Monitoring Automated
Temperature, pH and conductivity logged around the clock, plus every outlet state and power draw.
The journey
Building the Dutch 90P
This is not a finished showcase. It is the complete record of the build: the setup, the mistakes, the measurements, what changed, and what comes next.
Earlier scapes lived mostly in memory
This aquarium has been torn down and rebuilt several times. Each version taught me something, but almost none of it was recorded well enough to carry forward. Equipment choices, plant failures, lighting changes and maintenance decisions survived mostly as impressions — and a handful of photographs.
Both of those are hardscape layouts. The wood and the stone set the composition, and the plants are grown around a structure that already exists. Years of building them is the reason the current tank is what it is — and the reason it is a departure.
A Dutch layout has no wood and no stone. There is nothing holding the composition up except plant groups, their colours and the discipline of trimming them into shape. That change is what made the light start to matter enough to measure: hardscape does not care whether the ends of the tank are dim, and twenty plant streets do.
The fourth rescape is different in one other way. This time the tank keeps a record.
Building the Dutch 90P
The current aquarium is a twenty-species Dutch-inspired street layout in a 90P. It was planted on fresh ADA Amazonia over Power Sand and started with the usual problem of an active substrate: the tank looked finished long before the water was stable.
The early weeks were dominated by ammonia testing, heavy water changes, plant adaptation and the slow work of establishing shape.
The layout on planting dayPhotograph to be added
Planting dayEvery species and the light measured at it · The measured ammonia curve · The tank now
Pearl began as a record before it became automation
Pearl is the software behind the tank. It reads the controller, probes, camera and maintenance history, and turns them into a record I can inspect later.
Continuous logging came first. Automation came afterward.
That order matters. Automating a system before understanding what it is doing only makes mistakes repeat more consistently.
It is worth being plain about what that picture shows, because it does not match how the tank runs today. Pearl held the CO₂ loop for a while and it worked. It was still handed back to a fixed schedule: on water with almost no carbonate hardness the pH signal swings further than the dissolved CO₂ behind it, and a steady window turned out to beat a responsive one. A screenshot of a system doing something it no longer does is exactly the kind of thing a record is for.
How each system is built · Why the CO₂ loop is open · Neptune Apex
I bought a PAR meter to confirm the light was dying
For roughly six years I assumed the original ADA Solar RGB was becoming too weak. Red plants near the ends of the aquarium were not colouring as expected, while the fixture had already survived several rescapes and daily use.
I expected the meter to confirm that the light needed replacing.
Twenty underwater readings gave me a different answer.
The center still produced strong readings. Several outer plant positions received far less. On this aquarium, under this mounting configuration, the limiting issue was distribution rather than center output.
50lowest readingµmol·m⁻²·s⁻¹
176highest readingµmol·m⁻²·s⁻¹
20positions measured
Raising the light made the outer readings worse
The obvious no-cost fix was to raise the fixture and widen its footprint. I tried it.
The outer readings decreased, so the fixture went back to its previous height.
The comparison was not preserved as a complete standalone survey, which means it remains an observation that changed a decision rather than a fully documented result. It will be repeated under the current logging standard.
Built, manual and still planned
Running automatically5
Lighting Automated
Photoperiod held by the controller rather than a timer on the fixture, 9:30am – 4:30pm at full output.
CO₂ Automated
Solenoid driven by the controller on a fixed daily window, with a pH floor, a permitted time window and a leak sensor enforced in hardware. Nobody opens a valve by hand.
Top-off Automated
Automatic top-off replaces evaporation, so conductivity drift reflects the tank rather than the water line.
Camera Automated
A frame captured every hour the light is on, with a vision model reviewing a few of them a day for algae and plant health.
Still done manually3
Dosing By hand
2Hr Aquarist APT, dosed by hand. The Red Sea ReefDose is on the tank but not yet set up to drive it.
Water changes By hand
RODI mixed and changed by hand, roughly 20 gallons at a time, logged against the conductivity trace.
PAR By hand
Measured by hand with an Apogee MQ-510 at twenty repeatable positions.
Planned2
CO₂ under pH control Planned
Built and run once: Pearl held injection against a target drop off each morning's degassed baseline. It is currently switched off in favour of the fixed schedule.
Automated dosing Planned
Drive the ReefDose on a fixed daily schedule so fertilisation stops being a variable.
Current chapter
Current
Grow-in, first structural trims and stabilizing the routine
The planting is filling in, the system is producing a usable record, and the next work is less about adding hardware than making the routine repeatable. Dosing still needs to become consistent, future PAR surveys need complete depth metadata, and the developing plant streets need to be shaped rather than simply allowed to grow.
Working on now
- Establish the Dutch street shapes through repeated trimming
- Make fertilisation repeatable
- Repeat the mounting-height comparison as a preserved survey
- Add canopy-level PAR as the stems mature
Next actions
- Document the next major trim
- Publish a dated growth comparison
- Configure or deliberately defer automated dosing
- Repeat PAR measurements with per-position depth
Planned work. None of this is finished, and it is listed here so the difference is visible.
Journal4
Dated entries on what happened and what it changed.
What is actually automated, and what is notThe tank has a controller, a Raspberry Pi, probes and a camera. That does not mean every routine runs itself.Why the fourth rescape has a recordThe previous three left photographs and memories. This one is being documented well enough that its decisions can survive the next teardown.Raising the light did not fix the endsThe free solution was the first one worth testing. It failed, and the incomplete record changed how future surveys are logged.Twenty readings changed the question I was asking about the lightI expected an old fixture with weak output. What I measured was a strong center and uneven distribution.