CB Sounddesign
Controller QA
Operator's manual · v1.0
Operator's manual · v1.0
CB Controller Test Bench
A test rig for a MIDI controller. Play every key, sweep every knob, press every pad, and it tells you which ones are dead, which are noisy, which are stuck and which have lost resolution. Then it writes a report. For buying second-hand gear, for a warranty claim, and for the annual check that a working controller is still working.
Keybed
Coverage, velocity, stuck keys
Controls
Range, resolution, stability
Report
A written record, exportable
How to use it
There is no start button. Connect the controller, select it, and then exercise it: every key, every control through its full travel, every pad several times. The panels fill as you go. The test is your thoroughness, not the tool's.
There is no start button. Connect the controller, select it, and then exercise it: every key, every control through its full travel, every pad several times. The panels fill as you go. The test is your thoroughness, not the tool's.
Chapter one
Setup
Input
The controller. — no devices — means nothing is visible; use Rescan, then Request MIDI access if the permission was never granted.
Web MIDI
Needed, over HTTPS or localhost. Running the offline file directly from a folder may be refused — serve it with python3 -m http.server 8000 and open it from there.
Load map / Save map
A description of the controller's layout so the report names its controls instead of numbering them. generic is the fallback and is perfectly usable.
Reset
Clears all collected data. Do this between two different controllers — otherwise the second one inherits the first one's coverage.
Chapter two
Keybed
What the keys did. Play every key at least twice, once softly and once hard.
Keys hit
Coverage. The figure to read against the controller's key count — anything short means a key did not register, and the note range below says which end.
Note range
The lowest and highest note received. A 61-key controller reporting a 60-note range has one dead key at an end, which is easy to miss by feel.
Velocity distribution
A histogram of how hard you played. A healthy keybed produces a spread; a failing sensor produces a spike at one value, and a dying one produces a spread that does not reach the top.
Min vel / Max vel
The softest and hardest registered. If max never reaches 127 no matter how hard you play, the sensor or the curve is worn.
Note ons / Retriggers
Total presses, and how many fired twice for one press. Any retrigger count above zero is a contact fault and will get worse.
Stuck
Keys currently held with no note-off. A non-zero reading here with your hands off the keyboard is the fault you were looking for.
Chapter three
Continuous controls
Knobs, faders, wheels. Sweep each one slowly from end to end, twice.
Range
The span of values seen. A control that only reaches 118 of 127 has either not been swept fully or has a worn track — sweep it again to find out which.
Resolution
How many distinct values it produced. A 127-step knob showing 40 distinct values is quantising, either by design or by wear.
Steps
The count of value changes. Read with resolution: many steps and few distinct values means the control is jumping back and forth rather than travelling.
Stability / Jitter
Whether a control at rest stays put. Jitter on an untouched control is dirt or a dying pot, and it is the fault most likely to be dismissed as a software problem.
Rest
The value the control sits at when released. For a spring-loaded control this should be identical every time.
Chapter four
Pitch bend and aftertouch
Value / Min / Max
Live position and the extremes reached. Bend should reach both ends and return; a wheel that reaches 0 but not 16383 has a mechanical stop out of alignment.
Rest offset
How far from centre the wheel sits when released. Anything other than zero means every bend you play starts slightly out of tune, which is audible on a held note and almost impossible to diagnose by ear.
Aftertouch type
Channel or polyphonic. Worth confirming: controllers are frequently advertised as having polyphonic aftertouch and turn out to send channel.
Keys touched / resting
Aftertouch coverage, and what it reads at rest. A resting value above zero means the keybed is under constant pressure — a common fault in older gear.
Chapter five
Pads
Pressed / Presses
Which pad is down and how many times each has fired. Hit each pad ten times: a pad that misses one in ten is a pad that will miss on stage.
Bounces
Double triggers from one hit. Like key retriggers, any count above zero is a fault, and unlike key retriggers it is often audible immediately as a flam.
Chapter six
The report
Report
Everything collected, as text, with the device name and the date. Written to be read by someone who was not there — a seller, a repair shop, a warranty department.
What to do first
Reset, then work through the controller methodically in one pass: keys low to high, then every knob and fader, then the wheels, then the pads. A report from a haphazard session shows gaps that look like faults.
Buying second-hand
Take this on a phone or laptop to the seller. Fifteen minutes of methodical exercising produces a document that either confirms the description or does not, and it is far more convincing than an opinion after the fact.
Take this on a phone or laptop to the seller. Fifteen minutes of methodical exercising produces a document that either confirms the description or does not, and it is far more convincing than an opinion after the fact.
CB Sounddesign · Controller Test BenchOperator's manual · v1.0