View Single Post
Old May 12th, 2008, 23:57   #1
Delierious Designer of Dastardly Detonations
MadMax's Avatar
Join Date: Dec 2001
Location: in the dark recesses of some metal chip filled machine shop
GBB parts fitment: Why does my gun look hawt but suxxors when I have to bust a cap?

I often get asked what upgrades to a GBB are good performance and reliability upgrades and which ones should be avoided. Unless I've got direct experience with a particular build that was successful, I usually cheap out with some liability dodging response:

"I have no idea, because I haven't made many of that particular build myself, I cannot comment on whether or not it will be durable, reliable, effective, or propane tolerant."

I dislike having to give such a vague and useless answer but to tell you the truth, it's very difficult to work out how well upgrade parts will function in a host GBB especially when serveral brands of upgrade parts are to be used.

The short answer is that many upgrade parts manufacturers do not appear to understand the functional principles behind a host GBBs design so they can miss critical dimensions. These parts may look very similar to the original parts, but these small differences can cause significant fitment and functional problems especially when they're used in conjunction with other subtly incorrect parts. In short, looking like isn't the same as working like.

The long answer is a work in progress. I'll be touching on many aspects of product design that affect how well your upgrade plans may turn out. There's a lot of ground to cover so I'll be adding to this opening post periodically as I write additional material. My mission plan is to start with a fancy pants higher level understanding (general principles) and finally try to drill down to nitty gritty and illustrate some of these principles with practical specific GBB examples.

On Tolerance Stacking?

In complicated assemblies, several parts will interact to perform a function. When manufacturing parts, it's typical to produce parts which are not exactly identical to one another. For instance, if an automatic machine were set up to manufacture 1" long nails, you'd get an assortment of parts with some slightly longer than 1" and some shorter than 1". When it comes to 1" nails anything between 0.95" and 1.05" is probably quite acceptable so a super high speed machine could be used that bashed out nails in a rapid fashion. Your manufacturing process would crank out nails which were typically +/-0.05" long at an average length of 1". This is still a simplification of what you'd actually see. If you were to measure every nail and record it's length in say MS Excel, you could calculate an average length as well as a standard deviation of lengths. The STDEV would give you an idea of how many nails would fit within a given range of lengths (e.g. 98% in the 0.985"-1.025" range with 2% outside said range). Instead of diving into that stats class you dropped, I'm going to skip it in this discussion, but you can see how finite math actually plays a role in determining the percentage of working assemblies you could manufacture with a given quality of manufacture.

If I were to make a simple assembly of three nails stuck end to end, then the total variability in assembly length would be +/- 0.15" which would be three times the error of an individual component.

This phenomenon is called Tolerance Stacking and it can cause significant variations in the functionality of assemblies. One example of this in GBBs is the interface between a slide to an outer barrel (OB) to the inner barrel (IB). The gap between a slide and the OB is necessary so the OB can pass through the slide when it recoils. The small gap between the OB and IB is necessary so the OB can move around and clear the ejector port in the slide. The small gaps between all these parts assure that these parts don't end up locking together in most assemblies. If the design of these parts were meant to be line to line (e.g. pass in slide same avg diameter as the diameter on the OB) then some slides would have holes that would not be able to pass an OB that was slightly too large.

Unfortunately this necessary gap also means that there is some play between the IB and the slide. If your slide is forward (in battery) you'd be able to jiggle the IB around a bit relative to the slide. Since your GBB sights are mounted on the slide, this looseness can allow the IB to move relative to your sights which would increase your shooting group size. This is a simple example of tolerance stacking in GBBs affecting shooting performance.

When you build a GBB with two brands of upgrade parts, you're hoping that three manufacturers who do not cooperate and share information will manufacture parts that will work together. Different manufacturing processes will produce parts with different variability in diameters or lengths. If manufacturers don't communicate, (e.g. TM does not give it's engineering drawings to Guarder or RayRax) it's difficult to tell what the average diameter TM expects of it's parts or the necessary consistency between parts.

So how should one duplicate upgrade parts?

If one was to manufacture upgrade parts by duplicating stock parts and skip the messy part of actually having to figure out how things actually work, measuring a single parts set (buying one GBB to copy) would not provide sufficient material to duplicate. Because of the variations between manufactured parts, the parts in a single GBB would consist of parts hovering around the design average, so duplicating these parts would yield parts that were duplicated around the wrong average! The only way to figure out what design average TM was working around would be to to a statistical survey of say 50 GBBs and do something called a tolerance analysis. This would be a methodical measuring of critical dimensions of each part which would be enough data to calculate a meaningful average as well as allow the statistical analysis of TMs manufacturing processes. If you want to make parts that fit TM GBBs, you have appreciate what kind of variability you will see in the host build in order for you to design your own parts that will fit in almost every TM GBB.

Wha? The calipers they do nothing!!? Parts are too complicated to do statistical analysis!

GBB parts can have strange shapes so it's difficult to know what critical dimensions to duplicate. If you do not understand all of the subtleties that go into making a GBB work well, then it's difficult to figure out what are critical dimensions for your survey. In this case, it appears that many crap copyshops throw the 3D digitizer wrench and bang out a hastily scanned solid model of one part and hope that their duplicated parts just work because fate is friendly and they copied every feature so it should all work out fine right?

This is why some upgrade builds just seem to work just fine while other identical (brand wise and model number) builds get infested with fitment poltergeists that can't be exorcised with any amount of siliconeoil/files/dremels/pliers/wrench/hammer/screaming/GBBClassifiedsSection.

In the build which worked like peach schnapps and sex, you got lucky (damn lucky) and happened to have parts that fell from heaven with the right kind of manufacturing variabilities to cooperate happily. With that poltergeist gun that looks great, you just didn't have the right physical chemistry.

What happened was that you had a replicate part which was a copy of a working original part designed around an expected average. This part probably wasn't exactly average so some important feature may have been 0.002" longer than expected, but still okay. Then that critical dimension got measured with a tool with a 0.001" imprecision (measuring tools aren't perfect) which brings up the possible error to 0.003". Sometimes the measurement tool is wielded by a complete tool who doesn't believe in NIST tracable standards. In this case, a company can have measurement tools wielded by Tools that will all report different dimensions for the same part feature. I'm going to ignore this possibility because it's just snide. Then a mould was machined to make parts which shrink on cooling so the mould maker had to make some educated guesses on how to compensate for that and threw in an additional 0.002" uncertainty (0.005" uncertainty now). Finally parts got cast which each vary by 0.002" (if similar manufacturing techniques as the original were used) which brings the total error to a big 0.007".

0.007" is kind of like two sheets of regularish 8.5" x 11" bond paper thickness which doesn't look like much, but if you're going to mate three different parts together in an assembly then you get an uncertainty of 0.021" which is like a stack of 7 sheets which is often the time to whip out a single cut file or a dremel tool and hope that there's too much material instead of too little because neither of those tools add material.

In a 3 part assembly, the original manufacturer has a total of 0.006" uncertainty to work with, distributed around the intended design average while the copy shop has a whopping 0.021" of funk in in the trunk statistically distributed around the single (non average) copy they decided to photocopy.

So what now?

I'm not done yet, but I'll revisit this post as I gather my thoughts on writing new material. I'll probably get into how good GBB manufacturers might be achieving even better fitment between parts, then there's the advantage of actually knowing the design intent when manufacturing something. Design intent is that wierd ethereal essence that often seems replicable with direct measurement but sometimes is an uncapturable wisp that silently farts through your non NIST traceable calipers. Maybe when this damn grenade launch is out of the way and I have more time I'll get some help with some diagrams and pictures too.


Improving Fitment on an OEM level:

Engineering design usually starts with a somewhat messy stage of proof of concept building where designers start to hack together early concept models (POC) to prove that they may have a concept which is viable. Quite often the project definition gets altered as these POCs teach a designer more about their project. Sometimes it's difficult to even tell what features are even necessary at the outset. POCs give designers a practical understanding of how users may interact with their product that cannot be discovered with sketches or even solid models. Sometimes a POC serves to altogether shoot down a product idea "What was I thinking?! This is a TERRIBLE idea!".

In the case of fairly well developed ideas (say TM wants to make another 0.45cal GBB), most of the design principles are pretty much already worked out. For instance, the TM 1911 uses pretty much the same operating principles as their 5.1 Hicapa. Developing variants on a proven theme allows designers to take a more significant shot at their early POCs. They're probably designed from the beginning as a solid CAD model because they already plan to use many of the functional bits and pieces from a previous model.

After a first bunch of prototypes are made and proven, an assembly design is broken apart into individual parts which are encoded into a data file which goes to an injection mould maker who starts cutting a mould to shoot plastic into.

There are a lot of subtlties that go into an injection mould design which compensate for some of the strange behaviors of plastics when they're melted and squirted into a mould and left to cool and solidify. For instance, plastics display a fair bit of shrinkage when they cool. This shrinkage is primarily due to crystalization and not thermal contraction. Most materials expand when they're warmer, but plastics expand a lot when they're melted. This expansion is due to their molecules getting all loose and liquid. They're much less compact when they're not intertwined around each other so you get a lot of resultant shrinkage when a slug of plastic goo freezes.

To compensate for this change in density, an injection moulder has to take an educated guess at the size of a mould cavity to produce a part which has features of the correct size as planned in the original CAD design. A mould maker also has to choose where to shoot the plastic into the mould to prevent if from freezing too soon (before the part has fully filled). The entry point is called the gate, and it's placement is quite important as it also affects how a part might warp as it cools. Because the plastic starts to cool as soon as it touches the mould you get a range of varying temperature gradients as plastic flows into a mould which causes different areas to start to freeze and shrink at different times.

What I'm trying to get at is that making an injection mould is more difficult than making a bunch of complicated holes in a block of steel and shooting plastic in like a jello mould. A significant amount of planning goes into making a mould that pops out parts that end up shrinking and warping into the desired shape. This requirement to INTENTIONALLY make a cavity which isn't the planned shape of the final product in order because you PREDICT that the misshapen parts will shrink to the desired shape throws in a bunch of difficulty in injection moulding a heap of parts that end up having to relate to one another in a complicated working assembly.

Sadly your first shot at making an injection mould often isn't bang on perfect. Luckily there's usually some adjustments that can be made to a mould to make things right. Moulds can be reworked by eroding a bit of metal out here and there (to make things bigger). You can even microweld steel back in and machine it down to shape to make things smaller. The nice thing about steel is that you can add to it. When it comes to a hunk of mould that might have a $100k invested in it, expensive techniques become worthwhile to make it work just right.

After finding errors in injection moulded parts, a mould can be reworked to slightly adjust parts so they fit better. Mould rework is unfortuately not cheap. Much of the time spent in rework is usually blown in gauging and setup. When a mould is made most of the features are made in one go. Not much time is spent on setup. In most cases a robotic CNC mill works away all night without supervision as it does all the cutting (automatic tool changers rock!). With rework, the toolmaker has to painstakingly setup the mould block before starting the rework as well as plan cut paths. A lot of the economy of setup and toolpath work is lost when you do little bits of rework piecemeal.

A willingness to inspect parts and assemblies and plan expensive mould rework goes a long way to producing final product that fits just right. It's an expensive bitch, but if you're launching a sucessful product, you're more likely to amortize the costs of your design and mould over all the rocking products you sell.

Fitment After OEM:

If you're in it short term (say a copy shop) you might not appreciate how parts are supposed to fit because you don't really understand the intention of a design. It can be difficult to properly assign effort to functionally bad fits in an assembly if you don't know what to look for. If you're making upgrade parts, you have to keep track of the base build. If you make parts for a WA SVI, you have to periodically buy bunches of their product and plan measurements to make sure that WA hasn't changed something without giving you a heads up. Making upgrade parts that are dependant on another companies product can be a bitch because as their moulds start to wear, their part dimensions may change. The OEM may be monitoring this wear to make sure that it doesn't affect the fitment of their own parts, but this wear could affect the fitment of aftermarket parts before it's a problem with the OEM components.

What's Next?: Material selections and their impact on function. How making something feel more awesome might make something work less awesomely.
Want nearly free GBB gas?

Last edited by MadMax; May 20th, 2008 at 02:11..
MadMax is offline   Reply With Quote