Importing Bluetooth speakers from China

A speaker that claims 20 W of output from a 1,200 mAh battery cannot deliver them continuously: 1,200 mAh at 3.7 V is 4.4 Wh, and at 20 W that energy is gone in thirteen minutes. The end buyer does not compare spec sheets: they turn the Bluetooth speaker up to maximum, hear it distort or run out of battery sooner than the box promised, and send it back. Before paying a deposit you have to review three documents — the UN38.3 report for the exact cell and pack model, the MSDS and the Bluetooth SIG declaration — and measure two things on a sample: maximum sound pressure level and sustained power under load.

Why 20 W with 1,200 mAh is not a sustainable figure

The arithmetic does not admit interpretation. 1,200 mAh at a nominal 3.7 V is 4.44 Wh; divided by 20 W that is 0.22 hours, in other words thirteen minutes, and with the real efficiency of a class D amplifier you are left with about eleven.

A bridge-tied amplifier fed from a 3.7 V cell delivers no more than about 2 W per channel into 4 Ω. The maximum RMS voltage it applies to the load is around 2.6 V, and power goes with the square of voltage: 2.6² ÷ 4 = 1.7 W. To reach a real 20 W you would need a supply of about 13 V, and stepping up to that voltage from 3.7 V means drawing more than 6 A from the cell: a 1,200 mAh cell run at five times its nominal capacity degrades within weeks.

A 20 W claim comes from three places: the IC datasheet, measured at a voltage the speaker never gives it; the instantaneous peak the chip tolerates for milliseconds; or the sum of the channels under the label “maximum power”.

The question that replaces the spec sheet is one: sustained power per channel, at what supply voltage and with what declared distortion. If the supplier cannot give you the part number of the amplifier IC, they do not have that figure.

How two speakers are compared: measure, do not read

A spec sheet does not replace a measurement because the sound pressure level does not depend on the amplifier alone: it depends on the transducer’s sensitivity, the volume of the enclosure and the tuning of the passive radiator. Two speakers with the same IC can sit 6 dB apart if one mounts a 91 dB/W/m transducer and the other an 85 dB one, and 3 dB is double the acoustic power.

Distortion is not assessed at low volume, where every speaker sounds good. Nearly all budget models carry a compressor in the DSP that cuts the output to protect the transducer: below the threshold it sounds clean and just above it the dynamics are flattened. The honest figure is not the amplifier’s power, it is the maximum level before the limiter acts.

The battery decides the transport route

Because the cell sits inside the equipment, the speaker ships under packing instruction 967 (UN3481, battery contained in equipment), which is the most permissive configuration. That does not exempt you from anything: without the UN38.3 for the exact cell and pack model and without the MSDS — the chemical safety data sheet — the cargo is rejected at the carrier’s receiving desk.

Typical portable speakers run from 7 to 37 Wh and fly as cargo with that documentation; models with a 20,000 mAh cell exceed 70 Wh and fall into a different bracket, with state-of-charge limits. A shipment does not fail because of capacity: it fails because the factory changed cell supplier and the report still names the previous model. The thresholds and the mAh-to-watt-hour conversion are set out in our guide to shipping lithium batteries.

Bluetooth SIG: verified before the deposit

The Bluetooth SIG declaration is mandatory to market any speaker with Bluetooth legally, it is made by the manufacturer and it is checked in the public qualified-products database, not in the PDF the supplier sends you. Many 1688 resellers cannot produce it: they sell factory product under a name that is not theirs, and the declaration stands in someone else’s name.

Four data points to cross-check: the identifier of the finished product and not the module’s, because they are two declarations and the supplier shows you the one they already had; the holder, because if it is the factory and you sell under your own brand you need to know whether the listing covers you; the commercial name and model, compared with the box you are going to import; and the date and the module, because a listing tied to the previous chip does not cover the lot.

Water resistance: what an IPX rating tests and what it does not

The IPX rating describes a specific test, not a general capability: IPX4 is splashing, IPX5 a pressurised jet, IPX7 immersion at 1 m for 30 minutes and IPX8 a depth agreed with the manufacturer. A product rated IPX4 sold as submersible generates returns, and the failure is almost never in the shell: it is in the USB port cover and in the acoustic mesh that covers the transducer.

An IPX rating does not say three things: it says nothing about dust unless it carries the digit 6; it does not cover salt water or chlorine, so a unit that passed the laboratory can fail after a pool season; and it is not valid after a drop, because the test is performed on an intact unit.

What to test: immerse three units taken from the lot to the declared depth and for the declared time, dry them, play at maximum volume for thirty minutes and then open the transducer cavity. Test the cover with the cable in and with the cable out, which is where real sealing fails.

Firmware and pairing: the failure the user notices first

The first failure the buyer perceives is not sound quality, it is the connection: that it does not pair first time, that it cuts out at three metres or that it drops the link when you change room. It does not appear in a visual inspection and is only detected by testing with several source devices.

The protocol over two or three units from the lot:

• Ten to twenty pairing cycles with at least three phones, two versions of Android and one of iOS.
• Reconnection on coming back into range and after fully draining the battery.
• Two units of the same model switched on in the same room: this is the test that fails on modules that handle coexistence badly, and it is the real situation of a warehouse or a shop.
• Latency in interactive content, not in video: video applications compensate for the delay, games do not.

Alongside the firmware comes who publishes it, whether an app exists, whether the factory supplies signed firmware and the recovery tool, and whether it commits in writing not to change chipset. A speaker whose DSP tuning lives in firmware that only the factory holds cannot be corrected or replicated.

The cluster: here the geographic advantage is real

A portable speaker is an assembly of parts that in Guangdong are made less than an hour apart: the transducer, the passive radiator, the injection-moulded shell, the amplifier board and the Bluetooth module. Shenzhen provides the radio module and the final assembly in Bao’an and Longgang; Dongguan, the injection moulding and much of the transducers, in Changping, Qingxi and Tangxia.

The consequence is not price, it is iteration. Matching a transducer with a sealed enclosure and a passive radiator requires testing, and testing is only cheap here: a transducer change that from Europe is a three-week cycle is resolved here in days.

MOQ and why the enclosure mould raises the floor

A catalogue speaker starts at 300 to 1,000 units per reference. A custom enclosure raises the minimum to several thousand per model, and it is not a commercial decision: an enclosure is three to six injection parts — front grille, shell, passive radiator frame, port and battery cover — and each one carries its own mould.

Tooling cost depends on the part size, the number of cavities, the steel and the finish, is paid separately from the order and is not recovered if the model does not work. There is a second cost that gets forgotten: changing the enclosure invalidates the acoustic tuning, so the DSP has to be set up again and new samples validated.

For a first order, the sensible route is the catalogue product with customisation limited to colour and packaging: it stays within the scope of the certifications already issued.

Certification by market: Mexico, Argentina and Brazil

In Mexico, the NOM on commercial information requires labelling, instructions and a warranty in Spanish, and the test is performed on the finished product: the procedure starts once a manufactured unit exists and finishes before shipment.

In Argentina, the requirement that always appears is ENACOM homologation, per model and with a local holder. Unlike Mexico and Brazil there is no single body that combines radio and electrical safety: safety arrives by way of the power supply. A speaker supplied with only a USB cable does not incorporate a power supply and its scope is smaller than one with a mains adapter in the box. The Argentine regime changes often, so the specific scope is confirmed with the customs broker before producing.

In Brazil, ANATEL homologation is mandatory before importing for sale and the holder must be a Brazilian legal entity. The 2.4 GHz radio tests follow Ato 14.448/2017 and the electromagnetic compatibility tests Ato 1.120/2018. The SAR test of Ato 1.630/2021 applies to equipment used against the body and does not apply to a speaker, unlike a pair of earbuds. Inmetro covers the power supply when the product includes a mains adapter; if it only carries a cable, the scope is reduced.

MeliPrep coordinates destination certifications and pre-shipment inspection: AQL is billed at USD 299 per inspector-day, with ISO 2859-1 general level II sampling and AQL 2.5 / 4.0, and laboratory and authority fees are passed through at cost. A supplier who delivers the UN38.3 with the cell and pack model, the MSDS, the Bluetooth SIG identifier and the sustained power in writing is a supplier you can build a speaker line with. One who answers with a photo and the figure from the box is warning you where the problem will appear.