Directed (fast) advertising timeout and scanning at low power

RE: Directed (fast) advertising timeout and scanning at low power

Eliot Stock — Wed, 14 Dec 2016 18:40:55 GMT

@petter Thanks. Converted one of your comments to an answer.

RE: Directed (fast) advertising timeout and scanning at low power

Petter Myhre — Wed, 14 Dec 2016 14:46:15 GMT

I can't tell you how you can be reasonable sure that a connection is established in a certain amount of time. You will have to tune the scan window, scan interval, advertising interval to fit your application, I don't know your power budget, I don't know how long a connection establishment can take before the customer thinks it's too long, I don't know how much interference and packet loss there will be in the environment where the device are going to connect.

RE: Directed (fast) advertising timeout and scanning at low power

Petter Myhre — Wed, 14 Dec 2016 14:46:12 GMT

I would at least see if the minimum advertising interval (20 ms, but 25 ms averaged) with normal advertising is sufficient to fulfill the requirements of your use case. It seems you are tied to low power scanning (I'm not sure why), but with 500 ms scan window and 5000 ms scan interval, and with 20 ms advertising interval the central should receive 500/25 = 20 advertisments every 5 seconds. With an advertising interval of 60 ms (65 ms averaged) the central should receive 500/65 = 7.7 advertisments every 5 seconds. Here I assume that there is no packet loss or interference, which there always is. So the smaller the advertising interval the higher chance the central has of receiving a higher number of advertisments.

RE: Directed (fast) advertising timeout and scanning at low power

Petter Myhre — Wed, 14 Dec 2016 14:45:44 GMT

High duty cycle directed advertising the advertising interval is less than 3.75ms, and is usually used just after a link loss, because then you know that the scanner is actually scanning, so that you don't waste power spamming advertisments when there is noone listening. It can be used in other scenarioes, but it is an extreme measure, where latency is important, but power consumption is not.

RE: Directed (fast) advertising timeout and scanning at low power

Eliot Stock — Wed, 14 Dec 2016 14:01:51 GMT

Yes: can you confirm that directed (fast) advertising is not helpful in this situation? And do I really have to advertise for 15s in order to be reasonably sure of a connection?

RE: Directed (fast) advertising timeout and scanning at low power

Petter Myhre — Wed, 14 Dec 2016 13:51:45 GMT

I know that the scanner does not. Do you have any more questions?

RE: Directed (fast) advertising timeout and scanning at low power

Eliot Stock — Wed, 14 Dec 2016 13:22:58 GMT

I understand that the peripheral uses all three channels each time. But the scanner does not - look at the top of the same picture. As for the point about two requiring two advertising packets, I guess that's an argument for either advertising for longer or decreasing the advertising interval, or both.

RE: Directed (fast) advertising timeout and scanning at low power

Petter Myhre — Wed, 14 Dec 2016 13:14:55 GMT

Every advertising interval the peripheral sends out three advertising packets, one on each of the advertising channels, it doesn't stick to one channel. You can see that in the third picture in the blog you linked to. Also be aware that typically a central has to receive at least two advertising packets, one to tell the central what kind of device it is, and one so that the central can send connetion request to it. This is because the central can only send a connection request to the advertiser in a small RX window that is open just after an advertisment is sent.

RE: Directed (fast) advertising timeout and scanning at low power

Eliot Stock — Wed, 14 Dec 2016 11:42:56 GMT

I end up with 15s because it's 5s per scan interval for the central, and I need three of them. I'm assuming I need all three channels to be reasonably sure of a connection. So, for example, I miss the connection on channel 37 and again on channel 38 but I get it on channel 39. The advertising interval doesn't affect this calculation, as far as I can see, but I've tested with 30, 60 and 150ms. As long as I'm sending an advertising packet at least every 500ms, there'll be one there during every scan window.

RE: Directed (fast) advertising timeout and scanning at low power

Petter Myhre — Wed, 14 Dec 2016 11:14:31 GMT

High duty cycle directed advertising shall not last more than 1.28 s (but you can restart it after the timeout). How do you end up with 15 s? And what advertising interval are you using in this calculation?