps you want to just build something yourself instead of buying it from a store. You can learn how to build a great HDTV Antenna from coat hangers or an automated Nerf Sentry Gun to protect the office.

Enjoy! (and let me know how you made out)

Disclaimer: Umm, it’s the internet, so you can’t always believe what you see, even in a video. Use your head.

The Scale

We chose an Acculab 210.4 scale because it was accurate enough (.0001 gram resolution, which is about a tenth of a microliter for room temperature water) and because it has a serial interface that we can call from Visual Basic to easily collect data and put into a spreadsheet. The data stream is limited to about 8Hz, but it was fast enough for our purposes.

Scales with more resolution often come with damping mechanisms to keep them stable. Although the glass cover keeps the air movement down, this one didn’t have a means to damp vibrations, so we needed to isolate the scale ourselves. We used a small table with some rubber feet as the foundation. Then we used one of those heavy engineering textbooks we have so many of on some thick foam and put the scale on top. Considerate use of flexible beams, soft materials, and masses meant the scale was rock-steady without any software averaging required.

The Fluid Connection

To get the fluid onto the scale, we attached tubing with a luer-lock fitting to a long needle inserted into the scale cover through a hole in the top plate and held in place with a thumb-screw. By using a needle instead of just inserting the tubing into the container, we eliminated any contact with the container or scale. we found that any such contact threw the scale out of whack, especially when the fluid flowed through the tubing with momentum.

The question came up as to whether the needle should be above the water-level or below it. If the needle is below the water-level, surface tension on the needle can reduce the weight of the water. However, if the needle is above the water-level, a drop can form at the tip instead of depositing the small volume onto the scale. We needed fine-resolution flow-rate data, so we opted to ignore the surface tension, which we found to be below the resolution of the scale.

Evaporation of the Water

Over the duration of a test, water will evaporate. The rate is dependent upon temperature, relative humidity, and the surface area of the container. Many suggest adding mineral oil to the water to create a barrier layer to supposedly eliminate evaporation of the water. Data showed that the oil layer did reduce the rate of evaporation (from 0.13 g/hr to 0.08 g/hr), although the oil ruined the plastic connections of the tubing by making them brittle and causing them to leak and break. Instead, we drilled a 1/4” hole into the cap of the container for the needle to fit through, which dropped the rate of evaporation by two orders of magnitude. We collected data for various configurations of the water, oil, and lid and plotted them below. So, without using any oil, we were able to drop the rate of evaporation from 0.127 g/hr to 0.002 g/hr.

Volume and Flow Rate

So, how does a measure of weight relate to volume and flow-rate? Well, the density of water is pretty well defined based on temperature (and weight / density = volume). Accounting for the error of this density value, as well as errors in the scale resolution and time resolution provided excellent measurement results by which to evaluate the devices.