Peltier power supply and integrated PID controller

I am preparing to build a beer-brewing tank that will be temperature controlled. The goal is to keep the vessel at a very steady 72*F. Depending on the time of year, this might require heating and/or cooling 24 hours per day, or heating during the night and cooling during the day. Since the only heat load on the vessel is the heat leakage through its insulation, the total amount of heat that needs to be pumped is fairly low. This might be a good application for a Peltier heat pump since the heat load is low and the heat pump direction might need to be changed often.

As I mentioned in my aquarium chiller post, the Peltier module must not be controlled by PWM (pulse width modulation) or on/off thermostatic control. During the 'off' part of the cycle, heat will flow backward through the Peltier device and decrease efficiency almost to zero. Ideally, the PWM output should be smoothed with an L-C (inductor-capacitor) circuit to provide clean DC power to the device.

I happened to disassemble a thermoelectric refrigerator (and replaced its guts with a refrigerant-based system in a previous post) and had a 115VAC Peltier power supply and Peltier module. The power supply uses a TL494 IC to control a 115V->12V high-frequency transformer. The output of the transformer is smoothed with an L-C circuit. The TL494 is controlled by the output of a thermistor and digital thermostat, thus sending more power into the transformer when the fridge's temperature exceeds the thermostat set-point, and reducing power when the temperature falls below the set-point. I couldn't figure out the analog side of the circuit which had a few op-amps. Instead, I discovered that I could throttle the power supply by putting voltage on the TL494's DTC (dead-time control) pin, which wasn't used in the original circuit. Cool. So now I have an efficient power supply that I can throttle from 0V to 12.5V (full power) to the Peltier.

Next, I need a thermostat to control the power supply and also reverse the polarity of the Peltier device (to switch from heating to cooling). I thought about using an Arduino to do the whole thing: sense the temperature, provide a UI with LCD and buttons, process the PID loop, and send output to the power supply. My biggest worry was the UI. There are actually a lot of parameters in a PID loop, and I didn't feel like writing code to make all of them user-selectable. I also had another Eurotherm 2132 PID controller (same as the aquarium project) which I really like. I decided to use the Eurotherm and build a circuit to convert its output into a control signal for the power supply.

The Eurotherm has two outputs, a relay and a driver output for an external solid-state relay. I removed its internal relay and directed the relay's coil wire connections to the Eurotherm's rear terminals. I then used an optoisolator to interface the Eurotherm's floating outputs to the rest of my signal-converter circuit. The Eurotherm can be configured so that its output 'cycle' for each output is 1 second. Thus, it essentially outputs a PWM signal at 1Hz for cooling, and another 1Hz PWM signal for heating. I wrote a little code for an Atmel AVR ATMega8 that reads the two control signals from the Eurotherm and outputs a voltage that controls the Peltier power supply. It does this by generating high-frequency PWM and smoothing it with a simple R-C low-pass filter. The current draw is very low. The AVR also controls an NPN transistor that drives a DPDT relay. The relay will reverse the polarity of the Peltier device.


All of the above-mentioned circuitry is crammed into an acrylic box that I made for this project (left). The Peltier module has a small heatsink and a large heatsink with two fans (right).





I modified the power supply (by removing a fuse, and adding a fan), and tested it with two Peltier devices hooked in parallel. It seemed fine at 12.5V and 7A.

The next step is to mount the Peltier devices onto the brew tank and test it out.

The tank will hold 5 gallons. For thermodynamic analysis, I will assume it's water. So, that's about 20 liters, and water has a specific heat of 4.2J/g*C, so the tank will require 4.2(20)(1000)=84KJ to change 1*C. If the Peltiers are consuming 12.5(7)=88W, and are 50% efficient, they will be pumping about 44W (in cooling mode), which is 44J/s.

In order to change the tank's temperature by 1*C, the system will need 84000/44 = 1900 seconds, or 30 minutes. This is good and bad. The good news is that the tank will remain very stable, as it is not in danger of being quickly influenced by the Peltier. The bad news is that the tank must start out fairly close to the target temperature, or else it will take a long time to be regulated by the system.

Stainless TIG welding update

OK, I have learned a few more things about welding stainless steels with TIG. I am far from an authoritative source on the subject, but I hope to share what I do know:

* Keep the weld bead width as thin as possible. This is really key to everything else, and making a thin weld bead will force the welder to do everything else correctly (ie keep the current to a minimum, keep the electrode close to the surface, etc).

* Minimize the heat input to the work by any means necessary: Lower the amperage, bring the electrode closer to the surface, use copper chill blocks, stitch-weld 1" sections and allow the work to cool between welds.

*Keep the tungsten electrode sharpened to a needle-point. The length of the taper should cover about 2.5 times the electrode diameter, and I normally leave the tip fully sharp for low-amperage welding. I've never had a problem with the tip breaking, and it helps direct the arc.

* The part fit-up must be superb. I always try to tell myself "sure, I can fill that gap", but no. With stainless, filling the gap means dripping lots of filler rod, which will be very hot and leave the weld in very bad condition. The excessive heat will "burn" the chromium out of the alloy.

* The electrode should be as close to the surface as possible. This is not an exaggeration. I mean really get the electrode as close as you can without touching the puddle. A small, hot arc is much more effective than a large cool arc.

* Use thin filler rod. If you're welding 1/16" thick stainless, I'd say the weld bead should be 1/8" wide or less. This will be a problem to weld with 1/16" rod, because it is difficult to dispense a small enough amount of rod to keep the bead thin enough. Additionally, since the electrode is being held very close to the surface, there isn't much room to stick the rod into the puddle. Using .045" and .035" rod is a HUGE help.


* Don't let the arc melt the rod! When adding rod, move the torch backward a tiny bit so that you can dip the rod into the puddle, then pull the rod out of the puddle, and advance both of your hands forward in the weld direction. I can tell that I've messed up when I see the rod form a ball on the end, and the arc changes color. If the weld is important, I would cut the end off the rod and start again.

* When the weld is all done, check the colors to see how corrosion-resistant the weld will likely be:

Gray, pinkish-gray, black = Not good. The stainless has been chemically changed at the surface
Dark blue, purple, turquoise = Borderline OK. Acceptable for some situations.
Gold, silver = Perfect.



* The most common fill rod for welding stainless is 308. As near as I can tell, 308 can be used on anything, but it would be unwise to use it on higher grades of stainless like 316. I tried both 308 and 316 fill rods on 304 base metal. They behave pretty similarly, but 308 seems to melt a little more easily.

* Most rods and base metals can be obtained with an "L" designation, like 304L or 308L. These are low-carbon alloys and are preferable over the non-L alloys because welding can cause carbon to precipitate in the metal. The upshot is that welding 304 will have a more deleterious effect on its corrosion resistance than welding 304L. How important is this effect? I don't know. All of the information that I found on the net was purely anecdotal and poorly documented.



This was attempted with 1/16" filler. Very inconsistent bead width.


This was done with .045" filler. It's not perfect, but much better. Both welds were stitch-welded in 1" sections.

Painting a motorcycle

This is the second time that I have painted my motorcycle and I wanted to share a few things that I have learned:

Almost all spray-can paints will not resist spilled gasoline very well. Some are better than others, but all pale in comparison to factory paint.

Almost all spray-can paints will leave a finish that is soft compared to factory paint. This is a problem because dust will get embedded into the finish, and the surface will lose its shine.

Getting a good-looking paint job at home with spray paints is not difficult, but the paint job will not last as many years as a factory paint job.

I briefly explored the possibility of spraying two-part (2-K) urethane paint which is what most car and motorcycle manufacturers use. There were two options: 1. Mix the paint myself and shoot it using an air sprayer (eg an HVLP gun). 2. Get SprayMax 2K in a can: http://www.repaintsupply.com/pd_2_part_2k_aerosol.cfm I've heard good things about this paint, but it's very expensive. I would need at least three cans, and for a just a little more money, I could buy a quart of real 2K paint and a cheap HVLP gun from Harbor freight.

In the end, I decided to just use Krylon spray cans for $2.99 each. The motorcycle itself is not the most beautiful machine in the world, and repainting it every 5 years is an acceptable level of maintenance.


The original color was red, and I painted yellow over it without stripping last time. I used Jasco methylene chloride stripper on the steel gas tank.

The tank has a big dent that I filled with bondo before the last paint job

I've shot a few coats of Krylon primer, and used some glazing putty to fill in the last surface imperfections. I wet sanded the primer with 320 grit between coats. This was the only sanding that I did on the gas tank.
I shot about 3 coats of color on the tank, letting each coat dry for a day. Krylon can be re-coated within an hour or after 24 hours. These directions must be followed or the paint will lift and look terrible. It will have to be re-sanded and painted again. I was shooting pretty heavy coats, and I wanted to make sure everything was set before I loaded on more paint.



The plastic fairings on the bike have another problem: The methylene chloride stripper will dissolve the plastic fairing as well as the paint on it! I learned this the hard way the first time I painted the bike. A good alternative is to use regular DOT3 brake fluid on the plastic parts. It doesn't work nearly as well as paint stripper, but it's a lot better than nothing, and it doesn't attack the fairing plastic. Use really coarse steel wool to scrape off the old paint. I sanded starting with 80 grit, and worked up to 320. I did not use primer on the plastic parts -- just shot color right on the plastic.



I shot three heavy coats of color over the course of three days.

After drying another day, I added black stripes. The stripes were masked with blue 3M painter's tape, and the black paint went right over the yellow.

I did not use a clear coat over the color coats.

Lighting a BBQ with liquid oxygen

http://www.youtube.com/watch?v=1bjvj5FjUPE
After watching the famous liquid oxygen BBQ lighting video on the internet, I knew it was time for me to try it myself. In this video, I am pouring 1.5 liters of liquid oxygen on a bunch of smoldering charcoal briquettes. The oxygen rapidly turns to gas and accelerates the rate of combustion at the surface of the briquettes. Some of the charcoal fractures and shoots off tiny pieces that burn with a bright light.

I made the liquid oxygen by passing gaseous oxygen through a copper tube that was submerged in liquid nitrogen. The liquid nitrogen is colder than the boiling point of oxygen, so it causes the oxygen to condense and be ejected as a liquid from the other end of the tube. Liquid nitrogen can be bought at welding supply stores, however I made a batch myself with a cryocooler.
http://benkrasnow.blogspot.com/2008/08/diy-liquid-nitrogen-generator.html

Improved automatic cat feeder

I recently set up my automatic cat feeder (http://benkrasnow.blogspot.com/2009/02/automatic-cat-feeder.html) for a long-weekend vacation. This was probably the sixth time that I have used the feeder with great results each time. I also setup a motion-activated webcam that posted frame captures to a web server so that I could see the feeder (and cat) while I was away.

One day during the trip, I checked the web server and noticed the feeder had dispensed a huge amount of food. I've never seen my cat eat his fill and walk away from a dish with food in it, but he did this time so there much have been a truly massive amount. When I returned home, I saw what had happened: The cat had knocked over the darkroom enlarger timer, and by strange chance, the timer knob was turned from 3 seconds (an interval that dispenses a single-serving of food) to 20 seconds! The feeder emptied the entire storage chamber of food into the serving dish. It's not possible the cat knew what he was doing by messing with the darkroom timer, but it's also very strange that this happened by mere chance.

I never liked the timer approach to the cat feeder because it assumes the motor turns at a constant speed, and the darkroom timer is not easy to adjust accurately in the range of 2-4 seconds. I thought about adding a shaft encoder to make the system closed-loop. I also thought about using a mechanical means to control the shaft (ie a Geneva Wheel). Finally, I decided to use a strong stepper motor in open-loop control. I also wanted to keep the same gear reduction box that was originally coupled to the food dispenser. It's torque amplification is helpful, and it has also become a structural part of the feeder, so not easy to change.


I removed the old DC motor and pulled the armature off the shaft. I cut the shaft, then built a press-fit delrin coupler to mate the stepper motor shaft to the remaining piece of original shaft (with an integral helical gear on the end). I then made some spacers out of delrin and mounted the motor to the gearbox.


I used an Atmel AT-Mega8 AVR to turn the stepper motor through four shaft rotations when power is applied. I used PN2222 transistors to switch the unipolar motor windings. The transistors are driven through resistors by half the D port in the AVR. The gear reducer's output will spin about 1/6 of a turn, which will advance the "butterfly valve" in the feeder enough to dispense a single serving. I coded the Atmel chip using Atmel's IDE in C with WinAVR. I used PonyProg to download the code into the chip via a simple parllel port interface. I would normally use an Arduino or similar microcontroller for projects like this, but this one is so simple, a full-fledge microcontroller would be overkill and too expensive. The AT-Mega8 is about $6 and needs a $1 crystal. The Arduino is around $30.




My cat stuck his head into a pile of plastic shavings the garage and managed to get a couple scraps stuck in his whiskers.