Reference guide · Solar PV monitoring hardware

Find the broken link. Then send the truck.

A vendor-neutral troubleshooting guide for the hardware behind a solar monitoring portal: LTE modems, data loggers, RS-485/Modbus wiring, meters and CTs, inverters, single-axis trackers and weather sensors. Written for remote technical support work and interview practice.

About this pageReference guide with sample scenarios for support practice. It is study material, not a record of field work. Device names and values are generic or fictional; real behavior varies by model, so always follow the manufacturer’s documentation and the site’s safety procedures.

12 sections ~38 min read 34 fault cards 12 animated diagrams 6 interactive fault simulators 12-question quiz

Pick any stop, or scroll in order. Progress is saved only in this browser.

Comms or value?First decide whether data is missing or wrong. That one question halves the search.

One card per faultSymptom, likely causes, ordered checks, customer wording, escalation triggers and the evidence to attach.

Safety is the boundaryRemote support guides looking and photographing. Anything inside an energized enclosure is for qualified people.

Start here: follow the data chain

ORIENTATION · ~3 min

A monitoring portal is the last stop of a long chain. Support finds the broken link by walking that chain backwards, from the portal to the device.

A solar monitoring site is a small data network bolted onto a power plant. Field devices (inverters, meters, tracker controllers, weather sensors) answer questions on a field bus. A data logger asks those questions on a schedule, stamps each answer with a time, stores it, and uploads it through a cellular modem or the customer’s network. The cloud stores the data and the portal draws charts and raises alerts.

When a customer says “my system shows zero”, that is a symptom seen at the very end of the chain. The cause can sit at any link: the device, the bus, the logger, the backhaul, or the cloud side. Good remote support names the link before anyone drives to site.

Every link in this chain can fail on its own. The portal only shows the end result.

What support can usually see remotely

Last check-in time of the logger and of each device. A single timestamp often tells you which link stopped.
Logger health: uptime, reboot count, storage use, clock, upload errors.
Modem metrics: signal quality (RSRP, RSRQ, SINR), registration state, data usage.
Device data: live register values, status/fault codes, and the shape of daily curves.

What usually needs a person on site

Reading LEDs and displays, taking photos, confirming that a breaker or switch is in the expected position.
Anything inside an electrical enclosure: that is work for a qualified technician, never for the customer on the phone.

Field device

Inverter, meter, tracker controller or sensor. Holds values in registers.

Field bus

Usually RS-485 with Modbus RTU, or Ethernet with Modbus TCP.

Data logger

Polls devices, timestamps, buffers (store & forward), uploads.

Backhaul

LTE modem or site LAN that carries uploads to the cloud.

Portal

Charts, alerts, reports. Shows only what arrived.

Store & forward

If the link drops, the logger keeps data and backfills later.

Interview line“I trace the data chain from the portal back to the device. The last good timestamp at each link usually tells me where to look first.”

Next: Safety →

Safety first: customer-side checks only

NON-NEGOTIABLE · ~3 min

Remote support can guide looking, photographing and safe power cycles. Anything inside an energized enclosure belongs to a qualified, licensed person following lockout/tagout.

Solar sites are unusual: PV modules produce voltage whenever light hits them. Turning off an AC breaker does not make the DC side safe, and inverters can hold stored energy in capacitors for minutes after shutdown. Trackers can move on their own. A support call must never turn a customer into an unqualified electrician.

Do not ask anyone to open an energized enclosure, remove covers, touch terminals or re-land wires.
Customer-side checks only: read LEDs and displays through windows or from the outside, take photos, read labels, confirm a known external switch or plug-in power supply.
Lockout/tagout (LOTO) and work inside panels are for qualified technicians and licensed electricians, following the site’s procedures and local code.
Open CT secondary danger: a current-output CT (for example, a 5 A secondary type) with its secondary opened while primary current flows can develop dangerously high voltage. Only qualified people handle CT wiring, and they short or properly terminate the secondary first.
Trackers: stay clear of rows and pinch points; a row can start moving at any time (tracking, stow commands).
Rooftops and ladders: fall hazards are real. A photo from the ground is often good enough.

If in doubtStop, document what you see, and escalate. A truck roll costs money; an injury cannot be undone.

Interview line“My checks stay on the customer side of the enclosure door. If the next step is inside a panel, I document the evidence and hand it to a qualified technician.”

Next: Triage →

Master triage: comms or value?

METHOD · ~4 min

Two questions settle most tickets fast. Is the data missing, or present but wrong? Is it the whole site, or one device?

Comms problem: data stops, goes stale, or arrives late. Value problem: data arrives on time but the number is wrong, negative, scaled, or shifted. The fixes are completely different, so decide first.

Then check scope. If every device on the site stopped at the same minute, look at what they share: site power, the logger, the modem, the network. If one device stopped while its neighbors keep reporting, look at that device, its wire and its settings.

Walk the layers in order: power → link → protocol → configuration → data validation. Lower layers are cheaper to prove and break more often. A register map is worth checking only after you know the device is powered and talking.

Left: data is missing or stale → walk the layers. Right: data arrives but looks wrong → check meaning before hardware.

Pattern → most likely layer

Pattern in the data	Most likely layer	First check
All devices stale at the same minute	Logger, backhaul or site power	Logger last check-in, modem signal, outage notices
One device stale, others fine	That device’s power, wire or settings	Modbus timeouts for that ID; photo of its display
Data back-fills after a gap	Backhaul (store & forward worked)	Modem history; usually no device fault
Values negative	CT direction or sign setting	Compare meter with inverter total
Values exactly 10×, 100× or ½	Scaling or CT ratio	Scale factor, CT ratio setting
Curve shifted by one hour	Clock, time zone or DST	Logger clock and NTP status
Flat line that is not zero	Stale or frozen value	Register timestamp, device comms

Interview line“First I decide whether data is missing or wrong, then whether it is site-wide or one device. Then I walk power, link, protocol and configuration, and I finish by validating the value against a reference.”

Next: Modem →

Cellular modem / LTE gateway

BACKHAUL · ~5 min

The modem must see a tower, be allowed on the network, open a data session on the right APN, and reach the cloud. Each step fails differently.

Think of a cellular connection as four gates: radio (can it hear a tower?), registration (does the carrier accept this SIM?), data session (did it get an IP address on the correct APN?), and reachability (can it reach the cloud endpoint?). Modems usually expose enough status to tell which gate is closed.

Generic behavior only. Real LED colors and blink codes differ by model — always confirm in that model’s manual.

Signal metrics (LTE rules of thumb)

Metric	What it measures	Good	Fair	Poor
RSRP	Received strength of the reference signal	> −90 dBm	−90 to −105	< −105 dBm
RSRQ	Quality: signal vs. load and interference	> −10 dB	−10 to −15	< −15 dB
SINR	Signal vs. interference and noise	> 13 dB	0 to 13	< 0 dB

Thresholds vary by carrier, band and modem. Trends matter more than one reading: a sudden drop suggests something physical changed.

Generic LED meanings

Power: off = no supply; solid = powered.
Network/Status: blinking = searching or registering; solid = registered. An error color often means SIM or registration trouble.
Signal bars: rough RSRP indicator. Fine for a phone call with the customer, not for a ticket.
Data/Activity: flickers when traffic moves. Registered but never flickering suggests no data session.

Power cycling safely

One controlled power cycle, done through a known external switch or plug-in adapter, can clear a stuck session. Record the time, wait for a full boot (often several minutes), then check again. Do not cycle repeatedly, and never during a firmware update. Repeated cycling hides evidence and can corrupt storage.

Fault cards 4

Each card: symptom → likely causes → checks in order → what to tell the customer → when to escalate → evidence to attach.

4.1No signal or weak signalRSRP/RSRQ poor, drops at certain times

Symptom

Logger goes offline for hours, or flaps on and off. Modem history shows RSRP below about −105 dBm or SINR below 0 dB. Data usually back-fills when the link returns.

Likely causes

External antenna cable loose, damaged or water in the connector
Antenna inside a metal enclosure or moved during other work
Carrier tower maintenance or an outage in the area
New obstruction (building, equipment, vegetation)
Band or technology retired by the carrier (for example a 3G shutdown)

Checks in order

Read signal history: sudden drop (physical change) or slow decline (environment)?
Check carrier outage notices for the area and time
Ask for photos of the antenna, its cable and connector, from outside the enclosure
Compare with other sites nearby on the same carrier
If the antenna is external and loose, the customer may hand-tighten it; recheck metrics after 10 minutes

What to tell the customer suggested wording

“The monitoring link has a weak cellular signal, so uploads are delayed. Your system is still producing, and stored data should fill in when the link recovers. We may need a technician to check the antenna.”

When to escalate

Signal stays poor after the antenna is confirmed secure
Physical damage, water, or a cable that needs replacing
A carrier band or technology retirement requires new hardware

Evidence to attach

RSRP, RSRQ, SINR history with timestamps
Photos of antenna and connector
Carrier outage reference, if any
Model and firmware of the modem

4.2SIM inactive or data plan problemregistration denied or data stops mid-month

Symptom

Signal looks fine but the modem never registers, or it worked until a certain date and then stopped completely. Sometimes it returns on the first of the month.

Likely causes

SIM suspended (billing, contract end) or never activated
Data cap reached, or a firmware/log upload used more data than planned
SIM moved to another device and the carrier locks it to the old IMEI
SIM not fully seated after maintenance

Checks in order

Look up the SIM by ICCID in the SIM management portal: active, suspended, usage
Compare usage with the plan and look for a sudden spike
Check the modem’s reported error: “SIM not detected” vs. “registration denied”
Confirm nobody swapped hardware or SIMs recently

What to tell the customer suggested wording

“The connection is blocked at the carrier account level, not by your equipment. We are checking the SIM plan and will update you once it is active again.”

When to escalate

SIM account changes needing billing or carrier access
“SIM not detected” after confirming the account is active (hardware)

Evidence to attach

ICCID and IMEI
SIM status and usage screenshot
Modem error text and timestamp

4.3Wrong APN: registered but no datasignal fine, network LED solid, no traffic

Symptom

Modem is registered with good signal, but it has no IP address, or it has one and nothing reaches the cloud. Often starts right after a carrier change or a configuration push.

Likely causes

APN typo, or the APN still points to the old carrier
Private APN requires credentials that are missing
A private network requires routing or a firewall rule that was not added
DNS servers missing, so the cloud hostname never resolves

Checks in order

Read the modem’s data session state and IP address
Compare the configured APN with the carrier’s record for this SIM
Check change history: was a configuration pushed or the SIM swapped?
If an IP exists, test whether the modem can resolve and reach the cloud endpoint

What to tell the customer suggested wording

“Your modem connects to the carrier, but a network setting is preventing data from getting through. We can usually correct this remotely.”

When to escalate

Private APN or VPN routing needs the network team
The configuration cannot be changed remotely

Evidence to attach

Configured APN vs. expected APN
Session state, IP address, DNS results
Time of the last configuration change

4.4Modem keeps rebootinguptime keeps resetting, link flaps

Symptom

Logger shows short online bursts. Modem uptime is always a few minutes. Problems may be worse on hot afternoons.

Likely causes

Power supply undersized or failing; voltage dips when the radio transmits
Loose power terminal
Enclosure overheating
Firmware bug or a watchdog reboot after failed connections

Checks in order

Read uptime and reboot reasons from the modem event log
Correlate reboots with time of day and temperature
Ask for a photo of the power supply label and LEDs
Check for a known issue with the current firmware version

What to tell the customer suggested wording

“The modem is restarting by itself, which interrupts uploads. Your production is not affected. We are narrowing down whether it is power, heat or software.”

When to escalate

Power or terminal checks inside the enclosure (qualified tech)
A firmware rollback or update is needed

Evidence to attach

Reboot timestamps and reasons
Supply voltage readings, if available
Enclosure temperature data
Firmware version

Interview line“I check the four gates in order: radio, registration, data session, reachability. Signal numbers and registration state usually tell me which one is closed before I ask anyone to touch the hardware.”

Next: Logger →

Data logger offline

SITE BRAIN · ~5 min

“Offline” in the portal means “no upload received.” The logger may be dead, or alive and unable to send. Find out which.

The logger has four jobs: poll devices on a schedule, timestamp every reading, buffer data locally, and upload it. Because of store and forward, a short backhaul outage often heals itself: data back-fills when the link returns. That is good news to share with a customer.

Many loggers report health metrics that make triage quick: uptime and reboot count, supply voltage, storage use, clock and NTP status, WAN address, Modbus poll statistics and the last upload result. Read them before suggesting any reset.

Key ideaIf the modem is online but the logger is not uploading, the logger (or its LAN link to the modem) is the suspect. If both are silent at the same minute, suspect site power or the modem first.

Fault cards 5

Each card: symptom → likely causes → checks in order → what to tell the customer → when to escalate → evidence to attach.

5.1No power or failing power supplycompletely silent, no LEDs

Symptom

Logger and everything powered from the same supply go silent at the same time. Customer reports no lights, or a dim, flickering power LED.

Likely causes

Upstream breaker tripped or switched off during other work
Power supply failed, often after heat or a surge
Loose supply terminal
Voltage sag: the supply works, but not under load

Checks in order

Check whether the modem went silent at the same minute (shared power?)
Ask for a photo of the logger LEDs and the power supply LED
Ask the customer to confirm, without opening anything, whether a labeled external switch or breaker is on
Check history for low supply voltage before the outage

What to tell the customer suggested wording

“The monitoring equipment appears to have lost power. This affects data only; your solar system can keep producing. A technician may need to check the power supply.”

When to escalate

No LEDs and no external switch to verify (inside-enclosure work)
Repeated supply failures (likely heat or surge cause)

Evidence to attach

Last check-in time of the logger and the modem
LED photos
Supply voltage trend, if logged

5.2Storage full (SD card or flash)uploads stall, old data, file errors

Symptom

Logger is online but data stops updating, or uploads arrive hours late. Health shows storage near 100% or file system errors.

Likely causes

Long backhaul outage filled the buffer
Debug logging left on
Worn or corrupted SD card
Retention set too long for the storage size

Checks in order

Read storage use and the buffer queue size
Look for recent long outages that explain the backlog
Check whether debug logging is enabled
Check for file system or write errors in the event log

What to tell the customer suggested wording

“The logger’s local storage is full, so it is struggling to keep and send new data. We will clear space or schedule a card replacement. Some older readings may be lost if the buffer overflowed.”

When to escalate

Card replacement or reformat needs a site visit
Signs of corruption; data recovery is needed

Evidence to attach

Storage %, queue size, event log lines
Outage history that caused the backlog

5.3Clock drift / NTP failurecurves shifted, data rejected, TLS errors

Symptom

Daily curves are shifted (production “before sunrise”), readings land in the wrong interval, or uploads fail with certificate/TLS errors.

Likely causes

NTP blocked by a firewall or unreachable
Real-time clock battery depleted, so the clock resets after a power loss
Wrong time zone or DST setting
Clock so far off that certificates look not yet valid or expired

Checks in order

Compare the logger clock with real time
Check NTP status and which server it uses
Look for a recent power loss (RTC reset)
Compare the curve with sunrise and sunset times for the site

What to tell the customer suggested wording

“The logger’s clock is off, so data is being filed at the wrong time. We are correcting the time source; past data may need to be reprocessed.”

When to escalate

Historical data needs re-timestamping
Firewall changes on the customer network are needed for NTP

Evidence to attach

Logger time vs. reference time
NTP status
Example interval showing the shift

5.4Reboot loop or firmware problemuptime minutes, repeated resets

Symptom

Logger appears for a few minutes, then disappears again. Started after a firmware update or a power event.

Likely causes

Firmware update interrupted or incompatible
Corrupted configuration file
Watchdog resets triggered by a hung process
Weak supply causing brown-out resets

Checks in order

Read reboot count and reasons
Check firmware version and the date of the last update
Check supply voltage at boot
Check for a known issue in release notes

What to tell the customer suggested wording

“The logger is restarting repeatedly. We are checking whether a software update or a power issue is the cause, and we can often fix software remotely.”

When to escalate

Firmware recovery that cannot be done remotely
Hardware failure suspected

Evidence to attach

Reboot timestamps
Firmware version and update time
Supply voltage at reset

5.5Network configuration: DHCP vs. staticlogger alive, no route out

Symptom

Logger is powered and polls devices, but uploads fail. Often starts right after the customer replaced a router or changed the network.

Likely causes

Router replaced: new subnet, so the static IP is now wrong
IP conflict with another device
Static gateway or DNS wrong
DHCP lease lost: the logger falls back to a link-local 169.254.x.x address
Outbound firewall blocks the upload port

Checks in order

Read the logger’s IP, mask, gateway and DNS
Ask whether any network equipment changed and when
Look for a 169.254.x.x address (no DHCP)
Confirm the required outbound ports and hosts with the customer’s IT

What to tell the customer suggested wording

“Your monitoring device is fine, but a network change is preventing it from reaching the internet. We will send you the settings your IT team needs.”

When to escalate

The customer’s IT must change firewall or DHCP reservations
A site visit is required to reset the static configuration

Evidence to attach

Current network settings vs. expected
Date of the network change
Upload error text

Interview line“Offline only means no upload. I check whether the logger is alive and buffering or truly down, using uptime, storage, clock and the last upload error.”

Next: RS-485 →

RS-485 / Modbus RTU bus

FIELD BUS · ~7 min

One twisted pair, one master, many devices. Most RS-485 problems are wiring and settings, not broken hardware.

RS-485 is the electrical layer: a differential pair (two wires carrying opposite signals) that tolerates noise and long runs. Modbus RTU is the language spoken over it: the master (client) sends a request to one device ID, and only that device answers. Only one device may talk at a time (half-duplex).

Simulation for study. Toggle a fault and watch which replies disappear or turn into errors.

Rules that keep a bus healthy

Daisy-chain topology: the cable goes device to device in a line. Avoid star wiring and long stubs; each branch causes reflections.
Termination: one 120 Ω resistor across A/B at each physical end of the trunk. Exactly two, not zero and not one per device.
Biasing (fail-safe): weak pull-up/pull-down resistors hold the line in a known idle state. Usually applied once, often at the master.
Unique IDs: every device needs its own Modbus address (1–247). Address 0 is broadcast.
Same serial settings everywhere: baud rate, data bits, parity and stop bits must match (for example 19200 8E1).
Cable: shielded twisted pair rated for RS-485, around 120 Ω characteristic impedance. Runs up to roughly 1,200 m (4,000 ft) are possible at lower baud rates; higher baud rates need shorter runs.
Shield and reference: ground the shield at one end only to avoid ground loops, and carry a common/reference conductor where the devices need one.
Load: a standard transceiver supports 32 unit loads; fractional-load transceivers allow more devices.

A/B naming trapVendors do not agree on which wire is “A” and which is “B”. Some use D+/D− instead. A new device that never answers while everything else works is a classic polarity suspect. Any rewiring is done by a qualified technician.

Reading Modbus symptoms

Symptom	Usually means
Timeout (no reply)	Wrong ID, wrong baud/parity, polarity swapped, device off, broken wire
CRC / framing errors	Electrical noise, missing or extra termination, no biasing, two devices answering
Exception reply	Device is talking, but the register address or function is not supported (map problem)
Intermittent, worse at midday	Noise rising with inverter power, cable next to power conductors

Fault cards 6

Each card: symptom → likely causes → checks in order → what to tell the customer → when to escalate → evidence to attach.

6.1A/B polarity swappedone new device never answers

Symptom

After a device is installed or replaced, that one device times out on every poll. Every other device on the bus is fine. If the swap is at the master, the whole bus is silent.

Likely causes

A and B landed backwards at that device
Vendor labels differ (A/B vs. D+/D−), so the installer followed the wrong convention
A replacement unit uses the opposite labeling from the old one

Checks in order

Confirm the device is powered (display or LED photo)
Confirm its ID and serial settings match the bus
Check the date: did it start right after installation or replacement?
Compare the vendor’s wiring label on both devices
Hand over to a qualified tech to swap A/B at that device and retest

What to tell the customer suggested wording

“The new device is powered but cannot talk on the monitoring cable, most likely because two wires are reversed. A technician can correct this quickly on site.”

When to escalate

Always: rewiring requires a qualified technician on site

Evidence to attach

Poll statistics showing timeouts for that ID only
Install or replacement date
Photo of the terminal labels (from documentation or a tech)

6.2Termination missing or extraCRC errors, worse on long runs

Symptom

Intermittent CRC or framing errors on several devices, worse at higher baud rates or on the farthest devices. Sometimes everything works at 9600 but fails at 19200.

Likely causes

No 120 Ω resistor at one or both ends of the trunk
Termination switched on at every device, loading the line
A device was added beyond the old end of the line, but the terminator stayed in place

Checks in order

Read error counters per device: are far devices worse?
Check the as-built drawing: which devices are physically first and last?
Check whether devices were added recently
Ask a qualified tech to verify exactly two terminators at the two ends

What to tell the customer suggested wording

“The data cable needs a small electrical adjustment at its ends so signals stop echoing. It does not affect power production.”

When to escalate

Physical changes to termination need a site visit

Evidence to attach

CRC/framing counts per device
Bus layout or as-built drawing
Recent changes to the device list

6.3Missing biasinggarbage while idle, random framing errors

Symptom

Errors appear even when the bus should be quiet, or the first reply after a pause is corrupted. Behavior changes when a particular device is unplugged.

Likely causes

No fail-safe biasing anywhere on the bus
Biasing applied at several points, which skews the levels
The device that provided biasing was replaced with one that does not

Checks in order

Check the logger or master documentation for its biasing setting
Ask whether a device was replaced recently
Compare error rates while idle and while busy

What to tell the customer suggested wording

“The monitoring cable needs a setting that keeps it steady between messages. We will arrange the change with a technician.”

When to escalate

Hardware jumper or resistor changes on site

Evidence to attach

Error counters vs. time
Device replacement history

6.4Duplicate slave IDsvalues flip, CRC errors on one ID

Symptom

One ID shows values that jump between two different devices, or the reply for that ID is corrupted. Another device seems to be missing from the list.

Likely causes

A replacement device kept its factory default ID
Two installers chose the same address
A device was reset to default after a firmware update

Checks in order

Compare expected IDs with responding IDs
Look for values that alternate between two plausible numbers
Ask for a photo of the communication settings screen on each suspect device
Change one device’s ID (on site or via the device menu), then retest

What to tell the customer suggested wording

“Two devices are answering to the same address, so their data mixes. A quick settings change on one of them will fix it.”

When to escalate

If the device menu requires a site visit or a vendor tool

Evidence to attach

ID list (expected vs. seen)
Example of flipping values
Photos of the settings screens

6.5Baud or parity mismatchdevice after a reset never answers

Symptom

A device that worked for months goes silent after a firmware update, a power event or a factory reset. Its display shows it is running normally.

Likely causes

Device reset to default serial settings (for example 9600 8N1) while the bus runs 19200 8E1
Parity or stop bits changed during configuration
Installer used a different standard than the rest of the site

Checks in order

Check whether the device had a recent reset or update
Ask for a photo of the device’s communication settings
Compare with the documented bus settings
Set the device to match the bus (do not change the whole bus)

What to tell the customer suggested wording

“One device lost its communication settings, probably after a reset. Once they match the rest of the site again, its data will return.”

When to escalate

Settings require a technician or vendor software

Evidence to attach

Documented bus settings
Photo of the device settings
Date of the reset or update

6.6Topology, cable length or noiseintermittent errors that follow production

Symptom

Errors rise around midday and fall at night, or appear only on certain devices. The bus may have been extended over the years.

Likely causes

Star wiring or long stubs
Cable run in the same conduit as AC power or next to drives
Shield grounded at both ends (ground loop) or not at all
Run too long for the baud rate
Damaged cable (rodents, water, UV)

Checks in order

Correlate error counts with production and time of day
Review the as-built cable route and length
Try a lower baud rate as a test, if the site allows it
Have a qualified tech inspect the routing, shield bonding and cable condition

What to tell the customer suggested wording

“The monitoring cable picks up electrical noise when your system is working hardest. The fix is a cabling improvement, scheduled with a technician.”

When to escalate

Re-routing, re-cabling or shield changes

Evidence to attach

Error rate vs. production chart
Cable length and route
Baud rate in use

Interview line“Timeouts point me to addressing, serial settings, polarity or power. CRC errors point me to the electrical side: termination, biasing, noise, or two devices answering at once.”

Next: Meter/CT →

Revenue / production meter and CTs

MEASUREMENT · ~6 min

Meters multiply voltage and current. If the current transformer (CT) is backwards, on the wrong phase, or set to the wrong ratio, the meter calculates a confident, wrong answer.

A meter measures voltage directly and current through CTs clamped around the conductors. For each phase it calculates power: roughly voltage × current × power factor. That only works if each CT is on the same phase as its voltage reference, faces the right direction, and has the correct ratio configured.

kW is power right now (a speedometer). kWh is energy over time (an odometer). A 100 kW system running for 3 hours at full output produces about 300 kWh.
Revenue-grade meters meet a stricter accuracy class (for example, ANSI C12.20 classes in North America) and are used for billing or incentives. Production meters may be less strict.
CT ratio (for example 400:5) tells the meter how to scale the small secondary signal. Voltage-output CTs (for example 333 mV) are rated by primary current instead.
Interval data is usually energy per 5 or 15 minutes. Comparing an interval kWh to a kW nameplate without converting units creates false alarms.

Sample numbers. The pattern (sign flip, a clean ratio, a strange power factor) points to the cause.

Three-phase arithmetic worth remembering

If one of three CTs is reversed on a balanced system, the total is +P +P −P = P: the meter shows about one third of the real output. If a CT is paired with the wrong phase’s voltage (a 120° shift), that phase can read about −½ of its real value at unity power factor, with a strange power factor.

Open CT secondaryNever ask anyone to disconnect CT leads. A current-output CT with an open secondary under load can produce dangerous voltage. CT work is for qualified people who short or terminate the secondary first.

Fault cards 4

Each card: symptom → likely causes → checks in order → what to tell the customer → when to escalate → evidence to attach.

7.1CT reversed: negative kWproduction shows as consumption

Symptom

The meter shows negative production, or production “importing”. On three-phase sites one phase may be negative and the total about one third of the inverter sum.

Likely causes

CT installed with its arrow facing the wrong way
CT leads swapped at the meter terminals
Meter configured with the wrong direction or polarity

Checks in order

Compare meter kW with the sum of the inverter outputs at the same time
Check each phase separately: is one negative?
Check whether the meter software has a per-phase polarity setting
Check install or service history for recent CT work

What to tell the customer suggested wording

“Your system is producing normally. The meter is reading the direction backwards, so the numbers look negative. This is a measurement setting or installation fix.”

When to escalate

A physical CT correction is needed (qualified electrician)
Billing or incentive data must be corrected

Evidence to attach

Per-phase kW and PF screenshot
Inverter total at the same timestamp
Install date and service history

7.2Wrong CT ratio configuredreadings off by a clean factor

Symptom

Meter readings are consistently half, double, or another clean factor of what the inverters report, at every time of day.

Likely causes

Ratio entered for a different CT size (for example 200:5 instead of 400:5)
CT replaced with a different size without updating the meter
Voltage-output CT rating entered incorrectly

Checks in order

Compute meter kW divided by inverter kW over several hours; is the ratio constant?
Compare the configured ratio with the CT nameplate (photo from documentation or a tech)
Check the change history of the meter configuration

What to tell the customer suggested wording

“The meter is set for a different sensor size than the one installed, so it scales every reading by the same wrong amount. A configuration change will fix new data.”

When to escalate

Historical data must be corrected for billing
Nameplate verification requires a site visit

Evidence to attach

Ratio calculation table
Configured ratio vs. nameplate
Date of the last CT or meter change

7.3Phase mismatchodd power factor, low or negative phase

Symptom

One phase shows a power factor far from what the site normally runs, with low or negative power. The error changes with load instead of staying a clean ratio.

Likely causes

CT on L1 but the voltage reference for that channel taken from L2 or L3
CT channels swapped at the meter after maintenance
Voltage taps landed in a different order than the CTs

Checks in order

Check per-phase power factor: unity-PF sites should read close to 1.0 on all phases
Compare per-phase power with per-phase inverter output if available
Review the wiring diagram vs. the meter channel configuration
Ask a qualified tech to verify CT and voltage phase pairing

What to tell the customer suggested wording

“The meter is matching sensors to the wrong electrical phase, so its math is off. Production is fine; the reading needs correcting on site.”

When to escalate

Always: phase verification and correction are qualified electrician work

Evidence to attach

Per-phase V, A, kW and PF
Wiring diagram reference
Recent maintenance notes

7.4kWh vs. kW confusion“600 from a 100 kW system?”

Symptom

Customer reports impossible or worrying values that are actually correct numbers in different units, or Wh shown as kWh (a 1000× error).

Likely causes

Comparing daily energy (kWh) with system size (kW)
Interval energy shown where power was expected
Device register reports Wh, but the portal labels it kWh

Checks in order

Restate both numbers with units
Check the register units in the device documentation
Check whether the value is cumulative, per interval or instantaneous

What to tell the customer suggested wording

“The numbers you see are energy over the day, not instant power, so a higher value is expected. Here is how to read the two charts.”

When to escalate

A real unit or scaling error in the portal configuration

Evidence to attach

The values with units and timestamps
Register definition from the device map

Interview line“With meter tickets I look for the pattern: a sign flip suggests CT direction, a clean ratio like half or double suggests CT ratio, and an odd power factor suggests a phase mismatch. Then I confirm against inverter totals.”

Next: Inverter →

Inverter communication

DEVICE DATA · ~6 min

The inverter’s own display and the portal can disagree. The portal only knows what the logger reads from specific registers.

An inverter has two faces: its local display (status, fault codes, live power) and its data interface (RS-485 Modbus, Modbus TCP, or a plug-in comms card). The logger reads chosen registers on a schedule. If the registers, scaling or comms are wrong, the portal shows something different from the display.

Sample curves, not real site data. Dashed teal = expected curve.

Register map basics

Holding vs. input registers: different Modbus functions read them. Asking the wrong type returns an exception or wrong data.
Off-by-one: documentation often says “40001”, which many tools send as address 0. Reading one register too high gives a neighbor’s value.
32-bit values span two 16-bit registers. If the word order is wrong, numbers look huge or random.
Scaling: some maps store 1234 meaning 123.4 V, or provide a separate scale-factor register (value × 10^SF). Ignoring it gives clean 10× or 100× errors.
Sentinel values like 65535 or 32768 often mean “not available”, not a real measurement.

Night-time zero vs. real outage

Zero at night is normal. Many inverters power their electronics from the DC side, so they also stop answering after sunset: “offline at night” can be expected behavior. A real outage shows during daylight, while irradiance is high and neighboring inverters produce.

Fault cards 5

Each card: symptom → likely causes → checks in order → what to tell the customer → when to escalate → evidence to attach.

8.1Local fault code, portal shows zero or flatdisplay and portal disagree

Symptom

The customer sees a fault code on the inverter display, but the portal shows zero or a flat line and no alarm. Or the portal shows power while the display shows a fault.

Likely causes

Inverter really tripped; power is zero but the fault/status register is not mapped, so no alarm is raised
Comms froze, so the portal repeats the last value (flat but not zero)
Display and portal are reporting different times

Checks in order

Ask for a photo of the display, including the fault code and the time shown
Compare inverter output with the site meter and irradiance at that time
Check whether status and fault registers are in the logger’s map
Check the timestamp of the last fresh reading for that inverter

What to tell the customer suggested wording

“Your inverter reported a fault on its own screen. We are confirming whether it stopped producing and making sure the monitoring raises an alert next time.”

When to escalate

A confirmed inverter trip (vendor service or O&M)
Repeated faults of the same code
Missing alarm mapping (monitoring engineering)

Evidence to attach

Photo of the display with code and time
Inverter, meter and irradiance at the same timestamps
Logger map excerpt for status registers

8.2Comms card or interface failureinverter producing, data missing

Symptom

The site meter shows normal production, but one inverter’s data is missing or timing out. Often after a firmware update, a lightning storm, or work inside the inverter.

Likely causes

Comms card failed or not seated
Comms card firmware does not match the inverter firmware
Interface port disabled in the inverter menu
Surge damage on the RS-485 or Ethernet port

Checks in order

Confirm production with the meter (the inverter is working)
Check the timeouts or link state for that inverter only
Ask for a photo of the comms card LED, if visible without opening
Check the dates of recent firmware updates or storms

What to tell the customer suggested wording

“The inverter is producing power; only its monitoring connection is down. Your energy is still being counted by the site meter.”

When to escalate

Card replacement or reseating (inside the inverter, qualified tech)
Vendor firmware support

Evidence to attach

Meter vs. inverter data
Timeout logs
Firmware versions and event dates

8.3Wrong register map or offsetnonsense values after a swap

Symptom

After an inverter replacement or firmware change, values look absurd: voltage reads 0 or 6553.5, power shows a frequency-like number, energy jumps backwards.

Likely causes

New model or firmware uses a different map
Register address off by one (1-based vs. 0-based)
Wrong function code (holding vs. input)
Wrong word order for 32-bit values

Checks in order

Confirm the exact model and firmware of the new unit
Compare a few live register values with the inverter display (voltage, frequency, power)
Look for values that match a neighbor register (offset clue)
Look for sentinel values (65535, 32768)

What to tell the customer suggested wording

“The monitoring is reading the new inverter with the old inverter’s map. We will update the profile, and the readings will be correct from then on.”

When to escalate

A new device profile must be built or approved
Historical data needs cleanup

Evidence to attach

Model, firmware, replacement date
Side-by-side: display value vs. raw register

8.4Scaling factor errorclean 10×, 100× or 1000× error

Symptom

Power or energy is off by an exact power of ten. Shapes look right; magnitude is wrong.

Likely causes

Scale-factor register ignored or read from the wrong address
Unit mismatch: W vs. kW, Wh vs. kWh
Scale changed in a firmware update

Checks in order

Divide portal value by expected value; is it a clean power of ten?
Check the map’s scale factor and unit for that register
Compare with the nameplate: can this inverter even produce that much?

What to tell the customer suggested wording

“The data is being multiplied by the wrong factor in the monitoring setup. Your production is normal; we will correct the display.”

When to escalate

Profile change needed in the monitoring platform

Evidence to attach

Ratio calculation
Map excerpt showing units/scale

8.5Night zero vs. real outageis it dark, or is it down?

Symptom

Alert says “inverter offline” or “zero production”. The customer is worried.

Likely causes

Normal night shutdown (expected)
Early morning wake-up threshold not reached (cloudy, snow on modules)
Real daytime trip or grid outage
Comms loss during daylight

Checks in order

Check the timestamp vs. local sunrise and sunset
Check irradiance: is there enough light?
Check neighbor inverters and the site meter
Check whether the utility reported a grid outage

What to tell the customer suggested wording

“Your inverters shut down at night as designed. Production resumed this morning as expected. (Or: we confirmed a daytime stop and have opened a service ticket.)”

When to escalate

Daytime zero with good irradiance while neighbors produce

Evidence to attach

Time vs. sunrise/sunset
Irradiance and neighbor data
Grid outage reference

Interview line“Before blaming the inverter, I check whether the portal is reading the right registers with the right scaling, and whether the zero is at night or in daylight next to producing neighbors.”

Next: Trackers →

Solar trackers (single-axis)

MOTION · ~8 min

Trackers turn rows of modules east to west through the day. When one row stops following the sun, it leaves a clear fingerprint in the production curve.

A single-axis tracker rotates rows of modules around a north–south torque tube, facing east in the morning and west in the afternoon. Typical rotation limits are roughly ±50° to ±60°, depending on the design.

The parts (generic)

Row controller (often called a TCU): computes the target angle and drives the motor.
Motor / actuator / slew drive: turns the torque tube.
Inclinometer (angle sensor): tells the controller the real angle so it can close the loop.
Limit switches or software limits: stop the row at the ends of its range.
Power: AC-fed, or self-powered by a small dedicated PV panel and a battery.
Network controller (often called an NCU): talks to many rows over a wireless mesh (Zigbee-type or proprietary radio) or RS-485, and connects to the site network.
Weather inputs: wind sensors and sometimes snow or flood sensors that trigger protective stow.

How it decides where to point

Astronomical tracking: sun position calculated from time, date and GPS location. A wrong clock means a wrong angle.
Backtracking: early and late in the day, rows turn back toward flat so they do not shade the row behind them. It depends on row spacing (ground coverage ratio) and terrain slope settings.
Stow: a protective position. Wind stow is often flat or a vendor-specified angle; snow stow is often steep so snow slides off. Stow is a safety feature, not a fault.

Watch the ends of the day: rows turn back toward flat (backtracking) so they do not shade the row behind. Simplified geometry for study; real curves also depend on season, clouds, clipping and terrain.

How a stuck row looks in the data

Curve shape vs. neighbors	Likely situation
Narrow bell, lost morning and evening shoulders	Row stuck flat (behaves like a fixed-tilt array)
Early peak, steep afternoon fall	Row stuck facing east
Late peak, weak morning	Row stuck facing west
Dips only early or late, on certain rows	Row-to-row shading: backtracking misconfigured
Whole site narrow on a windy afternoon	Wind stow (check wind data first)
Row starts late every morning in winter	Self-powered controller battery drained overnight

Compare a row with its neighbors on the same inverter or combiner, on the same day. Weather affects everyone; a stuck row affects one.

Fault cards 7

Each card: symptom → likely causes → checks in order → what to tell the customer → when to escalate → evidence to attach.

9.1Tracker controller offlinerow or block missing from the portal

Symptom

One row, a group of rows, or every row on a network controller shows “offline” or a stale angle. Rows may still be tracking on their own.

Likely causes

Network controller lost power or its uplink to the site network
Wireless mesh link weak: a row too far from its neighbors, an obstruction, or a radio fault
RS-485 tracker bus problem (see the RS-485 section)
Self-powered row shut its radio down to save battery

Checks in order

Scope: one row, one block, or all rows on one network controller?
Check the production curve: is the row still tracking even though comms are down?
Check the network controller’s own status and uplink
Check battery or supply voltage history for self-powered rows
Look for changes on site: new equipment, vehicles, construction

What to tell the customer suggested wording

“We have lost monitoring contact with some tracker rows. Production data suggests they are still following the sun. We are checking the tracker network.”

When to escalate

Rows not tracking and not reachable
Network controller hardware fault
Repeated mesh drops on the same rows

Evidence to attach

List of affected rows and controller
Last-seen times
Production curves of affected vs. neighbor rows

9.2Stow mode activerows flat or steep for hours

Symptom

All rows, or one block, are held at the stow angle. Production is lower than expected for the irradiance.

Likely causes

High wind detected (expected protection)
Snow stow active
Stow command left active after an event or after maintenance
Faulty wind sensor reading high, triggering stow on calm days

Checks in order

Check wind and weather data for that time
Check the stow reason or source in the tracker system
Compare the wind sensor with a nearby weather source
Check whether maintenance requested a manual stow

What to tell the customer suggested wording

“Your trackers are in a protective position because of high wind. This is designed behavior and lowers production temporarily. (Or: the wind sensor seems faulty; we are checking it.)”

When to escalate

Stow active with calm conditions confirmed
Rows do not return to tracking after the event clears

Evidence to attach

Wind speed data
Stow reason/source and time
Production impact estimate

9.3Motor or actuator faultrow stuck at one angle, error code

Symptom

One row stays at a fixed angle while its neighbors move. The controller may report overcurrent, motor timeout or a stall.

Likely causes

Mechanical bind: debris, damaged bearing, ice
Motor or gearbox failure
Overcurrent protection tripped
Fuse or supply issue at the row

Checks in order

Compare reported angle with neighbors over the day
Read the row’s fault code and history
Check whether the fault started after a storm or freeze
Ask for a photo of the row from a safe distance

What to tell the customer suggested wording

“One tracker row has stopped moving. It still produces power, but less than its neighbors. A technician needs to inspect it.”

When to escalate

Always: mechanical and electrical repairs need a qualified tech on site

Evidence to attach

Fault code and time
Angle history vs. neighbors
Production loss estimate
Photo, if taken safely

9.4Inclinometer / angle sensor errorreported angle does not match reality

Symptom

The reported angle looks normal but the curve says otherwise, or the row hunts back and forth, or it stops early saying it reached its target.

Likely causes

Sensor drift or calibration offset
Sensor loose on its bracket
Sensor failed and reading a fixed value
Wrong sensor orientation after replacement

Checks in order

Compare the reported angle with a photo taken at a known time
Look for a constant offset vs. neighbors
Look for a reading that never changes
Check replacement history

What to tell the customer suggested wording

“The row’s angle sensor seems to be giving wrong information, so the row does not point where it should. A technician will check and recalibrate it.”

When to escalate

Recalibration or sensor replacement on site

Evidence to attach

Angle history
Photo with timestamp
Neighbor comparison

9.5Backtracking misconfigureddips early and late on some rows

Symptom

Morning and evening production dips on certain rows while noon looks fine. More visible in winter when the sun is low.

Likely causes

Wrong row spacing or ground coverage ratio in the settings
Terrain slope not accounted for
Backtracking disabled after a software update
Wrong clock or location, so the sun position is off

Checks in order

Check whether the dips align with low sun angles
Compare rows on slopes vs. flat ground
Verify controller time and location settings
Check recent software changes

What to tell the customer suggested wording

“Some rows are shading each other early and late in the day. The fix is usually a settings adjustment by the tracker team.”

When to escalate

Parameter changes need tracker engineering approval

Evidence to attach

Curves showing timed dips
Affected row list
Controller time and location settings

9.6Self-powered controller battery drainrow offline overnight, starts late

Symptom

A row disappears at night or early morning and comes back late in the day, mostly in winter or after cloudy days. It may miss stow commands.

Likely causes

Battery aging and losing capacity
Charging panel dirty, shaded or covered with snow
Charge controller fault
Extra radio traffic draining power

Checks in order

Read battery voltage history: does it sag every night?
Compare wake-up time with neighbors
Ask for a photo of the charging panel
Check the battery age

What to tell the customer suggested wording

“One tracker controller is running low on its own battery, so it wakes up late. Cleaning its small panel or replacing the battery usually solves it.”

When to escalate

Battery or charge controller replacement
Row missed a stow command (safety relevance)

Evidence to attach

Battery voltage trend
Wake-up times vs. neighbors
Charging panel photo

9.7Limit switch faultrow stops short or reports limit hit

Symptom

A row stops before its normal range, or the controller reports “limit reached” at an angle that is clearly not the end.

Likely causes

Limit switch stuck, damaged or misaligned
Wiring fault reads as a permanently triggered limit
Software limits set incorrectly

Checks in order

Compare the maximum angle reached with neighbors
Read the limit status flags
Check recent mechanical work on the row

What to tell the customer suggested wording

“The row thinks it reached the end of its travel too early, so it stops short. A technician will inspect the stop switch.”

When to escalate

Always: on-site inspection

Evidence to attach

Max angle vs. neighbors
Limit flags and fault codes

Interview line“For trackers I compare the row’s curve with its neighbors. A narrow bell means stuck flat, a skewed curve means stuck east or west, and a site-wide change on a windy day is usually stow, not a fault.”

Next: Weather →

Weather station and irradiance sensors

REFERENCE DATA · ~3 min

Expected energy is calculated from irradiance. If the sensor is wrong, every performance number is wrong with it.

GHI (global horizontal irradiance) is measured by a sensor lying flat. POA (plane of array) is measured in the same plane as the modules; on a tracker site the POA sensor rides on a tracker row.
Pyranometers need a clean dome, correct leveling or alignment, no shading from the mast, and the right calibration constant.
Temperature: ambient sensors sit in a radiation shield; back-of-module sensors are glued to a module and can fall off.
Wind sensors may also drive tracker stow, so a bad wind reading can cost production.

Sample curves for a clear day on a tracker site.

Fault cards 3

Each card: symptom → likely causes → checks in order → what to tell the customer → when to escalate → evidence to attach.

10.1Irradiance reads lowperformance looks too good

Symptom

Performance ratio rises above what is plausible, or irradiance reads several percent low every day with a normal shape.

Likely causes

Soiled dome (dust, bird droppings)
Wrong calibration constant entered
Shading from a mast or new equipment at certain hours

Checks in order

Compare with a nearby reference or satellite estimate
Look for a constant percentage error (soiling, calibration) vs. a timed dip (shading)
Ask for a photo of the sensor from a safe position
Check the calibration constant against the sensor certificate

What to tell the customer suggested wording

“The sunlight sensor seems to read slightly low, which makes performance look better than it is. Cleaning or recalibration will fix it.”

When to escalate

Cleaning or recalibration on site

Evidence to attach

Comparison chart
Sensor photo
Calibration constant vs. certificate

10.2POA vs. GHI mix-up or misalignmentPOA looks like GHI

Symptom

The “POA” channel looks like a horizontal sensor: narrow curve, much lower in the morning and evening than expected on a tracker site.

Likely causes

POA sensor mounted flat or on a fixed bracket
Channels swapped in the logger configuration
Sensor knocked out of alignment

Checks in order

Compare the POA and GHI curve shapes
Check channel mapping in the logger
Ask for a photo of the mount

What to tell the customer suggested wording

“One of the weather sensors is labeled or mounted differently than expected. We will correct the setup so expected energy is calculated properly.”

When to escalate

Remounting on site

Evidence to attach

POA vs. GHI chart
Mount photo
Channel map

10.3Temperature sensor faultmodule temperature equals ambient, or jumps

Symptom

Back-of-module temperature tracks ambient air too closely, reads a fixed value, or jumps when it is windy.

Likely causes

Sensor detached from the module
Broken cable or connector
Ambient sensor shield damaged, so it reads high in the sun

Checks in order

Compare module and ambient temperatures at midday (module should be warmer in the sun)
Look for fixed or impossible values
Ask for a photo

What to tell the customer suggested wording

“One temperature sensor is giving unreliable readings. It does not affect production, only how we calculate expected output.”

When to escalate

Re-attachment or replacement on site

Evidence to attach

Temperature chart
Photo

Interview line“Before I trust a performance ratio, I sanity-check the irradiance sensor: POA vs. GHI, the shape of the curve, and whether it reads low every day by the same percentage.”

Next: Ticket →

Ticket and escalation template

HANDOFF · ~3 min

A good escalation lets the next person act without calling you. Evidence, scope, steps tried, and one clear ask.

Sample escalation (fictional site, practice only)

# SAMPLE — fictional site and values, written for practice
Title:     [Site: Sample-Solar-07] Row 14 tracker stuck flat since 2026-10-03
Priority:  P3 · production impact ~2% of site (est.)
Scope:     Single row (row 14 of 40), network controller NCU-2
Symptom:   Row 14 curve is a narrow bell; neighbors 13 and 15 normal.
           Reported angle fixed at 0.4° since 2026-10-03 09:12 (site time, UTC−7).
Ruled out: Wind stow (wind < 15 km/h); comms OK (row reports every 5 min);
           clock and location verified on NCU-2.
Evidence:  1) curves, rows 13/14/15, 2026-10-03 to 10-05   2) angle history
           3) fault code "MOTOR_TIMEOUT" at 09:12   4) customer photo, 10-05
Customer:  Informed 2026-10-05 14:20 (UTC−7): "row still producing, reduced output".
Ask:       Field tech inspection of row 14 motor/drive. LOTO required.
Owner:     Tier 2 support → O&M dispatch

Good ticket

Times with time zone
Scope stated (one row, one site)
What was ruled out and how
Evidence attached, numbered
One clear ask and owner

Weak ticket

“Tracker broken, please check”
No timestamps
No scope or neighbors compared
Screenshots without context
Several asks mixed together

Who owns what (generic tiers)

Support (remote)

Triage, customer communication, safe customer-side checks, evidence.

Remote

Senior support / NOC

Config changes, device profiles, deeper log review.

Remote

ENG

Engineering

Firmware bugs, map/profile builds, platform issues.

Remote

FLD

Field / O&M technician

Anything physical: wiring, CTs, antennas inside enclosures, tracker repairs.

On site

VND

Vendor / carrier

RMA, warranty, SIM and carrier issues.

External

Interview line“When I escalate, I include scope, timestamps, what I ruled out, numbered evidence and one clear ask. The next person should not need to call me.”

Next: Quiz →

Check yourself: 12 questions

PRACTICE · ~5 min

Pick an answer to see the explanation. Your score stays in this browser only.

0 / 12 correct · 0 answered

1Every device on a site went stale at 14:02. What do you check first?

2A modem shows good signal and a solid network LED, but no IP address. Most likely cause?

3Which LTE readings indicate poor signal?

4Where should 120 Ω termination resistors be on an RS-485 trunk?

5A newly replaced device never answers, but every other device on the bus works. Which is a classic suspect?

6Modbus replies have CRC errors and a value for one ID flips between two numbers. Most likely?

7On a balanced three-phase meter, one of three CTs is reversed. Roughly what does the meter show?

8The meter reads exactly half of the inverter total, all day. Best first suspect?

9Portal power is exactly 10× the inverter display, with the right shape. Likely cause?

10A tracker row shows a narrow bell curve while its neighbors have broad, flat-topped curves. Likely?

11All rows went flat on a windy afternoon and production dropped site-wide. What is the most likely explanation?

12The customer is on the phone next to an inverter showing a fault. What may you ask them to do?

One-line summary“Comms or value. Site or device. Power, link, protocol, configuration, validation. Evidence first, safety always.”