Dragon's Eggs

WARNING AND DISCLAIMER:  If you are underage, then consult with your parents or guardians before attempting any of this.  You are on your own - I'm not responsible for your actions or harm you may bring to others because of your actions.  Making the items described below  can result in injury or death to you or people in your vicinity. Some things mentioned here may be illegal to make in your city, county, state, or country so check the laws that apply to you before you attempt anything described here. These notes are not complete on purpose. If you are reading them and new to pyrotechnics, then you are making a mistake. Stop now - this page is not for you. Get a beginning book on fireworks (see Skylighter or American Fireworks News (very quick shipping)  for a start) and read up. You can't make any of this work without more information so read up or join a club or ask someone to help you.


Mike Swisher - arguably one of the greatest living experts on pyro - gave me permission to publish some of his posts on this website. Thanks, Mike. Lloyd Sponenburgh - another heavy hitter - also contributed a post he made on rec.pyro.  Thanks, Lloyd.

Mike's posts are underneath the pictures (below). To go straight to Mike's posts, click HERE. Lloyd's comments are just after that.  To go straight to Lloyd's comments, click HERE . For a summary of a possible problem area that Lloyd discovered, go HERE.

Before we get to Mike's and Lloyd's comments, here is my way of making dragon's eggs - pictures  follow the comments.



I press them in a BP matrix and they respond well. I have not tried a hammer test with them but my understanding is that you can get them to pop if you hit them with a hammer.  I also roll them and use them in breaks. I have used a rolling pin to break up the sliced eggs - after they are fairly dry. However, I usually don't wait too long. When they are completely dry, they are tough to break up and it gives me the jeebies to crunch them hard enough to snap them into pieces.

Best Practices

There has been much - very much - posted on this. However, there has been a lot of misinformation about them. Almost everything stated about nitrate leeching and how to prevent it is, at the least, skewed and probably incorrect. You don't need to wax them, they will last at least two years once made and primed (I tested specifically with prime, without prime, with and without wax). The best formulas are the simple ones. My recommendation is to keep it simple until you know what the basic product will do.

Try this formula first if using the bismuth products:

bismuth trioxide 37.5%
copper(II) oxide 37.5%
magnalium 25% (finer MgAl means quicker pops - less smolder phase)

+10% NC

Here is an alternative formula from AFN III

Bismuth Trioxide                                  81.8
Magnalium, granular, -100 mesh            9.1
Copper(II) Oxide, black                          9.1

(add +10 or so NC)

You can use bismuth subcarbonate instead of bismuth trioxide, 1:1 This formula has also been successfully made using lead oxide/trioxide but I bound the mix in a BP matrix for a ground display. That means I mixed 10% eggs with meal and pressed it in a tube - here is a recent test using lead tetraoxide instead of bismuth, note the delay with some of the eggs (they were a variety of sizes). Lead is a little harder to get right but it gives louder pops:

Here is a test with the bismuth eggs surrounded by Fuch's crackle:


Roll the dragon's eggs dough between two pieces of baggy and between two popsicle sticks. Cut in 1 to 2 mm squares. I usually run a noodle cutter up and down them and then carefully run a razor blade between the hatched noodle cutter lines. You can do it any way you want, though. Just cut them small.

I bind with a solution of NC (double base smokeless). It doesn't matter how thick the solution is but it does matter that you add about +10 parts to the formula. Prime with green mix and a metal - I used green mix +10 silicon +5 dextrin. Note that if you are having difficulties, the first place to start fixing things is here. Many people shirk on the NC solution. Note that the eggs have an added 10% NC.

In the above baggie is a mixture of 37.5 lead tetroxide, 37.5 copper oxide, and 25 -325 MgAl  with thick NC lacquer made from double base smokeless powder

Roll the mix out on a pan between two baggies. You can use popsicle sticks (ice cream treat sticks) as borders to get a uniform patty.

You can use a noodle cutter or a sharp knife (or both) to make the 1-2 mm cross hatching in the patty. Let them dry for a while and then break it all up with your fingers or a rolling pin.

Other possible problem areas

See last section of this page for a reprint of a post by Lloyd Sponenburgh on a problem he encountered at work. The essence of this post is that all MgAl is not the same and that some problems with the smolder phase may be due to larger than normal aluminum content in the MgAl.  One way to offset it is to add just a touch of fine Mg to the mix. Experimentation is necessary.  Click HERE to go directly to that article.

About Mike's posts

I started looking them over and decided I didn't want to spoil the discussion so I'm just leaving the posts intact. These posts are out of context of larger threads (in most cases) - I won't add many comments because, even out of context, they are quite good and to the point.

Comments by Mike Swisher (used with permission):

Posting 1:

<OP>, the original 'dragon eggs' I examined in 1988 were made using granules bound with a phenolic resin and matrix composition bound with collodion. Collodion is a low nitrogen content NC that is soluble in a mixture of alcohol and ether. Camphor is added as a plasticizer. The reason a plasticizer is necessary is that the pharmaceutical use of collodion is as a wound sealant. Collodion USP is sold under trade names such as ''Nu-Skin'' for application to minor cuts and abrasions. It has been around for years.

My impression is that the Chinese chose it simply because the solvents would not attack the phenolic binder of the granules, so that they would remain distinct from the matrix. If the binder of the granules were soluble in the solvent of the matrix, the granules would disintegrate and blend with the matrix composition, and the effect would be spoilt.

Posting 2:

I never had much problem cutting the crackle granules when using a solution of NC in pure acetone. MEK will, as noted, slow down the drying but I'm not sure this is an advantage. My technique is to weigh out perhaps 12 oz. of the dry composition, add NC solution, and work it in till the consistency of the mix was about that of pie crust - then, placing this on a sheet of paper, and using an aluminum roller, roll the ball of composition out to 1/8'' or perhaps a bit less thick - then dice the rolled out composition quickly with a long relatively straight-edged knife. I do not bother trying to break the granules apart or to dust with prime while they are damp. I set the sheet of paper and granules to dry and break them up later, sorting them by sieving. Any that do not pass 8X8 hardware cloth are broken up till they do. Any that are not retained on window screen are re-powdered and may then be damped with acetone alone, then are again rolled out and cut, etc.

It is possible to roll out and cut quite a large amount of composition in an hour simply by repeating the above process with small batches of the size indicated above. Admittedly, if one wanted to achieve greater precision - as, e.g. for cores to be used in round stars - cutting would take more time. Since I have used my crackle granules embedded in a matrix mixture for relatively large pumped comets, this has not been an issue for me.

Posting 3 This postig was amid comments as to the fear of using lead oxides instead of bismuth - one poster said it wasn't as bad to use lead oxides as everyone was making it out to be. Simply use gloves and a respirator and wash your clothes afterwards. Make it outside if possible.

I will agree with you about overstatement of lead toxicity.

The principal hazard of lead poisoning appears to be to small children who often ingest gross amounts of it from chewing on old lead paint chips, recent Chinese toys, etc.

Good housekeeping and personal hygiene make sense in handling any pyrotechnic chemicals; disposable gloves, a good respirator, washing utensils and clothing used during mixing, and promptly washing yourself after working with them should be sufficient precautions for an adult. I employed hot-metal typesetters for 20 years, some of whom had been in the trade all their lives, and none ever suffered lead poisoning. A few of them had problems with ethanol toxicity but that is another matter!

I will also observe that one of the world's great problems is not lead poisoning but lead deficiency. Many individuals would benefit immensely from a dose of 240 grains, administered cranially.

Posting 4: This thread dealt with whether crackle can be rolled in a star roller

There's no reason why crackle couldn't be rolled, I just haven't bothered trying to figure out how to do it.

I suspect there may be some differences in performance because the ratio between surface area and volume/mass of an angular granule is significantly greater than that of a sphere. This will certainly affect ignition and probably the rate of the reaction once ignited.

Posting 5: This thread was about whether water could be used as a binder

In my experience NC helps the heating or smoulder phase of the crackle reaction. Other binders do not have this property.

It is also important that the crackle granules be bound using a solvent and binder that are not the same as the solvent and binder employed for the matrix composition in which the granules are distributed. The purpose of this is to preserve the heterogeneity of the granules. If an aqueous binder were used for both, the granules would soften and blend together with the matrix, destroying the crackle effect.

I note that the original 'dragon eggs' imported c. 1988 contained granules bound with some sort of phenolic resin and a matrix composition bound with collodion (low nitrogen content NC + camphor, dissolved in a mixture of alcohol and ether). The camphor smell gave the binder of the matrix composition away. A splash of acetone quickly dissolved the binding of the matrix, leaving granules that did not soften or dissolve in the acetone.

Thus, NC/acetone for the granules and dextrine/water for the matrix is not a necessary combination. Any two binding systems that use binders soluble in different solvents may be used, in order to preserve the heterogeneous nature of the combined granules and matrix. This, however, does not tell the whole story, as the use of NC to bind the granules yields definite benefits in their performance.In my experience NC helps the heating or smoulder phase of the crackle reaction. Other binders do not have this property.

It is also important that the crackle granules be bound using a solvent and binder that are not the same as the solvent and binder employed for the matrix composition in which the granules are distributed. The purpose of this is to preserve the heterogeneity of the granules. If an aqueous binder were used for both, the granules would soften and blend together with the matrix, destroying the crackle effect.

I note that the original 'dragon eggs' imported c. 1988 contained granules bound with some sort of phenolic resin and a matrix composition bound with collodion (low nitrogen content NC + camphor, dissolved in a mixture of alcohol and ether). The camphor smell gave the binder of the matrix composition away. A splash of acetone quickly dissolved the binding of the matrix, leaving granules that did not soften or dissolve in the acetone.

Thus, NC/acetone for the granules and dextrine/water for the matrix is not a necessary combination. Any two binding systems that use binders soluble in different solvents may be used, in order to preserve the heterogeneous nature of the combined granules and matrix. This, however, does not tell the whole story, as the use of NC to bind the granules yields definite benefits in their performance.

Posting 6:

I have made hundreds of pounds of crackle comets using NC-bound crackle granules and a water-bound, nitrate-based matrix composition. I never experienced any failure of the granules to crackle on account of moisture or nitrate.

The purpose of using NC to consolidate the granules is twofold. First, NC helps the preliminary heating or smoulder phase of the crackle reaction. My experiments in 1988 showed that granules consolidated with NC in acetone performed better than those consolidated using starch and water. Second, because different binding systems are necessary to maintain the heterogeneity of the granules and the matrix when the two are mixed together. We want the granules to be distributed in the matrix to make our crackle stars in the same way gravel is distributed in cement to make concrete, ot as raisins are distributed in rice pudding. If an aqueous binder were used for both the granules and the matrix compositions, they would blend together on dampening, and the effect would be lost.

I should note that the original Chinese dragon's eggs used granules that were bound with some sort of phenolic resin that was insoluble in alcohol, acetone, or water. The matrix composition was bound with collodion (a solution of low-nitrogen NC in alcohol and ether, to which camphor is added as a plasticizer). The use of collodion was given away by the camphor aroma of the consolidated mass. What is signoificant is that any two binding systems may be used for granules and matrix respectively, so long as the solvent used for the binder of the latter does not dissolve the binder of the former. The important thing is to keep the granule distinct from the surrounding matrix.

I believe based on what I have read here that the people who are experiencing problems are insufficiently consolidating their granules. One wants a hard, dense granule. I dampen my composition with NC solution, mixing them to form to a stiff dough, and roll it out to a thickness of perhaps 1/10 inch. I then score this finely using a sharp knife and when it is dry, break it up and sieve it. I use what passes 8X8 hardware cloth and is retained on 20-mesh. Whatever does not pass 8X8 is broken up till it does; whatever is not retained on 20-mesh is dampened and re-consolidated as previously described. This process takes more time than granulating the dampened mass through a screen, but results in much denser granules. It also does not leave the granulating screen covered with a sticky mess.

Posting 7: This thread was about degradation of dragon's eggs by leeching of the nitrates into the mix and causing the pops to stop.

I am puzzled by the reports of degradation.

I have found that crackle granules made with a solution of NC keep well, and for an indefinite time, without the necessity for waterproofing, wax coating, or other such steps. To those who are having problems - with what do you bind your granules? How do you consolidate them? how do you handle them after granulation?'

Posting 8: In a response about whether magnesium could be used instead of magnalium

Magnalium is generally used, though Shimizu has published some compositions that rely on litharge and aluminum alone. I do not think magnesium alone would work.

The crackle or dragons' eggs effect is best understood as a specialized strobe composition in which the normal strobing sequence of (smoulder/flash/smoulder/flash... ) is unable to continue because the flash phase is so violent that the first flash precludes further propagation of burning. The idea is to have a sufficiently small piece of consolidated composition that the whole mass of it becomes semi-reacted in the smoulder phase before it proceeds to the flash phase. Shimizu showed that the magnesium in the magnalium is consumed during the smoulder phase, and the aluminum then yields the flash (explosive) phase.

Posting 9: A follow-on to 8

The magnesium strobes work because they have a two-oxidizer system - the smoulder, or low-temperature, oxidizer is ammonium perchlorate, whereas the flash, or high-temperature, oxidizer is a sulphate (typically alkaline earth, ee.gg. barium, strontium, magnesium).

The barium nitrate/magnalium strobes work because they have a two-fuel system - the smoulder, or low-temperature, fuel is magnesium, whereas the flash, or high-temperature, fuel is aluminum. Typically the Mg and the Al are alloyed, but a mixture of powdered pure metals works. According to Stanbridge, Brock had such a composition called ''Goldern Shimmeron'' in the early 20th-c., which was displayed in the famous peace display of 1919.

Posting 10: More on using only NC as a binder

The benefit of nitrocellulose is that it actively encourages the smoulder phase of the crackle reaction. It does so because it is internally oxygen-balanced and actively burns by itself.

When I first investigated this effect in 1988 I found that binders that were fuels only (e.g., dextrine, starch, gum arabic, shellac) were not nealy as helpful to the effect as was NC. Cellulose acetate would behave more like a starch, gum, or resin than like NC.

Another method of making dragon's eggs (from rec.pyro postings - author Lloyd Sponenburgh)

Here's a re-preach from an earlier question:

The formula is no secret.  The real secret to getting good crackle is just getting enough NC in the mix.  You'll understand lower down in this email.

This formula has been circulating on the internet for years:

71  Bismuth trioxide
14  Black copper oxide
10  Magnalium (see notes)
 5  fine atomized aluminum (-300 mesh, like Service Chemical "X-Fine")

The magnalium for "regular" crackle is 200-mesh.  The coarser the magnalium mesh, the longer the delay between ignition and explosion.   At 80-mesh, it's about three seconds, and pieces as large as ¼" will explode in a single explosion.  The finer the magnalium, the smaller the pieces must be.  At 200-mesh, 1/8" dia. stars are about as large as you can make them.  Otherwise, they just spall small chunks, and
the body remains un-reacted.

The mixture is screened well, then mixed into a wet, sticky dough with 10% w/v nitrocellulose lacquer made from double-base powder (like Green Dot or Blue Dot or Bullseye shotshell powder).

There's a secret to making the lacquer that I've published over and over, and folks STILL think it takes a couple of days to make.  In reality, it takes about 30-minutes.

Measure out the acetone in a vessel at least four times larger than the volume you're working with.  I use a deep stainless steel mixing bowl from a kitchen mixer.  With a kitchen whisk in your strong hand, and the weighed NC powder in the other, start slowly but uniformly pouring the powder into the acetone while you whisk your arm off.  Mix FAST, and don't stop until the whole mass thickens up to about the consistency of heavy cream.  Work lumps down off the sides of the bowl as you go.  Don't _ever_ let any undissolved powder settle to the bottom.

Now, cover the bowl and let it sit for about 30 minutes.  Come back and whisk again for about three or four minutes, and your lacquer is ready.  You can ignore any "soft lumps"; they'll dissolve during the
kneading process. In another large, shallow bowl, "crater" your crackle mix, and start adding NC lacquer, a little at a time - just like making bread dough.

Knead it thoroughly.  At first, it will become "mealy" like making pie dough.  As you add more NC, it will become more and more dough-like.   Knead it until it's perfectly smooth and homogenous, and quite a bit
on the sticky side of too wet.  It should end up almost impossible to knead properly any more because it's too sticky to handle.  It should be about the consistency of a very thick batter, rather than a dough.

This is hard work.  Wear stout rubber gloves.  Scrape your gloves down into the wet mix to get the dried chunks back into the mass.  I know this might sound silly, but DON'T allow sweat to drip into the dough…
don't.  (you will be sweating)

It's important to the loudness of the crackle to get enough NC in there.  If the dough ends up too sticky to handle, that's OK… Just mush it out flat and let it dry a little.  As it dries (and it dries pretty fast) it will lose its stickyness.  Keep working the wet mass until it returns to the consistency of a stiff dough.

Once the dough is back to the right consistency, more "push" than "scrape" it through standard ¼" hardware cloth onto trays lined with paper.  You'll get ¼" "worms" about two inches long or less.  Don't
layer them more than a couple deep, or they'll all stick together.  If they do, let them dry a little, knead them back into a lump, and repeat.

Let them dry slowly in the shade without molesting them JUST until you can "diaper" the mass without having the worms stick to one-another.   The goal is to get them partially dry, but not so dry you cannot
further process them for smaller grains.

Now, working with a small amount at a time, push these worms through 6-mesh screening.  6-mesh gives 1/8" stars.  Especially now, don't layer the mass; you must keep the stars from adhering to one-another.

As they dry a little more (the "knack" part comes now), put them back in the 6-mesh screen, and gently ROLL them back through the screen with the flat of your gloved palm.  Avoid scraping or pushing straight
down.  What you're doing is basically rounding off the chunks and forcing them to size in one operation.

Let these stars dry thoroughly, occasionally diapering the mass to get damp ones to the surface.   In the sun, they should dry completely in an hour or so.  You should detect no acetone odor, but you will smell
a "sharp" odor from the nitroglycerine in the NC lacquer.

Finally, re-screen the entire mass to remove chunks larger than 6- mesh, and sift out any "fines" smaller than about 10-mesh.  The big pieces and the fines may be re-processed later simply by adding acetone and re-kneading.  You can make really small stars for crackling gerbs with the -10+20 pieces, if you wish.

Now, prime them.  Make up a saturated solution of potassium dichromate as your rolling solvent. DO NOT USE plain water - the stars will react with water, and can actually ignite from the heat.

Start a fairly large mass of the dried stars rolling in your star roller.  They must tumble freely.  They're so heavy that a small mass will just slip around in the drum.  In fact, they may not tumble at all until you start to dampen them with solvent (gives a little "bite" on the surface of the drum).

Wet them with the solvent until they glisten, but still tumble freely.   Lay on a layer of simple rough-mix powder (75/15/10 +7% dextrin). Do not mill this… you want it coarse; just mix it by screening repeatedly
through 20-mesh.  Just like star rolling, add dry mix until they'll not take up any more, but try to keep the layer smooth, without excess dry mix in the mass.  

Do only a couple of layers.  The goal is NOT to build them up in size, but just to evenly coat all surfaces of each grain.  There shouldn't be any grey crackle showing, but the grains should be only marginally
larger than when you started.  Finish up with a large excess of dry powder to make the grains' surfaces "dusty" for easier ignition.   Uniform ignition over the entire surface of the star is part of the secret to getting it LOUD.

Dry these again in the sun and gently sift out the excess priming mix, and you're done - with straight crackle.

Add up to 10% by weight of fine titanium sponge or flake to the original mixture (60-100 mesh) to get the "brocade" burst effect.

Around five pounds per batch is about as much as one person can handle comfortably at one time.

Although this sounds like more work than making cut stars, it's lots faster, and turns out a good product.  Even for cut stars, you have to do everything but the screening phases, and cutting uniform 1/8" stars
is difficult.

Once you've made five _finished_ pounds of these, call me up, and we'll get you some therapy for your arms.

Let me know when you've achieved a forty pound day!

A possible problem area:

Reprint of a post by Lloyd on Passfire (with permission). 

We had an experience last week that taught me quite a bit more about how crackle does (and should) work than I knew.

A number of folks here have had problems making crackle that would work in the air. They often related that it would work on the ground, unprimed, but then fail after priming, or when shot in a mine or as star cores.

Mine and a couple of others' responses have pretty much been "What's the problem? Crackle is EASY!", along with instructions on how to get more NC in the mix, how to process it mechanically, and other well-known but not-very-helpful information.

I think I have some insight into those problems now, and hope our experience last week can help everyone. What follows is a look into a purely empirical solution to a problem. It doesn't require a lab; only careful observation.

To whit:

Over the last month, we've noticed the performance of our crackle diminishing; primarily in the time-domain, as it went from a glorious CRASH (all at once) to a ragged, irregular train of explosions. It didn't fail all of a sudden, but got worse and worse over a month, until its performance was just plain unacceptable.

Our first assault on solving the problem was to review our process. It's easy for a process to "evolve" when staff do it continually. So we watched them making the material, and saw no problems. Next I did it myself. Since I have not been making crackle for over three years, I had nothing to use but our original documentation. The product turned out the same.

Many times, I've told folks, "it's the NC... get lots of it in your mix". We hadn't changed the amount, but realized that a couple of months prior, we'd exhausted our supply of authentic Hercules Green Dot powder, and were now working with the Swedish-made Alliant equivalent -- but it's not "green dot". It has no "green dots", and it's obviously been re-badged by Alliant, with their labels applied over the original Swedish labels. So I tracked down a couple of pounds of "the real stuff", and made a batch. Darn! No change.

It was unreasonable for me to only realize this at this point in time, but there was something really basic we'd missed. The crackle didn't just suddenly fail. It deteriorated in performance over time. Wha...?? Oh. We re-process fines from the granulation process. There was, at the beginning of this cycle, probably 100lb of 'old' fines that had been recycled into batches. As the old fines were slowly being replaced by 'new' fines, the performance deteriorated more and more.

That insight didn't solve the problem, but it pointed to something being wrong with one of the four constituent chemicals. We prayed it wasn't the bismuth trioxide, because we had a LOT of it - three hundred pounds - and the last shipment arrived about the same time this problem began to show itself. But we received new MgAl on that same truck.

We buy all of ours from the same source, and it has varied in appearance, smell, and performance pretty much on every shipment. That was the material we chose to suspect first, but we don't have the facilities to assay magnalium quantitatively.

So... how to tell?

We had some old crackle available as a control. Burning it on the ground resulted in a smoulder phase of about 1/2 second, with a rapid and sure transition to the explosion phase. The new crackle smouldered for almost two seconds. On the ground, it all exploded, but in the air (in crackle mines), quite a bit of it was coming down as burnt-out clinkers... it had smouldered then gone out.

We started some formulation experiments. Our crackle has bismuth trioxide, magnalium, copper oxide, and additional aluminum (plus, of course, the NC).

First we tried an OLD bucket of 60-mesh MgAl, knowing it had always given us about the same results we were seeing now. That is, the smoulder phase was too long, and didn't reliably transition to a pop when being cooled by flying through the air. When we made a batch, it performed almost exactly like the bad material made with -200 mesh new magnalium. That was a revelation, and we were beginning to form an hypothesis: The 200 mesh magnalium was somehow defective.

As a first look, we added a few percent of free magnesium to the formula. The smoulder phase went from around two seconds to almost zero! Wow! Well... that makes sense. The magnesium is the fuel consumed during smoulder, and normally it must be "robbed" from the MgAl alloy. Here, it had a source of free magnesium to burn, and it certainly did. The problem with that as a solution for the problem was that the amount was tiny, and not easy to control. Even a scant percent added to the formula tended to make the smoulder so fast that the particles exploded with spits and snaps instead of single loud pops. It had approximately the same effect as replacing the 200mesh MgAl with -325mesh material. Hmmm... finer particles expose more magnesium to the reaction... Hmmm...

But... our formula contains excess aluminum. So we started backing down on that, and replacing what we removed with more magnalium. Sure as can be, within one or two steps of binary search for a "correct" quantity, we'd found an amount that restored the smoulder to around a half-second, and brought the reliability in the air back to normal.

Running the figures backward, we calculated how much magnesium and aluminum _should_ have been in the mixture with the original formula, _assumed_ that same amount must be in the new modified formula, since it now worked correctly, then evaluated the MgAl on that basis.

It looks like the MgAl was 65% aluminum and 35% magnesium (which is a standard alloy), instead of the labeled 50/50. In hindsight, that's to be expected. Our vendor does not assay chemicals. They receive drums of MgAl from China, re-stencil them with their proprietary markings, and ship. It could have been zinc, and the label still would have said, "50/50 magnesium aluminum alloy".

So, our crackle problem is fixed. How does this relate to the guys who are having trouble with theirs?

1) examine the smoulder phase of your un-primed granules (about 1.5mm-2mm size). If they smoulder for a longer time than a second, they may not be reliable when moving through the air. (larger granules will smoulder longer)

2) to effect a change in this, you can add TINY amounts of magnesium, reduce any excess aluminum in the formula, seek an MgAl alloy higher in magnesium, or decrease the particle size of the MgAl (perhaps only replacing part of the coarser material with finer), which has the effect of raising the amount of magnesium exposed to the oxidizers.

3) changing the ratios among oxidizers and metals can have beneficial effect when you don't have an easy way to increase the magnesium content.

As I said before, this wasn't a highly technical trouble-shoot using quantitative analysis in a lab. It was empirical "shovel-full of this and spoon-full of that" experimentation with some careful observation. It's the sort of thing one can do without a lab.