diff --git a/README.md b/README.md
index 29e1e5d..26b3c92 100644
--- a/README.md
+++ b/README.md
@@ -8,3 +8,58 @@
 - Also requires **NumPy** and **SciPy** if you want to do analog range calibration
 - Having **Pandas** is useful for wrapping capture results for further processing
 
+## Longer Notes
+
+### Why have I written this?
+
+BitScope helpfully provide applications and libraries for talking to their USB
+capture devices, so one can just use those. I wrote this code because I want
+to be able to grab the raw data and further process it in various ways. I'm
+accustomed to working at a Python prompt with various toolkits like SciPy and
+Matplotlib, so I wanted a simple way to just grab a trace as an array.
+
+The BitScope library is pretty simple to use, but also requires you to
+understand a fair amount about how the scope works and make a bunch of decisions
+about what capture mode and rate to use. I want to just specify a time period,
+voltage range and a rough number of samples and have all of that worked out for
+me, same way as I'd use an actual oscilloscope: twiddle the knobs and look at
+the trace.
+
+Of course, I could have wrapped the BitScope library in something that would do
+this, but after reading a bit about how the scope works I was fascinated with
+understanding it further and so I decided to go back to first principles and
+start with just talking to it with a serial library. This code thus serves as
+(sort of) documentation for the VM registers, capture modes and how to use them.
+It also has the advantage of being pure Python.
+
+The code prefers the highest capture resolution possible and will do the mapping
+from high/low/trigger voltages to the mysterious magic numbers that the device
+needs. It can also do logic and mixed-signal capture.
+
+In addition to capturing, the code can also generate waveforms at arbitrary
+frequencies – something that is tricky to do as the device operates at specific
+frequencies and so one has to massage the width of the waveform buffer to get a
+frequency outside of these. It can also control the clock generator.
+
+I've gone for an underlying async design as it makes it easy to integrate the
+code into UI programs or network servers – both of which interest me as the end
+purpose for this code. However, for shell use there are synchronous wrapper
+functions. Of particular note is that the synchronous wrapper understands
+keyboard interrupt and will cancel a capture returning the trace around the
+cancel point. This is useful if your trigger doesn't fire and you want to
+understand why.
+
+### Where's the documentation, mate?
+
+Yeah, yeah. I know.
+
+### Also, I see no unit tests...
+
+It's pretty hard to do unit tests for a physical device. That's my excuse and
+I'm sticking to it.
+
+### Long lines and ignoring E221, eh?
+
+"A foolish consistency is the hobgoblin of little minds"
+
+Also, I haven't used an 80 character wide terminal in this century.
diff --git a/test.py b/test.py
index 6f7e5bc..8a2ccf3 100644
--- a/test.py
+++ b/test.py
@@ -1,26 +1,12 @@
 
-import numpy as np
 from pylab import figure, plot, show
 
 from analysis import annotate_series
 from scope import await_, capture, main
-from utils import DotDict
 
 
 await_(main())
-
-# o = 400
-# m = 5
-# n = o * m
-# samples = square_wave(o)
-# samples = np.hstack([samples] * m) * 2
-# samples = np.hstack([samples[100:], samples[:100]])
-# samples += np.random.normal(size=n) * 0.1
-# samples += np.linspace(4.5, 5.5, n)
-# series = DotDict(samples=samples, sample_rate=1000000)
-
-data = capture(['A'], period=20e-3, nsamples=2000)
-series = data.A
+series = capture(['A'], period=20e-3, nsamples=2000).A
 
 figure(1)
 plot(series.timestamps, series.samples)
@@ -35,6 +21,7 @@ if annotate_series(series):
         print(f"Found {waveform.frequency:.0f}Hz {waveform.shape} wave, "
               f"with amplitude ±{waveform.amplitude:.2f}V and offset {waveform.offset:.2f}V")
 
-    plot(waveform.timestamps + waveform.capture_start - series.capture_start, waveform.samples * waveform.amplitude + waveform.offset)
+    plot(waveform.timestamps + waveform.capture_start - series.capture_start,
+         waveform.samples * waveform.amplitude + waveform.offset)
 
 show()