About Me

My Photo

I previously worked on Virtual Reality and other hardware at Valve.  I currently work at Google[x].

Prior to starting at Valve, I built computer peripherals such as keyboards, mice, and joysticks that were designed to be used inside MRI machines.  My company, Mag Design and Engineering, sold these devices directly to researchers at academic institutions who used them to publish scientific papers in peer-reviewed journals.

After work, I spend time on many different types of projects that usually involve circuit design, machining, material selection, and general fabrication/hacking.  My favorite place to be is my home workshop.

ben dot krasnow at gmail

http://www.youtube.com/user/bkraz333

http://www.linkedin.com/pub/ben-krasnow/4/6a9/679

http://www.twitter.com/BenKrasnow

Tuesday, July 21, 2009

Fixing a misfire during acceleration in a 1992 Honda Civic

I took some packages to the post office today, and was surprised to find my 1992 Honda Civic was having a very bad day. It was misfiring during acceleration at low-mid RPM. Idle was smooth and cruising was fine, but any normal amount of acceleration from a stoplight or stop sign produced an extremely loping, uneven amount of power. The car was drivable, but just barely.

I took out a couple spark plugs, and was surprised to find the gap had become huge -- around .075"! The plugs appeared to be worn out. I figured this had to be problem. I got some new NGK plugs, and unfortunately the problem persisted. I checked the compression in each cylinder, and they were all 150 psi +/- 5 psi. After a short consultation with my dad, I decided to replace all of the common ignition parts with real Honda replacements: distributor cap, rotor, and plug wires. This did the trick! The car was back to normal, and running strong. The original cap and rotor were Honda parts, but the plug wires appeared to be aftermarket (Prestolite). Dad and I agreed that aftermarket plug wires should be avoided altogether. Just bite the bullet, and buy OEM wires. They'll last much longer.

Even though it is difficult to see in this photo, the old plug (right) has a gap almost twice that of the new plug (left). This too-large gap in all four plugs might have contributed to the destruction of the plug wires.

Monday, July 20, 2009

Finishing the refrigerator conversion project

In my last post, I had just finished charging the "new" refrigerator with R134a. I was surprised that it worked at all, and it was still working the next day, so something must be correct. I used some expanding foam to fill up the large square hole left by the thermoelectric module in the bottom of the fridge. The clear vinyl tube is a hose that I connected to a drain on the inside of the fridge. If (when) the evaporator defrosts, I am sure there will be a fair bit of condensate dripped out. I put a small stainless tray on top of the compressor to catch the water.

I also used the fridge's existing thermostat (the white tube on the right). The manufacturer setup the thermostat so that it would regulate the temperature on the evaporator itself. The thermostat has a capillary tube that is clipped onto the evaporator and a large knob with numbers 1 to 7 (like any refrigerator). I also filled up the hole through which the refrigerant lines and thermostat enter the fridge with expanding foam.



While the existing thermostat works well, the table has a nice cutout for a two-digit 7-segment display and four buttons. The thermoelectric fridge came with a digital thermostat, and I made the cutout in the tabletop specifically for it. I would like to keep the digital thermostat feature, so I am building one with an Arduino microprocessor.

Here is the rear of the completed table with white thermostat control knob at lower left. I've already written some Arduino code for the digital thermostat, and will probably be implementing that soon. In the meantime, my table refrigerator is now frosty cold, and using less power than the thermoelectric model.

Friday, July 17, 2009

Retrofitting a thermoelectric refrigerator with a conventional compressor system

A couple years ago, I built a nice wooden table that housed a small refrigerator. The idea is that it keeps drinks handy in the living room for parties and guests. It's also cool, and fits with my idea of what "functional furniture" should be.

I chose a refrigerator meant for storing wine because it had a nice glass door, was pretty small, was very quiet, and could operate facing upwards because the thermoelectric device doesn't care about its orientation. I knew that thermoelectric refrigerators generally suck at actually refrigerating, and this one was not exception.
On really hot days, the fridge would get up to 5o-60*F, which is cool, but not nearly as nice as drinking 40*F beverages. The other problem is that it draws 70W, essentially constantly. This comes out to 613 KWh/year. Most small conventional refrigerators use around 300 KWh/year. Also, I like to tinker and wanted to mess around with a refrigerant system.



I bought a small 1.8 cu ft Haier fridge off craigslist for $15. Geez, can't beat that! My first task was to pull out all of the system components from the insulated box. This required draining the system of refrigerant through a small hole that I drilled in the compressor fill tube. Before doing this, I checked the label, and it indicated the system used 1.6oz of R134a. I can buy R134a at the auto parts store, so I will be able to refill the system.


I de-soldered the tiny capillary line from the accumulator / dryer and de-soldered the suction line on the compressor. I used an oxy-acetylene torch to heat the joint, then I just pulled the tubes apart when the solder became molten.

Here is the evaporator liberated from the fridge.



Luckily, with one flap unfolded, the evaporator fit perfectly in the thermoelectric cooler (TEC) fridge.


I mounted the compressor underneath the table (also getting very lucky that there was enough clearance). I cleaned the copper tubes carefully, then re-soldered them with silver solder and some paste flux. I had to extend the suction tube by a few inches, and so I just got some tube that fit around the existing line and soldered it to the outside.


I soldered the capillary tube back into the dryer, and put a valve on the compressor fill tube. The system was sealed at this point. Now, I connected my vacuum pump (Welch 1400) with a micron vacuum gauge and a tee that connected to a can of R134a. I positioned the can on a sensitive scale so that I could meter out 1.6 oz of refrigerant.

I pulled a 375 micron vacuum in about 15 minutes or so. I even ran the compressor while the system was under vacuum. It raised the pressure just a couple hundred microns, then it settled back down quickly, so I felt the system was dry and sealed.


After I metered out 1.6oz, which only took a few seconds with the valve just cracked open, I tightly closed the valve that I added to the fill tube, and was very pleased to see feel the evaporator getting very cold.

Tomorrow, I will re-insulate the TEC fridge and hopefully give it a final test.

Thursday, July 16, 2009

Experimenting with a "liquid lens"

A couple months ago, I started to investigate building an auto-focus mechanism for cameras that could operate in high magnetic fields. This was for a work project (where I build MRI-compatible devices). Inside an MRI machine, the base magnetic field is usually 1.5 or 3 Tesla. At these fields, any motor or solenoid that relies on manipulating a local magnetic field will not work. Unfortunately, manipulating magnetic fields is one of the easiest ways to convert electricity into physical movement, so most auto-focus assemblies are built this way. There are some
alternatives, though:

Piezo "ultrasonic" ring motors. Many high-end DSLR (and SLR) lenses have silent ring motors that vibrate at ultrasonic frequencies to create rotation. Great! The only problem is that they are built from steel, and the steel cannot be used in the MRI environment. They are also too big -- I need a tiny lens about 1/4" in diameter, not 2" like a DSLR lens

Piezo wiggler motors. Looks interesting, but getting a sample is near impossible. The screw assembly is probably steel, and the system requires a fairly rigid and complex slide system

Liquid lenses. These lenses are built by sealing a small amount of water and oil inside a shirt button-shaped capsule. The walls of the capsule are metal with glass windows on the faces of the button. The water and oil will not mix, and the boundary between the fluids will bend light because the two liquids have different refractive indices. If the walls of the capsule are charged with electrical current, the shape of the water droplet will change because of electrostatic attraction. (check this out: http://www.youtube.com/watch?v=p1f6zLysilU ) The oil is made up of non-polar molecules, so it does not share this same attraction. By controlling the voltage on the capsule walls, the shape of the water can be controlled and thus the focal length of the lens formed by the water-oil interface. Neat!


The liquid lens is the metallic button-shaped thing on the left. I built a holder for the lens out of 1" squares of copper clad board.

There are only one or two manufacturers of liquid lenses: Varioptic, Philips, and maybe Siemens. Philips and Siemens would not discuss the liquid lens and would not sell any samples. Varioptic would send samples -- only $340 for each lens! What?! The evaluation kit was a mere $4000, while the full developer's kit was $12,750. I kid you not. What kind of company would charge a potential customer $12k for the privilege of testing their products? Insane!

Anyway, since I couldn't deal with any of the manufacturers directly, I set about finding an off-the-shelf product that contained a liquid lens. After many hours of searching, I found an actual product that had the Varioptic lens in it: Digitus DA-70817. There were many other prototypes and press releases for other liquid lens products, but it seems the Digitus camera is the only mass-produced device right now. Unfortunately, it is not available in the USA. It's not on eBay. Amazon.de has it, but will not ship to the USA. After many more hours of searching, I got in touch with a German distribution company called Assmann (no joke). I exchanged a few emails with a friendly person there, and she sent me five of these cameras (about 25 euro apiece with the usual outrageous international shipping charge).

I removed the lens from one of the cameras and mounted it between two pieces of copper clad board. To make the lens change focal length, I need to supply 0-70V AC to the lens. AC is required to prevent the liquid in the lens from polarizing over time and losing effectiveness. The voltage needs to be fairly high to deform the water drop sufficiently, and also needs to be changed rapidly for quick changes in focal length. Fortunately, there are two companies that make ICs specifically designed to directly drive liquid lenses. Unfortunately, one of the companies, Maxim, will not sell samples of their chip, and they will not even give out datasheets for it. Take a look: http://www.maxim-ic.com/req_full_ds.cfm?action=request&id=5949 How rude is that? WTF?! The other company that makes liquid lens drivers is Supertex. They will sell samples of their HV892 at a reasonable price, but the chip is only made in a 10-lead 4mm x 4mm DFN package with .65mm pitch. Apparently, this SMD footprint is so unusual, there are no DIP converters available from anywhere. Geeeez! I ended up designing my own PCB for the chip and having them made at http://www.expresspcb.com/. Here it is:

I interfaced the chip to an Arduino Duemilanove. The Arduino sends a value between 1 and 255 to the HV892 via the I2C bus. The value controls the HV892 output voltage, and hence the lens's focal length. I am not sure what the mathematical relationship is





I used a laser pointer to show the different focal lengths that the lens can achieve.

Friday, July 10, 2009

Sharpening planer / jointer knives

In short, sharpening planer and jointer blades is more difficult than it looks, and is probably not worth it. Get them sharpened or buy replacements if you value your time. I originally was inspired by this article:
http://www.ablett.jp/workshop/oak/doug_abbott.htm

I built a similar jig that looked like this:

I had used the sandpaper-on-glass technique for other projects, so I felt it could work well for sharpening knives. It works, but it's SLOW. My knives were in admittedly bad shape (very bad shape -- see picture) and needed a lot of material removed. I resorted to using a power sander to take off the initial material. Also my jig is not as good as the one shown in the link because tightening the screws will make the jig not be perfectly straight anymore.


Before and after

Yeah, it's sharp! The blades appear to be bi-metallic. I am not sure if all planer blades are like this, but there are definitely two different types of steels joined together.

Compressed air dryer / nitrogen generator

In my last post, I described designing and welding a pressure vessel for a compressed air dryer. I'll describe the completed dryer in the this post.

The purpose of this project is to take dirty, oily, damp air from a compressor, and provide very clean, very dry nitrogen to the liquid nitrogen generator.


The airflow path is as follows:

Air inlet -> mechanical filter -> carbon filter -> air dryer #1 -> air dryer #2 -> carbon filter -> mechanical filter -> humidity sightglass -> nitrogen membrane -> flow valve -> output

Oil vapor is a big problem when using conventional compressors with systems that rely on having clean air. Mechanical filters are not able to capture all of the oil odor, so that is why I also included activated carbon filters. I made these filters out of Harbor Freight pneumatic oilers. I gutted the oilers, added some filter material, rearranged the airflow path and threw in some activated carbon. Even though the photo shows the two filters joined together, this is just a mechanical connection -- each filter is actually plugged off. The flow goes between the bottom of the bowl and either the left or right outlet.

The whole system is fitted with 1/4" compression fittings.

Machining and welding an aluminum pressure vessel (air dryer)

I recently decided to build a compressed air dryer for use in my liquid nitrogen generator project. I will be using silica gel beads (the same stuff as found in those ubiquitous "do not eat" packets) to soak up moisture from the air. In order to do this, the silica gel must be contained in a vessel that will withstand the pressure of the compressed air. In this case, I designed the cylinders to have a working pressure of 150 psi. The system will normally operate around 100 psi. The vessel should also be fairly long and narrow to ensure the air flowing through it has enough time to make good contact with the silica gel.


I started with some basic engineering equations for a thin-walled cylindrical pressure vessel.
http://en.wikipedia.org/wiki/Radial_Stress

I already had some aluminum pipe that I felt would be suitable and checked it with these equations. The pipe is 3" in diameter and has a .0625" wall.

The tangential stress is = (150) * 3)/(2 * 0.0625) = 3600 psi
The axial stress is = (150 * 3)/(4 * 0.0625) = 1800 psi
The radial stress is = -(150) / 2 = -75 psi (negligible, the negative indicates compression)

In order to determine if this amount of stress is going to break my aluminum cylinder, I used the Von Mises stress calculation for multi-axial loading:
http://en.wikipedia.org/wiki/Von_Mises_yield_criterion

The Von Mises stress in the walls of my cylinder is:
= sqrt[ ( (3600-1800)^2 + (1800-3600)^2 + (3600 - -75)^2 ) / 2 ] = 3161 psi

In this case, the Von Mises stress is actually lower than the tangential stress component alone. This is because the walls of the cylinder are being pulled in two orthogonal directions, thus reducing the amount of shear that would be produced if the cylinder wall were being pulled in only one axis. In order to be as conservative as possible, I'll use 3600 psi as the load stress.

The cylinder is made from aluminum alloy 6061, which has a yield stress of at least 8000 psi. It's likely much higher with T4 or T6 heat treatments, but I will be welding this material, and I'm not sure what effect that will have on the yield stress, so I'll be very conservative and stick with 8000.

Clearly the 3600 psi load is much less than 8000, and this design has a safety factor of 2.2. Working backwards, the tank will hold 330 psi before suffering permanent damage. Again, these figures are likely to be very conservative.

I also calculating the plate deflection for cylinder's end caps, and it was insignificant.


I cleaned up the cylinder by turning it on the lathe and running some sandpaper over it.


The end caps are 1/8" thick and have a step turned on their edge to make placement and welding easy.


I made some bosses that will be threaded later. Luckily, the diameter of my horizontal belt sander drum matched the cylinders' diameters perfectly. The boss will sit flush up against the cylinder wall for easy welding.

Welding!


I milled a flange for the pressure vessel to hold an O-ring the in groove and the holes will be tapped for 1/4-20 bolts. The flange will be bolted to a 1/2" solid aluminum plate. This was done so that the vessel could be removed from the plate, and the silica gel could be replaced easily.



I tested the vessel, and....... it leaked! I had a tiny pinhole leak in one of my welds. It was so tiny, I could barely see the imperfection. I repaired the leak, and pumped the tank up to about 220 psi. Nothing was leaking or breaking, so I considered it a success.

Wednesday, July 8, 2009

Comparing different brands of acrylic and cement to get the best joints

Getting really clear, bubble-free joints in thick acrylic is not easy. I've done quite a lot of practicing with various methods and eventually did some testing on the materials and cements that I was using. The link below shows the results. Since McMaster plastic is a lot less expensive than TAP plastic, I tend to use McMaster to supply the plastic and I use TAP green label cement on it. I've never tried IPS #3, since I've been told it's identical to TAP green label. I'm not convinced about this and may have to try it some day.


http://www.magconcept.com/acrylic/

Tuesday, July 7, 2009

Replacing a Honda Civic condenser fan without draining the freon

My 1992 Honda Civic recently developed a strange problem where the engine would idle badly and sometimes stall when the air conditioning was being used. Oddly enough, the problem turned out to be a non-functional condenser fan. Presumably, the compressor would pump the high side of the system up to an abnormally high pressure because the system was unable to dissipate heat from the condenser. Eventually the mechanical load from the compressor was so great it would cause the engine to stall. I would have guessed there was a high pressure limit switch that would deactivate the compressor, but maybe not.

In any case, this is an R12 system, and I did NOT want to open any of the freon lines. It had been working great ever since I bought the car (until this recent problem), and I did not want to mess with it. Luckily, it's possible and not too difficult to change the fan motor without opening the freon lines.
The fan shroud is bolted to the condenser with two 10mm bolts near the top facing into the condenser. There is also a small bracket on your left that has two 10mm bolts to hold the bracket to the condenser and shroud. I removed all four of these bolts, two bolts that held the shroud to the car's frame at the top, a bolt that holds a freon hose to the shroud, and also a bolt that holds a relay to the frame (on your right).



The shroud can be lifted up and out of two pockets that are formed into the metal of the condenser. It cannot be fully removed from the car because there is a freon hose in the way, and the hose cannot be moved. Instead, I removed the plastic rock guard on the underside of the car, and slipped the shroud out through the bottom. There are a few wire-guides and connectors on your right that need to be removed before the shroud will clear the car. I didn't even need to jack the car up.



Once the shroud (and fan) were out of the car, I switched the motor with the new one. Do NOT forget to also move the tiny washer from the old motor to the new one, which tends to stick to the old motor shaft and fits so tightly it almost looks like part of the shaft itself. I made this mistake and it nearly ruined my day. Without the washer in place, tightening the nut will cause the metal hub of the fan to be pushed in a bad way so that it is no longer engaged with the shaft. Since there is no other way to grab hold of the shaft, you are left with trying to get the nut back off and no way to stop the shaft from turning. After fiddling with it for a long time, I remembered my dad recommended the use of a pneumatic impact wrench in situations where a nut must be removed from a free-turning shaft. The idea is that the impact wrench hits the nut so hard and so fast, it is able to back it off while the inertia of the shaft holds the part still. Lo and behold, it zipped the nut right off. Thanks, dad!


The rest of the job was pretty straight-forward, and now my air conditioner works like a champ and doesn't stall the engine.

The old motor appears to have died of old age. There aren't any catastrophic problems, but the brushes and commutator show heavy wear. The motor case was full of carbon dust from the brushes.

Monday, July 6, 2009

Repairing a wireless keyboard

Last night, I sat down to use my home theater (a PC connected to a DLP projector) and had some problems typing in the name of a movie into Netflix. I thought I had mistyped the word, but I soon realized that my wireless keyboard had some non-functional keys on it. The letter 't', numbers 1 and 3 among other keys were not working. Since the failure affected more than one key, and pressing harder did not help the situation, I reasoned the problem was one of the matrix lines in the keyboard's circuit and not a bad connection in the button itself. The rest of the keys worked fine, so the microprocessor and transmitter were probably OK.

I took the keyboard apart and located the matrix lines between the microprocessor and the array of buttons. I used an oscilloscope to see which lines were pulsed (outgoing from the microprocessor) and which were switched (incoming to the microprocessor). I also compared the signals from the working "2" key to the broken "1" key. The signal was present but weak for "1". I tested the total circuit resistance for "1" and "2". "2" (working) was a couple hundred ohms, but "1" was 10k ohms. The circuit had a bad connection somewhere.
I used my meter to trace the bad connection to this spot on the flexible circuit board beneath the keys. It looks fine visually, but there is an electrical discontinuity right in the very center of this photo where the trace becomes narrow and passes above the middle rectangular cutout.

I used some conductive paint (marketed for repairing automotive defrost grids) to cover the bad area of the trace.

I was tempted to use conductive epoxy for the repair, but it can be very brittle and has low adhesion qualities. We'll see how the defroster repair product holds up.


Saturday, July 4, 2009

Woodworking with a piece of raw tree trunk

My friend recently had a tree die and fall over on his property. He cut it up with a chainsaw, and offered a few of the pieces to me for use in woodworking. I have never started a project with just a raw log, so I figured it would be a fun learning experience. The tree died in the winter, and it was soaked with rain water, so I put the pieces on a few bricks to keep it off the ground and covered it with plastic on rainy days, leaving it uncovered on dry days. After the rainy season was over, I removed the plastic, and left it sitting in the sun for few months. It's dry now, but badly cracked. I have a feeling it may have been cracked even before the tree even fell over, but I would be interested in hearing from anyone who has experience drying logs.

I chose one of the smaller logs and sliced off a piece with the bandsaw.


Next, I clamped a 4x4 to the bandsaw table to act as a crude fence. It's set to a little over 1" of thickness from the blade.

I jointed the exposed log face before cutting each 1" slice with the bandsaw. I then jointed the other side of each board, and also jointed one edge. I then planed each board to exactly 1" and used a table saw to square up the other side. I now have flat, square stock ready for the project. The wood had lots of cracks, but the figure was really pretty. I think this was an almond tree.


I ended up cutting the boards into 1"x1" x 12" long strips. I decided to make a napkin holder, since it was something that I could build with a small amount of wood, and would be useful. I used standard yellow wood glue and only used tape while drying -- no clamps.


I used a 1/8" radius round-over bit in my router table to smooth the edges, then sanded with 150 grit on a random orbital sander, and also did some hand sanding. I applied a Tung oil finish (my favorite finish) in a few heavily-rubbed coats.