Our Signal Integrity, Power Integrity, EMC Philosophy.
As a consultant, I must deliver "right the first time" results on wide ranging technology. Every client's design has unique requirements. They often use technology that is new to me. I need to follow a process that results in reliable, fully functional, products that can pass EMI regulatory tests on the first pass. I have done that consistently for the last 15+ years.
As an instructor, I need to teach that process in a form that Engineers and PCB Designers (who are not SI/EMC experts) can use to make first time correct designs on a repeatable basis. This is not a difficult process for the vast majority of the designs I encounter. It is a process that will work well for anyone who chooses to be careful and stay within the limits of predictable structures. If you are guessing about how a structure behaves, you have two choices. Get a simulator or build a test board. Design guides are a mixed bag. Some are great and some are not. If you follow a bad one you will go through the pain, misery, and brutally expensive cost of a board spin development process I call "Whack-A-Mole."
An average board spin is $6,000 out of pocket and 6 weeks of project delay. The loaded labor cost of an engineer is roughly $20,000 per month. If one hardware engineer marks time for six weeks waiting for the next board spin, that is $30,000. If a software engineer is hindered by 50% because of inadequate hardware, that is another $15,000. This totals to about $51,000 per board spin using commonly accepted cost accounting principles. In other words, spinning a board is brutally expensive. Planning a project with multiple board spins is a "going out of business plan."
My mission is to explain complex interrelated mechanisms in easily understood, but functionally correct, terms. In other words, I want to explain things in a way that ordinary people can understand. This business is not inherently difficult as long as you are careful to pay attention to the details. It is, however, mind numbingly complex to those who do not understand what is important and what is not. Modern tools can make the nasty problems clear, but you still need to understand the cardinal principles or you can mis-use the tools and fail to get a solid design.
...A tale of two projects.
Most of my consulting is done on a fixed price basis. The fixed price covers the complete project regardless of the effort necessary to complete that project. A key part of the consulting is teaching the team the basics of high speed design...ie. I teach them the Jump Start class first. That is available either on YouTube or under the On-Line Tutorials tab. After they learn the basics of the process, we work through the design analysis as a unified team. Net result is that we get a solid working system on the first try. Everyone wins.
In November of 2011, I started a project which I fully expected to be completed by early February of 2012. I had been doing right the first time design for such a long time it never crossed my mind that this project would be any different. The first red flag was when they informed me that the team did not have time to take the training due to the tight project schedule. The second red flag was when they told me they planning on three prototype spins before the product was scheduled to go into production the following fall. The third red flag was when they refused to make critical changes in the power and grounding strategy because it would entail changing a connector and it was contrary to some pet theories about grounding. To make a long story short, we did the three spins and finally fixed the EMI problems with a set of dead bug fixes in July of 2012. The dead bug fixes addressed issues that were raised in December of 2011. It was a very painful process for everyone involved.
In August of 2012 the same client wanted me to consult on a second project similar to the first. I met with the design team. Everyone agreed on a pre-layout design strategy in late September. I participated in the post layout review in early December. After the December review, I did not hear from the team until March of 2013. I sent them an email stating that the contract was timing out and asked if there were any problems or questions. In reply, I received an email with a set of beautiful compliance scans that looked like they came out of a text book.
"Right the First Time" design means you will spend a little more time planning, simulating, and validating. That effort is quickly rewarded when the hardware proves to be solid, reliable, and quiet. The software developers can work at peak efficiency. The move to manufacturing will be much smoother. Money will flow because you have a finished product that can be shipped. Everyone wins.
...Symptoms of Signal Integrity Power Integrity and EMC failure.
Symptoms of SI, PI & EMC design failures include flaky boards, hardware software wars, poor manufacturing yield, high warranty return rate, FCC & CISPR regulatory failure, extra pre-production board design turns, established products suddenly becoming hard to manufacture, excessive EMI shielding, and generally bizzare strangeness. Designing with poor SI EMC hygiene is like leaving food on the kitchen counter in South Florida. A seemingly infinite number of roaches will converge on your kitchen. The purpose of our training is to teach methods of design that will keep the electronic roaches out of your product.
...Sort Signals into Interfaces and Signal Categories within those Interfaces
Each signal category has a critical timing and wave shape requirements. You need to write down those numbers and define what success would look like if you saw it on an oscilloscope or spectrum analyzer.
...Develop design rules that will meet those timing, signal shape, and EMI / EMC requirements
The design process should produce stack-up,routing topology, termination, and power delivery rules that are adequate to ensure success for both signal integrity as well as radiated emissions. In other words, you must intentionally plan for a solid design with adequate noise and timing margins. If you do not plan for success, don't be surprised when the product hass issues. Proper planning, analysis, and verification will result in a fully functional product with a high manufacturing yield, low warranty return rate, and an acceptable FCC / CISPR radiated emissions signature. Remember, if you fail radiated emissions, you can not ship an otherwise properly working product. If you can not ship the product, you do not get paid and life gets ugly.
...Check your work
You need a pre-layout design review prior to laying out the system. You need a post layout design review prior to fabrication. Finally, you need a post fabrication design review if you did not get a fully successful system on the first try.
...The Foundation is Board Stack Up & Power Delivery.
The foundation of proper SI EMC design is the board stack up and power delivery system. Digital designers often think only in terms of signal current since that is the signal path which appears on the schematic. Return current is the other half of the circuit. If you can not point at your design and determine precisely where the return current is flowing, you are in trouble. Nothing else will be predictable. There must be a clear purpose for each layer in the stack-up. You must clearly document the purpose of each layer prior to starting critical signal routing.
... Perspectives on Radiated Emissions
There are three basic ways of controlling radiated emissions. You can eliminate noise at it's source or you can attempt to contain it with shielding and filtering. If you have noise the only other choice is to eliminate efficient antenna structures. Containing noise is generally expensive and it can be quite unpredictable. If you suppress noise at its source, you do not need to chase it. Even if you have an antenna, with no noise there is no radiattion. Suppressing noise at the source is the classy way to go. The same types of things one does to suppress radiated emissions at its source are in most cases, identical to the things one would do to attain good signal integrity performance. The methods used will save money in board turns. It will also make the entire process more predictable.
Generally, it is far easier to get a board to work reliably than it is to get the board to be quiet. The only way to succeed on a regular basis is to solve noise problems at their source and eliminate efficient antenna structures.
Ancient history from the early 90's
A major telecommunications manufacturer documented their experience in trying to solve this problem. When they began measuring their process, they found that 94% of the time they needed to redesign the PCB in order to solve high speed problems. After implementing a new process with appropriate tools, they reduced the redesign to less than 1%. They also maintained that rate over a base of 1300+ PCB designs. At the end of the study, they had so few rework problems that it became standard practice to convene a formal process review any time a board failed. FCC certification became routine. Design cycles and product shipments became far more predictable. Redesign and rework costs were poured into new product development. The benefits to the organization were massive.
I had a very long conversation with the director of this program in the mid 90's. It totally revolutionized my perspective on high speed system design....and I never looked back....TFox Aug 30, 2013