Computer Aided Design …and a few bits of wisdom
Designing efficiently with CAD is more than just having engineering knowledge and learning the functionality of a CAD program. It requires a good understanding of the various machine tools used in today's manufacturing environment and how to interact with the shops that use those machines. It also requires a certain level of discipline regarding using the many features in a modern CAD package.
This document will go over some, but by no means all, of the basic design philosophies and good practices to improve skill and efficiencies when designing with CAD, and designing for CNC manufacturing. This information will only cover non-exotic manufacturing methods such as the lathe, the milling machine, and sheet metal. It is also for small to medium production and not multi-hundred thousand quantities.
Don’t be afraid to ask.
Kelly Johnson, one of the greatest airplane designers of all time and the creator of thousands of jobs, had an ironclad Rule of 10: his engineers (that is to say, his "creative" people) could not be located more than 10 steps away from the manufacturing floor. Reason being that only the constant, minute-to-minute interaction of engineering and manufacturing could understand and solve problems rapidly and creatively.
Never hesitate to ask a manufacturing or design question to a sharp, seasoned machinist that has a good understanding of current manufacturing processes. If you have the slightest doubt about a design, material, process, feature, you should e-mail, call, walk, or drive to the shop that will most likely get the work and ask. You will almost always get a real good answer, which will then let you proceed with your design in a logical direction. But do not, make it a habit of calling the good shop for information and sending the work somewhere else. Also, let the shop know you’re open to suggestions. You don’t know it all. If you did, you would be on you own private island right now looking at your Grumman Mallard anchored next to your Hans Christian.
Good communication is crucial as designs become more complex. Design reviews can be scheduled, or can be on an as needed basis. Letting the engineers schedule their own design reviews should be more productive as sometimes weeks may go by without the need for one, or, during a good week, several may be required. Attendance should be creative in that not only engineers and a manager should be present, but also consider machinists, assembly technicians, purchasing department, and sometimes someone from the shipping department. Try to keep upper management out, as they will only slow things down with questions. Design reviews are not to challenge designs but mainly to let everyone who is on the project know what is going on and if there are conflicts with space, control systems, power, weight, etc. Catching these conflicts early will save much grief later. Which brings up another point. In a design review, don’t just present what you have done; discuss what you are planning to do. Like using turn signals while driving.
Machine Shop Etiquette
There is a lot of variation in machine shops. You will find small mom and pop operations, MBA graduates who liked the business model, and big professional companies with a long history. None of the aforementioned criteria will indicate a good or bad shop. If you don’t already have several shops on your list, picking the correct shop for your part is no easy task. And I do mean part. Most shops have some area of manufacturing that is their specialty. Finding the best shop for your part can be crucial to a good part and your budget. What may be good for the mom and pop may not be good for the MBA dude. Ask questions, look at their equipment list, and if possible, drop by for a meeting. If they have six new VMC’s and only one old CNC lathe, they might not be the best for that complex turned part. One approach is to get recommendations from a local anodizer for example, other non-competitive companies, and of course, the web.
Treat your shop with respect as they are not just a bunch of blue-collar workers "cutting metal"* for your convenience. Most shop employees are professional specialist in their field with a bit of passion to boot. You will find that some machinists can't do any thing else because they "can't" do anything else... they must make things. But, as in all businesses, a few might be there just for the summer pushing buttons.
Always try to work with the shop. Occasionally, you get some parts back that don’t meet print, or with some other problem. You can of course refuse them, but the wise thing to do is to see if there is a work around. Even if you can’t find a way to accept them, the shop will appreciate your attempt. Let the shop know that work-arounds apply to mistakes during manufacturing. Imagine going to pick up your 500 parts 2 weeks late and you notice another pile of 500 on the floor by the scrap bin which, on the last machining operation, they missed a dimension on a surface which you could care less about because it only touched air. This stuff happens. It also never hurts to put you cell phone number on the print and tell the shop not to hesitate to use it.
And if everyone looks up tight, and you really need to get everyone’s attention in the machine shop, grab one of your pretty titanium parts and say, “ Hey, why did you make all these out of titanium?”
* It's not a good idea to use the term "cutting metal" when talking to a shop, as in... "We just want you to hurry up and start cutting metal." Bad... really bad.
Metric or English, that is the question.
There is no simple answer for whether to design in Metric or English as the answer depends on many factors and may still not be correct. It is a real travesty that the USA did not adopt the Metric system when the rest of the world did. And as time progresses, it may prove increasingly difficult to switch. This is not because of our learned intuition of distances, volumes, mass, etc, but the basic infrastructure in this country. Stock availability, hardware sizes, existing component, and many other factors present substantial roadblocks in “switching to Metric”.
So that brings up the real issue at hand on designing with CAD. One might say, “just design in metric and let CAD convert to English for manufacturing”. Yea, right. So what have others done-they use both English and Metric. I’ve heard that there are dimensions driving around on Mars that are in increments of 25.4mm. Gee, I wonder why.
If your parts will be made in a Metric country, Metric is the obvious choice. But if your parts will be made in one of the three remaining Countries where English is standard, Liberia, Burma, and the USA, one must consider using all English, both, and in only certain circumstances, all Metric. This decision is of course, design dependent. Let’s look at an example of a design for an optical device, which will be mounted, to an optical bench. The optics are Metric, the housing and all of the other parts will be fully machined from solid stock so Metric is OK, and the small hardware is available in Metric. But wait! What about the customer’s optical bench? Well after a several e-mails and hours later, you discover that their optical bench was made in California and it has 1.000” spaced .250-20 holes. Get the picture? This is why there is no easy answer to the Metric English question. So, making blanket rules such as “we are now totally Metric” is just not going to make you any more friends.
Top Down? Bottom Up?
Designing a fully parametric, top down CAD model is a difficult task. It requires a good preliminary knowledge of what you intend on doing and the desired relationships. An example would be a company that manufactures custom laser cut titanium dog crates. The model would be fully parametric so when dog gets measured, the dimensions drive the variables, which then create a custom assembly model of the crate. All parts are then sent to the laser. Titanium cuts best with helium.
The opposite approach is when you neighbor shows up at 9:00 pm dangling some cheap beer in one hand and some broken parts for his kid’s go-cart in the other. In that case, hopefully one-time occurrence, you knock out some re-designed parts, quickly stitch together an assembly if needed, generate some un-toleranced prints, burn them on a CD, and send him on his way because your lathe is broke, right? This is definitely not a situation for Top Down modeling.
The typical designer using CAD, after fully understanding the problem and the need for the design, will first think of the design in his head. They might then sketch up a few things on paper or just start with some part that should not change much, such as a base plate. Other parts that will obviously not change much in size, can just be part modeled and inserted into the assembly. Then as you design, you find parts or assemblies that will likely change as the design progresses. These will be moved, tweaked, stretched, featured out, and edited in all sorts of ways. These are the ones that should have some measure of Top Down parametric design. Sometimes it is a good idea to constrain some parts or assemblies to a plane or axis. That way when a line of a sketch is modified, there won’t be a constraint to crash.
Another thing to consider is that a complicated model which has many features tied to many other features, while nice during the initial design phase, will be a total pain later when the design has been frozen for years and parts have to be modified for some reason. Even if the same engineer is tasked with this, there is no way to remember how everything is tied together and what to change. Sometimes, after the initial design is frozen, some or most dependencies can be removed if they will never be needed again… your call.
It is good practice to group parts into sub assemblies and insert these into the main assembly model. We all know that having 472 parts all in one main assembly is bad practice. But, more times than not, you will wish you had put some group of parts in an assembly before you inserted them into the model.
Always remember, no matter how hard you try, or how many classes you take, you will never get it all right the first time. You may not even get it all right even once. Just try to get most of it right, most of the time.
General Considerations for Good Design
“Any intelligent fool can make things bigger and more complex… It takes a touch of genius – and a lot of courage to move in the opposite direction.” Albert Einstein
A single complicated part that is chocked full of features could be more expensive that two parts that share these features and are assembled. A single complicated part that is chocked full of features could be less expensive than two parts that share these features and are assembled. It depends on the design, the material, and the manufacturing method. Keep an open mind and ask the shop.
When designing parts, try to keep it’s manufacturing plan in mind. Most parts will not be designed for efficient manufacturing unless the design engineer has a considerable knowledge of the way that particular part could be manufactured. Always think about the material availability, the machines that will make it, the cutters that will cut it, the people who will have to make it, and the people who will have to inspect it. As a part gets increasingly complicated and feature count goes up, the shop’s risk increases for each additional setup for additional features. This increased risk will increase the cost and has to be sent somewhere and it’s not going to be the machine shop. An example would be a part that starts off with two operations in the lathe, and several complicated setups in a VMC, some square broached feature you just had to have, anodizing, and then engraving. One slip in the broaching, anodizing, or engraving could scrap the whole box. Try to avoid these complicated feature rich parts when possible. If your not sure, let the shop look at it. It might be faster, better, cheaper to assemble two less complicated parts together, or change the design to incorporate operations.
Try to let a design drive itself. Don’t spend too much time forcing a design in a direction it does not want to go.
Parts designed for the Lathe, Milling machine, and Laser are usually cheaper than parts designed for Grinding, Sink EDM, or Wire EDM. Try to design out complicated processes.
Once you’ve wasted a few hundred hours looking around for stuff lost in the feature tree, you will wish you had taken the time to label parts, features, planes, etc. when they were created. Also, try to label your parts and features with names that make sense like Top Plate, Main Revolve, Cap Bolt Pattern, and Outer Housing. Part names like 351-287-B-7, and features auto-named by the program will be more trouble than a sick horse.
Always model to nominal dimensions and use bi-lateral tolerancing. If a hole is going to accept a .9995” OD precision ground shaft, don’t model the hole to 1.000” -.000” + .005”. Model it to 1.0025” ± .0025”. This is important if the model is imported into CAM and also when the dimension shows up on the drawing. In all but a few special circumstances, the machinist will target the middle of the tolerance. Make life easy for them and model to nominal dimensions. This will save them time and you money. It is also is important if you send your model out for rapid prototyping. If you don’t agree with bi-lateral tolerancing, just go ask a shop.
When modeling a part in CAD, when appropriate, don’t just dimension the sketches. Enter tolerances and surface finishes at that time. If you wait until the drawing, there is no chance you will remember your initial design intent months later. Another option is to start the drawing at the same time as the part. But this takes time and it may be only necessary for critical parts. If you have a model of a CAD part has be hacked, patched, twisted, booleaned, and it looks like it’s been put through a defective transporter, just give up and re-model it. Another indication that your part is doomed is if the model tree starts to look like it’s from a rain forest. It’s just not worth it to keep dealing with a troubled part.
Request that the programmer and machinist add notes to the drawing as to improved dimensioning, missed dimensions, proper G D & T, additional views needed, and any other comments they may have. Request a copy of the machinist’s working copy returned to you so you can see his notes. This will help you understand what you could have done better on the drawing.
One part per print is easier to track of that a group of parts on the same print. But there may be exceptions.
Try to record all information, either in the model notes area or on a separate spreadsheet, about specific details of each part. Some CAD programs also allow notes on discrete features. Material, optional design ideas, which shop should manufacture it, calls to the shop about making it, etc. It may take months to release the first part and there is no way your going to remember all of the details and intent of its design.
When selecting sizes of parts, consider the standard material sizes. If you make the part 1.000” ± .005”, then the shop will have to start with 1.125” or 1.250” diameter material. But if you designed it with a diameter of .990” ± .005”, then 1.000” material can be used. Realize that sometimes, it helps to be in the correct mood to design efficiently. Musicians sing about it. Writers discuss it. It exists, and you will find sometimes it just flows, sometimes it does not.
And… this one will get you every time…. perfunctory \pur-FUNGK-tuh-ree\, adjective:_1. Done merely to carry out a duty; performed mechanically or routinely._2. Lacking interest, care, or enthusiasm; indifferent.
Special Considerations for Milled Parts
Here are some things to consider, if possible, when designing parts that are obviously going to be milled.
When designing a square shaped part, study it carefully to see which way it will be better to start the extrude. But do understand you will never get it right every time and there may not be a right or wrong way.
If your part has internal milled corners, use the largest radii possible and this radii should be around 5% larger than the nearest standard end mill. This will allow sweeping the corner and will prevent chatter. An exception would be if the part would be prototyped on a manual milling machine in which case the corner radii should match a standard end mill size. The same holds true with slots. Think about the endmill that will cut it. A .250 end mill will not be used to cut a .250” ±.003” slot.
Realize that there is a limit to deep side milling with an end mill. Machinability matters. Study the standard length and diameter end mill sizes to get a feel for this. A 2.000” deep pocket with .230” radius corners in 316 SS is just not going to work well. Call the shop when you are not sure.
Simple square or rectangular parts that can easily be held in a typical milling vise should be dimensioned with the 0,0 datum in the upper left corner. This way it can be referenced against the fixed jaw and stop in a standard vise setup in a VMC. Some of the time this is not possible, but worth keeping in mind.
Sculpted and profiled parts should not be feared, but they will require a special setup. There are several ways to accomplish this. One is when half the part is machined from a block in a standard vise, then flipped over and grabbed in a set of soft jaws profile milled to fit the part. Another interesting method, typically used on plate type parts, is to rough fixture it and machine all the holes and features, then another setup uses bolts in these features to hold the part on a fixture plate so the perimeter can be machined. There are also other ways to hold easy machinable parts such as glue, double stick tape, etc.
Allow enough drill depth at the bottom of blind tapped holes. Chips are can be controlled when tapping in blind holes by using a tap that pushes the chips up and out of the hole through the flutes. So you don’t need a very large volume under the bottom of the tap but it does not hurt. Drilling and tapping through is always better.
If you really need a flat-bottomed hole, an end mill slightly less than half the diameter of the hole making a circular pass will most likely be used. If it has a smaller hole in the center such as a counterbore, it’s easier because the end mill can be larger. Think about the aspect ratio of the end mill that will machine these holes in a circular pass. These holes and counterbores need not be any common end mill size. It is actually better if it is just larger than the closest end mill diameter as mentioned before. Flat-bottomed counterbores are available and most shops have the common sizes used for socket head cap screws.
Reamers make nice accurate holes. That is what they get paid to do. Reamers come in many sizes with small increments. Custom ones are reasonable but can add time and some extra cost to the manufacturing. So a hole feature can be real accurate in diameter for a reasonable cost when reamed.
Small corner radii down in the bottom of pockets should always be selected from standard end mill data but custom grinds on end mills can be done inexpensively. Look it up.
Visualize the way the part is being held for each setup when rounding those corners or adding features. What may just be a single click of the mouse may end up requiring another setup. If it does require another setup, make sure the feature in question is necessary. But additional features or external corner rounds are practically free if they are machined during an existing setup.
Try to design your parts to have all the critical features machined in one setup. Good accuracy between features machined without removing the part is generally no problem. Modern machines are well capable of locating inside of .001” or better, so one only has to be concerned with tool deflection. Tool deflection can be an issue though, and features such as drilled holes can wander around the mark a bit. Be reasonable with these tolerances unless necessary.
Tight tolerances between features done on different setups are another issue. Depending on the configuration, it may be unwise to ask for anything real tight unless of course, it is necessary.
Dimension between features machined on a single setup will always be tighter than these features to an edge. This is because of rough part size variability.
But what about “giving” the shop a few really loose tolerances just to “make it a cheap part”? In most cases, it’s not likely to save much. By default, a CNC machined part that is done in one setup will be accurate to .001” or better and if done in several setups, it will be within .002” or so. So one would be dreaming to toss in a few ± .050” thinking it will reduce the cost. There is another reason to not drastically loosen up non-critical feature or surface tolerances. Fixturing for machining, future modifications, and tooling used in production assembly may need these features or surfaces one day.
Baseline dimensions are nice for patterns and features if possible. But only use one baseline. If a side is not good for an origin, create centerlines or use an existing feature. Then reference that centerline or feature to an edge with a looser tolerance if possible. Again, use the top left if it’s looks like it will go in a vice in a standard VMC.
Special Considerations for Turned Parts
Here are some things to consider, if possible, when designing parts that are obviously going to be turned.
When modeling a turned part, you may find it easer to sketch the anticipated profile and do a revolve, than extrude a round bar, and remove the material in a revolve cut to create the features. It seems to make editing easier in the future. Also, when edges and surfaces of the revolve are used for references or constraints, they might not crash the model up. It is also easier to look at the entire sketch of the part at one time as if it were a drawing cross section. But there are limits and with complicated sketches, it may be easier to separate some sketches such as O-Rings, internal machining, external machining, etc.
Turned parts can be lower cost compared to square, milled parts. Sometimes, it is better to start out with a turned part and square it off later. But where this really makes sense is when your shop has a CNC lathe with live tooling. Think of a live tooling lathe as a standard turning lathe in which the spindle can also be the 4th axis on a VMC, and that 4th axis can flip up through 90 degrees too.
Tight tolerances in one setup are normally no problem. But when turned part is flipped around to do the other side, watch out for concentricity and angularity. As expected, it will be hard to consistently maintain concentricity better than .001” on a flip. Angularity could vary a bit, depending on the part, but it should be very good.
Try to design your parts to have all the critical features machined in one setup. Sometimes a part can be completely machined in one setup and then cutoff. Nice and cheap!
Inside corner radii should be at least something. If it has to be real small, do an undercut. Most lathe operators like to see a minimum corner radius of .016” to .031”
Outside corners can be specified as “break all corners”, or just model in a corner radius. An outside corner radius looks nice and should not add to the price because the tool is already there.
O-Ring grooves, as per Parker Handbook, have a side angle tolerance of 90 degrees –0 +5. Uhg!, Uni-lateral!!! But what they are trying to say is because there has to be a tolerance, make sure the sides of the groove is not undercut at all for proper sealing. Rest assured you groove would be machined closer to the 90 than 95. Probably spot on 90 degrees. Don’t forget about the inside and outside corner radii. Special cutters are available for O-Ring grooves so look them up or just refer to the Parker Handbook on the parts drawing. Usually, just putting “O-Ring groove per Parker tolerance” on a drawing will suffice. The shop knows what to do.
Outside and inside threads should always have an undercut next to the shoulder. This is a place for the tool to rest before pulling out. The undercut will probably be done with an O-Ring grooving tool so round the inside corners appropriately.
Special Considerations for Sheet Metal Parts
The laser is your friend. Nothing rips though sheet metal faster than a million dollar multi-kilowatt laser. It will cut almost anything except copper and brass, which have limits because those alloys are too reflective at the laser’s wavelength. If you’ve never seen one cut, look it up on the web. Accuracy is around ± .005 but usually a bit better. It depends on the machine and even the operator. Hole diameter in 300 series Stainless can typically be as small as the thickness of the material.
The best way to set the K-Factor is by reverse engineering some test samples from you sheet metal shop, bent in the way they will bend your parts. But to get it close, use K-Factors in the .4 to .5 ranges. This is for air bending.
The minimum inside bend radius should be the thickness of the material. But this is a minimum, try to keep it two or three times the thickness.
Keep the bend radii the same thought the part and if possible, the entire project. This will reduce cost by minimizing setup changes when bending.
When you get the prototype parts back from the shop, don’t fuss too much if the bend radii are not perfect. If possible, adjust the model to match the actual parts. This is a lot easier than the shop tweaking it to match your drawing. Just tell the shop to use the same tooling each time for this part. Your parts should be very consistent if it is a laser shop with a CNC press brake.
Short flanges and features close to bends can cause problems. Check with your shop as it depends on what lower dies they have.
Treat Drawings like a contract.
Geometric Dimensioning and Tolerancing is a wonderful tool. No question about it. Interpreting it in a shop is one thing. Using it correctly in CAD in a real tough assignment. It requires a through understanding of the G D & T concepts to use correctly. Using it partially can be done, and that may be the best way if a complete understanding is not had. If you do not fully understand G D & T, but use it in your drawings either fully or partially, do let the shop know your limited knowledge to avoid issues. The machinist and programmer will be glad to hear it and will most likely help you learn it. Ask them to G D & T a copy of your drawing to get an idea of how to do it.
Blanket tolerancing can cause problems and should be used carefully. It is all too easy to toss an “all dimensions .005” unless otherwise noted”, which will make one not give every dimension the scrutiny it deserves. So it might be better to just leave it out of the title block and force yourself to detail every dimension. But if you must use blanket tolerances, just make sure to go over the drawing carefully and look for features that should deviate and mark them correctly.
Also, if the programmer and machinist see ± .0005 on a bunch of surfaces that touch air, you are not going to get much respect.
Get a surface finish comparator and use it.
Bi-lateral NOT Uni-lateral tolerancing should be used on the drawing and model to nominal.
Fractions… What? Still living in caves?
Watch out for the angle default of ± 30 minutes. It will show up on an outside chamfer and make you look silly.
If you color your CAD model per the final look, your black anodized assembly will be difficult to work with. Use colors for visibility when working and possibly change it later if you have to make a presentation.
Try to orient a turned part on the drawing in the position of the first operation to be done. Try to create additional views for the flipped setup in the lathe. Dimensions should, if possible, be baseline and started from the faced off end sticking out of the chuck. The same is true for milled parts. Try to create additional views for each setup for milled parts if you can figure it out. Just think about it in the machine and give it your best try.
Reference dimensions are a good idea. Try to guess where the programmer and machinist will pencil them in on the drawing and put them there to help them. The programmer in his nice quiet office will not likely make a mistake on a calculator, but the machinist in a loud shop with coolant dripping off of his finger could.
Perpendicularity of features in a single setup will always be very good. It’s when it is on its second or third setup that perpendicularity errors will surface. Don’t forget to show all perpendicularity and concentricity on the drawings. If a turned part has a series of steps toleranced super tight, it should be obvious to all that all these steps should be machined in one setup. Use this to your advantage.
Always try to send a solid model to the shop. But always send a detailed drawing also to convey the necessary information about the manufacturing of the part. We are not there yet.
Modern Engineering plastics can be a good alternative to some materials. An example could be using black acetal instead of black anodized aluminum. Plastics can also be used when a flexible part is necessary. But resist using tight tolerances on plastic parts. Hey, it’s plastic!
Hesitate to use the exotic metals such as monel or beryllium. But if these are required to solve the problem, well…. be prepared.
Don’t specify 316 Stainless if 303 will do. Check the machinability before specifying materials.
Steel rusts. Painted steel chips, then rusts. Stainless does not rust. Price it out. But watch the tensile yield of stainless. It is not what you think. But if you need a strong, non-corrosive, fairly easy to machine material, use the truffle oil of materials, Titanium!
Regular anodizing will typically add some material. Hard anodizing will add a tiny bit more. Call your anodizer for exact numbers. Hard anodized aluminum can be post lapped with diamond for a very accurate but costly finish.
Cast aluminum tooling plate is very flat and stable. It machines nice and will anodize correctly if you let the anodizer know what it is. If it is anodized normally, it will come out splotchy. The same holds true for 2000 series aluminum, let the anodizer know what it is.
Dreamy materials to machine are 2000 and 6000 aluminum alloys, free cutting brass, 660 bronze, acetal and most plastics, Stressproof steel, 12L14 (Leadloy) steel, and 303 stainless- (almost).