Table of Contents
A note on mistakes:
Product requirements
The requirements of a device are the list of things the device must be able to do.
A sample (abridged) set of product requirements might be:
Product specifications
A product specification lists the technical requirements needed to achieve the product requirements.
Corresponding product specifications:
List of qualities that a good software requirement should have by Leanna Rierson’s book “Developing Safety-Critical Software: A Practical Guide for Aviation Software and DO-178C Compliance”:
Visiting a dump
As hardware engineers, we build physical things. That means at some point, everything you design and manufacture is going to end up in a dump. That doesn’t mean you should think of engineering as a nihilistic exercise in futility. It does mean that you should think about the materials you design into your product, the reliability of the product, and the repairability of the product. Choose materials and processes that are as efficient and ethical as possible.
As requirements grow, completeness decreases
| LOC | Pages of Requirements | Requirements Completeness |
|---|---|---|
| 1000 | 14 | 97% |
| 10000 | 115 | 95% |
| 100000 | 750 | 80% |
| 1000000 | 6000 | 60% |
Building contingency
Add footprints for parts that you can use if something doesn’t work, add sub-circuits that you can rely on if you need to, and give yourself plenty of opportunities for easy rework in case it’s necessary.
Delays
Many projects and even entire companies have failed because of an unexpected production delay or a delay in getting a component. Source your components as early as possible and always have a backup supplier. Don’t forget to take into account any necessary long delays like certification or testing.
Tendency to underestimate
One of the leading reasons for this is an inaccurate or incomplete understanding of the manufacturing process. If you’ve never gone through the whole process of production for a significant quantity of a new product, it can be easy to underestimate.
Summary of manufacturing
Do!
Do not!
Reel vs Cut tape
If you’re going to be populating parts with a pick and place machine, you’ll need them on a reel… If you need very small numbers of parts or are assembling by hand, you should order cut tape.
How many extra components to buy?
| Component size | Extra % |
|---|---|
| 0201 | 50% |
| 0402 | 25% |
| 0603 | 15% |
| 0805 | 10% |
| 1206 | 10% |
How many extra components to buy by size?
| Component size | Extra % |
|---|---|
| < 0.6mm | 20% |
| 0.6mm - 1mm | 10% |
| 1mm - 1.6mm | 5% |
| 1.6mm - 2mm | 10% |
| > 2mm | 5% |
Know the failure mode of parts
Knowing the failure mode of each part you select is also im-portant. This depends both on the component itself and how it will be used. Will it fail open or closed? Will it get hot? Will it outgas anything? Don’t choose a component with an unacceptable failure mode.
Part size
For most passive components,
0603is a good balance between size and ease of assembly. They’re large enough to be easily hand soldered, but small enough to not take up a lot of real estate on your PCB… Unless space is really at a premium, don’t go smaller than0402.
PCB thickness and connectors
Most PCBs are
62 milsthick by default, but you can find edge mount con- nectors that fit several different board thicknesses. If you want to use this style of connector, it’s best to try to keep your PCB thickness to the standard 62 mils.
Distributors
Parts research website
Rather than deploying each iteration (like in software), you iterate on less reliable prototypes in your lab before deploying a single reliable design that incor- porates all of your prototype iterations at once. The important thing is to not try to optimize for both reliability and speed at the same time.
Agile development
Modules and eval boards
Modules and evaluation boards usually have schematics and layout available and are a great resource for implementing the device on your custom PCB since you have a known working design to start with.
When not to use breadboards
As Hobbs puts it in his book “Building Electro-Optical Systems”, “… at least resist using white protoboards above 50 kHz, 100mA, 50V, or when a randomly sprinkled 0.1 ohms or 100 nH in series or 10 pF and 10 megaohm in shunt will screw you up.”
Using Arduino
Because Arduino is really just a bootloader, you can use the Arduino development environment on lots of different chips, not just the standard Atmel ATMega328p that comes on the Arduino Uno. You can also install the bootloader on more powerful chips with different ar- chitectures, like the STM32 family or the more power efficient MSP430 family. It’s possible to deploy a product that runs Arduino code, but for a more reliable and feature complete codebase, you’ll need to develop in an embedded C environment. You can often reuse parts of the Arduino libraries though, since they are written in C.
Protecting dev boards with GPIO
It’s easy to accidentally damage or destroy development boards or single board computers. Many of them lack protection circuitry, since it adds to the BOM cost. Make sure you don’t ever set a GPIO as a high output and then short it to ground. You should also never short a GPIO pin set high to another GPIO pin that is set low. In both of these cases, you can cause an GPIO overcurrent condition. GPIO pins can only source a certain amount of current before they’re damaged. This maximum current can be found in the datasheet and you need to make sure you don’t exceed it.
RF Prototyping challenges
RF prototyping is significantly more difficult than prototyping circuits at lower frequencies. You can’t use a breadboard, because they have too much stray capacitance and inductance… Another useful tool for RF prototyping is a Software Defined Radio (SDR). SDRs allow you to digitally control a radio to transmit or receive over a large frequency range with a fairly large bandwidth.
Why keep a lab notebook?
It is a good enough record of what you tried and designed that someone else with no prior knowledge could understand, replicate, and contribute to your work using only that notebook. It’s not uncommon for a team to gain and lose people as time goes on, so it’s very likely that your notebook will be used in this way at some point. Another purpose your lab notebook will have is to prevent knowledge from being siloed.
Notebook entry
Each entry should have a title, date, the hypothesis that you’re testing or motivation for what you’re doing, and a brief summary of any relevant background. The meat of the entry should have the materials you used, the conditions under which the test was performed, the equipment used, and setup of the experiment, the results, and an analysis and conclusion of the results. Include raw data, screenshots, plots, and any other rele- vant information. If your data is too big to fit, explain where to find it.
Advantages of keeping a lab notebook
A good notebook can almost act like a mind map, which will help you and others who read your note- book understand your train of thought and your motivations for your designs and experiments. It will help you stay focused, keep things organized, and see the bigger picture. Writing more like a journal will also make it much easier to get back into the same frame of mind after a long break.
block diagrams
Aside from the actual schematic, it’s usually very useful to draw a high level block diagram of your circuit. If you use hierarchical blocks in your schematic, you can combine the two so that your high level schematic acts like a block diagram showing where signals go, and then opening the schematic that is represented by each hierarchical block shows you the details of how it’s implemented.
Capacitor placement
when placing the capacitors for a particular IC near the chip itself in the schematic, smaller value capacitors can be placed next to the IC symbol and larger capacitor values can be placed further away. This visually represents that capacitors should be placed the same way during layout. However, do not make the schematic look like the desired layout at the expense of schematic clarity.
T junctions
It’s better to use “T” junctions instead of “+” junctions so that it’s easier to find where the EDA software mistakenly forgot a junction dot and thus didn’t make an electrical connection.
Schematic sheet information
For neatness and completeness, put the name of the circuit, its version, and the project on each page. Schematic sheets typically have a small box in the lower right corner for this purpose. When you print schematic pages out and have them lying all over your desk, you will be thankful that you can tell which page goes where and if it’s an old version or not.
Schematic symbol pinout types
When drawing the schematic, it’s typical practice to keep inputs on the left side, outputs on the right side, power on the top, and ground on the bottom. As usual, the exception to this is if there’s another way to do it that results in a more legible and clearer schematic.
Clearer TX and RX annotation on UART lines
Consider naming nets something more descriptive, like
TX_DOforTXdata out andRX_DIforRXdata in.
Pull-up or Pull-down for flexibility
Sometimes designers will put both a pull-up and pull-down resistor on a reset line so they can pick if they want the device to be default enabled or disabled, but forget to mark one as DNP.
LED
Always include at least one LED on the board. This will be really helpful during your first “hello world” test to see that the microcontroller got programmed correctly and has control of at least one GPIO.
Maximum rating
Don’t use parts at their maximum ratings. That includes maximum voltage, the upper end of their tolerances, maximum current, speed, etc. This will prolong the life of your components, and frankly you’re just asking for trouble if you use parts barely within their rated range.
Bypass capacitors
So while in theory a 10 μF capacitor in parallel with a 1 μF capacitor would be the same thing as an 11 μF capacitor, it would actually behave completely differently in real life. Most datasheets and application notes will tell you the recommended bypass capacitors to use (how many, what value, and where to put them). Follow those recommendations to get the best performance.
Using ferrite beads
If you use a ferrite bead to help reduce power supply noise, it should be placed in front of the bypass caps (that is, current should flow through the bead before the bypass caps). If the bypass capacitor goes before the ferrite bead, the IC will be unable to draw energy for current spikes from the capacitor because the inductive part of the bead will oppose any changes in current.
Placement
Good layout is driven by good placement. Good routing should naturally fall out of your placement choices. Beginner designers will often place parts in a fairly haphazard way and spend the majority of layout time on routing traces. Instead, you should spend more time thinking about where to place parts, what rotation they should have, which side of the board they should go on, etc.
think about how hard it will be to replace components by hand if something needs to be fixed. How good do your tolerances need to be? Can you fit a soldering iron down in there? How many screws do you have to remove? Do wires need to be desoldered to get at anything? If you assemble your engineering prototypes by hand, you will quickly discover any shortcomings in maintainability and repairability.
Mechanical constraints
In general, be careful about placing components too close together or in such a way that they will interfere with tools that need to be used on the PCB. There are also mechanical placement requirements to think about. Of course, there’s the obvious examples like component height, connector overhang, and mounting and screw holes. There are also more subtle things, like how far the leads on the bottom of through hole parts stick out. One good way to detect and avoid these problems is to use a CAD package with 3D modeling support.
Manufacturer
Use the same PCB manufacturer that you will use for the final production PCB if possible. At least use a manufacturer capable of producing boards in the same class. Most PCBs are, by default, class II per the IPC standards body. Absolutely do not do a production run without fabricating a few prototypes first, even if you’ve proven the design with a different fabricator. Not all PCB manufacturers are created equal.
Silkscreen
To improve silkscreen resolution, use two different layers of different color solder mask instead of standard silkscreen. You’ll need to talk to your PCB fabricator about this, but solder mask is placed at much tighter tolerances and has sharper edges than normal silkscreen. This is the technique that Arduino PCBs use to achieve the fine visual detail that make their boards look great.
Silkscreen text directions
Make sure all designators and text can be read in one of two directions. This will keep you from having to rotate the board constantly when trying to read the silkscreen. Keep silkscreen off of vias and pads so that it doesn’t get cut off. Be sure to label the polarity of any polarized parts like electrolytic capacitors and diodes.
Silkscreen notes
It’s a good idea to put the PCB version and date in silkscreen so you don’t confuse multiple iterations of the board. You can also put a small rectangle of silkscreen in a corner somewhere so you can write notes or label the PCB with a marker (e.g., a serial number).