Mathematical and Practical Roadblocks Oh My!
We’ve completed not one but two rounds of accuracy testing and we’ve hit a bit of an obstacle. Initially, I wanted to have the data out in a “one or two week” timeline from the posting date but, for reasons we’ll get into below, I’m not comfortable releasing it yet.
The problem in my accuracy tests, is accuracy. I’m not getting as much as I should be getting. Take a look at the below comparison of X-component deviation* in the baseline groups from Punkworks and our first shooting session. Punkworks is getting a final result that is more than twice as accurate.
- Punkworks: 1.725 inches
- First Round: 3.6 inches
Obviously I was missing something… I came across the FS Centrifuge post on M. Carter Brown and I thought that it was a very likely explanation for my accuracy problem. The paint in bulk boxes doesn’t sit uniformly so the asymmetric mass distribution of odd settled paint throws off the accuracy. Seems legit right? I grabbed a round and forcefully tapped it against a level surface to settle the paint against the nose of the round. Along the same lines of how you would force a bottle of ketchup to settle at the spout. I visually inspected the round and the paint seemed to be pretty uniformly settled in the nose. To be safe I allowed the rounds to sit nose down overnight before the next round of testing and re-tapped them before firing. The comparison between Punkworks, the first round, and the second round is below.
- Punkworks: 1.725 inches
- First Round: 3.6 inches
- Second Round: 3.55 inches
No statistically important difference. I racked my brain trying to explain the difference, and while looking at the 8 shot group distinctions in the Punkworks data set, I remembered that their methodology artificially centered each magazine’s group. It follows then that the normalizing might also have a damping effect on the natural randomness of the rounds. The
most likely (confirmed) way they did this was to average each groups X and Y values, and then subtract that average from the actual values. This results in three groups whose average values (i.e. centers) are all at the same (0,0) point.
Edit: Ha, I knew I remember seeing a video about this: Tiberius First Strike Test Thoughts by cockerpunk. It touches on the damping effect that correcting the groups has.
(ostensibly) the same mathematical methodology to my groups and it made a difference. We got about an inch off the group, which is pretty significant. Definitely closing the gap but certainly not all the way there.
- Punkworks: 1.725 inches
- First/Second Rounds: approx. 2.6 inches
Now we just need to shore up that last inch.
Thinking about it some more, it could be the barrel setup. There is a short stretch of smoothbore before the rifling in my MR5 setup that could be throwing off the accuracy. We know that smoothbores don’t do much to the spin of the round; in some cases it can even rotate in the wrong direction. It’s possible that the smoothbore presection is allowing some rounds to get a counterproductive spin before they hit the rifling and that those rounds aren’t being stabilized as well as they otherwise would be.
It could also just be that my paint settling methodology didn’t work as well as long-term nose down or centrifugally settled paint. I might have just grabbed and visually inspected a couple of rounds that were naturally settled right.
Another possibility is that I’m botching something about the mathematical theory here.
So basically I need to shoot a group that has been centrifugally settled, shoot a group from a system that doesn’t have the smoothbore presection, and
mess with the numbers more/shoot a larger group size/get confirmation on the Punkworks math.
Once I figure out what’s going on with my accuracy I’ll feel better tackling my 121 accuracy/drag data.
The issue going forward is money. It’s a finite resource for most of us, and I definitely include myself in that number. It would be a serious kick in the dick to shell out money for the only tested rifled barrel that works natively with the MR5. That would of course be the Hammer 7 threaded Tiberius/Lapco barrel. The other less dick kickish option is to buy the threaded Hammerhead back for Ion/Impulse threads and do a test group on my Ion. There are a couple of concerns there but it’s a decent solution from a trouble shooting standpoint. Anybody have the Lapco barrel or a Hammerhead back in Ion/Impulse they’d be willing to loan out? I’d of course be willing to put up collateral through a third-party for the full replacement value. I can float the money easy as pie, but I’d rather avoid sinking it when I need to restock my paint. Other than that, I’m gonna move some gear on craigslist and buy paint as sales allow.
Oh and the studio/warehouse owner shot some behind the scenes stuff when we did our first shooting session. Bro did it gratis and I think it turned out pretty cool. Check it out here.
*I prefer to compare accuracy using just the X-component rather than the vector. I’ll get into the specific reasoning in
a future this post, but for now bear with me.
UPDATE (04-01-2014) : The x-component deviation numbers above are the average of the rifled barrel tests for the Lapco and the Hammerhead from the Punkworks test. My posted numbers are averages for the entirety of my testin using the Hammerhead. The first thing I tested was the manufacturing consistency with a sample size of 500+. It’s over here. Comparing it the previous works on the subject, my data suggests that the manufacturing process, on the ammo end, has actually improved. it’s possible that the barrels have gotten poorer. The twist rate is too slow and the length too short to reliably measure using conventional firearms methodologies (rotating handle barrel patch).
I dont think the data will be too useful at this point. There’s a lot going on in terms of variables and refinement between the tests but, I’ll put out a rough set of sheets later today or tomorrow. I have a lot of annotations and qualifications to make in the data so that they make proper sense, but I’m out in Daytona Beach picking up some machining equipment (lathes, mills, and hydraulic rams oh my! – I’m tooling up for some experiments with .300 BLK) and I wont be back till late. For now I’m just sitting out rush hour traffic.
In related news, I drank too much coffee yesterday afternoon and couldn’t sleep last night. So what I did do was jot up my argument for favoring the x-component. I’ll post it after this over here.
In terms of my platform, it’s an unmodified spyder MR5. I’ve rigged it to run on a palmers stab (as you can see in the “behind the scenes” pics). The MR5 uses a method of barrel attachment simialr to the T8x/T9x series of markers and Hammerhead doesnt stock a back for it. The MR5 did however come with an early run of an aftermarket standard threaded barrel adapter, so I’m using it with that. That’s where my concern for the smoothbore presection comes from. In addition to that I’m firing the marker upside down, and clamped to mobile contruction table/clamp setup. It’s stable and super conveinent for firing – I cut the bottom off of the mag, and feed the rounds in hopper-style now. It’s way faster to run the tests and I dont have the shifting POA problem.
In the first round of tests we measured in CMs out to roughly 0.1 resolution. That took way too long however, so I went out and bought one of those vinyl DND battle mats. It’s a premade 30×40 inch grid that makes assessing the rounds (and cleaning up afterward) super easy and fast. This we measured out a rough 0.1 resolution. I ran the numbers though, and the extra resolution doesnt add anything to the final SD calculation, so next time well just go for a resolution of 1 or 0.5 inches.
I did look over my first round of data today though and there was one outlier. When I shot my first baseline group with the first round of testing, I got an x-component SD of 2.48, which when accounting for the mathematical centering actually beats the Punkworks value. At the time, I thought the value was kinda crazy so I ran it again to be sure – I wasnt able to reproduce it. Thinking about it, I explicitly recall grabbing only rounds that were nose down at the top of the bulk box. It was easier to extract the rounds by just clamping the skirts with my fingertips. It’s possible that I grabbed a bunch of rounds that were naturally settled nose down for that one test. I think it’s even more likely considering that punkworks shot their rounds from the old stacked tubes.
UPDATE (04-07-2014) : And… heres the rough cut of the data. Be sure to look at the below before you dive in.
In the first round, my process with the variables was to create a baseline using a speculative “most accurate” combination of factors and then test each factor one at a time, while the others remain constant with the baseline. In the second round, I wanted the test to be more “apples to apples” with the Punkworks test so that the two sets could be readily compared or used together. Hence the difference in variable schema between the first and second round.
The chrono in the first few groupings of the second test was malfunctioning due to a dying battery. Those cells have been highlighted in a darker orange. The battery was replaced, and the reading became more reasonable and consistent with previous testing.
The rigid variable test group in the second test suffered from a shifting POA every four shots due to operator error. The groups are separated by dashed lines. I might artificially center them later (or let someone with excel skills take care of it). Edit: I ran the numbers and unsurprisingly the more groups you break a population into, the more exaggerated the corrective effect of the mathematical centering.
Some quick notes on the variable terminology:
- Rigid vs Free-Float: whether or not the barrel was braced against the edge of the firing bench/clamp to pretension the barrel and prevent flexing.
- Unsorted/Ratioed vs Sorted: whether or not the ammo lot was sorted to size. If the ammo was unsorted it was still made to reflect the size ratios found in the Manufacturing Consistency phase of the test.
- Untapped vs Tapped: whether or not an attempt was made to settle the paint into the nose of the round by a forcefull tapping.
- Tipless vs Tip: whether or not the barrel had the factory tip during firing.
UPDATE (05-11-2014) : News on the trouble shooting front. I ran a test last Wednesday, and eliminating the smoothbore presection helped accuracy (or it could be that closed bolts are naturally more accurate, but this seems less likely). The problem remains that we’re still no where near the right level of accuracy even after getting a very minor boost from artificially centering three groups of eight. At this point I think I might just have a bum HH barrel. Before running out of the door for a game in Tampa this weekend, I looked down the barrel and saw what appeared to be slight hill in the bore. Similar to what you might expect the negative side of a ding to look like. There wasnt a corresponding ding on the outside of the barrel though. Once I’m back in jax, Ill take another look to confirm, but I’m fairly certain there was slight bump in the bore (a manufacturing error maybe?) and that it might be whats frustrating my accuracy.
In related news, I also tested centrifuged rounds. Turns out that they are more accurate than plain jane rounds. My impact plot data for that test was messed up, so I dont have hard standard deviations, but what I do have is the ability to compare down field chrono sweetspots.
Edit: I also fixed my chrono problems. If you recall, I suspected that the light source I used was washing out the chrono’s ability to detect peripheral passes. I bought some rope light and stretched them across the plastic light diffuser that sits above the eyes, and got much better sensitivity.
Basically there is a trapezoid area above each eye of the chrono, that a round has to pass through in order to properly register at that eye. The machine counts the amount of time that passes between each register and calculates the speed from that. If the round goes through the chrono but is outside both of these areas, then you wont get a read (referred to later as a ‘no-read’). If it passes through the first trapezoid, but not the other, then you get an ‘error’ reading. In the plain jane test, I got several ‘no-reads’ and one or ‘two errors’. This means that a lot of rounds were so wild that the chrono was essentially blind to them, and at least a couple that were less wild but still bad enough the the chrono had no idea how fast they were going. The centrifuged rounds, had one ‘error’ reading and no ‘no-reads’. No hard vector based standard deviation, but still more accurate.
Edit: For the sake of completeness, the data was recorded but it was very rushed and I’m not confident that I didnt switch some of the positive/negative signs in my haste. I caught myself doing it once or twice and needless to say, this is something that can inflate the final standard deviation. I’ve added this third run to the previous spreadsheet.
UPDATE (05-14-2014) :
The culprit. I usually check bores from the chamber end and this bore defect is small enough that you’d only ever notice it if you were holding the barrel at the right angle to a bright light source while looking down the muzzle. Something I dont do as a matter of habit, and something I only did as a matter of expediency. It sits very close to the muzzle of the barrel and is probably not doing my accuracy any good. I’ve emailed Hammerhead, so hopefully well be able to get it switched out and tested soon.
UPDATE (08-22-2014) : Lack of updates right? Well I lost my shooting space. The good news is that Hammerhead sent a new barrel.
Once my replacement made it in, I called up Bubble Wrap only to find that they had just finishing moving to a new space. Cleaner, brighter, taller ceilings! Great for photographers. But just a bit under 50 feet at it’s widest. Bad for me. That essentially put me in shooting range homeless limbo.
I know I’ve mentioned it elsewhere but finding indoor spaces in the sub-tropical south is a nightmare. Indoor sports arent really a big thing down here, on account of it never getting that cold, and even when you do find a space big enough, you still have to talk the owners into letting you shoot things inside it. I spent about a month searching with nothing to show for it. I almost scored an office complex with a long hallway, but the building owner was (understandably) skittish about his windows. On a lark I re-emailed the local hack space JaxHax. Previously, they had showed interest in the project but for reasons I cant remember, they weren’t suitable at the time. But as luck would have it, they just purchased a second story in the multi-level warehouse complex they occupied. A very large first floor space with over 100 feet of clean and clear linear distance. KO.
The best part is, the price they were willing to let me shoot for. My photography studio had charged $50 an hour, which compared to other hourly studio rates, is a bargain. JaxHax was willing to let me have unlimited 24 hour access for the breath taking price of $75 a month. Perfect KO.
I’ve been getting the tests moving again and I cant tell you how sweet it is there. The lack of time pressure and hard deadlines makes everything so easy and convenient that I cant explain all the ways it manifests into an improved data product.
But, more to the point. A few days ago I completed what I consider to be a near perfect testing run with the new barrel and I can safely say that my previous accuracy troubles were barrel based. I wont get into the final numbers yet, since the test is ongoing, but there are already some interesting accuracy implications about manufacturing quality OR closed bolt markers. Since it was the only apparent flaw in the barrel, I feel confident that the micro-minuscule bump near the crown of the muzzle was my issue. I’ll write more about it at some point, but I’m more than a little miffed with Hammer Head at the moment. That bad barrel cost me several hundred FS rounds to diagnose.
Anyway, testing is live again. And just to titillate a bit more, our first session yielded some solid dual chrono numbers for rifled barrels. Mmmmm, impending BC answers…..
UPDATE (09-19-2014): This should be the final trouble shooting update. After confirming that the barrel was my main accuracy problem. I set about testing different markers, the Empire Tracer worked great but for some reason, the velocity on my MR5 wouldnt go higher than 260. Obviously, I cant test with that. Figuring I’d hand load the rounds into the barrel, I switched to my old school Ion, but for some reason the board was malfunctioning and it wouldn’t let me shoot with the eyes off (and yes, I was using a fresh battery). A couple of other guns in my inventory later and I still hadn’t found a suitable shooter.
At this point, I figured I’d have to buy a new gun to get the test moving in a timely fashion. After putting some feelers out on craigslist I found a pawnshop had listed an old Spyder clone for the paltry sum of $30. I snapped it up figuring I’d make modifications to it to get it running up to spec. Imagine my surprise to find that the Spyder is just about perfect out of the box. Its one of the older half-tank breeds with flat sided heavy duty frame, meaning that I can clamp the ever living shit out of it for great stability. The feed neck slides off the frame and allows me to easily load FS rounds directly into the breech – no further milling or dremeling needed. Finally, as an added bonus the bolt has a groove in the nose that perfectly held an oring after the ball detent was removed, and now the gun is obnoxiously efficient. In short an unexpected unicorn.
You can find the community discussion of this post at the following forums: