diff --git a/amodem/common.py b/amodem/common.py
index e771428..ef0581c 100644
--- a/amodem/common.py
+++ b/amodem/common.py
@@ -8,15 +8,6 @@ log = logging.getLogger(__name__)
 scaling = 32000.0  # out of 2**15
 SATURATION_THRESHOLD = (2**15 - 1) / scaling
 
-LENGTH_FORMAT = '<I'
-
-
-def to_bits(bytes_list):
-    for val in bytes_list:
-        for i in range(8):
-            mask = 1 << i
-            yield (1 if (val & mask) else 0)
-
 
 class SaturationError(ValueError):
     pass
diff --git a/amodem/framing.py b/amodem/framing.py
index a57a826..62b291a 100644
--- a/amodem/framing.py
+++ b/amodem/framing.py
@@ -1,47 +1,105 @@
 ''' Reed-Solomon CODEC. '''
-from reedsolo import rs_encode_msg, rs_correct_msg
-
 from . import common
+import bitarray
 
+import functools
+import itertools
+import binascii
+import struct
 import logging
 log = logging.getLogger(__name__)
 
-DEFAULT_NSYM = 10
-BLOCK_SIZE = 255
+_crc32 = lambda x, mask: binascii.crc32(x) & mask
+# (so the result will be unsigned on Python 2/3)
 
 
-def end_of_stream(size):
-    return bytearray([BLOCK_SIZE]) + b'\x00' * size
+class Checksum(object):
+    fmt = '>L'  # unsigned longs (32-bit)
+    size = struct.calcsize(fmt)
+    func = functools.partial(_crc32, mask=0xFFFFFFFF)
+
+    def encode(self, payload):
+        checksum = self.func(payload)
+        return struct.pack(self.fmt, checksum) + payload
+
+    def decode(self, data):
+        received, = struct.unpack(self.fmt, data[:self.size])
+        payload = data[self.size:]
+        expected = self.func(payload)
+        if received != expected:
+            raise ValueError('invalid checksum')
+        return payload
 
 
-def encode(data, nsym=DEFAULT_NSYM):
-    chunk_size = BLOCK_SIZE - nsym - 1
+class Framer(object):
+    block_size = 1024
+    prefix_fmt = '>L'
+    prefix_len = struct.calcsize(prefix_fmt)
+    checksum = Checksum()
 
-    for _, chunk in common.iterate(data=data, size=chunk_size,
-                                   func=bytearray, truncate=False):
-        size = len(chunk)
-        if size < chunk_size:
-            padding = [0] * (chunk_size - size)
-            chunk.extend(padding)
+    EOF = b''
 
-        block = bytearray([size]) + chunk
-        yield rs_encode_msg(block, nsym)
+    def _pack(self, block):
+        frame = self.checksum.encode(block)
+        return struct.pack(self.prefix_fmt, len(frame)) + frame
 
-    yield rs_encode_msg(end_of_stream(chunk_size), nsym)
+    def encode(self, data):
+        for _, block in common.iterate(data=data, size=self.block_size,
+                                       func=bytearray, truncate=False):
+            yield self._pack(block=block)
+        yield self._pack(block=self.EOF)
+
+    def decode(self, data):
+        data = iter(data)
+        while True:
+            length, = self._take_fmt(data, self.prefix_fmt)
+            frame = self._take_len(data, length)
+            block = self.checksum.decode(frame)
+            if block == self.EOF:
+                return
+
+            yield block
+
+    def _take_fmt(self, data, fmt):
+        length = struct.calcsize(fmt)
+        chunk = bytearray(itertools.islice(data, length))
+        if len(chunk) < length:
+            raise StopIteration()
+        return struct.unpack(fmt, chunk)
+
+    def _take_len(self, data, length):
+        chunk = bytearray(itertools.islice(data, length))
+        if len(chunk) < length:
+            raise StopIteration()
+        return chunk
 
 
-def decode(blocks, nsym=DEFAULT_NSYM):
+def chain_wrapper(func):
+    @functools.wraps(func)
+    def wrapped(*args, **kwargs):
+        result = func(*args, **kwargs)
+        return itertools.chain.from_iterable(result)
+    return wrapped
 
-    last_chunk = end_of_stream(BLOCK_SIZE - nsym - 1)
-    for block in blocks:
-        assert len(block) == BLOCK_SIZE
-        chunk = bytearray(rs_correct_msg(block, nsym))
-        if chunk == last_chunk:
-            log.info('EOF encountered')
-            return  # end of stream
 
-        size = chunk[0]
-        payload = chunk[1:]
-        assert size <= len(payload)
+@chain_wrapper
+def encode(data, framer=None):
+    framer = framer or Framer()
+    for frame in framer.encode(data):
+        bits = bitarray.bitarray(endian='little')
+        bits.frombytes(bytes(frame))
+        yield bits
 
-        yield payload[:size]
+
+@chain_wrapper
+def _to_bytes(bits, block_size=1):
+    for _, chunk in common.iterate(data=bits, size=8*block_size,
+                                   func=lambda x: x, truncate=True):
+        yield bitarray.bitarray(chunk, endian='little').tobytes()
+
+
+@chain_wrapper
+def decode(bits, framer=None):
+    framer = framer or Framer()
+    for frame in framer.decode(_to_bytes(bits)):
+        yield frame
diff --git a/amodem/recv.py b/amodem/recv.py
index b0e7285..22dde3c 100644
--- a/amodem/recv.py
+++ b/amodem/recv.py
@@ -224,15 +224,16 @@ class Receiver(object):
         filt = self._train(sampler, order=11, lookahead=5)
         sampler.equalizer = lambda x: list(filt(x))
 
-        data_bits = self._demodulate(sampler, freqs)
-        self.bits = itertools.chain.from_iterable(data_bits)
+        bitstream = self._demodulate(sampler, freqs)
+        self.bitstream = itertools.chain.from_iterable(bitstream)
 
-    def decode(self, output):
-        chunks = framing.decode(_blocks(self.bits))
+    def run(self, output):
+        data = framing.decode(self.bitstream)
         self.size = 0
-        for chunk in chunks:
+        for _, chunk in common.iterate(data=data, size=256,
+                                       truncate=False, func=bytearray):
             output.write(chunk)
-            self.size = self.size + len(chunk)
+            self.size += len(chunk)
 
     def report(self):
         if self.stats:
@@ -265,15 +266,6 @@ class Receiver(object):
         self.plt.title(title)
 
 
-def _blocks(bits):
-    while True:
-        block = bitarray.bitarray(endian='little')
-        block.extend(itertools.islice(bits, 8 * framing.BLOCK_SIZE))
-        if not block:
-            break
-        yield bytearray(block.tobytes())
-
-
 def izip(streams):
     iters = [iter(s) for s in streams]
     while True:
@@ -296,7 +288,7 @@ def main(args):
     try:
         signal, amplitude = detect(signal, config.Fc)
         receiver.start(signal, modem.freqs, gain=1.0/amplitude)
-        receiver.decode(args.output)
+        receiver.run(args.output)
         success = True
     except Exception:
         log.exception('Decoding failed')
diff --git a/amodem/send.py b/amodem/send.py
index 7ec890e..fc3e329 100644
--- a/amodem/send.py
+++ b/amodem/send.py
@@ -47,15 +47,13 @@ class Writer(object):
         self.write(silence)
 
     def modulate(self, bits):
+        padding = [0] * modem.bits_per_baud
+        bits = itertools.chain(bits, padding)
         symbols_iter = modem.qam.encode(bits)
-        symbols_iter = itertools.chain(symbols_iter, itertools.repeat(0))
         carriers = modem.carriers / config.Nfreq
-        while True:
-            symbols = itertools.islice(symbols_iter, config.Nfreq)
+        for _, symbols in common.iterate(symbols_iter, size=config.Nfreq):
             symbols = np.array(list(symbols))
             self.write(np.dot(symbols, carriers))
-            if all(symbols == 0):  # EOF marker
-                break
 
 
 def main(args):
@@ -73,9 +71,10 @@ def main(args):
     log.info('%.3f seconds of training audio', training_duration)
 
     reader = stream.Reader(args.input, bufsize=(64 << 10), eof=True)
-    data = itertools.chain.from_iterable(reader)
-    encoded = itertools.chain.from_iterable(framing.encode(data))
-    writer.modulate(bits=common.to_bits(encoded))
+    data = list(itertools.chain.from_iterable(reader))
+    bits = list(framing.encode(data))
+    data_ = list(framing.decode(bits))
+    writer.modulate(bits=bits)
 
     data_size = writer.offset - training_size
     log.info('%.3f seconds of data audio, for %.3f kB of data',
diff --git a/tests/test_framing.py b/tests/test_framing.py
index 197c63b..29f3bf1 100644
--- a/tests/test_framing.py
+++ b/tests/test_framing.py
@@ -1,32 +1,35 @@
 from amodem import framing
 import random
 import itertools
-import reedsolo
+
+import pytest
 
 
-def concat(chunks):
-    return bytearray(itertools.chain.from_iterable(chunks))
+def concat(iterable):
+    return bytearray(itertools.chain.from_iterable(iterable))
+
+r = random.Random(0)
+blob = bytearray(r.randrange(0, 256) for i in range(64 * 1024))
 
 
-def test_random():
-    r = random.Random(0)
-    x = bytearray(r.randrange(0, 256) for i in range(64 * 1024))
-    y = framing.encode(x)
-    x_ = concat(framing.decode(y))
-    assert x_ == x
+@pytest.fixture(params=[b'', b'abc', b'1234567890', blob, blob[:12345]])
+def data(request):
+    return request.param
 
 
-def test_errors():
-    data = bytearray(range(244))
-    blocks = list(framing.encode(data))
-    assert len(blocks) == 2
-    for i in range(framing.DEFAULT_NSYM // 2):
-        blocks[0][i] = blocks[0][i] ^ 0xFF
+def test_checksum(data):
+    c = framing.Checksum()
+    assert c.decode(c.encode(data)) == data
 
-    i = framing.DEFAULT_NSYM // 2
-    try:
-        blocks[0][i] = blocks[0][i] ^ 0xFF
-        concat(framing.decode(blocks))
-        assert False
-    except reedsolo.ReedSolomonError as e:
-        assert e.args == ('Too many errors to correct',)
+
+def test_framer(data):
+    f = framing.Framer()
+    encoded = concat(f.encode(data))
+    decoded = concat(f.decode(encoded))
+    assert decoded == data
+
+
+def test_main(data):
+    encoded = framing.encode(data)
+    decoded = framing.decode(encoded)
+    assert bytearray(decoded) == data
diff --git a/tests/test_full.py b/tests/test_full.py
index b27a065..545df35 100644
--- a/tests/test_full.py
+++ b/tests/test_full.py
@@ -51,8 +51,13 @@ def run(size, chan=None, df=0, success=True):
         assert rx_data == tx_data
 
 
-def test_small():
-    run(1024, chan=lambda x: x)
+@pytest.fixture(params=[0, 1, 3, 10, 42, 123])
+def small_size(request):
+    return request.param
+
+
+def test_small(small_size):
+    run(small_size, chan=lambda x: x)
 
 
 def test_error():
diff --git a/tests/test_recv.py b/tests/test_recv.py
index 296a9ad..d2dca80 100644
--- a/tests/test_recv.py
+++ b/tests/test_recv.py
@@ -51,7 +51,3 @@ def test_find_start():
         start = recv.find_start(buf, length*config.Nsym)
         expected = offset + len(prefix)
         assert expected == start
-
-
-def test_blocks():
-    assert list(recv._blocks([])) == []