PD Fire Alarm v0.31.0

735 rules written from those standards in our own words, each referenced to its clause: 151 from BS 5839-1:2025, 116 from HTM 05-03, 65 from BS 5839-6, 54 from BS 7273-6, 53 from BS 5839-8, 48 from BS 7827, 40 from BS 9991, 30 from BS 8300-2, 28 from Approved Document B, 24 from BS 5839-9, 24 from BS 7273-4, 21 from BS 8519, 18 from BS 8629, 16 from BS 5839-3, 15 from BS 9999, 8 each from BS EN 54-23 and the Fire Industry Association (FIA) visual alarm guide, 6 from BS 7346-7, one rule that cites three of the BS 5839 parts together, and 9 from manufacturer guidance, trade correspondence and PD project precedent, each flagged as such where it is used. HTM 05-03 and Approved Document B are Crown copyright, reused under the Open Government Licence v3. The standards themselves are not reproduced here: to read a clause in full you need your own copy.

30/09/2026 (0.31.0): The lines on healthcare audibility, battery capacity and call point delays now match the cited clauses, and the BS 5839-1 battery capacity is now in A·h. For phased evacuation, the fire control centre now has its own NOTE line (BS 9999:2017 Clause 24). There is one new rule on evacuation lifts, and the intelligibility rule is now three rules. Other wording was revised, with the meaning unchanged. No calculation, figure, limit or other verdict changed; the browser tests (3,454 free and 3,473 subscriber scenarios) were re-run with no findings; the served files matched the release. 29/09/2026 (0.30.0): PD Fire Alarm is the tool's new name. Eleven lines about a missing or wrong input are now marked CHECK in amber. Abbreviations are now written out on first use in each view, the printed report and the exports. Both battery formulas now have a legend. The BS 9999 line now quotes Approved Document B paragraph 0.9. The printed report's small text is now 7.1 pt. No calculation, figure, limit, clause reference or other verdict changed; the browser tests (3,454 free and 3,473 subscriber scenarios) were re-run with no findings; the served files matched the release. 26/09/2026 (0.29.1): about 60 lines written in the tool and 18 rule texts in the requirements lists shown to subscribers now use the verb of the clause they cite, "should" (or "recommends") where the clause says should and "shall" where it says shall, and the verb of every changed line was read from the clause in the library copy of its standard. Three rule texts keep "must", because their source says must (Approved Document B volume 2 paragraphs 1.16 and 1.17, and the FIA guide on visual alarm devices). Two lines now state what their clause says: the note on the cause and effect matrix says that sample testing should not be permitted, where it had said so only for complex buildings (HTM 05-03 Part B:2024 4.63, BS 5839-1:2025 37.4), and the note on lifts says that, for an evacuation lift, the building designer, the system designer and the evacuation lift provider should agree whether a recall signal and a suspend-service signal are provided, where it had said both signals were required (BS 7273-6:2019 9.4). In the subscriber view, the word on the Show and Hide button for the requirements list (step, section or check) no longer changes to "stage" when the list is opened or closed. The note shown to free users now ends "Every answer above shows its own clause", where it had said "on this step" even on the Design Summary and Design Check tabs; the introduction on the Design Summary tab and the disclaimer now open with a full sentence, and the introduction says "steps" where it said "stages". The printed design basis and the exports show the new wording wherever a changed line is part of them, and a saved report that is reopened and then printed or exported without a change keeps the wording it was saved with. The green for a pass, the amber for a point to check or a warning, and the grey for a neutral mark are now darker, so that any white text on them meets the 4.5:1 contrast that the Web Content Accessibility Guidelines 2.2 require at level AA for text of 12 px; this change applies only to the colours, not to any calculation, result, wording or layout. The exit doors in the drawings and the word PASS in the printed design basis are in the new green, and the amber is not used in this tool. A new click test uses every control in the tool once, at desktop width in the subscriber view and at a 390 px phone width in the free view (818 uses in total); it fails when a fault is put into a copy of the tool on purpose, and it found the fault on the Show and Hide button; the 25 hand-computed cases, the 66 comparison checks and the full set of browser tests in both tiers (3,454 free and 3,473 subscriber scenarios) were re-run on 0.29.1 with no findings, and so were the edition reference and wording checks; no calculation, figure, limit, verdict or clause reference changed; the served files matched the release and the page was read. 26/09/2026 (0.29.0): the tool now looks like the PD Power Distribution tool, with the same font and colours, and the same style of buttons, fields, tables and verdict lines. The tool name and version are shown in a new dark bar at the top of the tool. The three tabs are on a grey strip, with the selected tab underlined in blue. Above each design step, the phases and the list of steps are shown as buttons; the tabs work as before, and every step can still be reached. On each verdict line in the tool, PASS is shown in a green tag, FAIL in a red tag and NOTE in a blue tag, and the clause reference at the end of the line in a grey tag. The drawings are in the new colours and font, in the tool and in the printed design basis, as are the drawing pictures in the exports. Manual call points are now drawn red where they were copper, and exit doors green where they were teal; no text, figure, position or scale in any drawing changed. The printed design basis and the printed cause and effect matrix are in the new colours; apart from the font of the drawing labels, their text, layout, font, font sizes and page breaks are unchanged. When the page is under 768 px wide, as on a phone, the text in the tool's fields is 16 px, so a phone does not zoom the page when a field is tapped. The panel for saving and exporting, below the results, keeps its present colours for now. The saved reports, the report emailed to free users and the text of the exports are unchanged, and so is the wording shown in the tool, apart from the new version number in the dark bar. A new automated check reads the colours, fonts and layout of the tool and its drawings in a browser, with the site's own stylesheets loaded, at desktop width and at a 390 px phone width, and fails on the 0.28.2 build. A second check prints the design basis and the matrix for two project profiles to PDF, compares the page count, the text on every page and the position of every line of text with 0.28.2, and fails when one line of text is moved; the 25 hand-computed cases, the 66 comparison checks and the full set of browser tests in both tiers (3,454 free and 3,473 subscriber scenarios) were re-run on 0.29.0 with no findings, and so were the edition reference and wording checks; no calculation, rule, limit, verdict or clause reference changed; the served files matched the release and the page was read. 15/09/2026 (0.28.1 and 0.28.2): the standards references were read back against the current editions. Where a rule quotes another standard's reference to an earlier edition of BS 5839-1, the reference is kept as that standard prints it and the BS 5839-1:2025 clause is added beside it, for example "BS 5839-1:2017 Clause 45 (now BS 5839-1:2025 Clause 43)", each read from the library copies of the 2013, 2017 and 2025 editions; the HTM 05-03 references state the 2017 edition they cite. On the cables step, emergency lighting fed from a central power supply system is referred to BS 5266-1:2025 8.2.2, "Standard 60" (60 min) or "Enhanced 120" (120 min), where the note had given the fire alarm equivalences of 30 min and 120 min. The wording of the rule texts and the tool was revised in plain UK English, with the meaning unchanged, and four lines that quote a standard word for word now show the quotation in quotation marks, one of them corrected to the clause exactly (BS 8300-2:2018 20.7.4). Two new automated checks read every edition reference and the wording of every program file, in the tool and in the report, and fail on the earlier builds; the 25 hand-computed cases, the 66 comparison checks and the full set of browser tests in both tiers (3,454 free and 3,473 subscriber scenarios) were re-run on 0.28.2 with no findings; no category, figure or verdict changed; the served files matched the release and the page was read. 10/09/2026 (0.28.0): three profile inputs that only a minority case reads were moved to their case. The number of storeys, read only by a phased evacuation strategy, shows only for one; a property protection category can now be declared alongside a life safety category, so a strategy declaring L2 and P1 is recorded as L2/P1 in the design basis with the P1 coverage lines added and labelled; and a zone left without a search distance prints a request for it instead of a silent default (BS 5839-1:2025 clause 12.3). The automated tests gained checks for all three, and their links to the fields of the tool were rebuilt to find each field by its label, after the new control was found to shift the earlier position-based links without any warning; the 25 hand-computed cases, the 66 comparison checks and the full set of browser tests in both tiers (3,454 free and 3,473 subscriber scenarios) were re-run on 0.28.0 with no findings, and the same tests fail on the 0.27.0 build; no rule, limit or verdict changed; the served files matched the release and the page was read. 10/09/2026 (0.27.0): a check across every PD tool for a control asked of every user that only a minority case reads. Five were found and each now shows only for its case: on the zoning step the open plan column only on a call points only zone, and the addressable with location text question only once a search distance exceeds the limit; on the alarm devices step the means of warning only for a bedhead space, a phased evacuation strategy or a venue, opening on voice alarm where it is mandated; on the cables step the two enhanced cable triggers, remote areas that stay occupied and a risk assessment determination, under an exceptions section that is closed until opened, with the standard cable verdict now stating the basis it used; and the networks and radio notes naming their condition so the design basis never reads them as unconditional. The automated tests gained a check that each control is hidden for the case that does not read it and shown for the case that does, that the exceptions section is closed until opened and its trigger works once opened, that the standard basis line prints, and that every networks note states its condition; every release check, the 25 hand-computed cases, the 66 comparison checks and the full set of browser tests in both tiers (3,449 free and 3,465 subscriber scenarios) were re-run on 0.27.0 with no findings, the six controls listed since 0.15 as possibly having no effect were cleared along with the open plan column, and the same tests fail on the 0.26.0 build; no rule, limit or verdict changed; the served files matched the release and the page was read. 10/09/2026 (0.26.0): a check across every PD tool for a figure presented as the designer's that the designer never entered. The design basis was found to work every step the designer had not opened from the tool's default figures and to print the results as the design basis: a detector grid for a default room, a call point verdict for a default route, a visual alarm device pass for a device nobody had declared, a battery capacity from default currents. A step that was not opened is now recorded in the design basis as not entered, with the inputs to enter named, and is never worked from defaults; the cables and containment step is worked from the project profile with its two assumptions stated; steps that take no input are worked from the profile as before. Visual alarm devices are declared on the alarm devices step before a coverage verdict prints. The automated tests gained a check, run first on a fresh page, that every input-taking step shows the not-entered line and no verdict until it is opened, that no cause and effect table exists before the interfaces step is opened, that no visual alarm verdict prints for an undeclared device, and that an opened step is still worked in full; the 25 hand-computed cases, the 66 comparison checks and the full set of browser tests in both tiers (3,548 free and 3,564 subscriber scenarios) were re-run on 0.26.0 with no findings, and the same tests fail on the 0.25.2 build; no rule, limit or verdict changed; the served files matched the release and the page was read. 04/09/2026 (0.25.2): the same check completed on the two height fields the first pass could not reach. The ceiling height, asked on the detector selection and siting steps, is the height of the ceiling above the floor, which BS 5839-1 Table 3 limits, not the storey height and not the height of the detector below the ceiling (clause 21.2.4); neither field said so and the siting field had no help. The visual alarm device mounting height is the installed height of the device, which BS EN 54-23 clause 4.3.1 rates a device to, not the figure in the category code; the field had no help and the verdict did not print the height it used. All three fields now show that help and the visual alarm verdict prints the mounting height it was assessed at. Every release check, the 25 hand-computed cases, the 66 comparison checks and the full set of browser tests in both tiers (3,554 free and 3,557 subscriber scenarios) were re-run on 0.25.2 with no findings; no rule, limit or verdict changed; the served file matched the release. 04/09/2026 (0.25.1): a check of whether every input the standard names by a term of art is defined in the tool, printed back and questioned where a wrong entry can be seen. The total floor area on the project profile had no help, and the BS 5839-1 clause 12 zone limits apply to the building total, not to one storey. The field now says so, and the zone step prints a check where the building total entered is below the zones already entered, which cannot be the whole building. Every release check, the 25 hand-computed cases, the 66 comparison checks and the full set of browser tests in both tiers (3,554 free and 3,557 subscriber scenarios) were re-run on 0.25.1 with no findings; no rule, verdict or clause changed; the served file matched the release. 03/09/2026 (0.25.0): a new audit of whether every answer is reached through the input the standard derives it from. Eight defects were found: standby battery capacities were not computed for emergency voice, voice alarm and Grade A systems although the currents were asked for; a Grade A standby was printed as 24 h from an assumed attended premises; a beam or aspirating detector was passed against the Table 3 ceiling limit without asking the height; every point heat detector was treated as Class A1; a residential care home took the HTM complex hospital cable rule; every healthcare space took the HTM 45 dB(A); a design check run could replace the drawing in the design report; and a blank input could produce a figure. All eight are corrected in 0.25.0 with new questions for the missing inputs, the automated tests gained a sweep for this kind of defect, and the full set of browser tests was re-run in both tiers (3,554 free and 3,557 subscriber scenarios, 66 comparison checks) with no findings; after upload, the page was read. Claims audit 11/08/2026: 192 statements shown to users were extracted from the code and checked one by one against the rule each cites, and every rule reference was verified against the rule base with none missing. Scenario audit 10/08/2026: 7,069 input combinations were driven through the tool and checked against independently written calculations, 3,536 in the subscriber view and 3,533 in the free view, covering every building use, evacuation strategy, risk profile, premises class and declared category, room and corridor geometry from zero to 1,000 m, ceiling heights across the Table 3 limits, all 28 BS 8519 Table 1 systems, the battery load matrix, and cause and effect matrices from 1 to 40 levels. Thirteen defects were found and corrected in that audit, every one of them in how a result was adjudicated, cited or presented rather than in a calculation. Source read-back, completed 11/08/2026: all 732 rules have been read back word for word against the source itself rather than against our own notes, from the clause-cited answers in the tool through every figure, every prose rule, the commentary and guidance, and the PD project precedents; the final two were written directly from their sources the same day to close coverage gaps found in the reading. The read-back found 329 defects. Almost every one was a missing qualifier rather than a wrong number: an exception the standard allows, a condition that limits when a figure applies, a relief written as a prohibition, or a note's example presented as the requirement. All 329 are corrected and included in release 0.22.0. The rule base also increased from 714 to 732 during the reading, as recommendations that had never been turned into rules were written from source and rules that covered several clauses at once were split so each cites exactly its own clauses. Every calculation is compared with a separate reference calculation of the same rule, written in Python: 25 cases computed by hand (battery Annex E sizing, detector spacing and coverage, corridor intervals, zone limits, call point travel, sound levels, visual alarm coverage, cable category) all match, and 66 comparison checks, re-run in the browser on every release, confirm that the tool returns what the reference calculation returns. Detector spacing reproduces the Apollo pocket guide layout. Two real project fire strategies were used as the reality check on the workflow. The full set of browser tests was re-run on 12/08/2026 at the 0.24 interface releases: 3,533 free and 3,536 subscriber scenarios and all 66 comparison checks, both tiers, with no new findings. They were re-run again on 02/09/2026 after the 0.24.3 wording revision: every release check, the 25 hand-computed cases, the 66 comparison checks and the full set of browser tests in both tiers, with no new findings; no rule, verdict or clause changed.