Start at the cable, not the software

A sending card with power and no input signal still answers on the network. A card with no power obviously does not. So before opening any adapter settings, look at the LEDs: is the card powered, is the patch lead in the port marked Ethernet rather than the one marked DVI, and does the link light come on at both ends.

On a 2 m sphere this takes twenty seconds, because the card sits in the base plate behind one panel. On a 6 m build it usually means opening a service hatch, which is why we label the card bay before crating. If the link LED is dark, no software setting will find that card, and you can stop reading here.

Put the laptop on the card's subnet, and mean it

NovaStar sending cards leave the factory on a fixed address in the 192.168.0.x range; the exact number is printed on the card and stated in the manual for your model. Your PC needs an address in the same subnet, with a different host number. Two things break this quietly: a DHCP lease that lands on 169.254.x.x because the adapter never got an address, and a subnet mask typed as 255.255.255.0 when the card expects a wider range.

Set the adapter manually for the duration of the job. Turn Wi-Fi off at the same time — a laptop with Wi-Fi enabled will happily route discovery traffic out of the building and back to you as 'nothing found'.

Windows drops the broadcast without saying so

NovaLCT discovers cards by broadcast. The Windows firewall drops broadcast traffic on a Public-profile adapter, and most third-party security suites drop it on every profile. If the adapter shows as a Public network, switch it to Private before blaming the card. If the machine runs a security suite with a network-protection module, disable that module for ten minutes and retry.

This is also why the same card that is invisible on a client's IT laptop is found immediately on a clean machine. The card was never the problem, and the IT department will not thank you for proving it politely.

Virtual adapters are the quiet killer

A laptop running a VPN client, Hyper-V, VirtualBox or VMware carries several interfaces that exist only in software, and some of them sit on 192.168.0.x simply because that range is popular. NovaLCT binds to one adapter and broadcasts on that one only, so it may be shouting into a virtual switch while the physical port sits idle.

Open the adapter list and disable everything except the Ethernet port you actually plugged in. On a Mac with a USB-C dongle, confirm the dongle's interface is the one holding the address, not the Wi-Fi.

Ping first. It splits the problem in half

Ping the card's address from a command prompt. If the ping fails, the fault is cable, switch, power or addressing, and no software change will touch it. If the ping succeeds and NovaLCT still shows an empty device list, the fault is in software: firewall, subnet mask, a second instance of NovaLCT already holding the socket, or a firmware build the installed version does not recognise.

We keep a screenshot of a good ping in the commissioning folder that ships with each crate. Arguing with a client's IT department is easier with a picture of a reply.

Sometimes there is no card to find

On processor-based systems — the VX series, the J6, the N9 — the sending function lives inside the processor. There is no separate card, and you configure it in V-Can or the processor's own interface, not in NovaLCT. We have been called out to 'missing' cards that were never ordered, because the processor was already doing the job.

The same applies to hot-backup pairs. Two cards configured as primary and backup answer as a single device. If you expected two entries and see one, that is the configuration working, not a card that failed to appear.

Version mismatch, and why flashing is the wrong fix

A card flashed with firmware newer than your NovaLCT build can fail to enumerate. The symptom is fairly specific: the card answers a ping, other cards in the same rack appear, this one does not. Install the current NovaLCT from NovaStar's own download page rather than a copy from a supplier's drive, then power-cycle the card with the receiving cards already connected.

Do not flash firmware to fix a detection problem. If a card is not enumerating, a flash has a real chance of leaving it in bootloader with no way back except a return to the factory, and that costs two weeks of shipping instead of ten minutes of settings.

What we do before crating, so this call does not happen

Every sphere leaves the floor having been addressed, mapped and driven from a laptop on our own bench, and the configuration is saved as an .rcfgx file. The card's address, the software version used and the file name go on a label inside the base plate.

If you inherit a screen, or buy one second-hand, that label is the fastest route back to a known-good state. If it is missing, send a photo of the card and the receiving-card batch code and we will tell you what the build should have been.