Semi Custom Power Supplies: Build or Buy
Feb 5, 2019
This whitepaper seeks to enable engineers to make the most informed decision when answering the "Build or buy?" question. Read on to learn more.
Get the whole Whitepaper now
1. Introduction
“To build, or to buy?” That is the Question
When determining the system power budget needs and individual power rail requirements for a new design, the evaluation will inevitably converge on whether to design each power solution in-house or look to external vendors to procure turnkey (a.k.a. – off-the-shelf) solutions. Even forgoing the extensive technical proficiency and experience required to design one’s own power solution, the sheer amount of resources, processes, and qualifications/certifications involved can be mind boggling.
Many of the factors touched upon in this whitepaper are pertinent to the overall evaluation process, but we do not deep-dive on all considerations here. Each of these factors could be its own, full whitepaper or even multiday training session accompanied by years of practical experience. This whitepaper seeks to enable engineers to make the most informed decision when answering the “build or buy?” question. The decision is not always straightforward and requires a bigger cost-benefits/total cost of ownership (TCO) analysis that extends well beyond just the bill-of-materials (BOM).
Many of the factors touched upon in this whitepaper are pertinent to the overall evaluation process, but we do not deep-dive on all considerations here. Each of these factors could be its own, full whitepaper or even multiday training session accompanied by years of practical experience. This whitepaper seeks to enable engineers to make the most informed decision when answering the “build or buy?” question. The decision is not always straightforward and requires a bigger cost-benefits/total cost of ownership (TCO) analysis that extends well beyond just the bill-of-materials (BOM).
Many Application Spaces and Markets, All with Their Own Nuances
The typical areas of focus for design priorities when considering the strategic approach to power solutions are on the Size, Weight, and Power (SWaP) factors, sometimes also known as SWaP-C (including. the Cost element) factors. SWaP is particularly important in applications for power systems such as mobility (especially e-mobility), MIL-Aerospace, portable medical and high-end transportation such as railway applications. As we shall expand upon, the optimization of these SWaP factors can often be in conflict so there are no universal rules for maximizing each, which makes finding the right balance for your system and application more of a creative art than a rote process.
In many situations, design priority conflicts must contend with physics outside of an engineer’s control. This can be doubly challenging when trying to meet business and performance goals, where simplicity reigns supreme. As such, the benefits of simplified decision processes and costing models can be somewhat at odds with a more accurate, informed process. For example, take the dollar/watt ($/W) metric, which attempts to benchmark power supply costing by evaluating the basic ratio of unit cost to max (continuous) rated output power.
Design engineers throughout the industry are asked to utilize this metric to evaluate power supply designs and to drive the selection process of solutions. However, this metric can become particularly challenging as the power train (and therefore max power rating) of a solution is commonly not the linear driver of TCO costs.
For applications with specialty needs for spacing/safety, hermetic sealing, chassis materials/coatings, filtering, or even connectors – a single one of these specialty components can cost nearly as much as the rest of the power supply BOM and therefore wildly skew the $/W metric. A simple example of this is shown in the figure below, which compares a handful of 300 W, ac-dc solutions.
Fig. 1: Comparison of Multiple Versions of 300 W ac-dc Power Solutions with $/W Metric [1]
A sampling of some application-specific design nuances and challenges can be found in the table below. We will address how to approach some of these design tradeoffs in the following sections of this whitepaper.
In many situations, design priority conflicts must contend with physics outside of an engineer’s control. This can be doubly challenging when trying to meet business and performance goals, where simplicity reigns supreme. As such, the benefits of simplified decision processes and costing models can be somewhat at odds with a more accurate, informed process. For example, take the dollar/watt ($/W) metric, which attempts to benchmark power supply costing by evaluating the basic ratio of unit cost to max (continuous) rated output power.
Design engineers throughout the industry are asked to utilize this metric to evaluate power supply designs and to drive the selection process of solutions. However, this metric can become particularly challenging as the power train (and therefore max power rating) of a solution is commonly not the linear driver of TCO costs.
For applications with specialty needs for spacing/safety, hermetic sealing, chassis materials/coatings, filtering, or even connectors – a single one of these specialty components can cost nearly as much as the rest of the power supply BOM and therefore wildly skew the $/W metric. A simple example of this is shown in the figure below, which compares a handful of 300 W, ac-dc solutions.
Fig. 1: Comparison of Multiple Versions of 300 W ac-dc Power Solutions with $/W Metric [1]
A sampling of some application-specific design nuances and challenges can be found in the table below. We will address how to approach some of these design tradeoffs in the following sections of this whitepaper.
| Example Application Specialized design / support requirements | |
|---|---|
| MIL-AEROSPACE |
|
| TRANSPORT / RAILWAY |
|
| MEDICAL IMAGING |
|
Table 1: Simple Summary of Specialized Requirements for Some Key Application Focus Areas
Thinking about incorporating wide bandgap (WBG) components?
Given the relentless pressure to constantly improve power supply efficiency and address SWaP metrics, there has been much buzz and advancement in the last 10-12 years over incorporating WBG solutions (typically gallium nitride or silicon carbide, GaN or SiC, respectively) due to enhanced figures of merit for FET/diode characteristics such as blocking voltage, channel on resistance, gate charge, thermal conductance, switching speed, and package size.
Fig. 2: Radar Chart Benchmarking Figures of Merit for WBG Materials vs. Si [2]
Fig. 2: Radar Chart Benchmarking Figures of Merit for WBG Materials vs. Si [2]
It should be very clearly noted that in nearly all comparable applications (we should at least confine to non-RF switching power supply applications for the context of this whitepaper), WBG components are NOT drop-in replacements for their Si contemporaries. Take the GaN high-electron-mobility transistor or HEMT, for instance.
The greatly decreased switching energy and high electron mobility of a GaN HEMT compared to a Si FET can allow for transitions in the ones of nanoseconds range, but this extreme current transition can now cause previously innocuous parasitic loop inductances to now cause a catastrophic voltage overshoot as shown in the simple calculation examples below. Parasitic inductances of only a mere, few nanohenries may be considered negligible with current slew rates in Si-based designs, but be catastrophic to a GaN-based design.
The greatly decreased switching energy and high electron mobility of a GaN HEMT compared to a Si FET can allow for transitions in the ones of nanoseconds range, but this extreme current transition can now cause previously innocuous parasitic loop inductances to now cause a catastrophic voltage overshoot as shown in the simple calculation examples below. Parasitic inductances of only a mere, few nanohenries may be considered negligible with current slew rates in Si-based designs, but be catastrophic to a GaN-based design.
Fig. 3: Representation of Current Flows (red/yellow/green) with Parasitic Inductance in Boost DC/DC Topology [2]
Fig. 4: Calculation of Parasitic-inductance-induced Voltage Overshoot by Common Device Packages and Characteristics [2]
For All the Equations and Physics Involved, EMI Filter Design is an Art
It does not matter how efficient, dense, robust, and cheap your power supply is if you do not meet electromagnetic compatibility (EMC) requirements for mitigating electromagnetic interference (EMI) in the grand majority of applications since this is a legal requirement in most geographical deployment regions.
We lack the space in this whitepaper to go through all the equations (i.e. – Middlebrook Stability Criterion, filter resonance/component calcs, etc.) and design guidelines to consider for filter design but want to make the point that filter design can be surprisingly challenging. Filter components are commonly on the critical path for development because they are critical to passing EMC, yet can also be some of the largest and/or most expensive components in the BOM. Do not underestimate this aspect of power supply design and optimization.
Fortunately, there are some free resources on the web to help with EMI design, for example, the RECOM DC/DC and AC/DC Books of Knowledge [3b]
We lack the space in this whitepaper to go through all the equations (i.e. – Middlebrook Stability Criterion, filter resonance/component calcs, etc.) and design guidelines to consider for filter design but want to make the point that filter design can be surprisingly challenging. Filter components are commonly on the critical path for development because they are critical to passing EMC, yet can also be some of the largest and/or most expensive components in the BOM. Do not underestimate this aspect of power supply design and optimization.
Fortunately, there are some free resources on the web to help with EMI design, for example, the RECOM DC/DC and AC/DC Books of Knowledge [3b]
2. Decision-making process
Should I start with Google/YouTube or should I reach out by email/phone?
Now that we have covered many of the concerns to focus on when defining new power solution requirements, it is time to start doing the research and determine the best path forward. This typically takes one of two approaches, which is to either open your internet browser to start looking for some guidance/advice/solutions or pick up the phone or send an email to reach out to an established resource. There is no right answer here and one can even argue the best success comes with a hybrid approach in which some fundamental research of design techniques, power topologies, and established solutions is performed to supplement reaching out to experts.
The hybrid approach is good because even well-established power vendors/experts are subject to their own biases, particularly in terms of trying to sell you their own product. Mining resources such as webinars and technical papers can yield lots of great information, but must be taken with a grain of salt. It is important to understand most technical papers come from academic resources, which can be very good at solving a pain point in a very controlled setting, but may not port well to industrial and/or high-volume applications or be financially pragmatic.
There are vast amounts of resources from both these categories, ranging from calculator/simulation tools and whitepapers/app notes/webinars to HW/SW emulation tools for evaluating components such as evaluation boards/kits.
Digital power solutions will typically have a GUI or some kind of SW app to facilitate interface/setup of the HW. Nothing compares to building and testing real boards. If investigating specific components, then be sure to pay attention to supply availability along with the technical specs. Major distribution partners (i.e. – Mouser, Arrow, Digi-Key, etc.) are great for this because they have extensive component lists and the availability of stock can be a key indicator of global supply. Remember, a perfect part is irrelevant if you cannot order it (or receive it when you need to).
The hybrid approach is good because even well-established power vendors/experts are subject to their own biases, particularly in terms of trying to sell you their own product. Mining resources such as webinars and technical papers can yield lots of great information, but must be taken with a grain of salt. It is important to understand most technical papers come from academic resources, which can be very good at solving a pain point in a very controlled setting, but may not port well to industrial and/or high-volume applications or be financially pragmatic.
There are vast amounts of resources from both these categories, ranging from calculator/simulation tools and whitepapers/app notes/webinars to HW/SW emulation tools for evaluating components such as evaluation boards/kits.
Digital power solutions will typically have a GUI or some kind of SW app to facilitate interface/setup of the HW. Nothing compares to building and testing real boards. If investigating specific components, then be sure to pay attention to supply availability along with the technical specs. Major distribution partners (i.e. – Mouser, Arrow, Digi-Key, etc.) are great for this because they have extensive component lists and the availability of stock can be a key indicator of global supply. Remember, a perfect part is irrelevant if you cannot order it (or receive it when you need to).
Key Considerations for Internal Power Supply Development Strategy
As with most things in life, success comes with good communication and power supply development is certainly no exception to this. So many of the typical trappings (i.e. – field failures, schedule creep, over budget, poor manufacturing yield, etc.) can actually be root-caused to communication (or lack thereof) with key stakeholders. Even something as simple as ensuring the power stakeholder sits down with the owner of the product/marketing requirements document (PRD/MRD) can have massive effect on the chances of success at mitigating these common pain points. Making sure there is a clear understanding of:
- what the power solution/system needs to do (e.g. – functional specification);
- This is not just about min/max/nominal set points for the supply, but also understanding the functional requirements of the system it powers. First-order needs are meeting the static and dynamic (transient) power needs of the load, but there are many second-order needs to consider such as the overall thermal profile to support (compatibility with system environmental characteristics as well as power supply cooling solutions) and what kind of SW/FW interaction must exist between the supply and system (i.e. – fault handling).
- its operating environment (e.g. – safety/testing/qualification requirements);
- Aside from ambient temperature/humidity, one must consider other factors that can impact both design and performance. Supporting a wide elevation profile translates directly to a wider operating temperature range at sea level. Perhaps the system exists in a highly corrosive environment such as an oil and gas deep-well application or subject to more extreme shock, vibe, or radiation.
- One should also pay special attention to the impacts of temperatures at the extremes. Batteries will have different capacities or perhaps even have a hard time delivering cold-crank power. Capacitor life can ...
Want to read the whole Whitepaper?
Applications
