After getting gas quotes and talking to Doug Jones, I decided to abandon the idea of using Ethane/GOX for the igniter and decided to go with IPA/GOX, eventually moving to NOS like Andrew Case. The ethane was too expensive at $200 for 80 cubic feet. I decided to start with GOX as it is cheap and simplifies the plumbing a bit. I started putting the plumbing together recently and learned a lot about 37 degree flare fittings in the process which I will write up separately. I bought an inexpensive cart from Enco ($35) that I am going to mount everything on. I also bought a 1/2" aluminum plate from Metal Connections to bolt onto the cart top and use as a stiff mounting point, but the plate hasn't shown up yet. Metal Connections will cut it to size and includes shipping in its pricing. Not a bad deal. I am still acquiring plumbing, gas bottles, regulators, etc. but it is all coming along. I bought a 32" x 1.75" copper 145 remnant from Yarde Metals for only $74. I am trying to not use to much surplus stuff as that starts making it harder to reproduce, in whatever fashion, my results. But I'll take a deal if I can get it. The new Rev. B High Voltage Driver boards came in and they are perfect except that the bottom-side writing is mirror-imaged. Jeez, it's always something. I also found that Aircraft Spruce sells shielding braid for sparkplug wires that is a lot cheaper that the buck a foot stuff I had been using. Aircraft Spruce's is $0.35 a foot, P/N 11-02463.

My biggest accomplishment of the month was to create a design spreadsheet. I used the booklet How to Design, Build and Test Small Liquid-Fuel Rocket Engines . Something I have wanted to do for a while is to try and come up with closed form expressions for various design parameters like ISP and chamber temperature as a function of O/F ratio and chamber pressure. In the past, to get a handle on those sorts of issues, I would make a bunch of runs in PROPEP and then laboriously edit and graph the data in MS Excel. That took a lot of time and only gives you a two dimensional view of the data. My idea was to try and curve fit the PROPEP data using a commercial curve fitting program. I used a demo copy of Table Curve 3D and it really worked out well. TC3D is a fantastic program that is very easy to use. I think I had to look up one thing in the manual in a week of using it. Since the complete spreadsheet with all of the graphs and data is almost 7M bytes, I also created a slimmer version that just has the minimum necessary for the motor design section. When I compare the equations used in the booklet to Humble and Larson they seem simplified, so I'd be interested in hearing how useful and realistic everyone finds the booklet design procedure to be. I am sure there are errors here and there in the spreadsheet, so if something looks funny let me know and I'll look into it. By the time the spreadsheet gets to Rev C or so I hope it will be getting truly useful.

The Tea Cart

This is an example output of the curve fitting process for IPA + GOX chamber temperature showing the equation, coefficients, and R squared (a measure of how well the generated equation fit the curve). An R squared of 0.995 is pretty good!

Small Motor Design Spreadsheet Full Rev. A (6.9MB)

HV Igniter Driver board, schematic, and library files (50KB zipped)

Bitmap picture of the Rev. B schematic (4.8MB)

Some pictures:

The completed coil, sparkplug, and driver assembly.
The coil and sparkplug are mounted on a plate, which is viewed edge on.
The driver is sitting on the stool.

HV Igniter Driver compared to a Quarter

Top view showing the connectors.

Side view showing the IGBT, and standoff. You can also see the two jumpers for selecting Controller mode or Timer mode.

Side view showing the heatsink mounted on the IGBT. You can also see where I milled the inside of the case a bit so that the heatsink would make better contact.

HV Igniter Driver Rev A parts list

Bitmap picture of the board artwork

Rev A Eagle format board file (Try a right-click and Save Target As if clicking the link doesn't work)

Rev A Eagle format schematic (Try a right-click and Save Target As if clicking the link doesn't work)

Screen shot showing glitch
Click to enlarge

The prototype as it exists today. See the braided shield?
Click to enlarge

I think this performance is good enough at this point to go ahead and layout the board, keeping in mind that all the electronics will need to be in a metal box for shielding. I also decided to put the coil driver IGBT on its own board so that it can be located away from the main PCB, keeping the HV as far away as possible from the main board. Since the two boards (main and driver) can be fabbed as part of the same job, there is no additional cost other than I'll get less total boards from a minimum lot. No big deal since the first batch is a prototype.

I also built a thermocouple amplifier to try out the idea of using a thermocouple as an ignition sensor. The response time of the exposed junction K-type thermocouple I tried out was quite fast, way less than 100 milliseconds. Check out the plot made with my trusty Dataq DI-194RS:

Screen shot showing glitch
Click to enlarge

I generated the thermal step function by holding the thermocouple against an ice cube and then plunging it into a hot cup of coffee. The y-axis thermal scale is reasonably accurate. Omega makes a thermocouple, the Pipe Plug Probe for $39 that has the exposed junction option needed for a fast response time. It has a stainless steel overbraid and is good to 2500 psi, although the max temperature rating is only 480°C. Hopefully the thermocouple won't melt above that temperature, but we'll see.

Thom McGaffey moved to Arizona which will be another set back as Thom had bought all the gas equipment. Thom may come off as a bit gruff on Arocket, but in person he is a great guy to work with and I'm going to miss him :(

Here's the hardware block diagram
Click to enlarge

Since I was faced with that setback I decided to build a small prototype of the AVR microcontroller circuit and check out the noise situation that I was so worried about last month. I also learned how to use interrupts and use the AVR onboard timers so it was a fun little project. I didn't bother to write all the code (or get Jamie to do it) and just set up the 16 bit timer1 to put out a 1kHz square wave. Once the timer is programmed it runs independently of any other code that is running so it is like mini-multitasking in the microcontroller. To imitate the other code (for solenoid control, ignition detection, etc.) that will be running in the main loop I programmed two output pins to drive two LED's that blinked in an on/off sequence. That way it would be easy to tell if the AVR died or freaked out when the spark plug was sparking. Well, the square wave came out and the spark plug sparked , but the LED's went wacky. When I removed the coil 12V (oh yeah, I used the old sparkplug/coil prototype Thom and I had used before except the coil and plug are rigidly mounted to a small aluminum plate. The coil is connected to a large 12V gel cell and the controller to a 9V battery), the microcontroller went right into the main loop like nothing had ever happened. If there was a lot of EMI I would expect the program counter and register contents to be scrambled and the whole thing would just hang. Somehow the controller was able to recover quite nicely! Obviously there is some kind of EMI issue that needs to be solved so I am first going to try adding a snubber on the IGBT to reduce the flyback dV/dt, second I'll shield the coil/plug and/or processor, then third add optical isolation to the driver and isolated power if absolutely necessary. To design the snubber I really need to know what the coil HV waveforms look like so I need a HV scope probe. I have been avoiding this issue since HV scope probes are kind of expensive ($300-$600). I found a few used ones on Ebay for about $100 but decided to try and build one first. There is a nice explanation and example at Basics of High Voltage Probe Design and Homemade High Voltage Probe (> 40kV) . I wasn't able to find any of the required high voltage resistors at the usual online sources so I called Caddock directly. I talked to a very friendly applications engineer who recommended the MX440 11kV 10Meg resistors as a place to start. When I asked for one sample he replied "I won't send you one but I will send you two. I never send one of anything". Nice guy! It will be interesting to measure the actual output voltage as I really have little idea of what it is.

On another note, as I mentioned in another Arocket post, I was fortunate enough to see Elon Musk of RocketX speak at Stanford University. In the course of his talk he mentioned that they were using ethernet for internal communications, which got me thinking. I had bought an AVR ethernet demo board from EDTP Electronics for another project ( Easy Ethernet AVR ). The EDTP board comes with C source code for an ImageCraft compiler. There is also source code for the free AVR Freaks gcc compiler (which Jamie plans on using for the igniter board) elsewhere on the 'net. I have a CodeVision compiler and didn't want to take the time to install the gcc compiler ( Winavr ). Fortunately it only took a few minutes to fix the compiler errors and produce working code (well, I was able to ping it) using my CodeVision compiler. A pretty complete solution (support for tcp, udp, icmp, arp) fits in only 5KBytes out of 16KBytes of flash memory. Pretty impressive! And that 5K includes some UART code for debug messages so I bet it could be reduced to only 3 or 4K for tcp support only. The ethernet controller (Realtek RTL8019AS) is also only $8. If CAN turns out to be a big hassle, ethernet looks like a pretty attractive alternative.

Thom McGaffey and I had worked on machining the igniter body about a month ago but the original design was too hard to accurately machine. Thom redid the design and has machined and welded a beautiful prototype from 304 stainless. I didn't get a picture of it but I will next time.

Decided to use the Tecumseh small engine coil. It is not as high performance as the Ford coil and it runs hotter, but it is $18 versus about $60 and has a nicer form factor. The Tecumseh coil has a square hole in the center that fits a 3/8" carriage bolt. It also fits over the sparkplug insulator which makes for some interesting packaging possibilities.

I have been talking to the guys at Portland Aerospace (PSAS) about their use of the CAN bus as I am designing in both CAN and I2C (and manual bit switching) interfaces. I hope to meet up with them at BALLS and get an up close look at their avionics architecture & implementation. They will be launching their airframe with both an N and O motor at BALLS.

Ford 1L2Z-12029-AA
Motorcraft 1L2U-12A366-AA

Some pictures:

NGK racing plug with a "surface gap" electrode
The spark covers half the electrode

NGK chainsaw plug BM6F

BM6F with a nice spark

Left to right: Chevy, motorcycle, Ford "coil on plug", Tecumseh small engine

Left to right: NGK R0045J-9 racing plug, RJ12C regular car plug,
NGK BM6F, NGK BM6A, Tiny NGK CM-6

Thom McGaffey's well equipped electronics area

High pressure and low pressure gauges for hydrogen and oxygen

Worked through the preliminary design of the controller board. It will have something for everybody (at least the prototype): AVR controller, CAN bus interface or parallel control lines, two channel temperature sensing (chamber and heatsink), two channel solenoid control, status read back via the AVR or parallel lines. Should have boards in two weeks. Question though: are there strong preferences for SMD or through-hole? Either will work and through-hole may have some advantage because it is easier to heatsink through-hole devices. But it will be smaller with SMD. Any opinions?

Thom has a design for the chamber and is starting on that. It'll be GOX and Hydrogen. The prototype will have outboard gas regulation so we can easily change pressure. The body will be a chunk of stainless steel with diffusion welded SS tubing to the gas ports. Holley solenoids for gas control. Looking at reusing large threaded CO2 cartridges as the flight gas storage. Will start testing the prototype chamber a week from Monday (July 7).

So far so good!

The tests were conducted by monitoring the voltage waveform on the series current monitoring resistor while varying the signal generator frequency. The IGBT heatsink temperature was monitored with a thermocouple temperature meter. Theoretically, if the secondary is open (ie, not in discharge), the coil will act like an inductor. If a fixed voltage is put across an inductor it will generate a linear current ramp. Once the inductor saturates (can no longer increase the energy storage in the magnetic field) the effective inductance will drop sharply and the current ramp should increase its slope sharply. Measuring the slope in the low slope portion gives us effective magnetizing inductance (L= V/delta I/delta t). The peak current that could be supplied in the linear region was also measured. Peak current plus magnetizing inductance gives use energy storage (E= 0.5*L*I^2). In reality, only one of the coils could be driven into true saturation. The rest would limit due to the circuit resistance plus internal coil resistance. One coil, the motorcycle coil, didn't produce a linear ramp but instead produced a single-time-constant-RC looking waveform. I'm not sure what that was from and will have to check into it some more later.

The results:

	Coil Primary Peak Current Magnetizing Inductance Energy Storage Energy Storage Density Chevy Coil 26.7 mA 96.2 mH 18.89 mJ 32.3 uJ/gm Motorcycle Coil 192 mA ?? ?? ?? Ford Coil-on-plug 753.3 mA 38.1 mH 10.81 mJ 85.9 uJ/gm Tecumseh 960 mA 17.6 mH 8.11 mJ 83.6 uJ/gm

The Chevy and motorcycle coils ran very cool as did the IGBT when driving them. Low peak currents pay off in thermal performance. The Tecumseh ran hot and the IGBT ran very hot. We decided that on all tests we would shutdown when the heatsink temp reached 150ºF. The Tecumseh had to be shutdown after about 10 seconds max. The Ford coil could be run for about a minute before the temperature max was reached. All coils performed (subjectively) best at about a 1KHz switching frequency.

Plugs (subjectively): CM6 was the wimpiest of the bunch. An anemic spark. Champion RJ12C was OK The BM6 (not sure if it was the A or F) produced a very bright, hot spark The NGK R0045J-9 was an arc welder. A full spark curtain surrounding the surface discharge electrode when driven at the optimum frequency. Very cool. Too bad racing plugs only come in 3/4" reach.

Coils (subjectively): Interestingly, subjective performance went up as size went down. Chevy, OK Motorcycle, somewhat better Ford Coil-on-plug, hot intense spark Tecumseh, noticeably hotter more intense spark.

However, they all, plugs and coils, worked. So I think that really any of them can be successfully used. At this point it looks like the racing plug is the most interesting, especially since it doesn't have a protruding electrode to erode, and is design for hot race car conditions. The Tecumseh coil is great, but it is such a current hog that it is at a thermal disadvantage. That would leave the Ford coil as the best looking one right now (Jamie has good instincts!).

The IGBT worked great with no problems through several hours of testing abuse without any clamp/protection/snubber circuitry. I did replace it at one point thinking it was shorted but it turned out I had shorted the test leads. Thom successfully lit his nitrous/butane microtorch several times in the racing plug discharge. No problems there! We had to kluge some coil to sparkplug leads for the Chevy and motorcycle coils using regular wire. That produced a ton of RF so that the temperature meter wouldn't work. That will have to tested/fixed some more before using a controller.

Next steps: A bit more testing to try and understand the motorcycle coil Take scope photos for documentation Try pulse-width modulating the IGBT Build a prototype PCB and ignition chamber

Thom bought a bunch of stainless stock and a bottle of hydrogen to use for the chamber. Should be fun!

Next, I measured the primary and secondary ESR's using a four-wire kelvin connection. I was able to decently resolve 1 milli ohm, but below that the measurements were a little noisy. The HP 4184 has the provision to increase the ADC integration time, but I couldn't figure out the menu system. I also couldn't find the manual (almost everyone had already left for the day). On a whim, I decided to measure the secondary with respect to both ends of the primary (+ and -). The secondary impedance was always higher with respect to the "minus" side of the primary than the "plus" side! I was surprised, because that means the + side is the common point with the secondary, just the opposite of what I would have thought (see attached schematic). To read the schematic you'll have to download the viewer from ExpressPCB. I seem to recall that automotive sparkplugs fire when the electrode is negative so electrons flow from the center electrode to the rim. Is that correct? If so, this picture is consistant with that scenario. But also if this is true, I should really redo the turns ratio measurements as I may have been driving the coil in such a way as to bias the secondary with the primaries drive level. It's a small error, but should be corrected.

I am not 100% sure of my results because of the noise problem I had with the voltmeter. Because of the noise I really only had 4 significant digits. I need at least one more to really be sure of these results. To reduce noise I either have to average some samples or increase the integration time (the voltmeter has a dual slope integration converter). I couldn't figure out how to do that by just fiddling around, I need to read the manual. Unfortunately Amicronix just sent the meter out for calibration so I can't use it again for a couple of weeks. Thom has a 6 and a half digit voltmeter so I am going to go over to his house on Wednesday and try again. If the IGBT's come in we can also do some inductance testing. Thom also reports he received the thermocouples for ignition detection.

I would also like to get some pictures posted this week!

I decided to use the International Rectifier IRGB14C40L IGBT to drive the coil. It's $3.75 to $3.96 from Digi-Key in single units depending on the package. I'll order some TO-220's for testing and, depending on power dissipation with the chosen coil, go with a surface mount device for the PCB. The IRGB14C40L is a 400V clamped device that will switch 20A. It was designed to drive coils so it has a built-in clamp as well as a 5V drive spec. Paul Kelley recommended the MJ10012 darlington. That is an old Motorola 400V 10A device that is hard to get in the US now. It also would require a more elaborate drive and clamping scheme, but it probably is pretty cheap. I would guess it is still being made by some mainland Chinese company. Infineon has a pretty cool device that has an internal current measuring resistor (BTS2140/45/65). Unfortunately I couldn't find a source for it.

I bought two NOS Powershot solenoids to test as igniter gas switches. I talked to Holley technical support and they told me the Powershot is NOT rated for continuous duty. The maximum on time is 14 seconds followed by a minimum off time of 20 seconds. I want to try using a current modulated switch to coax effectively continuous duty from the Powershot. Has anybody tried current or pulse-width modulating them?

Thom McGaffey gave me a NGK BM6A sparkplug which is much smaller than a car sparkplug. Bill Bullock sent along a REALLY small NGK BM6F chainsaw motor sparkplug. The BM6F doesn't have a crush washer though. I'm not sure if that will be a problem. Thom is going to machine a flame injector for testing.

Thom is also working on a design for trying a thermocouple for ignition sensing. I'm going to try the ion detection method and the glow plug. Andrew Case is working on optical detection.

Next step: as soon as the rest of the coils come in I'll measure the magnetizing inductance, leakage inductance, ESR, saturation current, turns ratio, and weight of each coil. Then add the switch and see what kind of HV they produce.