How to Apply Reference Designs in Projects: Start from Proven, Then Make It Yours

Starting a new design from a blank page is rarely a good idea, and a proven reference design is usually the better first step. However, a reference design is a starting point, not a deliverable. If you treat it like a finished product, you are likely to deal with costly redesigns and system failures. This guide is for hardware engineers who need to extract proven intelligence from a reference board and safely adapt it to their own system constraints.

Key Takeaways

• Qualify the reference design carefully before you use it, starting with its age, its source, its test data, and the lifecycle status of its BOM.
• Don’t copy anything until you check the reference board against your actual system limits.
• You need to remove the evaluation clutter, but be careful not to delete parts that are actually keeping the board stable.
• Before you route, run at least a short list of targeted checks on thermals, transients, and every part you swapped.
• Document every deviation from the original design so your team can safely review the changes.

Read our foundational guide, Smarter Circuit Builds: How Reference Designs Accelerate Your Project, to see the true cost of starting from a blank page.

Step 1 – Qualify the Reference Design Before You Mine It

Not all reference designs are worth your time. Before you download the schematics, look at the age of the design and the source. Identify what the board was optimized for. Was it built for the lowest cost, the smallest footprint, or peak efficiency at one specific load point?

You also need to check the global supply chain viability of the board. According to recent supply chain intelligence, over 620,000 electronic components were discontinued in 2025 alone, and 52% of them reached end-of-life without a formal Product Change Notification. If a reference design is three or four years old, some of its BOM may already be EOL or NRND. While you’re at it, check how many distributors stock each key part. One carried by a single distributor is a supply risk before layout even starts, and a part whose inventory history keeps dropping to zero stock is worth replacing now.

Finally, check if the vendor published real test data or a validated layer stackup. Be aware of reference designs that don’t have test data, as they may not have been properly tested. This simple filter saves you from debugging someone else's untested theory.

Step 2 – Extract the Assumptions, Not Just the Schematic

Before you copy anything, extract the assumptions behind the schematic. That is what keeps you from a blind copy-paste. You need a checklist to understand what environment the original designer expected.

Figure out what they assumed about:

• Input voltage ranges
• Load profiles and transient spikes
• Airflow and ambient temperature
• Duty cycles
• EMC targets
• Isolation
• Expected lifespan

What to Keep

Leave the core topology, control-loop compensation, feedback networks, protection, and filtering as-is. The original designers already solved the complex math and sensitive routing for these sections, so there is no reason to reinvent them.

Don't touch the critical layout patterns if you can avoid it. Look closely at how the original designer handled the switch-node loop area, sense lines, grounding, and thermal vias. For example, keeping that switch-node loop tight brings parasitic inductance down, and that controls your EMI ringing. These specific areas carry the most earned knowledge from the original engineers, so lean heavily on their work.

What to Re-Spec

Your constraints will always differ from the evaluation board. Take MLCCs as an example. Due to the DC bias effect, a Class 2 ceramic capacitor can lose up to 80% of its effective capacitance when operated near its maximum rated voltage. If the reference design uses a 10V capacitor on a 3.3V line, and you blindly copy that exact part onto a 5V line in your design, your capacitance drops and your regulator might become unstable. 

You must recalculate passive voltage, temperature, and ripple ratings for your specific ambient environment. You also need to re-spec packages and land patterns for your own assembly process, and change connectors to fit your mechanical interfaces. When a part has to change, look through its alternates and similar parts on Octopart for one in the same package first, since a new footprint means re-placing and re-routing that part of the board.

Rip out anything marked EOL or NRND, and take a second look at any NCNR part, since you can’t cancel or return that order if the design changes. Finding active alternatives for those specific parts keeps your BOM buildable for the life of the product. And if your main system constraints shifted during this process, make sure you swap out the MCU or regulator to match.

What to Strip Out

You definitely need to remove the evaluation board clutter. Rip out the debug headers, test points, onboard programmers, LED indicators, unused peripheral sections, and jumper-selectable rails. Leaving that lab-testing scaffolding on a production board just eats up board space and drives up your costs.

But before doing so, you need to know what breaks if you strip the wrong part. Sometimes a large bulk capacitor looks like it is just there for test-bench stability, but it is actually doing real work to suppress input ringing. If you delete it blindly, your board fails.

The Keep, Re-Spec, and Strip Decision Table

Here is a quick reference for evaluating circuit blocks.

Step 3 – Targeted Validation Before You Route

Simulating the entire board is recommended. At a minimum, run this short list of targeted checks before you route: 

• Run a thermal check for your worst-case ambient temperature.
• Simulate how the board handles load steps and input transients.
• Hand-calculate the safety margins for any part you swapped.
• Verify your footprints and land patterns.
• Run a loop-stability simulation, but only if you messed with the compensation network.

Skipping these checks usually leads to a redesign, which can add 4 to 12 weeks to a project's schedule.

Extra tip: Pay special attention to the thermal checks. Rapid temperature changes and mismatched thermal expansion rates between materials are major causes of microcracks, which account for up to 15% of PCB failures in some production batches.

Step 4 – Document Every Deviation

Documenting your changes is exactly what makes the design reviewable and handoff-safe for the rest of your team. Keep a simple log of every tweak you made to the original reference. Here is an example of how that looks:

What Goes Wrong When You Copy a Reference Design?

Even with a plan, accidents happen; teams can reduce risk by watching out for these common traps:

1. Copying layout without the stackup: Traces designed for a 6-layer board will have completely different impedance on a 4-layer stackup.
2. Inheriting a 3-year-old BOM: As covered in Step 1, blindly accepting an old BOM will likely leave you with unsourceable parts on day one.
3. Trusting lab-bench thermals: The reference board was tested on an open bench at 22°C. Your board goes inside a sealed plastic box at 50°C.
4. Dropping input bulk capacitance: Removing what appears to be "extra" capacitance often causes the regulator to become unstable during transients.
5. Missing why a protection part was oversized: The original designer might have used a massive TVS diode to pass a specific automotive transient test.

Executing Steps 1–4 with Octopart

Octopart already handles most of the part-level work in this process. When you sit down to qualify a design and swap out parts, up-to-date lifecycle status, distributor stock, datasheets, and CAD models are on each part page, so you aren’t checking a dozen sites to find them.

Here is what that looks like in practice. You open the reference design on Octopart and go straight to its BOM section, which already shows the lifecycle status of each part, so you know exactly which EOL or NRND parts need to be swapped (Steps 1 and 2). Each part page links to the datasheet and has CAD models (schematic symbols, footprints, and 3D models), so you can grab what you need for footprint checks and margin math (Step 3). When you pick an active replacement, its stock and pricing across distributors are sitting right there to note in your deviation log (Step 4).

Open your reference design on Octopart and check the lifecycle status of every part in its BOM before you route →

Frequently Asked Questions

How do I know if the compensation network is safe to modify?

If you change the output capacitance or the inductor value significantly, the compensation network is no longer safe. You must recalculate the poles and zeros or run a loop-stability simulation to ensure the regulator will not oscillate.

Should I keep the original thermal vias if my board is smaller?

You should keep the pattern, but you will likely need to adjust the quantity. A smaller board has less copper area to dissipate heat, so you may need to rely more heavily on thermal vias to move heat to inner planes. Always run a thermal calculation based on your actual board size.

What is the fastest way to check the BOM for obsolete parts?

Open the reference design on Octopart and check its BOM section. It shows the lifecycle status of every part, so you can spot NRND or EOL parts before you start routing. To re-spec them, export the design to Altium or download the BOM, then open it in the Octopart BOM Tool to edit.

What should I do if my board based on a reference design doesn't work?

Start with your deviation log. Compare each change against the original reference design, beginning with swapped passives, any changes to the compensation network, and layout differences such as a new stackup, since these cause most failures after adapting a design. If you still can't find the problem, contact the manufacturer's support or applications team and send them your deviation log, so they can see exactly what you changed.