What? Why do we say “develop”? It’s because it really is a competitive field!For many low-end BLE chips, the main function is simply wireless connectivity; being able to connect and transmit a few data packets is sufficient. Some even do not require a connection, as they can scan and advertise.So what is the biggest cost for such BLE chips? It is the die cost. This article analyzes how to create such a chip from the following aspects:
- Die Cost
- Processor, preferably RISC-V, abandoning ARM M0 to reduce royalty costs
- Memory, reduce or even eliminate SRAM, switching to a ROM + OTP solution, requiring precise evaluation of application code size
- External storage, either not using it or replacing Flash with a small capacity EEPROM
- Peripherals, just enough for the application, select based on specific scenarios, it is best to rewrite peripheral IP instead of using existing IP from Synopsys, etc. (too large!)
- Power supply, forget about DC-DC buck converters, use LDO instead
- GPIO pins, just enough for the application, select based on specific scenarios
- BLE digital IP, cut unnecessary features, such as long range, AES, etc.
- RF circuit, avoid IQ modulation, use direct two-point modulation, eliminate the mixer, lower RFPLL specifications, minimize RXADC bit depth, GFSK is sufficient for zero-cross detection, control transmission power, and PA can also be minimized
- Process, calculate die size and wafer cost, choose accordingly, generally 55nm or 40nm, which wafer foundry to use depends on personal capability
- Packaging
- QFN? BGA? Forget it, just go with SOP
- When EEPROM is not needed, use single packaging directly, no need for combined packaging
- Testing
- CP testing, after large shipment volumes, calculate yield, save costs wherever possible, as they will be sold as general materials
- FT testing, keep it as simple as possible, ideally RF loop testing should be done in-house, and later save costs based on yield
- PCB
- External crystal, save costs where possible based on the scenario; if no connection is needed, use the internal oscillator directly, only for beacon mesh
- RF matching network, SOP is inherently difficult to match, save costs where possible, or create a small matching network on-chip
- A 2-layer PCB is sufficient
Most people probably do not want to compete in this way and may think such chips are low quality. However, before there are high technical barriers or an ecosystem is formed, not competing is not an option. If you do not compete, you may really struggle to sell, as the market will make its choice.End of article!