SpudFiles

As you may know, I plan on doing a goodly number of spudgun tests this summer.

There are multiple reasons I created this thread:
1) To document my equipment
2) To keep my from backing out of the project
3) To get feedback on equipment/test ideas
4) To get the community involved

The completed test launcher (sorry, I don't feel like making multiple chambers at this time) will have variable and movable chamber features and a barrel with many photo(transistors?) to record the movement of the projectile.

Specifics of the chamber:
4" sch 80 threaded plug / 4" sch 40 threaded female adapter / ~8.5" sch 40 pipe / sch 40 coupler / sch 40 socket-weld bushing (4" -> 2" female NTP) / 2.5" long 2" sch 80 PVC nipple
Chamber volume, by water displacement to the top of the nipple, is ~2910 ml.
Chamber features are attached using nifty C-shaped PVC clips... this allows them to be moved easily.

Specifics of the barrel:
2" sch 40 female NPT adapter / 10' 2" sch 40 pipe
First 4' of barrel with have a photo transistor every 3"; last 6' will have a photo transistor every 12"... total of 22 photo transistors wired in series, if I counted correctly.
Work has not yet begun on the barrel, nor have most of the materials been gathered.

On July 23, I cobbled together a propane meter.
*****************

I plan on doing the following tests:
1) Optimum fuel-mixture test with propane - taters (common mass) and non-moist ammo (common mass).
Effectively confirms the latke C:B results, their fuel-air results, as well as finding the C:B ratio for imprecisely metered fuels (such as aerosols).

projectile mass tests
single spark placement tests
1/2/3 spark comparison tests
fan placement tests

Now, we might have issues with some tests, such as spark placement tests, because it will probably depend upon the fan placement - which will, likewise depend upon the spark placement.
Any ideas on how to get around that without shooting 1000 taters?

Make adjustable spark strips, with moveable points, sliders???

There are a couple of techniques that you can use to reduce the number of total tests that have to be done when you are testing a large number of variables.

In general, as long as every test parameter is independently varied against at least one other parameter you can usually extract the independent affects.

One simple example would be the affect of projectile mass. If you try to do say two spud mass (100g and 200g, corresponding to roughly a half spud and a full spud?) and you do it as a function of CB, number of spark gaps, spark placement, fuel ratio etc. then you've got a zillion tests to do. You can simplify (and omit) many tests if you do it correctly. You don't need ten replicates (like Latke did) for every change in any one parameter^*. Overall you need a fair number of test but you dont need (number of parameters)x(number of values/parameter)x(number of replicates) total tests.
+=+=+=+=+
^* The number of replicates per parameter that Latke did was a good idea. He was only changing one parameter at a time and ten test per change was great. Fewer tests, perhaps 5, probably would have been adequate. Of course, I can look back at his data and judge the point of diminishing returns in terms of replicates. Latke didn't have that luxury before he conducted the tests.
+=+=+=+=+
Instead, you change more than one parameter at a time and do a lot fewer tests. This can be tricky and, if not done right, you can end up with results that are uninterpretable.

Back to the spud mass example. If you record the muzzle velocity of each spud then you can use the kinetic energy of the spud as the "output" parameter. This partially uncouples the mass of the spud from the results. Same gun, same friction (a problem there), same sparks etc. but different masses will give pretty close to the same KE. So, in your replicates on the number or location of sparks you can also be varying the spud masses and then correct for those changes. (Heck, this is always the best thing to do with variable mass ammo like spuds.)

Like I said above, this approach can be tricky to get to work. It works best with variables that are independent and linear (or can be expressed as linear after a transform like taking the square root). I've used this approach in a system with 6 independent variables each of which can take on up to ~10 different values (these values were catagorical and not numeric). The total number of tests required to test all possible combinations, one time each, is ~60 million. I was able to build a pretty good model with data from only ~800 tests.



Have you given any thought on exactly how to wire up the photodetectors? IIRC the voltage drop across a photodiode or phototransistor is ~0.7V. With 22 devices in series you need a total voltage of ~16V. You could perhaps use a pair of 9V batteries (or a wall wart) and then use a PC sound card to record the voltage drop across the last (or first) gate in the chain. (The sound card will probably only handle voltages up to ~5V.)

<i>"It works best with variables that are independent and linear"</i>

I don't think there will be a way to express the KE or the MV as a liniar function of spark or fan position, so I don't think this tactic will work.

I think the best way to go is simply to minimize the number of options... something like 5x5 x 5 or 10 repetitions. (I'm partial to higher reps, but careful control of projectile masses might give me good results with only 5)
************

Perhaps one could wire the phototransistors/diodes in parallel, then?
I think it's obvious that I don't know much about electronics, but would this give you too weak of a signal?

If so, perhaps one could wire them in series AND parallel (ie, 7 batches of 3 objects)? Such would keep the signal stronger, but prevent the voltage drop from getting too large, right?

BoilingLB: You are probably right, to do the correct sampling reduction you need to have a pretty good idea of the form of all the relationships. Since we don't ...

5 reps sound like it should be enough. Even in Latke's spud studies if you take his first five replicates the results don't change much. Measuring the spud masses should really help as well.


If you wire the detectors in parallel then the total current draw will probably be too much if you plan on using a MIC input to power the detectors. Each detector will draw something like 1mA which is about all a MIC input will supply. In addition, with 22 detectors in parallel then when one is blocked the current change is only one part in 22 (~5%). The nice thing about the series design is that blocking one detector will give nearly a 100% change in the signal.

Using a combination of series and parallel might work.

But I think using a pair of 9V batteries as the power source and wiring the detectors in series is the easiest way to go. If you want, I can do more research on ways to get 22 or so detectors to work.

Anyways, I got a regulator put together. Much fun resulted, and I took some pictures:

<a href="http://smg.photobucket.com/albums/v611/ ... puds/">All Photos</a>

<a href="http://img.photobucket.com/albums/v611/ ... Chamber</a>

<a href="http://img.photobucket.com/albums/v611/ ... pg">Inside the chamber (right now)</a>

<a href="http://img.photobucket.com/albums/v611/ ... m.png">The fueling system</a>

<a href="http://img.photobucket.com/albums/v611/ ... n.jpg">The movable ignition points and a fan</a>

A couple points:
1) I'm going to replace that camping fan before I do the tests.
2) Yes, the movable ignition points have holes drilled in them.
3) Unscrewing the endcap is tedious; I might install a latke-style quick-vent.
4) The short 24" barrel can be attached to the 51" barrel, which makes for a fairly quiet cannon. (but not as quiet as I was lead to believe a .7:1 CBratio should be)

Anyways, the next project is the velocity-recording barrel.
I have to save up some money before that will happen, but in the mean time, jimmy, could you draw up a wiring diagram (or explain) how I would use 9v batteries to power the sensors?

Maybe something like this?

(Mic +voltage)(-9v+)(-9v+)(24xsensor)(Mic -voltage)

Boiling, I thought you had abandoned the project...

I started to work on a similar gun... oh well.

I'm worried that there won't be enough light in the barrel, and that there will be a lot more light in front of the spud versus behind it, so I've added IR LEDs to the circuit for illumination. I figure you could just drill a suitable size whole clear through the barrel (through both walls) to insure that the LED and detector pair is actually pointing at each other.

Here is my current design for a 16 detector barrel using infrared LEDs and phototransitors. Power is from a 12V wall wart. Since the LEDs have such a large voltage drop (~1.4V/LED) they are in two sets of 8 LEDs each. The second set of 8 LEDs just duplicates the first set.

I was going to do some fiddling around with a piece of 2" pipe and the phototransistors and LED this weekend. (I have a couple of each already but not enough for a complete barrel.)

The parts are from All Electronics;
IR LEDs are $0.33 each, part number ILED-8.
Phototransistors are $0.65 each, part number PTR-1.

I have not tested the circuits so I'm not sure it'll work correctly.

Negative, I just work slowly.

I think I will continue this project, though... I've already spent most of the money that is required.

What size chamber is your launcher using?

In your diagram, I take it PT1 is being used to increase the signal strength?

The IR LEDs are probably a good idea even if they arn't required; they, being (basically) point sources, should provide a 'sharper' signal than the diffuse lighting that we would otherwise get.

My chamber volume (3"x12") will be about 1/2 the size of yours, same with the barrel (which is 2" ID like yours).

In the diagram PT1 is exactly the same as the others, I was just trying to show that they are all connected in series and not in parallel.

The reason I've included the IR-LEDs is because there will be fairly little light hitting the detectors behind the spud, and a lot of light hitting the detectors near the muzzle (from light entering the muzzle). With the detectors in series the output is limited by the detector that sees the least amount of light. If no detectors are blocked, but the spud has passed one or more detectors, then the signal is limited by the ambient light in the barrel behind the spud. I'm a bit worried that that won't be enough light. When a detector is blocked by the spud then that detector limits everyone else and will probably drop the whole strings of detectors to basically zero current, which is good. I just want to make sure that the minimal signal with no detectors blocked, but several detectors passed, is strong enough to be easily detected.

I wonder if the detectors will pick up the flame front as it moves through the barrel?

I am toying with the idea of including three pressure sensors in the chamber (I've already got all three).
1. A piezo transducer to record the "pressure" profile during firing (uncalibrated), recorded with a PC sound card.
2. A standard 120 PSIG pressure gauge.
3. A 120 PSIG tire pressure gauge (peak pressure recording gauge).

Like your gun, I'm using a threaded coupler to attach the barrel to the gun so I thought I would replace the barrel with a threaded cap and do some closed chamber studies.

The standard pressure gauge (#2 above) will be used to calibrate the other two gauges. I'm going to use my standard Schrader valve + syringe setup to fuel the gun so I can use the Schrader to pressurize the chamber. This might let me get measurements of the static friction (by pressurizing the chamber slowly till the spud moves) and also do leak tests on the chamber. It might also let me measure how fast the pressure drops in a closed chamber due to cooling of the gases.


I really don't won't to stomp all over what you are doing, mostly because it was your idea to begin with. I think I'll focus mostly on the closed chamber stuff to start with.

No, no, stomping all over my turf is a good thing; it gives us more data.
But feel free to start with the closed chamber tests... not that it will do any good, considering the disparity of our build rates.

Hi, BLB

Way cool that you want to contribute data!

One thought; if you cannot get the phototransistors to work reliably, replace them by 100 turn coils of thin copper wire wound on some kind of narrow form, each with a 10kOhm series resistor. Hook up all the coils in parallel.

Blast an old hard disk drive with your biggest gun. There should be fragments of magnets remaining. A piece of, say, 2 x 2 x 2 mm in the projectile makes plenty of voltage in the coils.

I have actually tried all this; it works great! (and in any lighting conditions). The only reason I haven't written about it yet is that I have only made a sloppy prototype set-up, not suited for publishing I should wind some decent coils on PVC pipes that slip over the barrel.

Maybe a trigger coil of 250 turns right after the initial psoition of the projectile is a good idea too. It triggers the data recording (in one of those programs for oscilloscoping with a sound card).

Regards
Soren

Blast an old hard disk drive with your biggest gun. There should be fragments of magnets remaining.

Or you could just loosen the screws
No really, a lot of the stuff in hard drives is worth your time salvaging.
(disks, magnets, motor, various aluminum rings and such)

But yes, I know blasting a hole in it is more fun. One time I threw one in a forge while the disk was spinning (That was interesting). It started making weird noises immediately and then shut off after a couple seconds. The aluminum case melted off and the rest started burning blue.
Good times.

dongfang wrote:One thought; if you cannot get the phototransistors to work reliably, replace them by 100 turn coils of thin copper wire wound on some kind of narrow form, each with a 10kOhm series resistor. Hook up all the coils in parallel.
...

I have actually tried all this; it works great! (and in any lighting conditions). The only reason I haven't written about it yet is that I have only made a sloppy prototype set-up, not suited for publishing I should wind some decent coils on PVC pipes that slip over the barrel.

Maybe a trigger coil of 250 turns right after the initial psoition of the projectile is a good idea too. It triggers the data recording (in one of those programs for oscilloscoping with a sound card).

Regards
Soren

I've fiddled with this. Directly winding 4 coils of 30G wire on a 2"D piece of PVC pipe generates just barely enough voltage and current to be picked up my a microamp meter when you wave a powerful rare earth magnet (3/4"D by 3/16" thick) near the coil.

You should be able to wind the coils directly on the barrel. A little PVC glue or hot melt glue to hold them in place.

I think this would work well for a chronometer (a pair of coils near the muzzle) but I'm not sure if it will work for the research barrel.

Lets see, if you did use 100 turns/coil, on a 2"D pipe, 20 total coils;
(100turns/coil)(20 coils)(2Pi)=12600" of insulated wire

That's ~1000 feet of wire.

I wonder if the coils will interfere with each other? 20 coils, in series or parallel, will have a fair amount of inductance.

BTW, the wire needs to be insulated, "magnet wire", which has a very thin lacquer coating, would probably be cheapest.

I did some experimenting with a photo transistor (PT) to see if there is enough light inside a piece of 2" PVC for the detector to work without the illumination LEDs.

I mounted a PT on a stick and slid it to the center of a 6.5' piece of 2" PVC. I used an 8.3V battery in series with an 860 Ohm resistor and the PT, maximum current possible ~10mA. The PT passed 4.3mA of current with the pipe in full sun with rags in the ends. The rags were so the PT was just seeing the light that passed through the pipe wall, not light bouncing in from the end of the pipe.

Dropping a heavy cloth over the center of the pipe dropped the current to 0.3mA, a 14x reduction.

So it looks like there is enough light passing through the pipe wall for the PTs to see and the LEDs don't seem to be needed. The rest of the PTs are on order so it'll be another week or two before the barrel is ready.

Finished the chamber today; 12"Lx3"D, 3" cleanout adaptor, 3" to 2" reducer and 2" female threaded adaptor for the barrel junction, 3 spark gaps, chamber fan, piezo transducer in the chamber (to record Pressure vs. time), 0-200 PSIG gauge and a 120 PSIG tire pressure gauge to record the peak pressure.

Pressure tested to 40 PSIG and I'm having trouble getting the threaded plugs to seal. (I'm using a 2" threaded cap to close off the fitting where the barrel will go.) A couple layers of Teflon tape help a lot but the thing still looses pressure at the rate of a couple seconds per PSI. Not a significant amount of leakage for an actual gun but I would like to do a couple of closed chamber firings to see how quickly the pressure drops as the gases cool and to get an idea of what the actual peak pressure is for propane + air in a hunk of PVC.

I'm a little concerned about the chamber withstanding the closed chamber tests. The theoretical peak pressure for propane + air is ~120 PSIG (adiabatic). The 2" pipe should handle that without any problem but the threaded adaptors are not pressure rated. Any thoughts on the chances of blowing the threaded plugs out?

EDIT: The chamber also has a Schrader valve. There are a total of 13 holes in the chamber. All but the spark gaps are in the fittings.

I have personally taken 4" cleanout caps to 100 psig, although I don't do that sort of stupid shit anymore.
I would not expect your 2" fittings to fail... but it wouldn't be a bad idea to <i>not</i> stand inline with the fittings, based on the premise that they will most likely blow out in the direction the pipe is facing.

Heck, I don't stand behind by combustion launcher when I fire, and that's a sch 80 plug in a pressure rated sch 40 female adapter.

As far as the leaks go, I think it should be fairly simple to adjust the measured post-combustion pressure decline for the leakage rate, presuming of course that the pressure doesn't take a really large (ie, couple minutes) amount of time to thermally subside.