diff --git a/README.md b/README.md
index 96e345d..3a96d17 100644
--- a/README.md
+++ b/README.md
@@ -3,7 +3,7 @@
 This repository is a demo - it isn't a fully fleshed out system and isn't intended to be. Things might break. It also makes use of some 3rd-party code that we don't have a commercial licence for, so legally this can only be used for demonstration purposes, and only interally.
 
 # Getting started
-
+- Use a debian-based 64-bit system. No way around it; this project is not cross-platform.
 - Install the SDRPlay device driver & install the project dependencies.
   - `./SDRplay_RSP_API-Linux-3.07.1.run`
   - Follow the prompts, there is some audience participation getting the driver installed.
@@ -23,17 +23,67 @@ The NESDR Smart is a good starting place as it has a command line utility alread
 ## SDRPlay radio
 The script `./sdrplay_sample` was built to mimic the interface provided by `rtl_sdr`, and is nearly a slot in replacement but for the SDRPlay hardware. It is used for producing raw I/Q samples from the radio device, which (assuming you tune to a frequency where an FM radio station is available) can be piped into the FM demodulator `fm_demod.py`, which itself outputs raw PCM audio.
 
-To run the whole chain at once, tuning to a radio station frequency, demodulating the IQ samples to PCM and rendering to the system audio sink, run the following:
+To capture raw IQ samples from the device, follow the below example (omitting the comments).
 
 ```bash
-# FM broadcasting is in the VHF Band II space -
-# in Australia this is 87.5 MHz to 108 MHz. You can specify the
-# frequency as either a plan integer (87500000) or using a suffix:
-export station="87.5M"
+# FM broadcasting is in the VHF Band II space - in Australia this is 87.5 MHz to 108 MHz.
+# You can specify the frequency as either a plain integer (87500000) or using a suffix:
+station="87.5M"
 
 # You'll want to replace that number with an actual radio station.
 
-./sdrplay_sample -d sdrplay --format CS16 -f ${station} -s 384k | \
+./sdrplay_sample \
+  -v \                # Enable verbose output.
+  -d sdrplay \        # Connect to first radio of sdrplay driver;
+  --format CS16 \     # Sample format of complex signed 16-bit int;
+  -f ${station} \     # Tune to FM station frequency;
+  -s 384k \           # Sampling rate 384 kHz;
+  -n 4M \             # Collect 4 million samples (~10 seconds);
+  -o output.iq        # Save IQ samples to file.
+
+# To find the formats, frequencies, sampling rates etc that are supported by the
+# radio device, run:
+./sdrplay_sample -v --list-devices
+```
+
+To demodulate the received IQ samples into raw PCM audio, do:
+
+```bash
+./fm_demod.py \
+  -f CS16 \           # Sample format must match IQ samples;
+  -s 384k \           # Sampling rate must match IQ samples;
+  -d 32k \            # Output rate - see notes.
+  < test.iq \         # IQ sample input file,
+  > test.raw          # Write PCM to file.
+
+# The output rate is a bit finicky as it needs to satisfy multiple requirements.
+# - must be an integer divisor of the sampling rate;
+# - must be a supported input format for the system sound card.
+#
+# A good ratio to aim for is for sampling-rate / output-rate = 10. This might not
+# always be possible, but near enough will work. Higher and you're wasting samples
+# and too low you will face audio problems.
+```
+
+Finally, to play the PCM file through the system audio out, use the `sox` utility. Most of the parameters are required to be fixed values, so only the modifiable ones are explained.
+```bash
+sox \
+  -t raw \
+  -r 32000 \          # Sample rate (must match demodulator output rate)
+  -b 16 \
+  -c 2 \
+  -L \
+  -e signed-integer \
+  test.raw \          # Input audio file
+  -d                  # Output to default audio device
+```
+
+To run the whole chain in realtime without using intermediary files, tuning to a radio station frequency, demodulating the IQ samples to PCM and rendering to the system audio sink, run the following:
+
+```bash
+./sdrplay_sample -d sdrplay --format CS16 -f ${station} -s 384k -o - | \
   ./fm_demod.py -f CS16 -s 384k -d 32k | \
   sox -t raw -r 32000 -b 16 -c 2 -L -e signed-integer - -d
-```
\ No newline at end of file
+```
+
+There's an outstanding bug where running live causes a buffer under-run in alsa - still working on that.
\ No newline at end of file
diff --git a/sdrplay_sample b/sdrplay_sample
index e1cfb64..f782c05 100755
--- a/sdrplay_sample
+++ b/sdrplay_sample
@@ -164,15 +164,14 @@ if args.d:
 
         if args.verbose:
             print('Rx:', result.ret, result.flags, result.timeNs, file=sys.stderr)
-        if result.ret != samples:
-            continue
 
         if result.ret < 1:
             print('Stream read failed, exiting.', file=sys.stderr)
             keep_going = False
 
         if outfile:
-            outfile.write(struct.pack(PACKINGS[format] % len(buffer), *buffer))
+            received = int(result.ret * 2)
+            outfile.write(struct.pack(PACKINGS[format] % received, *buffer[:received]))
 
         if sample_count > 0:
             total_samples += result.ret