跪拜 Guibai
← Back to the summary

Bluetooth LE Connections: Roles, Events, and the Parameters That Keep a Link Alive

Device roles and connection knowledge map: roles determine connection division of labor, connection establishment lets both parties obtain communication arrangements, connection events carry bidirectional send/receive, parameters control work rhythm, applications also need to handle services, security and reconnection

Preface

To notify residents of a water outage, the community can post a notice on the bulletin board. Residents see it and know the outage time, without first establishing contact with the property management.

But if a pipe at home breaks and you need to report it to property management, explain the situation, and confirm the resolution, reading a notice is not enough. The two parties need to communicate individually.

Bluetooth Low Energy has a similar distinction. Devices can broadcast information, and nearby devices that receive it can use it; when continuous data exchange with a specific device is needed, a connection is usually established. For example, a phone seeing a smart lock in its list means it obtained a clue to discover the device; to check the lock's status or change settings, further communication with the lock is required.

So what exactly does a Bluetooth "connection" establish? Why can a phone and a lock stay in contact after connecting, without constantly sending and receiving data?

1. A connection lets two devices agree on how to communicate

In Bluetooth LE, a connection is a relationship between two devices that have shared communication parameters and send/receive data according to an agreed schedule. The two parties not only know the other exists, but also know when to communicate later, which channels to use, and how long without contact counts as a lost connection.

When people say a phone is "connected" to a sensor, they usually mean an LE ACL connection has been established. ACL is the transport mechanism that carries normal bidirectional data here; operations like reading status and changing settings usually go through it. It differs from isochronous communication used for time-sensitive transmission like audio, so not all Bluetooth transmission should be treated as the same kind of connection.

After a connection is established, the two sides can exchange data under common rules, without having to rediscover each other for every business message. But this does not mean they occupy a dedicated radio frequency or lay a "wireless wire" in the air that is always working. The radio can rest between communication gaps, and other devices still share the band.

Broadcast and connection each have their use:

What needs to be done Consider this way Reason
Let multiple nearby receiving devices learn a tag number Broadcast The same information can be received by multiple receivers without establishing connections one by one
Repeatedly exchange settings, status, and records with a specific device Connection Both sides have a common send/receive schedule, suitable for continuous bidirectional interaction
Both easy to discover and able to modify device settings Combine broadcast and connection Broadcast provides discovery clues, connection carries subsequent interaction

This is not "a reply means a connection is needed". Periodic broadcast with response, known as PAwR, can also arrange replies from receivers; but it uses broadcast grouping, time slots, and other mechanisms, different from point-to-point connection. Which to choose depends on receiver count, data volume, timeliness, security, and device capabilities, not just "whether power saving is needed".

2. Central and peripheral devices, the division of connection work

2.1 Who initiates, who accepts

The Central device is responsible for initiating the connection; the Peripheral device is responsible for accepting the connection. In the common connectable advertising process, the peripheral device publishes connectable advertising, and the central device initiates the connection after finding a suitable target.

When a phone connects to a smart lock, the phone is usually the central device and the lock is the peripheral. The phone selects a target and initiates the connection, and the lock accepts qualifying requests. Here "central" is not a performance ranking, and "peripheral" is not just passively waiting for commands.

After the connection is established, both ends can send and receive data. For example, the phone can send a request to modify settings, and the lock can also send messages to the phone when status changes. How to organize these messages needs application-layer agreement; the lock does not lose the ability to send data just because it is the peripheral.

Older materials often use Master, Slave, often written in Chinese as "主机、从机". Now Central, Peripheral are used. When learning, they can be correspondingly understood, but do not confuse the "master" here with the Host in the protocol architecture: Host is part of the protocol stack, Central is a role in the connection.

2.2 Four GAP roles describe the division of labor in communication

GAP is the Generic Access Profile, which specifies basic behaviors like device discovery and connection. Bluetooth LE defines four GAP roles:

GAP role Main division of labor A possible use
Broadcaster Publishes information without establishing a connection A tag publishes its own number to nearby devices
Observer Receives broadcast information A receiving device collects broadcasts from nearby tags
Peripheral Supports accepting connections A lock waits for the phone to establish a connection
Central Supports initiating connections A phone selects and connects to a lock

Broadcaster and Observer roles themselves are not responsible for establishing connections. If a product both publishes broadcast data and accepts connections, it also takes on roles like Peripheral; it cannot be fully described only as "Broadcaster".

Also, broadcasting does not mean only being a Broadcaster, and scanning does not mean only being an Observer. A Peripheral can wait for connections via connectable advertising, and a Central also finds targets by receiving broadcasts. "Broadcast, scan" are current activities; "GAP role" describes a set of communication division and capabilities.

For example, after taking the Central role, a phone can scan, select the lock, and initiate a connection. Before connecting to the lock, the phone is already acting in this role to find a connection target, but the two parties have not yet established a connection.

Once a connection is established, the link layer responsible for underlying radio transmission also calls the initiating end Central and the accepting end Peripheral. GAP roles indicate the division of labor in device discovery, connection establishment, etc.; link layer roles indicate identity in a specific connection. The two use the same names, but "taking the Central role" and "already connected to the other side" are two different things. GAP device roles

2.3 A device can take multiple roles

Roles are not labels that stay fixed after a phone or sensor leaves the factory. When supported, a phone can act as a Peripheral and accept connections; a gateway can also act as a Central to connect sensors while acting as a Peripheral to accept phone connections.

For a gateway, it is "acting as a Central in the connection toward sensors, and as a Peripheral in another connection toward the phone." It does not swap roles within the same connection due to send direction changes.

Whether a device can simultaneously scan, broadcast, maintain multiple connections, and the maximum number of connections, depends on the controller, protocol stack, system interfaces, and radio scheduling. You cannot infer an unlimited connection count from "supports Central role", nor assert that only one Central can connect based on "Peripheral".

2.4 Central is not the same as GATT Client

Apart from connection roles, application data has another common set of roles: GATT Client and GATT Server.

GATT is the Generic Attribute Profile, which organizes the data and functions a device exposes into services, characteristics, and other objects. For example, battery level can be a readable data item; a setting can be a modifiable data item. The side that provides these data and their access methods is the server; the side that initiates reads, writes, and other operations is the client.

Roles to distinguish Question they answer
Central / Peripheral Who initiates the connection, who accepts it; how the link organizes send/receive
GATT Client / Server Who initiates data access, who provides the accessed data and operations

When a phone reads the lock's battery level, the phone is usually both Central and GATT Client; the lock is both Peripheral and GATT Server. This is a common combination, not a binding rule.

If the application is designed that way, a phone acting as Central can also provide GATT services, and the other side acting as Peripheral can read them. The same device can also simultaneously act as GATT Client and Server.

A GATT Server can also send notifications or indications according to the rules, without waiting for the client to read first. Thus, "who sends this business data first" cannot be used to determine who is Central. GATT roles and configuration

3. What happens from broadcast to connection

3.1 Finding a target is not the same as connecting

The common process is: the lock sends connectable advertising, the phone receives the broadcast, the application selects a target based on device information, then requests a connection.

Receiving one broadcast means the phone heard the device's published information; obtaining a scan response means it got an additional piece of supplementary information. Neither process equals a connection.

Initiating the connection is a separate specific process. After the lower layer enters the initiating state, it waits for qualifying connectable advertising and sends a connection message at the appropriate time. It is not as simple as saying "connect" to a name left in the scan list.

This is also one reason why "it could be found just now, but now it won't connect": the list may retain old results, but the target has stopped connectable advertising, left the range, or currently does not accept this phone.

The application does not have to display a scan list first. If it already knows target information, it can directly request a connection, and the lower layer executes the necessary search and initiation process. "No device list popped up" does not mean the radio can skip the necessary air interaction.

3.2 The connection message brings a common communication arrangement

Taking traditional connectable advertising as an example, the phone responds to a suitable broadcast and sends CONNECT_IND. This packet not only identifies the two communicating parties, but also provides initial parameters needed for subsequent communication, such as:

Traditional connectable advertising connection sequence: peripheral sends advertising, central responds with CONNECT_IND and gives initial parameters, then both sides exchange data in connection events; traditional process has no extra connection acceptance response packet

In the traditional flow, after the peripheral receives a qualifying CONNECT_IND, it enters the connected state. There is no extra "accept connection" response packet following it. Then both sides exchange data on data channels as agreed. Drawing the traditional flow as "broadcast - connection request - connection confirmation - start communication" adds an extra step that does not exist.

Extended advertising connection is different: it uses auxiliary advertising to find the right timing and channel, and uses AUX_CONNECT_REQ and AUX_CONNECT_RSP to complete the request and response. This response does exist, and the two flows should not be mixed.

The protocol also distinguishes "entering the connected state" from "actually receiving the data channel packet from the peer". The former means the lower layer has started working according to connection rules; the latter can only show that both sides have met on subsequent channels. Even after the connection message is sent, if subsequent send/receive fails, the establishment process can still fail. The application should ultimately handle results based on the system interface report, not just display "connected" merely because a request was issued.

The diagram uses the common connectable advertising path. Devices supporting PAwR also have corresponding mechanisms to initiate connections from periodic advertising processes, so "Peripheral must first send ordinary connectable advertising to Central" cannot be generalized as the only rule for all LE connection methods.

4. After connecting, the two sides exchange data in connection events

4.1 A connection event is a round of send/receive opportunity

A Connection Event is a round of send/receive activity that occurs at agreed timing within a connection. Multiple packets can be exchanged in one event; it is not an application message, nor is it an application-layer "connection success callback".

Think of it as an agreed meeting time: both sides do not have to stay in the same place all day; they remember the next meeting time and meet at that point. Bluetooth also has to agree "which channel to meet on", and this arrangement is executed by the lower layer at precise times, not by an application timer calling send every few tens of milliseconds.

In a normal round of LE ACL send/receive, the Central sends first, the Peripheral replies, then both sides can continue alternating. If the Peripheral has data to send, it sends within this link timing, not grabbing the radio arbitrarily outside the agreed schedule.

Two connection events send/receive: each round starts with Central sending, Peripheral replying, and one round can have multiple alternating send/receive groups; next round selects channel based on common rules, gaps can be used for rest or other radio tasks

When there is no new business data, the lower layer can still use packets that carry no new application data to maintain the link and confirm send/receive. Therefore, a period without business callbacks like temperature or button events is not enough to conclude the connection is broken.

The start times of adjacent events are arranged by the connection interval. But which events actually participate is affected by peripheral latency, connection subrating, scheduling, and other conditions; the ideal diagram should not be understood as every event guaranteeing a successful exchange of packets.

4.2 Both sides switch channels according to common rules

A regular LE ACL connection can choose data channels from 37 general channels. One connection event uses the same data channel, not arbitrarily changing with each packet; for subsequent events, both sides select the corresponding channel according to the same rules, and the newly selected channel may be the same as the previous round.

The channel map marks which channels can be used. The Central can update the channel map based on the radio environment, and then let both sides apply the update at agreed times per the protocol; when supported, the Peripheral can also report channel conditions to help choose.

This is part of adaptive frequency hopping: try to avoid poor-performing channels and reduce the chance of repeated interference. It cannot make interference disappear, nor does it mean getting an exclusive frequency that no one else can use after connecting.

4.3 Link received is not business complete

For normal ACL data requiring reliable transmission, the lower layer uses sequence numbers, acknowledgments, and retransmission to handle reception failures. For example, if one side does not get the expected acknowledgment, it can retransmit; the receiver can also identify duplicate packets to avoid passing the same lower-layer data up twice.

These tasks are usually done by the Bluetooth protocol stack; the application does not need to write retransmission procedures for each radio packet. But if radio conditions keep deteriorating, the connection will still be lost; newer versions also provide optional mechanisms to discard specific real-time data after expiry. "Connection reliable" cannot be understood as all data waiting indefinitely and eventually arriving.

More importantly, the lower layer confirming data reception does not mean the peer application successfully executed the operation. The phone sending "modify lock settings" data only means the transmission reached some layer; the lock might reject it due to insufficient permissions or invalid parameters. If the business needs to know the final result, it must define corresponding success, failure, or status query mechanisms. LE connections, send/receive, and link control

5. Three parameters determine how a connection waits and works

5.1 Connection interval: how long between the start of two rounds

The Connection Interval is the time between the start times of adjacent connection events. Assuming a connection interval of 50 ms, in a simplified timeline, events can start at 0, 50, 100, 150 ms.

It is not "each transmission lasts 50 ms", nor "a business data item must be generated every 50 ms". A round may exchange only a few packets or multiple groups; an application measurement might update only once per second.

A shorter interval usually provides denser send/receive opportunities, which helps timely response; it also tends to increase radio wake-up and scheduling overhead. A longer interval usually gives more rest time, but new messages may wait longer. Actual throughput and power consumption are also affected by per-round transmission volume, the radio transmission method used, retransmission, and system scheduling; this parameter alone is not enough.

The available connection interval range also relates to features supported by both sides. When the shorter connection interval capability is not used, the baseline range is 7.5 ms to 4 s, in steps of 1.25 ms; Bluetooth Core 6.2 introduces Shorter Connection Intervals, providing shorter intervals, with optional extended values down to 375 μs, i.e., 0.375 ms. The new capability requires support from both sides and the corresponding protocol stack; claiming a version number does not mean any application can use the shortest interval. The 50 ms values in this section are only to illustrate the time relationship, not recommended parameters. Shorter connection interval capability and negotiation

5.2 Peripheral latency: allowed to skip a few rounds

Peripheral Latency indicates the maximum number of consecutive connection events that a Peripheral may skip listening to when the rules are met. In the basic case without connection subrating, it can be understood directly as consecutive connection events.

Assume the connection interval is 50 ms and the peripheral latency is 2. The peripheral can participate at 0 ms, skip the 50 and 100 ms rounds, and participate again at 150 ms. It can also participate earlier; it is not forced to appear "only once every three rounds".

Peripheral latency illustration: connection interval 50 ms, latency 2 allows skipping two rounds after participating in one, then participating at 150 ms; event schedule does not change, peripheral can participate earlier

Why is this useful? A button device with no new data most of the time does not need to wake its receiver for every communication opportunity. When it has a message to send, it can participate in a qualifying subsequent event; but new data from the Central may also wait longer because the Peripheral is skipping events.

In this basic case, if the full skip allowance is used, the gap between participation opportunities is:

(Peripheral Latency + 1) × Connection Interval

That is (2 + 1) × 50 ms = 150 ms.

This 150 ms describes the participation rhythm above, not the maximum delay for any business response. Messages may also queue, retransmit, or need application processing.

The number in peripheral latency is "number of events", not milliseconds; it is also not a network-measured response delay. Latency of 0 means no proactive skip allowance, not that the link will never drop packets or have scheduling conflicts.

5.3 Connection supervision timeout: how long without valid contact counts as disconnection

The Connection Supervision Timeout is the time threshold the lower layer uses to determine a lost connection. After a connection is established, each end uses a supervision timer to track valid link packets from the peer; receiving a valid packet resets the corresponding timer, and only after a long time without receiving one does it judge the connection lost per the rules.

It does not check "how long the application has not received a business message". Even if the application has no new data for ten seconds, as long as the lower layer has valid send/receive meeting the requirements, the connection can be maintained.

Assume the supervision timeout is 2 s. If the target suddenly powers off, the other end usually does not know at the moment of power-off, but judges the loss after continuously failing to receive valid packets. The time the application receives the disconnection notice also involves lower-layer judgment and system event reporting; you cannot infer the exact power-off time from a momentary UI.

This threshold must work with the connection interval, peripheral latency, etc. Otherwise, the protocol allows one side to rest, but the other side declares it lost before the rest ends, creating a contradiction.

In the basic case without connection subrating, the requirement is:

Connection Supervision Timeout > 2 × (Peripheral Latency + 1) × Connection Interval

Still using interval 50 ms, latency 2, the right side is 300 ms. The supervision timeout should be strictly greater than 300 ms; taking 2 s satisfies this constraint, taking 300 ms does not. Satisfying this formula does not mean the system will necessarily accept the configuration; numerical ranges, other protocol conditions, and platform policies must also be met.

A short timeout can detect a lost connection earlier, but is more easily affected by temporary interference or transmission gaps; a long timeout may make the application wait longer. It is not "the bigger the more stable" or "the smaller the more advanced".

5.4 Parameters can be updated, but the requested value is not the effective value

A connection does not have to use the same set of parameters throughout. A device can request denser communication when fast transfer is needed, and request more power-saving arrangements during long waits.

Both Central and Peripheral can request parameter changes when supported by the corresponding mechanisms. The lower layer must handle the protocol and let both sides use the new arrangement at a jointly determined time; peer capabilities, controller resources, and system policy may all affect the result. An application submitting a "faster" request does not mean the parameter has become effective as desired; it should check the interface result and actual parameters.

5.5 Connection subrating: keep short rhythm, work less on demand

Connection Subrating is an optional mechanism that, while retaining the base connection interval, lets both sides use only a portion of the connection events according to rules. It can reduce radio activity when idle, while retaining a shorter base rhythm for phases needing dense communication.

Assume base interval is 30 ms, the basic event time points are 0, 30, 60, 90, 120, 150 ms... If the subrating factor is 5, in a simplified arrangement ignoring continuation events and peripheral latency, use one of every five base events, i.e., 0, 150, 300 ms... for send/receive. The base interval is still 30 ms.

If you directly change the connection interval to 150 ms, it also looks like communication every 150 ms when idle. The difference is that after directly lengthening, the base schedule no longer has events at 30, 60, 90 ms; to use denser events, you need to update the connection interval. Subrating retains those base events and uses a dedicated rule to decide when to use them.

One of those rules is the Continuation Event: when qualifying data activity occurs, continue using subsequent base events according to pre-configured settings. For example, when data transmission begins at the 150 ms event, and continuation conditions are met, you can continue send/receive at 180 ms; if subsequent data activity still meets the conditions, you can continue using the corresponding events, without necessarily waiting until 300 ms.

Directly using a 150 ms interval has only base events separated by 150 ms; a 30 ms base interval with subrating factor 5 can work every 150 ms when idle, and when data activity occurs and continuation conditions are met, can use subsequent base events at 180, 210 ms etc.

A single event can carry multiple packets. Here we compare when subsequent events are available when communication spans multiple rounds; it does not mean every packet waits 30 ms or 150 ms.

This suits devices that usually have little data and occasionally need to exchange a continuous segment: reduce activity in normal times, increase activity by rules when busy. Subrating also has a dedicated update process that can switch work rhythm faster than a general connection parameter update. But it does not guarantee that a message suddenly generated during idle is sent immediately, nor can the application bypass rules and temporarily choose a skipped event to send.

The difference from peripheral latency: peripheral latency mainly allows the Peripheral to skip listening for a few rounds; subrating directly arranges for both sides which base events are normally needed. They can be combined; when combined, the earlier supervision timeout constraint must also count the subrating factor, not copy the simplified formula. Whether it can be used still depends on both sides' support and system interfaces. Connection events and timeout parameters

6. After successful connection, the feature must still be made usable

Getting a connected phone does not mean the other party understands the language you speak. The Bluetooth lower layer being connected only provides the basis for subsequent communication; applications need to use data protocols both sides support.

For common GATT applications, you usually also need to discover or confirm available services, find the needed data items, and check access requirements. To continuously receive notifications, you need to complete the corresponding notification configuration; connecting does not automatically push all sensor data to the application.

GATT function readiness flow: after link establishment, confirm required services and data items, handle security per access requirements; when continuous notifications are needed, complete configuration, then conduct business send/receive. Security checks may be triggered during data operations

Three things need to be distinguished:

Process Problem it solves
Connection Both sides send/receive data according to common link rules
Pairing Negotiate security methods and generate or distribute keys used for link encryption, etc.
Bonding Save keys, device identity, and other information for future reconnection

The key used for link encryption here is a set of secret data involved in encryption calculations. The sender uses it to protect transmitted content, and the receiver uses the corresponding key to restore the content. It is not a device name, nor a string a user manually remembers; the number entered or verified during pairing is part of completing the pairing process, not directly using that number as a long-term encryption key.

For example, when the phone first modifies the lock's protected settings, it can pair on the already-established LE connection. Both sides choose a method based on their display, input capabilities, and security requirements; it may require entering a pairing code or verifying a number, or may not require manual confirmation. Different methods provide different protection against impersonation; "pairing success" alone does not mean any method is equally secure.

Connection is not pairing, and pairing does not require bonding. LE can establish a connection first, then complete pairing when accessing a protected function, enable encryption, or meet corresponding authentication requirements; data that does not require protection may be accessed directly.

If both sides want to continue using these security information in future reconnections, they can choose to bond, saving the necessary keys, device identity, and other information. On the next connection, if the saved information is still valid, both sides can restore the corresponding security state without pairing again; if one side has cleared these records, re-pairing may be needed.

Therefore, pairing addresses how to establish the information needed for security operations like encryption, while bonding addresses whether to keep that information for later. Saving bonding information does not mean the device is currently online, already connected, or that all operations are authorized.

When troubleshooting "shows connected but cannot read data", continue to distinguish: whether the service exists, whether the data item supports the operation, whether security requirements are met, whether notification configuration succeeded, and whether business data is interpreted in the shared format. Repeatedly clicking connect will not solve all subsequent problems.

7. After disconnection, reconnection is a new connection

7.1 Active disconnection and loss of contact are handled differently

Both Central and Peripheral can actively initiate disconnection; it is not only the Central that has this right. For example, the user ends the session, or the device enters a certain working mode, which may trigger active disconnection.

The other case is sudden power-off, moving out of range, or sustained severe interference. At that time, there may be no chance to send a disconnection message first; loss of contact must be detected through mechanisms like the supervision timeout.

Therefore, when the application receives "disconnected", it cannot immediately conclude the signal is weak. It should combine the system-provided reason, user operations, device state, and logs to distinguish active termination, establishment failure, security failure, or connection loss; different systems provide different levels of reason granularity.

7.2 Bonding information can be kept, but the link must be re-established

After a connection loss, the original link will not automatically recover as the same connection just by "bringing the phone closer". You need to re-initiate and establish a connection, then restore the required application state.

Common re-establishment depends on several conditions: the Peripheral provides connectable advertising again, the Central searches and initiates according to policy, and both sides still meet filtering and security requirements. If the product does not re-advertise, has reached the connection limit, or the system temporarily disallows scanning, the user may bring the device back and still cannot connect.

The so-called "automatic reconnection" is the application, system, or controller automatically executing the reconnection process per configuration, not a natural guarantee of the wireless link. A reasonable strategy usually tries after an interval on failure, rather than endlessly dense scanning and repeated requests.

After reconnection, you cannot assume business has recovered just because local old data remains. You must re-confirm necessary state per the protocol: GATT service cache may still be usable or may need updating; whether notification configuration is retained depends on the bonding relationship and implementation; unfinished business operations should be checked for results to avoid duplicate execution. Saving bonding information is keeping material needed to re-establish trust, not keeping a never-dropping connection.

8. Summary

Bluetooth LE connection lets two ends share a set of communication arrangements: know when to send/receive, how to select channels, and how to judge loss of contact. It can rest during gaps and also exchange data bidirectionally in the same connection event.

The Central is responsible for initiating connections, the Peripheral for accepting them; this division differs from the GATT Client/Server division. Products can support multiple roles, but concurrent activities and connection counts are limited by actual capabilities and resources.

The connection interval sets the basic communication rhythm, peripheral latency allows the Peripheral to skip listening to some events per rules, and the supervision timeout is used to judge lower-layer loss of contact. They must work together; newer capabilities like shorter connection intervals and connection subrating also require support from both sides, not just a version number.

Broadcast suits providing information to nearby receivers, connection suits continuous interaction with a specific device. After link establishment, services, security, and application configuration must meet business requirements; after disconnection, a new connection must be established and necessary state restored.

9. Test questions

  1. If a phone sees a lock in the scan list, does that mean it is already connected? If it also obtains a scan response, does the conclusion change?
  2. A phone initiates the connection, a lock accepts it. Later, the lock sends a status change to the phone. Who is Central, who is Peripheral? Do the roles swap because of the send direction?
  3. A gateway connects multiple sensors and also accepts a phone connection. What roles might it take on? Can you infer the maximum connection count from just these roles?
  4. "Central must be a GATT Client, Peripheral must be a Server." Is this statement true? Why might a Server also send data to a Client?
  5. In traditional connectable advertising, is there a dedicated connection acceptance response packet after CONNECT_IND? How does the extended advertising process differ? Can issuing a connection request declare connection success?
  6. Connection interval is 50 ms, and four packets are exchanged in one event. Does this violate the rule "only one packet every 50 ms"? If there is no new business data for one second, can you conclude disconnection?
  7. Without connection subrating, connection interval is 50 ms, peripheral latency is 2. After using the full skip allowance, how long between participation opportunities? Does a supervision timeout of 300 ms satisfy the time constraint in the text? Can the peripheral participate earlier?
  8. A phone has connected to a device but receives no notification data. Besides reconnecting, what else should be checked? If the lower layer has confirmed receiving a data item to modify settings, can you conclude the setting modification succeeded?
  9. What do pairing and bonding do respectively? If a device has been bonded and then powers off and restarts, does it still maintain the original connection with the phone? What conditions and processing are needed for "automatic reconnection" to truly restore functions?
  10. Directly setting connection interval to 150 ms, versus base interval 30 ms with subrating factor 5: are they the same time arrangement? If continuous data transmission starts at 150 ms, why might the latter arrangement continue communication at 180 ms?