search

falkenber9 / falcon

FALCON - Fast Analysis of LTE Control channels
GNU Affero General Public License v3.0
242 stars 74 forks source link

Question regarding PRB utilization #4

Closed parklize closed 3 years ago

parklize commented 3 years ago

I'm interested in getting PRB utilization in TTI level for a cell (1 ms). I saw COLUMNS collected by FALCON contains 'nof_prb', # number of allocated PRBs, but not sure how to convert it to a cell PRB utilization/ms with other COLUMNS. Is it possible in FALCON ? if possible, how to derive the PRB utilization/ms?

falkenber9 commented 3 years ago

You can group and aggregate the lines of the DCI tracefile by TTI and direction (UL/DL), e.g:

import pandas as pd

DCI_TRACEFILE = 'example-dci.csv'

COLUMNS_FALCON_DCI = [
    'timestamp',    # unix timestamp in [s] as float, 1µs resolution
    'sfn',          # system frame number
    'subframe',     # subframe index {0,1,...,9}
    'rnti',         # the addressed RNTI
    'direction',    # 0 for uplink alloc., 1 for downlink alloc.
    'mcs_idx',      # MCS index of the first transport block
    'nof_prb',      # number of allocated PRBs
    'tbs_sum',      # total Transport Block Size (TBS) in [Bit]
    'tbs_0',        # TBS of first transport block (-1 for SISO)
    'tbs_1',        # TBS of second transport block (-1 for SISO)
    'format',       # index+1 of DCI format in array flacon_ue_all_formats[], see DCISearch.cc
    'ndi',          # new data indicator for first transport block
    'ndi_1',        # new data indicator for second transport block
    'harq_idx',     # HARQ index
    'ncce',         # index of first Control Channel Element (CCE) of this DCI within PDCCH
    'L',            # aggregation level of this DCI {0..3}, occupies 2^L consecutive CCEs.
    'cfi',          # number of OFDM symbols occupied by PDCCH
    'histval',      # number of occurences of this RNTI within last 200ms
    'nof_bits',     # DCI length (without CRC)
    'hex'           # raw DCI content as hex string, see sscan_hex()/sprint_hex() in falcon_dci.c 
]

dci = pd.read_csv(DCI_TRACEFILE, sep='\t', names=COLUMNS_FALCON_DCI)
dci['tti'] = dci['sfn'] * 10 + dci['subframe']

tti_ul = dci[['tti', 'nof_prb', 'tbs_sum']].loc[dci['direction'] == 0].groupby(['tti']).sum().reset_index()
tti_ul.plot(x='tti', y='nof_prb', title="Spectral Utilization")
tti_ul.plot(x='tti', y='tbs_sum', title="Throughput (kbit/s)")

tti_dl = dci[['tti', 'nof_prb', 'tbs_sum']].loc[dci['direction'] == 1].groupby(['tti']).sum().reset_index()
tti_dl.plot(x='tti', y='nof_prb', title="Spectral Utilization")
tti_dl.plot(x='tti', y='tbs_sum', title="Throughput (kbit/s)")
parklize commented 3 years ago

@falkenber9 Thanks for quick response. This is very useful!

Is it somehow possible to get the PRB utilization (%)? I guess something like the allocated PRB - "nof_prb" / total PRB.

falkenber9 commented 3 years ago

Yes, just compute 100 * nof_prb / tot_prb to get the percentage. The total number of PRB is printed to stdout upon cell synchronization after MIB has been decoded:

Example:

Decoded MIB. SFN: 100, offset 0