Jay's Technical Talk » Projects

Using refrigerator door bins and vegetable crisper drawers for garage organization

Jay — Mon, 23 Jan 2012 01:53:38 +0000

Cardboard boxes are fine, but refrigerator door bins and vegetable crisper drawers are nicer for storing things. The good ones I used here are made from clear plastic so you can see inside them easily and they don’t block light.

Of course, they are terribly expensive to purchase for this use, but with the (lack of) quality of Kenmore /Amana refrigerators manufactured after 2004 you can find free donor components relatively frequently. I also believe that with the addition of standard closet shelving support brackets the glass shelves from these fridges can be re-purposed into standard wall shelves.

Boost Converter Schematic

Jay — Sun, 22 Jan 2012 14:43:56 +0000

I have drawn up a schematic (click to enlarge) of the high current and sensing portions of my maximum power point tracking (MPPT) 2-phase boost converter battery charger circuit. The schematic does not include the micro-controller, MOSFET gate driver IC, and associated power supplies, as those items are on the (relatively) low-power side of things.

What do all of these things do?

L1, Q1, and D1 – These three components make up the heart of the boost converter. When Q1 turns on, power builds up in L1 as the current rises. When Q1 turns off, all of that power exits via the only available route (out past D1) and the voltage is boosted as the inductor (L1) resists the current change. If you turn Q1 on and off very quickly (under control of the micro-controllers’ PWM output via a MOSFET gate driver) it raises the output voltage higher than the input voltage.

L2,Q2 and D2 – See L1,Q1, and D1 above. These three guys act in the same way, but Q2 is turned on and off 180 degrees out of phase from Q1. So while one half of the circuit is drawing power in, the other half is pushing power out, and visa-versa. This doubles the power capacity of the circuit, and reduces the size of the filter capacitors that are needed, as they are shared between the two phases.

C1 & C2 – Ceramic & Electrolytic capacitors that work together to filter the input power (i.e. they provide short bursts of current needed when Q1 or Q2 turn on)
C3 & C4 – Ceramic & Electrolytic capacitors that work together to filter the output power (i.e. they absorb the short bursts of current leaving L1 and L2 when Q1 or Q2 turn off)

R1, R2, and C5 – Voltage sensing. R1 and R2 are a classic voltage dividing resistor network. Because R1 is so much larger than R2, the high voltage output (up to 200 volts) is reduced to a low voltage output on the “VOLTAGE SENSE” output (between 0-5 volts), suitable for a microcontroller input (A2D) pin. Because it is coming out of a switching boost converter, the output voltage has a lot of noise in it, so capacitor C5 helps to reduce this. I still had to make multiple sense readings on the microcontroller input pin and average them to get a reliable reading. I expect that if I made an analog low-pass filter on the voltage sense output it could fix this with a higher component count.

R3, U2, R4,R5, Q3 – These are the current sensing subsystem. R3 is a 0.1 ohm 5watt resistor that is acting as a current shunt. 1 amp flowing through this resistor equates to a 0.1 volt drop from one side of the resistor to the other. U2, the AD8212 chip amplifies this small voltage, and together with Q3, converts it to a low voltage signal suitable for a microcontroller A2D input pin. Currently R5 (at 100K) is set up for a 100x gain in the amplification, but I may reduce that as I start to test at higher power levels.

In reality, I have not yet soldered D2 and Q2 onto my pref-board, so the circuit is currently a single phase boost converter. This is fine, as I am not yet testing it at full power and my current microcontroller isn’t powerful enough to output a proper 2 phase control signal anyways. On the software side, I have gotten my micro-controller to regulate the circuit to output a (relatively) constant voltage regardless of the input voltage by controlling the PWM duty cycle. I have measured the current-sense output using a multi-meter and it is working, but I have not yet integrated current sensing into the control software. Here is a picture of the pref-board. The current sense module is still on a solder-less breadboard next to it.

My lack of progress on the software side is due to a lack of the proper development tool (PIC Kit 3) which is required to program the newer PIC microcontroller that I selected (PIC16f1824) to drive both phases of the boost converter. When I ordered the new PIC’s I didn’t read the specifications closely enough and just assumed that my eight year old PICKit1 would be able to program it. Silly me….Apparently Microchip has made some improvements in the last eight years. Of course, when I got the new chips, I ripped my code apart to make it work with the dual ECCP PWM modules and faster internal clock, and by the time I figured out that I couldn’t program the new chip I had already gotten the code to a state where I didn’t want to reverse all of my changes just to be able to play with the old microcontroller for a few days until the new programmer arrived. On the plus side, the new chip has an internally generated 32Mhz clock which will (eventually) let me run two PWM channels 180 degrees out of phase with a duty cycle of 125 kHz and 8 bits of resolution. It also has a UART, so I can easily use a serial LCD module for displaying status information.

The $25 Sous-Vide Egg cooker

Jay — Sun, 15 Jan 2012 03:38:36 +0000

I put together a $25 sous-vide egg cooker with a $20 PID temperature control module I purchased from ebay (which included a thermocouple and free shipping), a $2 immersion heater and a $1 extension cord (both purchased at a thrift store). I added a few crimp on connectors that I already had (but we’ll say you could get a few for $2 at the hardware store) and a free bowl and clothespin that I stole from the kitchen.

I’m happy to report that it cooked the egg just fine.

I’m not 100% certain I’m going to get into Sous-Vide cooking, but the technology has several alternative uses: unpasteurized milk heating for making cheese, automatic baby bottle warmer, etc…

A few downsides to my $25 setup:

The temperature of the water in the bowl would vary by about 5 degrees from the top to the bottom as I did not have any type of pump to circulate the water. I was hoping that convection currents would do something for me, but apparently the water just stratified.
It has a few exposed wires laying about, and a bowl of water sitting near them. For permanent use an enclosure will add to the overall cost.
It can only cook a single egg at a time. The immersion heater draws 125 watts at 120 volts (around one amp) which can easily be switched by the internal relay in the PID controller (rated up to 3 Amps). But, it won’t work with anything big (like the electric kettle I also purchased at the thrift store) until I add a bigger 10-25 amp relay (solid state or otherwise). If you plan on running a big 8-12 amp load from your PID module, I suggest getting one that has a 12 volt output designed to directly control a solid state relay.

Printer Ink, $9507 a gallon

Adventures with a boost converter

Jay — Sat, 14 Jan 2012 03:24:38 +0000

A boost converter is a dynamic electronic circuit that can “boost” a lower voltage up to a higher voltage. Current is of course reduced in the process, but if you design it right you only lose 5-25% of the overall power. I am in the process of building one that will convert the approximately 70 volt nominal output provided by my solar panels into 125-140 volts that will charge my trucks batteries. I also plan on adding a bit of maximum power point tracking and battery charge regulation to the control logic, making it a MPPT solar boost converting charge controller. Why do I have all of these wires and parts scattered across the dining room table? Because the market does not sell what I need. (Believe me, I’ve looked.) So I have decided to build what I need.

Not being an electrical engineer, I probably won’t be able to get 95% efficiency out of my system, but so far I’ve measured 85-90% efficiency when doubling voltage. [Test 1: 23 volts in at 0.28A for 6.44 watts, 46 volts out at 0.12A for 5.52 watts means 0.92 watts loss = 0.92/6.44 = 0.142 or 14.2% losses. Test 2: 34.8 volts in at 0.35A for 12.18 watts, 72.3 volts out at 0.15A for 10.84 watts, or -1.335 watts loss, which is 10.9% losses.]

I’m still running things off of a (relatively) low voltage and current supply (two/three 12V 5AH batteries) while getting everything working, and only boosting up to 50-72 volts when testing.

I had the circuit up to 95 volts (max of 16 watts) on the output side, but I really want to optimize it for a voltage doubling, and you lose a good amount of efficiency when going higher than a 50% duty cycle.

I have also not been able to find a commercially available boost regulating control IC that works with 100 volt inputs and produces up to 200 volt outputs. One probably exists, but digikey doesn’t appear to have it. I’m solving this problem by programming my own using a PIC micro-controller. At least the programmable nature of my controller should also allow me to add the battery charging maximum voltage/current and solar panel MPPT logic into the system. I am currently working on a single phase boost converter, but the plan is for the final system to be two phase (which basically means that I use two of them in parallel). This means that I need my single phase system to be able to support 100-175 watts maximum. Soon I’ll be moving from my 25 watt resistive load (25 watt lightbulb) to a 100 watt resistive load. Eventually, I may even borrow a 75 volt power supply and test it up to 140 volts.

While working on my boost controller, I’ve found the following reference links very useful:

I have also made a few mistakes that people making their own boost converter may want to avoid:

Within reason, the higher the switching frequency the better. I started off using a PIC12F683 chip which only has a single PWM output. Because I was planning on building a two phase system, I decided to code my own PWM loop that would support two phases with a 180 degrees difference. (A fancy way of saying that they would each be ON for the same amount of time, but that while channel A’s ON time would start at the beginning of the period, channel B’s on time would END at the end of the period. If the duty cycle went above 50% they would overlap for a bit in the middle.) After spending an hour on my fancy 2 phase PWM system and testing it with some visible LED’s, I took all of the delays out of it only to find that my 8Mhz internal oscillator could only support a period of 1.5-3 kHz. Sure, it sort of worked, but you could HEAR it switch. It also required a much larger inductor for the same power. So, I switched over to using the built in PWM subsystem, which can easily hit 73 kHz while still having 5-6 bits of precision. I now plan on switching over to a PIC 16F chip which has a dual channel PWM that I THINK can do an 180 degree 2 phase signal in hardware (if I’m reading the specs correctly…)
Most MOSFETs require at least a 10 volt gate signal. If you switch them with a 5volt logic signal directly from the PIC they will switch (at least mine did), but they will not carry current without burning a lot of it up as heat. I was wasting 80% of my power as heat in the MOSFET until I wired up a transistor off of a 12V rail to amplify the signal to 12 volts. I plan on using a IXDN604PI gate driver chip for my final design.

Using the Microchip PIC Kit 1 with Linux

Jay — Sun, 01 Jan 2012 22:15:40 +0000

The PICKit 1 is a combination programmer and development board for midrange PIC micro-controllers. In addition to being able to program FLASH pic devices, it also allows them to run and access 8 (charlieplexed) LED’s, a pushbutton switch, and a potentiometer. It’s a great little board for learning the basic of micrcontroller programing, but unfortunately it is not supported by Microchip’s new MPLAB X software (that is written in Java, and supports Linux/Mac in addition to Windows).

You CAN however use the PICKit 1 under linux. The piclab software is supposed to support it (I have not tested this). I use version 1.6 of the PicKit1 flash usb programmer for unix. Even the newest 1.6 version reports checksum errors after writing the hex file, but it does work correctly.

To make it work as an external program in MPLABX I had to select the “format HEX file for download” option under the Linker so that it would not fill all 2048 flash bytes when the program was smaller than that. I am able to manually run the usb_pickit command after building to flash the code, but it’s kind of annoying, as that program has a problem verifying the checksum and reports an error every time (which is interpreted as a build failure) plus, I have to run mplab X with root permissions to be able to access my USB port.

Ferrofluid Sculpture test 3 (a)

Jay — Sun, 27 Nov 2011 23:46:31 +0000

The main problem I had with my previous ferrofluid test was that the magnets would repel each other enough that I was not able to achieve a single, solid object. After (mostly) cleaning the ferrofluid off of my magnets, I carefully removed half of the magnets and a washer, replacing them with a bit of Sculpey that weighted exactly the same. I pushed half of the magnets into the Sculpey, and then used them to hold another 17 magnets “outside” of the main body. This spacing keeps them far enough apart that they do not repel each other, and the Sculpey fills in the areas between the magnets.

The Sculpey will probably lose some weight when it is baked, but I’m counting on the ferrofluid to allow me to adjust the weight of the hanging object suitably. I don’t want to test it with ferrofluid until after I bake the Sculpey and coat it with black spray paint. The punk potato has a magnet with a plastic cap on it, so I needed to determine the lowest possible temperature that will harden the Sculpey without demagnetizing the magnet or melting the plastic cap. (Currently, the “punk potato” is right at the lower weight limit for successful levitation.) As the lowest curie temperature I could find for common rare-earth magnets was 310 °C, the plastic (which I assume is ABS, with a melting point of 105 °C, (or 220 °F) is my limiting factor. I set the oven to 215 °F and wrapped some tinfoil around the plastic in an attempt to protect it.

When I showed it to my wife she closed one eye, tilted her head, and said “Weird”, so I know I’m getting close to the aesthetic I’m aiming for.

Solar Panels on the Electric Truck

Jay — Wed, 16 Nov 2011 10:16:13 +0000

I purchased two used GE Energy GEPVc-175 watt solar panels and mounted them to the bedcover of my truck using 10-24 machine screws and T cone washers as rubber vibration dampers. Each panel weighs 31 pounds, and is rated to deliver up to 175 watts of power in full sun (at around 36 volts each, or 73.4 volts at 4.7 Amps in series). I am still waiting on a group of Georgia Tech ECE students for the magic box that does maximum power point tracking (MPPT) and voltage boosting to charge my 120 volt battery pack, but I hope to be generating 1-3 miles of my daily commute from the sun soon. As my daily commute is 4 miles, this can be a significant percentage of my total energy usage.

I have calculated that in the summer the panels are far enough behind the cab that they will not be shadowed by it, even if I have to park facing south. In the winter and early spring / late autumn I need to park facing north to avoid shading a strip of the solar panels.
I still need to figure out a way to tilt the panels towards the sun to collect as much energy as possible. This is especially critical during winter, when the solar angle is way off of vertical. As the bed cover tilts, AND the bed of the truck can tilt (the other way) I figure I can work something out (with a few pieces of wood cut to the correct height, or linear actuators if I want to get fancy.

Ferrofluid Sculpture test 2

Jay — Mon, 07 Nov 2011 01:28:50 +0000

After buying some strong rare earth magnets I re-tried my ferrofluid levitating sculpture. The main problem is that the ferro-fluid moves to the poles of the magnets (top and bottom, with gravity favoring the bottom) and does not surround the sides of the magnets. It was also hard to get enough small magnets to fit under the main magnet (as they repel each other), so I had to include a screw in the middle to get the weight correct.

Note: Working with Ferro-fluid is dirty and messy. If magnets snap together they can throw dark oily fluid which stains just about anything. Be sure to have lots of paper towels on hand, jars/shields under/around everything, and don’t work in a nice room. Dish detergent and abrasive scrubbing will eventually get it off of most surfaces that are not porous. Wear disposable rubber gloves and use disposable pipettes. Wear disposable clothes. Don’t use any of your wife’s kitchen appliances or instruments that you can’t afford to replace.

Update: After floating under the magnetic levitation for 24 hours, the ferro-fluid became much less “spiky” and eventually the drops at the end of the magnets became spherical.

Also, cleaning ferrofluid from a magnet is almost impossible.

Controlling the Cricut Personal (desktop cutter) via Linux

Jay — Tue, 25 Oct 2011 00:35:25 +0000

I caught a deal on cricut.com where they were selling their original “Cricuit Personal Cutter” (refurbished) for $49.99 ($65 with shipping). This seemed like a very good deal to me, so I bought one. By itself, the Cricuit can cut specific shapes and letters from paper (it includes a free set of shapes/letters, and you can purchase cartridges to add more). I plan on using it to cut metal foil (one of the few things you can’t cut with a laser cutter, because it is reflective) and potentially make stencils. I may also be able to use it as a plotter using special attachments for pens. (Circuit board resist markers anyone?)

To make full use of it, I wanted to be able to cut arbitrary paths from my material, which means controlling the Cricut from my computer. Luckily it comes with a built in USB port that is recognized as a standard serial port. Also luckily, other people have already figured out the command and control protocol and built a library to control it.

I found the directions on this page very useful getting libcutter working. So far I have simply use the draw_svg utility program that is included with libcutter to cut out vectors created in Inkscape.

Tracking individual battery voltages

Jay — Sun, 16 Oct 2011 19:14:37 +0000

This graph shows the voltage (multiplied by 10, so 65=6.5 volts, and 50 = 5.0 volts) batteries 1-6 of my electric pickup while accelerating. My first battery (blue) is consistently 0.1 volt below the others, so I am keeping an eye on it. But this graph shows that although it is a 1/10th of a volt lower, it does not sink lower than the other batteries under load, so it appears to be holding up well so far.

The graph also shows off my new (to me) Pak Trakr system. The Pak Trakr system connects to each battery in your pack with small remotes that daisy chain together. Each remote monitors six batteries and transmits the voltage levels once a second to a display and optional serial data logger.