Planar vs. wire-wound transformers: a quantifiable engineering selection guide

AI-server power and the 800V HVDC architecture have pushed magnetics selection pressure forward — decisions once deferred to the prototype stage must now be made at the architecture stage. Yet the most common mistake in engineering discussion is treating 'planar is more advanced' as a default premise. In reality these two structures were never substitutes: planar buys lower leakage and a shorter thermal path with higher upfront investment and weaker current capability; wire-wound buys a mature supply chain and low NRE with greater height and process-dependent consistency. The real question is not 'which is better' but 'what are your constraints.' This article breaks that into six quantifiable criteria.

1. Ask the right question: not which is better, but what the constraints are

The 'planar or wire-wound' debate rarely concludes because it is framed as a question of technical superiority. Put the parameters side by side and each wins on almost every axis: planar is clearly better in height, leakage controllability and batch consistency, yet not necessarily better in high-frequency AC winding loss, and usually worse in current capability and cost. Wire-wound is the reverse.

So this article draws no 'choose planar' or 'choose wire-wound' conclusion. It gives a judgement order instead: check height first (hard constraint) → then frequency (drives AC loss) → then cost and lead time (drives commercial feasibility). The order cannot be reversed — height is a non-negotiable mechanical boundary, while cost can be covered by system-level gains. Reversing the order means rework late in the project.

One premise first: a planar transformer is not a 'more efficient transformer'. It has lower leakage (cutting switching loss and voltage spikes) but higher winding AC resistance (raising copper loss). Which yields higher net efficiency depends on operating frequency and duty cycle — above roughly 500kHz planar usually wins, while at the 100kHz scale a wire-wound design is often more efficient. Section 3 covers this.

2. Height and volume: the hardest dividing line

If one metric had to distinguish the two schemes quickly, it would be height. A wire-wound transformer uses round copper wire wound layer upon layer, so height grows with turns, plus the vertical space taken by bobbin and air gap. A planar transformer changes the winding to foil, PCB copper layers or flat wire, turning the winding from vertical stacking into horizontal spreading, which compresses height significantly.

Take a 100W flyback transformer as a typical case — the comparison is fairly direct (Table 1):

• Total height: conventional (EF25 bobbin) 22–28mm, planar (EQ25 core) 8–14mm;
• Winding height itself: 12–18mm vs 3–8mm;
• Window utilisation: 60%–70% vs 85%–95%;
• Relative volume: with the conventional scheme as 100%, planar is about 50%–70%.

The engineering meaning is clear: when the product height must be under 15mm, a planar transformer is often not the 'better choice' but the 'only feasible one'. Ultra-slim LED supplies, GaN fast chargers, 1U server supplies, telecom supplies and automotive OBC all favour planar primarily because of height, not efficiency. Fig 1 draws the two cross-sections side by side: on the left, round wire stacked layer by layer with a bobbin; on the right, PCB copper layers spread horizontally on a planar core, with the height scales making the difference obvious.

3. Frequency and AC winding loss: where the planar cost really sits

Height is planar's advantage, but planar's cost concentrates in one place — AC winding loss. Understanding it requires skin effect and proximity effect.

Current is not distributed evenly inside a conductor: as frequency rises, current crowds toward the surface and the effective cross-section falls — the skin effect. Copper skin depth scales inversely with the square root of frequency; the commonly used values are about 0.21mm at 100kHz, 0.093mm at 500kHz, 0.066mm at 1MHz. Above 100kHz, if conductor thickness approaches or exceeds skin depth, the extra copper contributes almost nothing to conduction while still absorbing eddy loss.

Even more influential than skin effect is the proximity effect: the field of neighbouring conductors and adjacent layers drives circulating currents inside a conductor. Dowell's classic analysis reduces the winding to equivalent foil layers and expresses the AC-to-DC resistance ratio (Rac/Rdc) as a function of the conductor-thickness/skin-depth ratio and the number of winding layers m. It yields a counter-intuitive but crucial result: Rac/Rdc rises rapidly with layer count, so a winding with many thin layers can perform far worse than one with fewer, better-proportioned layers.

This is exactly where a planar transformer needs care: to reach a target turns count within limited layers, planar designs easily become multi-layer stacks — and stacks are precisely the worst case for proximity effect. Conversely, primary-secondary interleaving reduces the magnetomotive force accumulated across the winding stack and therefore cuts both proximity loss and leakage — the single most important degree of freedom in planar design. The right side of Fig 2 shows the Dowell trend for Rac/Rdc against thickness ratio and layer count: more layers, steeper rise.

This also shows up in measurement: one study compared three conductors on the same core, taking the frequency at which AC resistance exceeds DC by 5% as a boundary. The FPCB (flexible PCB) winding was around 2MHz, versus about 8kHz for solid round wire and 20kHz for Litz wire. So planar/FPCB structures sit in the 'no AC penalty' region across the whole low-frequency range, while round wire starts paying from a few kHz — though the size of that penalty still depends on whether thickness and layer count are controlled well.

4. Leakage and parasitic capacitance: lower is not always better

Leakage is often treated as 'lower is better,' but that is the easiest misjudgement in selection. More precisely: the desired leakage depends on the topology.

• Hard-switched topologies (flyback, forward, two-switch forward): lower leakage is better. Energy stored in leakage releases at switch-off, creating voltage spikes, stressing snubbers and worsening EMI — so these topologies want low leakage.
• Resonant topologies (LLC, CLLC): leakage (sometimes together with an added resonant inductor) is part of the resonant design — a design variable, not a parasitic defect. What matters is 'controllable and repeatable'; blindly chasing ultra-low leakage shifts the resonant point and changes the gain curve.

Here the two structures differ clearly: planar fixes winding geometry on PCB layers, so leakage can be designed precisely via layer arrangement and interleaving, with good consistency; wire-wound leakage depends more on winding technique and varies more batch to batch. Common industry magnitudes are 3%–8% for conventional and 0.5%–2% for planar; in one measured comparison on the same architecture (RM10 core), planar leakage was about one fifth of conventional and AC impedance about one third, while window utilisation fell from 75% to 30%.

But low leakage is not free: greater layer overlap area raises primary-secondary parasitic capacitance, and that capacitance is a key path for common-mode noise. Industry reference values put primary-secondary capacitance at about 50–150pF for conventional and 10–50pF for planar, though the figure depends heavily on stack-up and shielding. In other words, low leakage usually has to be traded against parasitic capacitance, and EMI performance is ultimately set jointly by the switching node, current-loop area, Y-capacitors and PCB layout — the transformer cannot decide EMI alone.

5. Thermal path: planar cools more easily, but can form hot islands

Planar has a genuine structural advantage in cooling: large-area copper-to-core contact gives a short conduction path; heat reaches PCB copper and chassis directly; no bobbin blocks the path and heatsinks integrate more easily. Industry comparison figures (100W flyback, natural convection) are: hot-spot temperature 95–110℃ vs 70–85℃, and winding-to-ambient thermal resistance 25–35 K/W vs 15–22 K/W.

Two caveats matter here. First, a planar structure is not inherently 'cool'. Temperature rise ultimately depends on core material, copper loss, core loss, gap design, assembly pressure, potting material, air speed, nearby heat sources and mechanical integration — get any one wrong and the structural advantage disappears. Second, planar structures readily form a hot island: with skin and proximity effects more concentrated near the centre, plus CTE mismatch between core and PCB and uneven potting conductivity, measurements show the centre can run several degrees hotter than the edges.

So for high-power projects the criterion should not be 'the structure looks cooler' but measured temperature rise of a sample across load and ambient conditions. A wire-wound hot spot typically sits inside the winding window or between insulation layers, where heat must cross winding, insulation, bobbin, gap and air resistances in series — the hot-spot markers in the two cross-sections of Fig 1 correspond to exactly this difference.

6. Safety spacing: how IEC 62368-1 constrains the two structures differently

The safety difference is not in the standard itself but in 'how easily the required spacing can be achieved.' IEC 62368-1 gives minimum values for clearance (through air) and creepage (along a surface), depending on working voltage, pollution degree, insulation class and material group.

The magnitudes are worth remembering: at pollution degree 1 and altitude up to 2000m, minimum clearance at 800V peak is about 0.10mm; at 1500V peak, basic insulation about 0.5mm and reinforced about 1.5mm; at 3000V peak, basic about 2.0mm and reinforced about 3.8mm. Creepage also rises with working voltage: at pollution degree 1, about 2.4mm at 800V RMS and about 3.2mm at 1000V; but at pollution degree 2 or 3 with a poorer material group, the value at the same voltage multiplies, reaching several millimetres or even the low tens.

One frequently cited example: for 100–240V input, 800V peak working voltage, pollution degree 2, altitude 2000m, overvoltage category II requires about 1.5mm basic and 3.0mm reinforced clearance; overvoltage category III raises these to about 3.0mm and 5.5mm.

What this means for selection: a planar transformer can build creepage and clearance directly into PCB stack-up and slot design, with high geometric precision and consistent batches; a wire-wound transformer must achieve it through insulation material, bobbin structure and winding process — flexible, but more process-dependent. In high-voltage scenarios (800V HVDC, automotive OBC, industrial supplies), creepage is often not 'satisfied in passing' but 'decides the layout'; where winding process cannot guarantee it, a planar structure (or structural measures such as slots and barriers) becomes necessary. Note that no scheme may rely on solder mask alone for creepage.

7. Cost and delivery: ground wire-wound still holds

After five criteria planar looks dominant, but cost and delivery are precisely wire-wound's strongest ground. Again at the 100W scale and 10K order quantity, the cost structures differ (Table 2):

• Core cost: standard core ¥3–5 vs custom planar core ¥5–8 (about +60%);
• Winding material: round wire ¥1–2 vs foil + insulation film ¥2–4 (about +100%);
• Process cost: automatic winding ¥1–2 vs stacked assembly ¥2–5 (about +100%);
• Tooling: standard bobbin ¥0 vs planar core and stacking tooling ¥5,000–15,000 (one-off);
• Overall unit cost: ¥5–9 vs ¥9–17, about 80%–100% higher.

Delivery differs too: prototyping 5–7 days vs 10–15 days (incl. tooling); production lead time 2–3 weeks vs 3–5 weeks; MOQ 500–1000 pcs vs 2000–5000 pcs. More importantly, design-change response: a wire-wound design changes by parameter, while a planar change may touch the core and PCB stack-up, requiring new boards and tooling. And commercial planar designs typically top out around 2–3kW; beyond that wire-wound or Litz approaches are more practical.

So cost should not be judged from unit price alone but from the system account: the height saved by planar can buy a smaller enclosure, fewer fans or a shorter air path, and that system gain often covers the 80%–100% device premium — provided those gains genuinely matter for the product. If the product is highly sensitive to cost and lead time and height is not a constraint, a mature, production-proven wire-wound design remains an excellent choice.

8. Decision path: one judgement table and one hybrid option

Condensing the previous seven sections into an executable judgement order (Fig 3):

• Step 1, height: if the height limit is ≤15mm, lean planar — it is often not better but the only feasible option;
• Step 2, frequency: below 100kHz the two differ little and wire-wound is more economical; from 100–200kHz both can work, so compare loss, height, EMI and total cost; above 200kHz lean planar, and above 500kHz planar's advantage is clear;
• Step 3, cost and lead time: when highly sensitive to both and height is not a constraint, lean wire-wound;
• Throughout, leakage requirement: hard-switched topologies want low leakage, resonant topologies want controllable leakage — this sets layer arrangement and interleaving strategy and should be evaluated early with the magnetics supplier.

Note the overlap zone in particular: roughly 100–500kHz, 100W–1kW, 12–18mm height. There is no generic answer in this band — neither 'planar is always better' nor 'wire-wound is always cheaper' holds, and only measured comparison of samples under real operating conditions settles it. This is exactly where engineering judgement earns its keep.

If you do land in the overlap zone, there is a third option: a hybrid — a planar primary (low leakage, predictable parameters) with a foil or wire secondary to carry high current. It combines the strengths of both in high-density supplies at the cost of a more complex assembly process. For high-frequency designs using GaN or SiC, evaluate leakage, distributed capacitance and AC loss jointly with the supplier early, because these parameters directly affect system efficiency, spike voltage and EMI — late changes are expensive.

Closing

Planar and wire-wound transformers are not competitors but two tools for different constraints: height-limited, higher-frequency, consistency-critical designs lean planar; cost-sensitive, high-current designs needing flexible ratios or fast changes lean wire-wound. The effort is better spent ordering your own constraints as 'height → frequency → cost' and validating on that order — not arguing which technology is more advanced. Where is your current project stuck: height, frequency, or AC loss?