Minimum PCB Size for RF

Minimum PCB Size for RF: The Pre-Certified Module Misconception

This conversation comes up on nearly every compact wireless project I quote. The product needs to be as small as possible, it needs good range, and the plan is to use a pre-certified module because that removes the RF risk and the certification cost. All three are reasonable on a normal sized board. On a small board none of them hold.

The module is not the antenna. The board it sits on is most of the antenna, and the certification applies to the conditions it was tested under, not to the part number.

The Three Misconceptions

"The module is 8 x 8 mm, so it fits a small board." It fits mechanically, but that says nothing about whether it works. The module's antenna radiates by exciting currents on the host ground plane, and that ground plane is part of the radiating structure. If you shrink it you are changing the antenna, not just the packaging.

"It's pre-certified, so I skip the RF testing." Only while you meet the integration conditions the approval was granted under. Those conditions include the ground plane. On a board much smaller than the vendor's reference, you no longer meet them.

"The datasheet says 250 m, so I'll get somewhere near that." Those numbers are measured or estimated under conditions that have very little to do with a small product talking to a phone.

Your Board Is the Antenna

At 2.4 GHz the free space wavelength is about 123 mm, so a quarter wave is roughly 31 mm. Any board whose longest dimension is well under that is an electrically small radiator, and the Chu-Harrington limit starts setting the terms: as the smallest sphere enclosing the structure shrinks, minimum Q rises, bandwidth narrows, radiation resistance falls, and fixed losses in copper, dielectric and matching components take a larger share of the power.

The effect is far larger than most people expect. One chip antenna vendor publishes efficiency for the same antenna component on a range of board sizes, with the matching network re-optimised for each. Their round board figures are the most useful comparison I have seen published:

Going from a 30 mm disc to a 15 mm disc costs nearly 8 dB, which is a factor of about 2.4 in range. That is on bare laminate, with no battery underneath, no enclosure and no hand nearby. Every one of those adds more loss in a real product.

This is also why a board that measured well on the bench can disappoint in the product. The antenna did not change, the ground plane it works against did.

What Module Vendors Actually Publish

If you read the integration manuals rather than the product pages, they all say much the same thing in different words:

• One 6.5 x 6.5 mm BLE SiP: "For a large PCB, larger than one wavelength (61.5 mm), the shape and size of the ground plane is not critical. For smaller PCBs where the length and the width are below one wavelength (61.5 mm), the size and shape of the ground plane is an important factor." The same manual notes radiated power fluctuates increasingly as the ground plane shrinks, and its published total radiated power chart spans 0.40 to 1.60 wavelengths, roughly 49 to 197 mm. Nothing below that.
• A 10 x 15 mm module: "recommended to have the ground plane extending at least 10 mm on both sides of the module", which implies a host board of around 30 x 35 mm.
• A leading 8 x 8 mm SiP: the vendor's own EM simulation study of range versus ground plane size stops at 20 x 20 mm, the smallest board they were willing to characterise. At that size the module shows about −4 dBi peak gain and a return loss of only about 6 dB, which means it is significantly mismatched as well as inefficient.
• A recent 6.3 x 7.9 mm module: "the chip antenna is tuned when the module is sitting on a development board (host PCB) with size of 113 mm x 63.5 mm x 1.6 mm."

That last one is worth noting. The module is a few millimetres across and it is tuned on a board bigger than a credit card.

The headline: almost nobody publishes ground plane data below about 0.4 wavelengths, which is roughly 49 mm at 2.4 GHz. If your board is 20 mm across, you are off the bottom of every published curve and into territory the vendor has not characterised and will not support.

In Fairness: Modules Do Work on Small Boards

One vendor does show a working application at 7.5 x 12 mm including a coin cell, and quotes about 100 m range for it. So "you cannot use a module on a tiny board" is too strong a claim, and if you have looked into it you may well have found that example.

The number underneath tells you more. That board produces a total radiated power of about −8.4 dBm from a module whose conducted output is +4 dBm, which works out to roughly 6 % radiation efficiency. The chip antenna data above gives 13 % on a 15 mm disc. So a properly co-designed custom antenna on a small board is worth something like 3 to 4 dB over a module, which at these sizes matters a lot.

So it is not that you cannot do it. You can, but you need to know what it costs you.

Where Marketing Range Numbers Come From

Published range figures for 2.4 GHz modules routinely run into the hundreds or thousands of metres. They are not fabricated, but the conditions behind them rarely match a product:

• Module to module, not module to phone. Another module has better receive sensitivity and a better antenna than a handset held in a hand. That difference alone is commonly 10 dB or more.
• Long range coding. Figures quoted at 125 kbps assume coded PHY, worth about 8 dB of link budget over 1 Mbps. Worth knowing that not every phone platform exposes it.
• Outdoor line of sight with Fresnel zone clearance, both ends elevated, and typically no fade margin at all.
• A full size reference ground plane. Not your 20 mm board.

They are also often estimates rather than measurements. I have seen a single vendor document quote two range figures whose ratio would require about 8 dB of difference in radiated power, while the two boards differ by under 3 dB. So treat published range figures as a rough indication only and work out your own link budget.

The Certification Argument Collapses First

This is the part that surprises people, and it is usually the reason the module was chosen in the first place.

Modular approval is granted subject to conditions: the antenna it was tested with, the keep-out area, the placement, and the host ground plane. Vendors state this plainly. One integration manual puts it like this: "To take advantage of the certification, the customer is required to implement the specific ground plane design according to the reference design."

On a board far smaller than that reference you cannot meet the condition. The antenna environment has changed, which means the radiated characteristics have changed, which is exactly what the approval was about. Expect at minimum verification testing, and be prepared for a full intentional radiator assessment.

So on a small board you pay the module's unit cost, you give up board area you did not have, you accept a worse antenna, and you still carry the certification work. The one benefit that justified the module is gone. This is the part that usually needs explaining twice, because it is not obvious until you lay it out.

Practical Size Thresholds at 2.4 GHz

Combining the published data with what I see in practice, these are the thresholds I work to. Treat them as starting points for a link budget, not guarantees:

Note that length matters more than area. The ground plane is the other half of the radiator, so what it wants is roughly a quarter wave along the edge the antenna sits on. Long narrow boards do surprisingly well: published data shows a 40 x 5 mm board reaching 77 % efficiency, better than an 80 x 40 mm board with the same antenna. For a round board, the diameter is that dimension, which is why circular products hit the wall sooner than rectangular ones of similar area.

In a Compact Product, Everything Is in the Near Field

The near field extends roughly λ/2π from the antenna, about 20 mm at 2.4 GHz. In a product measuring 20 or 25 mm across, that region contains the entire device: the battery, the enclosure walls, any metal, the PCB itself, and whatever the product is clipped or strapped to.

A coin cell is the one that catches people out most often. It is a metal disc sitting directly under the radiator, and depending on geometry it can act as a parasitic element, a detuning load, or part of the counterpoise. It always has an effect. The same goes for a metal clip, a magnet, a battery holder or a user's body.

None of this can be solved by selecting a better part, because the interaction is geometric. It has to be co-designed: antenna, ground plane, battery position, enclosure and mounting, all tuned together. A pre-certified module takes that freedom away, because its antenna is fixed, its tuning is fixed, and its certification depends on both staying that way.

How to Decide

Below about 25 mm: a custom antenna is realistically the only viable route. Budget for EM simulation and tuning iterations, and accept that certification is yours to do regardless.

25 to 35 mm: a real trade. A module saves RF engineering effort and gives a smoother certification path, at a cost of roughly 1 to 3 dB against a well executed custom design. If the link budget has margin, take the module.

Above 35 to 40 mm: a module is usually the right commercial answer. You are in the region vendors characterise and support.

Whichever route you take, do the arithmetic before the mechanical design is frozen. Work backwards from the range you need, through a realistic receiver and a real fade margin, to the radiated power you require, and from there to the antenna gain and board size that delivers it. Increasing the diameter by a few millimetres early in a project costs nothing. Finding out after tooling that you are 8 dB short is expensive, and there is no firmware change or better chip that recovers it.

One last piece of advice for anyone specifying this work: do not let a supplier commit to a range figure without a defined test method attached. A number like "50 m" means nothing without stating the receiving device, the height of both ends, whether the product is in its enclosure, what it is attached to, and whether you are measuring first detection or reliable continuous detection. A supplier who commits to a bare number either has not thought about it or is planning to argue about it later.

Small wireless products are one of the most common places I see projects lose months, almost always because the board size was fixed before anyone ran a link budget. These principles run through all my RF design services, from the first size and range trade-off through to radiated measurement on production hardware.