rssi-1 measurements DECT NR+

[nrfuser2137]

Hello,

I am scheduling an RSSI measurement operation using DECT PHY firmware v1.0.0.
The parameters used for scheduling the operation are:
- start_time = 0
- carrier = 1677
- duration = 48 (in subslots)
- reporting_interval = NRF_MODEM_DECT_PHY_RSSI_INTERVAL_24_SLOTS

The operation scheduling is successful, nrf_modem_dect_phy_rssi() returns 0.

In the callback to receive RSSI measurements, void (*rssi)(const uint64_t *time, const struct nrf_modem_dect_phy_rssi_meas *meas)
I am printing the *time and the meas->meas_start_time.

My expectation was that the difference between *time and meas->meas_start_time for each operation would be roughly 10ms (in modem time units) as the rssi measurement duration is 48 subslots, and reporting interval is also the same. 

However, the output printed in rssi_cb is as follows.

[00:00:04.137,207] <inf> phycb: cbtime 249018865 and meas->meas_start_time 249010078
[00:00:04.147,735] <inf> phycb: cbtime 249746536 and meas->meas_start_time 250428979
[00:00:04.158,264] <inf> phycb: cbtime 250473924 and meas->meas_start_time 251847555
[00:00:04.168,792] <inf> phycb: cbtime 251201314 and meas->meas_start_time 253266159
[00:00:04.179,321] <inf> phycb: cbtime 251928707 and meas->meas_start_time 254684742

So, my question is, what exactly does meas->meas_start_time refer to?

[Charlie]

Hi,

Try to catch up with your calculation.

The meas_start_time interval is around 1418900 modem time unites. Considering the following definition in nrfxlib\nrf_modem\include\nrf_modem_dect_phy.h. The actual meas_start_time interval is around 1418900/69120=20.5ms.

#define NRF_MODEM_DECT_MODEM_TIME_TICK_RATE_KHZ 69120

The subslot duration is defined in ETSI TS 103 636-3 [2], clause 4.4. a slot duration is 0,41667ms. There are 48 subslots measurement duration and 24 slots reporting interval, but I am not sure if you will get a report if the measurement is not finished. You can also easily verify with a test like 72 subslots measurement duration. If this is true, 48*0,41667ms=20ms, still have 0.5ms mismatch. 

I will check with our modem developer anyway.

Best regards,

Charlie

[nrfuser2137]

Charlie said:

The meas_start_time interval is around 1418900 modem time unites

meas_start_time is an interval or an absolute time?

According to the comments in nrf_modem_dect_phy.h, I understood the meas_start_time is an absolute timestamp when the modem started the measurements. So, FT could take meas->meas_start_time value as the reference point to calculate start of a frame which can be further used by FT for scheduling, ..

To see how the PHY API works, I wanted to do the rssi measurement for 10ms and also have the information at what time the rssi measurement started. RSSI measurement operation needs duration and reporting_interval. I wanted to keep both of them equal to 10ms. According to the documentation, both params need to be in subslots. But the 

enum nrf_modem_dect_phy_rssi_interval reporting_interval seems to have fields in "slot" units (not in subslots), e.g., NRF_MODEM_DECT_PHY_RSSI_INTERVAL_24_SLOTS. The params i used to scheduled the operation are, duration = 48 subslots (10ms) and reporting_interval=NRF_MODEM_DECT_PHY_RSSI_INTERVAL_24_SLOTS (10ms). 

Now, looking at the logs printed, cbtime refers to *time param of the callback. The time difference between first two callbacks is 249746536-249018865=727671 or (727671÷69120=10.5ms). I would account 0.5ms for the debug logs overhead, etc... and the callbacks are async. MAC operations can be scheduled taking modem time into account. 

I was also expecting the (meas->meas_start_time - *time) roughly ~727671 considering the parameters to schedule the operation, but this is not the case. And except for the first log, meas->meas_start_time is greater than *time. How can meas_start_time be greater than the current modem_time? Seems like I misunderstood the measurement operation. 

While looking at the nrf_modem_dect_phy.h again, I found some inconsistencies. We can also get rssi measurements with rx operation. But there, the documentation does not mention whether the reporting_interval is in slots or subslots. The duration for rx operation is in modem time units (in contrast to rssi operation where duration is in subslots). 

Would be very helpful if you can comment on this...