Why waking up every conn_interval even with slave latency?

RE: Why waking up every conn_interval even with slave latency?

Ole Bauck — Fri, 04 Dec 2015 14:32:57 GMT

What was your previous supervision timeout? I thought you should get an error if the connection parameters where wrong, like described in the link.

RE: Why waking up every conn_interval even with slave latency?

Primož Kralj — Fri, 04 Dec 2015 14:11:33 GMT

I think I got the parameters somehow wrong. It now works with latency 60, conn interval [200,250ms] and supervision timeout BLE_GAP_CP_CONN_SUP_TIMEOUT_MAX (~32sec). Future readers check this answer too.

Ole, I will do more tests next week to pinpoint exactly what went wrong and will post it here. Thanks for now, Nordic support is amazing!

RE: Why waking up every conn_interval even with slave latency?

Ole Bauck — Fri, 04 Dec 2015 12:57:17 GMT

ok, then the central is accepting every connection parameter update requests. The sniffer log you gave me was too short, connection parameter update request usually happens some time after the connection. If you are using the ble_conn_params library, it happens FIRST_CONN_PARAMS_UPDATE_DELAY after the connection.

RE: Why waking up every conn_interval even with slave latency?

Primož Kralj — Fri, 04 Dec 2015 12:48:54 GMT

I am running heavily modified omnipresent UART example (I can upload it though). I have sniffed the link though, pcap log is here. I am not very keen with reading the packets yet, though.

The only thing happening on central in BLE_GAP_EVT_CONN_PARAM_UPDATE_REQUEST event is:

err_code = sd_ble_gap_conn_param_update(p_gap_evt->conn_handle, &p_gap_evt->params.conn_param_update_request.conn_params);

RE: Why waking up every conn_interval even with slave latency?

Ole Bauck — Fri, 04 Dec 2015 12:03:53 GMT

Are you running any example? Could you check what is done in BLE_GAP_EVT_CONN_PARAM_UPDATE_REQUEST event on the central? Or even better if you manage to sniff the connect request packet, see sniffer.

RE: Why waking up every conn_interval even with slave latency?

Primož Kralj — Fri, 04 Dec 2015 11:46:59 GMT

I have played a bit more and found out that slave latency param has no effect. I am setting the connection parameters on both devices (both nRF51) with ble_gap_conn_params_t; on central with sd_ble_gap_connect() and on peripheral with sd_ble_gap_ppcp_set(). For clock I am using NRF_CLOCK_LFCLKSRC_RC_250_PPM_4000MS_CALIBRATION and I can see it triggering properly.

But these spikes (wakeups) are the same as connection interval, always. Is it possible that central device discards the latency setting and sets it to 0 internally in softdevice? Do you have any other idea? Thanks!

RE: Why waking up every conn_interval even with slave latency?

Ole Bauck — Fri, 04 Dec 2015 10:32:17 GMT

Are you sure that all the connection parameters are accepted at the central? Some devices may not accept this combination of connection interval and slave latency (e.g. iphone).

What are the current measurements at this intervals? Can it be something else like a repeating task started by app_timer? If you are using RC LFCLK, the clock will be started regulary for calibration. Are you using NRF_CLOCK_LFCLKSRC_RC_250_PPM_250MS_CALIBRATION?