The Photograph You Take Before You Touch Anything
Why this matters
Some information exists only while a machine is in one particular state, and once you disturb it there is no document anywhere that can give it back. Which conductor was on which terminal. Where the adjustable stop was set. Which of two identical valves was the one throttled. The manufacturer never published it because it is not a property of the equipment, it is a property of this installation.
A photograph taken before you touch anything is not evidence for a dispute, though it can serve as that. It is a document you are authoring, and it is the only copy. Judge it the way you would judge any document: can a person who was not there act on it?
The call that made the case
A control module on a packaged unit was diagnosed as failed and replaced. Straightforward visit, about 90 minutes, and the unit ran on the way out.
Two days later it faulted in a way it never had before, and a second tech took the call. The new module was correct and the original diagnosis had been sound. The problem was the landings. Eleven conductors came off the old module. Four of them were unambiguous from the diagram adhered inside the compartment. The remaining seven were field additions, made at some point by somebody else, and the diagram did not show them at all - it showed a factory configuration that the installation had departed from years earlier.
The first tech had photographed the job. Three frames: the unit from the doorway, the failed module in his hand, and the new module installed and closed up. Not one of the three showed a terminal.
So the second tech de-energized and rang out the seven unknown conductors to identify what each one served, tracing each back to its device. That took 135 minutes and required a return trip of about 40 minutes each way to pick up the interlock schedule from the property manager's file.
Before ringing anything out, he opened the disconnecting means, applied his own lock and tag, and proved dead by testing his meter on a known live source, then the conductors, then the meter again (NFPA 70E-2021, 120.5). The general-industry requirement to de-energize and lock or tag before working on or near exposed energized parts is 29 CFR 1910.333(b)(2); the construction counterpart for lockout and tagging of circuits is 29 CFR 1926.417. This matters to the photography as well as the tracing: you do not reach a phone into an open energized enclosure to get a better angle on a terminal strip. Either the equipment is proven dead and you can work at the terminals, or you shoot from outside the working space and accept the wider frame.
The capture that would have replaced all of it would have taken about 3 minutes. One uncropped frame of the terminal strip with the module in shot, then two closer frames. Against 135 minutes of ring-out that is a ratio of about 45 to 1 in technician minutes, and because the return was a callback the shop absorbed, every one of those 135 minutes was unbilled.
Why the photos they had failed
The first tech was not careless. He took photographs. They failed for four specific reasons, and each one is a rule you can apply directly.
They were taken after the disturbance, not before. The installed-and-closed frame documents his work. It carries nothing about the state he inherited, which is the only state that could not be recreated.
They were cropped past their own reference. The module in his hand has no context: no cabinet, no orientation, no way to tell which of the two similar compartments it came from.
They had no chain back to a fixed feature. Nothing in the frame set connects the close view to the wide view, so even a good close-up would have been unmoored.
They photographed the object and not the interface. Techs instinctively photograph the part. The part is the least valuable subject in the frame, because you can buy another one. What cannot be bought is the pattern of how it was connected.
The overlap chain
The single habit that turns photographs into a usable document: every frame you take contains at least one element from the frame before it. Wide contains the room, medium contains a feature from the wide, close contains a feature from the medium. That chain is what lets somebody who was not there place a zoomed photograph of a terminal strip in a building.
A close-up with no chain is a picture of a terminal strip. There are millions of those. Yours is only worth something because it can be located.
Build the chain in this order, all before the first tool touches anything:
- Establishing frame. The equipment and what it connects to, including a permanent feature that identifies the location - a door, a wall penetration, a unit tag, the pipe or duct leaving the space.
- Identity frame. The data plate or nameplate square-on and legible, close enough to read the model and serial without zooming, with a piece of the cabinet still in shot.
- Interface frame, uncropped. The whole terminal strip, the whole manifold, the whole bank of takeoffs, in one image with the surroundings visible. This is the frame that gets skipped and it is the one that matters most.
- Detail frames. Individual landings, individual valve tags, individual switch positions, each including one feature from the interface frame.
- Setting frame. Every adjustable thing in the position you found it: switch banks, potentiometers, mechanical stops, throttled valves, damper positions, setpoint displays.
- Absence frame. Empty terminals, capped stubs, unused ports, a missing cover, a bypassed device. What is not there is documentation too, and it is the class most often lost, because there is nothing to point the camera at unless you decide to.
- Existing-label frame. Every tag and directory already on site, including any you believe to be wrong. If you correct one later, this is the record of what it said before, and the fact that it was wrong is itself a warning worth keeping.
Making the frames readable by somebody else
Shoot square-on, not at an angle. A nameplate photographed at 40 degrees is often unreadable at the letters that matter, and the letters that matter are usually the smallest ones.
Get light on it rather than zooming in the dark. Phone zoom on a dark subject produces a soft image that looks fine on the phone and fails on a monitor. A work light held to the side also gives you shadow, which makes stamped or embossed characters legible when flat light will not.
Include something for scale where size is the fact - a common hand tool, a coin-sized object, your glove. Do not include your hand in a frame where the hand covers anything.
Take the same frame twice from slightly different positions on anything reflective or behind glass. Reflection kills more nameplate photographs than focus does, and you will not see it on the phone screen in that light.
Photograph in an order you can reconstruct, because the file timestamps are your sequence. If you shoot the whole job out of order, nobody can tell which frames were before the disturbance.
When nobody photographed it
You will inherit the situation the second tech inherited. The reconstruct path, in order of what it costs:
Start with what is physically marked. Legends printed on boards, stamped valve tags, factory wire markers, colors that the trade uses consistently in that context. Physical marking that shipped with the component beats anything on paper, because it was printed to match that revision.
Then the site's own paper. The adhered diagram, the panel directory, an as-built the property manager holds. Treat each as a hypothesis with a revision date, not as an answer, and check whether it accounts for the number of connections you actually have. The four-of-eleven mismatch in the case above was detectable before any tracing, just by counting.
Then trace, de-energized and proven dead, one conductor at a time, and label each one as you identify it rather than at the end. A tech who traces all of them and then labels has to hold the whole map in their head, which is where the error enters.
Then ask the building. Somebody has watched this equipment for years. The maintenance person who told the second tech that "the freeze thing was added after the first winter" saved him the interlock hunt, and no document said it.
Whatever you reconstruct, photograph and write down before you close the panel, because you have just paid full price for a document that did not exist, and it should never have to be bought twice.
Telling a usable photograph from a comforting one
Photographs feel like documentation while you are taking them. The test is not how many you have, it is whether they answer a question. Pull up a set from a recent job on a monitor, not a phone, and check:
- Can you read the model and serial without guessing a character? If a 5 might be an S, the identity frame failed, and it fails silently because you can usually infer the model from the job and never notice.
- Can you tell which side is up and which compartment you are in? If not, the chain is broken between two frames and every detail below the break is unlocatable.
- Is there any frame showing the state before your first change? If your earliest frame already shows a cover off, you documented your work, not the installation.
- Could you land eleven conductors from these images without calling the tech who took them? That is the real bar. Everything else is a picture of a machine.
References
- 29 CFR 1910.333(b)(2), general industry, and 29 CFR 1926.417, construction, on de-energizing and lockout or tagging of circuits before work on or near exposed energized parts
- NFPA 70E-2021, 120.5, on the live-dead-live proving sequence
- See related: The As-Found Drawing and Why You Make One; How to Take Photos That Document the Before and After; The Existing-Conditions Photo Record