DFM is something by which we can differentiate an experienced engineer from the less experienced engineer. One thing is to make a design that works. However, one step further is to make a design that works and is easy to manufacture. What does it mean “easy to manufacture”? We will discuss this through several points.
The following points can make the lives of manufacturing engineers and technicians much easier (or make them wish they choose a different profession).
Bill of Materials (BOM) Optimization
If you are a fresh engineer, you probably go on designing your system one subsystem at a time. During this design phase, you don’t think about how many different components you add to the system. First, you set up a microcontroller, required sensors and carefully choose and place other passive components around them.
Once you are done, you proceed with the design of your power distribution system. Here, you have several DC/DC converters as well as some linear regulators. Of course, each of them has their own set of feedback resistors calculated to provide the exact voltage as desired.
This is all cool, meaning it will work. However, the result will be a monstrous BOM list with probably a couple of hundreds of components. In addition, the quantity of many of these components will probably be one.
Now imagine you are a procurement manager and must order all these components. For a single design, you need to source for example a hundred components. Also, if you are manufacturing let’s say fifty modules, you only need fifty of these components. This won’t get you a good price from the suppliers.
After that, we proceed to the manufacturing stage.
Manufacturing engineers will see the BOM and realize their pick ‘n’ place machines do not have that many feeders (feeders are holders for the components that feed the machine with them). So, in some cases it will be needed to go two times through the pick ‘n’ place machine. This greatly increases the manufacturing efforts (and price).
Now let’s look at a different design scenario, done by the more experienced engineer.
Once he designs a new subsystem, he tries to reuse the components that have already been used. Of course, it’s hard to consider all factors while designing a complex system. That’s why BOM optimization is something usually done at the end of the design (as a part of the design checklist).
Once you have your design that you think is nearly finished, generate the BOM. Now go through this BOM and particularly pay attention to the components present in low quantities, like 1 or 2 (we are talking about passive or active discrete components, not ICs), and try to substitute them with some of the components used on other parts of the schematic.
Sometimes, you can sacrifice the regulator voltage accuracy (5V rail does not have to be 5.000V) by replacing the feedback resistor with the value you already use on some other part of the system. Trust me, nothing beats the satisfied feeling when you see you’ve reduced the number of different BOM components by 15%.
The result of this optimization will be a more compact BOM – a BOM with approximately the same number of components (overall), but a much lower number of different components.
In other words, the procurement manager will have a lower number of different components to order (which reduces the sourcing time), but the quantity of each component will be higher (which will help with the price squeeze from the suppliers).
Furthermore, the manufacturing engineer won’t have a hundred, but 70 different components to place, and this will make it possible for him to place all required components at once to a pick ‘n’ place machine. All in all, everybody will be more satisfied. 🙂
Optimal Components Package Choice
Many of the components (we are talking primarily about integrated circuits) are available in a variety of packages. For example, you can get the same MCU in QFP, QFN or chip-scale package (CSP).
Many of the young engineers will be tempted to use the latter one since it is the smallest and it is automatically assumed it will make the design easier. However, think twice – this will:
- Make the PCB fabrication technology much more complex – it might be necessary to have a lower copper clearance, adding more layers to the PCB (to be able to fan out all signals), blind, buried vias and so on.
- Make the assembly much more difficult – it will be easier for the pins to get short-circuited during the soldering process.
- Make the debugging and any rework during prototype testing phase difficult or impossible.
Rather than automatically going with the smallest package, consider the complete design. Of course, if you are space constrained and can’t afford to place QFP package, use the smaller one.
However, if all other components are significantly larger and none of those has significant technological requirements, keep it simple. Go with the largest package that you can fit in the design. The assembly yield will be higher, and it will be much easier if you want to modify something on your prototypes.
Using Generic Components Where Possible
Let’s say you’ve found a specific IC that is tailored to your design needs, and it suits you perfectly! For example, let’s consider an LED driver. You require 16 LED channels and you’ve found a 16-channel LED driver IC with integrated current sinks, not requiring an external resistor. It is not too expensive and is available only from one manufacturer.
You think that’s OK because the market situation looks good now, and you are going forward with this solution. But what happens after six months when the manufacturer for some reason ramps down the manufacturing of this chip?
You are forced to redesign your module.
Additionally, with the hardware redesign, most often comes the software modifications also. This introduces a non-negligible development cost that was not foreseen! Instead of using such specific ICs, you could get away with a jellybean component like a shift register and external resistor array. These components are available all over the market and their price is significantly lower, even though the solution is a bit more complex (in a way it uses more discrete components).
What’s best – since these are standard components, it is practically impossible for them to disappear from the suppliers’ stock.
Additionally, we aim to use components in a standard package, as some manufacturers offer components in a specific package that is not offered by other manufacturers – and this breaks the whole idea of interchangeability.
Multiple Footprint Solutions
Any design engineer with a couple of years’ experience has received that mail at least once.
The manufacturing house wants the recommendation for alternative components because specific diode, IC or similar is not available on the market now. And that’s alright if a component in the same package exists. But if a specific IC (voltage regulator or similar) is not available anymore, chances are low that you will find a drop-in alternative.
This experience has taught us at Byte Lab to listen to the market trends.
We consider the desired life cycle duration of the product. If we anticipate that during this life cycle, certain components could become unavailable, we anticipate multiple footprints in the design for the same solution (e.g. communication interface transceivers, voltage regulators…).
This enables us to simply mount a different solution if the initial one is not available to source, and the final product remains the same.
PCB Panelization
The initial design is usually released as a single PCB (and manufactured in that way as well). However, once we reach the production phase of the product, the electronic modules assembly is always done on a panel of PCBs. This allows the components to be mounted much faster (to several individual PCBs at once).
PCB fabs offer a service of PCB panelization.
However, maximum panel size, technological edge size, mounting holes and other aspects of the panel vary with the specific machines used in the assembly line. Also, depending on the mechanical integration of our electronic module, there might be additional requirements on the PCB panel.
Due to all these variables, we at Byte Lab prefer to design our own panels, accommodated to our in-house machines.
Anticipate Production Testing
All assembled electronic modules need to go through the production testing procedure.
This is clear as day – during the production lifecycle stage. Not all engineers consider this as early as in the prototype design phase.
Either way, production testing is a procedure that requires the test points all over the electronic module. Now, if this is not considered during the prototype design phase, it might be much more difficult to add them afterwards. Not to mention it is possible to alter the design verification and pre-compliance measurements results.
If you are designing an electronic module that is planned to reach the production lifecycle stage eventually, start planning the production testing procedure from the very start.
This means you consider production testing procedure for each subsystem you design, as well as anticipating any required test points. We will discuss the production testing procedure in more detail in one of the following blogs.
Final Word
The first step of becoming a good engineer is to know how to design a product that works.
However, one step further is to know how to design a product that works and does not cost any more than it needs to in the manufacturing.
Use the above points to make your design more manufacturing friendly – in other words, do the DFM (design for manufacturing).
Also, remember that the best optimization can be achieved if you collaborate actively with the manufacturing house (EMS) – their feedback is of the greatest value for improving your design.
