Documentation
Concentration Chart maps failure data onto a picture of the board. Everything runs in your browser — there is nothing to install and no account needed to try it.
The fastest way to see it
Open the app and press Load demo board & data. That loads a synthetic board and 90 days of its failure history — three shifts, about 25 failures a day, from three different export formats — and generates the chart. The history has a story in it: a chronic problem, one that only happens on 3rd shift, one that stopped after a fix, and a one-day spike. Nothing is downloaded and nothing of yours is involved; the demo's records are kept in this browser, marked as demo data, and can be removed from the setup screen.
Using your own data
You need three things, all of which an assembler normally already has:
- A placement file — the pick-and-place / centroid export for the board. This is what lets the tool find each component without anything being positioned by hand.
- A board image — a top-down photo, or a CAD or Gerber render.
- Failure data — whatever your test, AOI, repair or quality systems export. You can load several files at once and mix formats.
Optionally, a small CSV marking which parts are critical and which are transparent to test. See Setting up a board for the details that matter most, and Supported formats for what is read.
What the numbers mean
| Tile | Meaning |
|---|---|
| Failures shown | Records currently on the map, and the total read from your files. The two differ when parts are excluded or filters are on — the difference is always accounted for. |
| Components affected | Distinct reference designators with at least one failure in the current view. |
| Top offender | The component with the most failures in view. |
| Transparent-to-test | Records on parts the tester cannot measure. Excluded from the map by policy, never from the count. |
| Data sources | How many distinct export formats were detected across the files you loaded. |
Scoping a report
By assembly
If an export covers more than one product, the app says so. Scope the report to a
single assembly before reading the numbers — U7 on one assembly is a
different part from U7 on another, and merging them produces totals that
look reasonable and are wrong.
By period and shift
Pick Today, Yesterday, the last 7, 30 or 90 days, or any from/to range of days, and optionally one shift. With both ends set, each component also shows its change against the immediately preceding window of equal length, in the same shift — red for worse, green for better — which is how to check whether a process change actually helped. The comparison window is named on screen.
Days and shifts follow your plant's clock and shift times. A night shift that runs past midnight counts toward the day it started, so "Yesterday, 3rd shift" is one whole shift. Nothing a window leaves out is dropped silently: records outside it are counted by reason — outside the period, no date, a date with no time of day (which cannot be placed in a shift), or another shift — and the counts add up to the total.
Work on these: chronic problems, not one-day blips
Above the board, Work on these lists the parts and failure types that keep coming back, chronic problems first and then emerging ones. A failure is chronic when it shows up in most weeks of the period and is still showing up; emerging when it has appeared on several of the last two weeks' days and much more often than before. A one-day spike is named as a blip, and a problem that recurred and then stopped is named as stopped — the answer to "did the fix work?". Each line shows the counts behind it, such as "30 failures on 30 of 90 days, in 13 of 13 weeks". It is counting, not AI, and the thresholds are part of your profile.
You can also narrow the board, the Pareto and the list to one component class (resistors, capacitors, ICs and so on, from the designator prefix), and show the top 5, 7 or 10 components.
When something looks wrong
| What you see | What it usually means |
|---|---|
| The board is empty and every record is "not on the map" | The two files spell designators differently (U7 vs U007). The report offers to match them — see Designators. |
| "N components fall outside the board image" | The board size or origin does not match the placement file. See Origin and units. |
| A file is listed with "0 records" | This should not happen — a recognized file that yields nothing is reported as an error instead. If you see it, please tell us. |
| Dates read as five-digit numbers | Also should not happen: Excel day serials are converted. If a date column comes through raw, send us a sample and we will add it. |
Anything that could place a part in the wrong spot is called out in the app rather than absorbed quietly. If you find a case where it is not, that is a bug worth reporting to info@engsecsolutions.com.