2016-07-20 08:59:39 +08:00

189 lines
6.1 KiB
Python
Executable File

##
## This file is part of the libsigrokdecode project.
##
## Copyright (C) 2012 Joel Holdsworth <joel@airwebreathe.org.uk>
##
## This program is free software; you can redistribute it and/or modify
## it under the terms of the GNU General Public License as published by
## the Free Software Foundation; either version 2 of the License, or
## (at your option) any later version.
##
## This program is distributed in the hope that it will be useful,
## but WITHOUT ANY WARRANTY; without even the implied warranty of
## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
## GNU General Public License for more details.
##
## You should have received a copy of the GNU General Public License
## along with this program; if not, write to the Free Software
## Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
##
import sigrokdecode as srd
'''
OUTPUT_PYTHON format:
Packet:
[<ptype>, <pdata>]
<ptype>, <pdata>:
- 'DATA', [<channel>, <value>]
<channel>: 'L' or 'R'
<value>: integer
'''
class SamplerateError(Exception):
pass
class Decoder(srd.Decoder):
api_version = 2
id = 'i2s'
name = 'I²S'
longname = 'Integrated Interchip Sound'
desc = 'Serial bus for connecting digital audio devices.'
license = 'gplv2+'
inputs = ['logic']
outputs = ['i2s']
channels = (
{'id': 'sck', 'name': 'SCK', 'desc': 'Bit clock line'},
{'id': 'ws', 'name': 'WS', 'desc': 'Word select line'},
{'id': 'sd', 'name': 'SD', 'desc': 'Serial data line'},
)
annotations = (
('left', 'Left channel'),
('right', 'Right channel'),
('warnings', 'Warnings'),
)
binary = (
('wav', 'WAV file'),
)
def __init__(self):
self.samplerate = None
self.oldsck = 1
self.oldws = 1
self.bitcount = 0
self.data = 0
self.samplesreceived = 0
self.first_sample = None
self.ss_block = None
self.wordlength = -1
self.wrote_wav_header = False
def start(self):
self.out_python = self.register(srd.OUTPUT_PYTHON)
self.out_binary = self.register(srd.OUTPUT_BINARY)
self.out_ann = self.register(srd.OUTPUT_ANN)
def metadata(self, key, value):
if key == srd.SRD_CONF_SAMPLERATE:
self.samplerate = value
def putpb(self, data):
self.put(self.ss_block, self.samplenum, self.out_python, data)
def putbin(self, data):
self.put(self.ss_block, self.samplenum, self.out_binary, data)
def putb(self, data):
self.put(self.ss_block, self.samplenum, self.out_ann, data)
def report(self):
# Calculate the sample rate.
samplerate = '?'
if self.ss_block is not None and \
self.first_sample is not None and \
self.ss_block > self.first_sample:
samplerate = '%d' % (self.samplesreceived *
self.samplerate / (self.ss_block -
self.first_sample))
return 'I²S: %d %d-bit samples received at %sHz' % \
(self.samplesreceived, self.wordlength, samplerate)
def wav_header(self):
# Chunk descriptor
h = b'RIFF'
h += b'\x24\x80\x00\x00' # Chunk size (2084)
h += b'WAVE'
# Fmt subchunk
h += b'fmt '
h += b'\x10\x00\x00\x00' # Subchunk size (16 bytes)
h += b'\x01\x00' # Audio format (0x0001 == PCM)
h += b'\x02\x00' # Number of channels (2)
h += b'\x80\x3e\x00\x00' # Samplerate (16000)
h += b'\x00\x7d\x00\x00' # Byterate (32000)
h += b'\x04\x00' # Blockalign (4)
h += b'\x10\x00' # Bits per sample (16)
# Data subchunk
h += b'data'
h += b'\xff\xff\x00\x00' # Subchunk size (65535 bytes) TODO
return h
def wav_sample(self, sample):
# TODO: This currently assumes U32 samples, and converts to S16.
s = sample >> 16
if s >= 0x8000:
s -= 0x10000
lo, hi = s & 0xff, (s >> 8) & 0xff
return bytes([lo, hi])
def decode(self, ss, es, data):
if not self.samplerate:
raise SamplerateError('Cannot decode without samplerate.')
for self.samplenum, (sck, ws, sd) in data:
data.itercnt += 1
# Ignore sample if the bit clock hasn't changed.
if sck == self.oldsck:
continue
self.oldsck = sck
if sck == 0: # Ignore the falling clock edge.
continue
self.data = (self.data << 1) | sd
self.bitcount += 1
# This was not the LSB unless WS has flipped.
if ws == self.oldws:
continue
# Only submit the sample, if we received the beginning of it.
if self.ss_block is not None:
if not self.wrote_wav_header:
self.put(0, 0, self.out_binary, [0, self.wav_header()])
self.wrote_wav_header = True
self.samplesreceived += 1
idx = 0 if self.oldws else 1
c1 = 'Left channel' if self.oldws else 'Right channel'
c2 = 'Left' if self.oldws else 'Right'
c3 = 'L' if self.oldws else 'R'
v = '%08x' % self.data
self.putpb(['DATA', [c3, self.data]])
self.putb([idx, ['%s: %s' % (c1, v), '%s: %s' % (c2, v),
'%s: %s' % (c3, v), c3]])
#self.putbin([0, self.wav_sample(self.data)])
# Check that the data word was the correct length.
if self.wordlength != -1 and self.wordlength != self.bitcount:
self.putb([2, ['Received %d-bit word, expected %d-bit '
'word' % (self.bitcount, self.wordlength)]])
self.wordlength = self.bitcount
# Reset decoder state.
self.data = 0
self.bitcount = 0
self.ss_block = self.samplenum
# Save the first sample position.
if self.first_sample is None:
self.first_sample = self.samplenum
self.oldws = ws