The hard part of double-sided assembly is not the placement. It is that the first side's solder joints have to go through the oven a second time.
Single-sided assembly has one thermal history: print, place, reflow, and the joints are final. Double-sided assembly adds a second one. When the other side runs the same sequence, the whole panel enters the oven again and every joint formed on the first side remelts. For those seconds above liquidus, the first-side components are hanging upside down beneath the board, held up by nothing but the surface tension of molten solder.
That single fact drives nearly every process decision on a double-sided board. Which side goes first is not a layout habit. Which components must move to the second side is not something the line can change. Whether SMT adhesive is needed was decided by the board design before the files were released. By the time the panel reaches production, only execution is left.
This article follows that causal chain: what the second reflow actually changes, how to decide side order, how to split components across sides, how to handle thermally and moisture sensitive parts, where inspection closes the loop, and what to supply at design release and RFQ.
Sunking PCB publishes double-sided and PCBA capability as follows: board thickness 0.2–6.0 mm, smallest placed component 01005, minimum pitch 0.2 mm for BGA and 0.3 mm for QFN and QFP, largest component a BGA or QFP up to 74 mm diagonal, assembly accepted to IPC-A-610 Class 3. Everything below stays inside that envelope.
1. What the second reflow actually changes
Looked at properly, the difference is not "doing it twice". It is three specific things:
1. The first side's joints see a second full thermal history. The intermetallic layer keeps growing through the second melt. That is not immediate failure for a single joint, but it consumes that joint's remaining rework margin. If something has to be reworked later, that joint is no longer in first-heat condition.
2. First-side components are inverted during the melt window. This is the one genuine risk in double-sided assembly. Whether a part drops depends on its mass relative to the total wetted area of its joints. Larger wetted area and lower mass mean more surface tension margin. It is a ratio, not a mass threshold.
3. The board goes through another expansion and cooldown cycle. Laminate and copper expand at different rates, so every oven pass is another chance to warp. On the second pass the board already carries a full side of components, so its stiffness distribution has changed and it does not warp the way a bare panel does.
Of the three, item 2 is the only basis for side order, and items 1 and 3 are why "can we avoid one oven pass" is a question worth asking seriously.
2. Side order: the ratio of mass to joint area
The standard practice is to populate the light side first and push heavy, large components to the second side. The reason is item 2 above: whichever side goes first has to hang inverted through another reflow.
Stated more precisely, what decides whether a part stays put is its mass divided by the total wetted area of all its joints. Lower is safer. So:
- Chip components (0402, 0201, 01005 and similar) are effectively never the problem. They are very light and have ample joint area for their mass, which makes them the typical first-side population.
- Fine-pitch BGAs depend on the package. A BGA with many balls of reasonable diameter has substantial total joint area and may be fine for its mass; a large package with concentrated mass needs its own assessment.
- Large connectors, shield cans, electrolytic capacitors, transformers, heat sinks and anything with a metal body carry high mass for their joint area and are the classic candidates for the second side.
- Through-hole parts are handled after both reflows in principle and never take part in the inverted pass.
The common misjudgement is reading absolute mass only. A heavy part with a large bottom thermal pad, such as a power package, can be safer than a light part standing tall on two small terminations, because the tall part has a high centre of mass and a long moment arm when inverted. Centre-of-mass height belongs in the same assessment.
3. SMT adhesive: when it earns its place
SMT adhesive is dispensed under the component before placement and cured to hold it mechanically to the board. It exists for exactly one reason: to substitute mechanical retention when surface tension does not have enough margin.
Cases where adhesive is worth discussing:
- A relatively heavy part has to stay on the first side for layout or thermal reasons.
- The first side carries a part with a high centre of mass.
- The panel is large and warps noticeably on the second pass, adding out-of-plane displacement at the component.
Most boards do not need it. If the first side holds only normal chip components and packages with ample joint area, adhesive is a redundant operation that introduces new problems: an extra dispense and cure step, components that can no longer be reworked normally, and pad contamination when the dispensed volume is off.
So adhesive is not a standard feature of double-sided boards. It is a remedy triggered by a specific side-split outcome. Try to solve it by moving components first and only then discuss adhesive.
4. Splitting components: what must go on the second side
Beyond the mechanical criterion, several categories must sit on the second side because they cannot take a second thermal history:
Thermally sensitive parts. Some connectors, plastic bodies, LEDs, crystals, certain sensors and battery holders have limited tolerance for heat and for repeat exposure. The datasheet normally states the permitted number of reflow passes, and that is the component maker's specification, not something the assembler can decide on the customer's behalf. Any part whose datasheet permits only one reflow pass must be on the second side.
Moisture sensitive devices. For parts classified under J-STD-020, floor life and bake requirements follow the device's moisture sensitivity level. An extra reflow is an extra thermal shock, which matters more for parts that have already exceeded their floor life. Bake conditions follow the component maker's specification and are not a generic number — this has to be confirmed item by item at BOM review rather than assumed from experience.
Parts that cannot be inverted. Some components have moving elements, directional openings or internal structures that shift when passed through the oven upside down. Putting these on the second side is a structural requirement independent of mass.
Parts with high rework probability. If a component has historically needed rework, placing it on the second side means it has seen only one thermal history when rework comes, leaving more margin.
These four plus the mass criterion are the complete input for the side split. They can conflict — a part that is both heavy and thermally sensitive can only go on the second side, which may already be full of other must-be-last components. That conflict has to be resolved at design release. Production cannot resolve it.
5. When both sides carry large components
Real boards frequently have components on both sides that cannot be moved. The order of remedies is:
- Go back to the design first. Can a heavy part move to the other side? Often the cost is routing effort, which is far cheaper than a process cost.
- Then assess mechanical margin. If the heavy part must stay on the first side, check whether its joint area can hold it. Packages with a large bottom pad have more margin.
- Then consider adhesive, as the remedy when step 2 falls short.
- Then split the process. In extreme cases some first-side parts can be moved to selective or hand soldering after the second reflow, at the cost of consistency, and this is not applicable to fine-pitch packages.
The order matters because step 1 is the cheapest and is only available before the design is frozen. Once the board is released, only steps 2 to 4 remain, and their cost rises in that order. This is why double-sided DFM has to happen during design.
6. The second reflow and board warpage
On the second pass the board is in a different state: one side is fully populated, stiffness is asymmetric, and thermal behaviour changes accordingly. Thin boards show it most — the thinner the board, the less it resists thermal warpage. Sunking PCB's thickness capability spans 0.2–6.0 mm, and boards at the low end of that range need warpage discussed on its own terms for double-sided assembly.
The consequence is degraded local coplanarity: the corners of a fine-pitch BGA can lift off their pads and open or produce a cold joint, and large components see uneven load at their ends. The remedies live on the design side (symmetric component and copper distribution, process edges where needed) and the tooling side (carriers and support through the oven). Board support and flatness control is its own subject and is not expanded here, except to note that it is coupled to the side split: making one side heavy and the other light is itself a contribution to warpage.
7. Verification: what to check after the second reflow
Inspection on a double-sided board is not the single-sided flow run twice. The emphasis is re-checking the first side after the second reflow:
- SPI after printing on each side, to confirm deposited volume before the oven. A joint short on solder also has less margin when inverted on the second pass, so on double-sided boards SPI doubles as the upstream gate for mechanical safety.
- Post-reflow 3D AOI after the second side, checking not only the new joints but also the first side for shift, tombstoning and dropped parts.
- X-Ray for BGA and QFN packages whose joints cannot be seen. On double-sided boards X-Ray has a task single-sided boards do not have: confirming that first-side BGA balls did not deform or bridge during the second reflow. Note that components on both sides superimpose in the image, so interpretation needs appropriate technique.
- ICT and FCT for final electrical confirmation.
On voiding, solder void rate is controlled to under 10% with nitrogen reflow; that is the ledger capability figure and its applicable conditions should be stated at engineering confirmation. Assembly is accepted to IPC-A-610 Class 3 and the bare board to IPC-6012 Class II and Class III. Sunking PCB runs 10+ inspection gates, including 100% electrical test by flying probe or fixture on all orders and 100% FQC inspection.
8. What to supply at design release and RFQ
The information gaps on double-sided assembly are consistently these, and closing them is what lets engineering review pass in one round:
- Assembly drawings and placement lists separated by side. Which parts are top and which are bottom; a merged drawing is not enough.
- A list of heavy and large components, with mass where available, package type, and whether they carry a metal body or heat sink structure. This is the direct input to the side split.
- A list of thermally and moisture sensitive parts, with the datasheet-permitted reflow count and the moisture sensitivity level. This is the most common cause of rework when missing: the part reaches the line, turns out to allow only one pass, and has been assigned to the first side.
- Notes on parts that cannot be inverted (moving elements, directional openings, internal structure).
- Thickness, dimensions, process edges and tooling holes. Thin and large panels need explicit mention because they drive warpage and carrier decisions.
- BOM completeness: the BOM limit is 1000 lines, and alternates and do-not-populate items must be marked.
- Test requirements: whether X-Ray, ICT or FCT are needed, and the acceptance class.
On expediting, Sunking PCB's stated expedited assembly terms are a 50-line BOM in 2 days or less and a 100-line BOM in 3 days or less. These are conditional commitments whose applicability depends on material availability and the specific board, confirmed per order — and double-sided assembly runs one more complete pass than single-sided, which has to be counted in the scheduling assessment.
9. Summary
Every constraint in double-sided assembly traces to one thing: the first side's joints get melted a second time.
What follows from it is settled. Populate the light side first, judged by component mass over total joint area with centre-of-mass height included. Move thermally sensitive, moisture sensitive, non-invertible and high-rework-probability parts to the second side. Treat adhesive as the remedy when the side split cannot solve it, not as standard equipment. When both sides carry large components, go back to the design before adding process steps. Discuss second-pass warpage separately for thin and large panels. Focus inspection on re-checking the first side after the second reflow.
All of these judgements are available at design release and none of them are available on the line. Handing over the side-split list, the thermal and moisture sensitivity list and the heavy component list at RFQ is the one place where double-sided assembly rework can actually be avoided.
FAQ
Q1: Is there a fixed rule for which side to populate first?
There is a criterion but not a fixed answer. The standard practice is to populate the light side first, because whichever side goes first hangs inverted through a second reflow. The accurate criterion is each component's mass divided by the total wetted area of its joints, lower being safer, with centre-of-mass height considered alongside it. A heavy package with a large bottom pad can be safer than a light part standing tall on two small terminations.
Q2: Does the second reflow damage the first side's joints?
First-side joints remelt during the second reflow, which is normal process behaviour rather than failure. Two real risks exist: whether the inverted component is held by surface tension during the melt window, and the intermetallic growth through a second thermal history that consumes rework margin. The first is managed by the side split and, where needed, adhesive. The second is why high-rework-probability parts belong on the second side.
Q3: Does double-sided assembly always need SMT adhesive?
No, and most boards do not. If the first side carries only normal chip components and packages with ample joint area, adhesive is a redundant operation: it adds dispense and cure steps, prevents normal rework, and can contaminate pads if the volume is off. It is justified only when a heavy part or a part with a high centre of mass has to remain on the first side. Solve it with the side split first and discuss adhesive only if that fails.
Q4: Which components must go on the second side?
Four categories. Thermally sensitive parts whose datasheet permits only one reflow pass (some connectors, plastic bodies, LEDs, crystals, certain sensors, battery holders). Parts with a moisture sensitivity level that makes an extra thermal shock unwise. Structural parts with moving elements, directional openings or internal structures that shift when inverted. Parts with high rework probability where preserving margin matters. Large components with high mass relative to joint area also go to the second side preferentially.
Q5: What bake conditions apply to moisture sensitive devices?
Those stated by the component maker for the device's moisture sensitivity level. Generic numbers do not apply. Classification follows J-STD-020, but floor life and bake conditions vary by device and must be confirmed item by item at BOM review. Double-sided assembly adds a thermal history, which makes that confirmation more consequential than on a single-sided board.
Q6: What needs attention for double-sided assembly on thin boards?
Warpage. The thinner the board, the less it resists thermal warpage, and on the second pass one side is already fully populated so stiffness is asymmetric and the board does not behave like a bare panel. The consequence is degraded local coplanarity, where fine-pitch BGA corners can lift off their pads. Remedies sit on the design side (balanced component and copper distribution, process edges where needed) and the tooling side (carriers and oven support). Sunking PCB's thickness capability is 0.2–6.0 mm; boards at the low end should be called out explicitly at engineering review.
Q7: What gets left out most often when quoting double-sided assembly?
The thermal and moisture sensitivity list. A part reaching the line, turning out to allow only one reflow pass, and having been assigned to the first side is the most common cause of rework. Next are the heavy component list (mass, package, metal body or heat sink structure) and notes on parts that cannot be inverted. Side-separated assembly drawings, thickness and dimensions with process edges, BOM completeness (1000-line limit, alternates and DNP marked) and test requirements should come with them.