Technical Manual · LTC Workbench
LTC Workbench — Methods, Equations and Limits
What every tool in the LTC Workbench computes, the equation it uses, the document the limits come from, and where the arithmetic stops being valid. The workbench is a hub and five standards modules (ASHRAE, Tier, TIA, NFPA, ISO) beside the liquid-to-chip system model.
▶ Open the LTC Workbench01 Purpose & scope
The workbench answers concrete engineering questions — is this inlet condition inside A1 at 1,200 m, can this UPS be taken out without dropping the load, how much agent does this room need, does this link fit its optics budget — with the equation shown and the source named. It replaced six long-form articles in v3.12.0.
It does not issue compliance findings. Limits are summarised from the cited documents; the governing edition adopted by the jurisdiction, the listed manufacturer's manual and the authority having jurisdiction decide.
| Module | Tools | Governing documents |
|---|---|---|
| Plant (hub) | Standards map · coverage matrix · commissioning lifecycle · question index | All of the below |
| ASHRAE | Air envelope · liquid classes · condensation · sensor log · rate of change | TC 9.9 Thermal Guidelines, 5th ed. (2021); Standard 90.4; ANSI/ISA-71.04 |
| Tier | Power and cooling path simulator · availability | Uptime Institute Tier Standard: Topology |
| TIA | Spaces · fiber budget · copper channel · port count · bonding | TIA-942-B, TIA-568.2-D / 568.3-D, TIA-607-C, IEEE 802.3 |
| NFPA | Agent quantity · detection plan · release sequence | NFPA 2001, 72, 75, 76, 13, 70 Art. 645 |
| ISO | KPIs · energy Sankey · EnPI baseline · monthly PUE | ISO/IEC 30134 series, ISO 50001 / 50006 / 50015, ASHRAE Guideline 14 |
02 One plant behind every tool
Since v3.14.0 every page reads one plant model, js/ltc-plant.js: a 100 MW liquid-to-chip campus by default, five halls, edited in one place — the plant strip at the top of every workbench page (Design · Site · Advanced; every advanced input auto until changed). The tools then ask only for what the plant cannot know: ISO needs nothing, ASHRAE, TIA and NFPA take their plant-derived inputs as auto values the reader may type over, the Tier simulator draws DH01 from it, and the model lab computes on it.
Plant map. An overview and three details — Power, Liquid heat, Air & rooms — each within the diagram engine’s nine-block budget, every block carrying its live number. Hover a block: it and its connectors stay lit, the rest fades, and a tooltip shows the numbers and the formula. A table compares the four plant configurations (air-cooled chillers; fluid, dry or adiabatic coolers with water-cooled chillers) at the same site and load.
03 Commissioning lifecycle and the DH01 thermal map
Lifecycle. L1 factory, L2 delivered, L3 started, L4 functional, L5 integrated, operate — each a gate with a script summary: what is tested, what is expected, the hold point that closes it, and the documents. L4 contains the heat load test: load banks stepped 25 → 50 → 75 → 100 % with liquid and air at the design split, each step held until temperatures settle, the TCS, FWS, CRAH and rack-inlet readings recorded, typical acceptance (TCS supply within ±1 K, flow ≥ design, rack inlet inside the recommended envelope, measured ṁ·cp·ΔT within ±5 % of load-bank kW), then CDU, pump, CRAH and CDU-feeder failures at 100 %. Acceptance values marked typical are set by the project’s commissioning plan.
DH01 thermal map. The hall in plan from the plant’s geometry, solved as a mass and energy balance and drawn as one continuous air loop: fan wall → end zone → cold aisle → through the rack → contained hot aisle → ceiling plenum → back to the fan walls. Rack airflow is set by its heat at the design ΔT; CRAH fans share one speed up to their rated flow (5 % over demand under pressure control); each fan wall feeds the aisle mouths in proportion to demand, less where its units are out; each cold aisle is marched from both ends, so the supply band on the plan ends where its air is used up; a rack’s unmet demand is pulled from its hot aisle and the plenum return (a fixed point); the CRAH coil is an effectiveness model fitted at design, so supply air rises when the coil cannot hold set-point. The test pins the balance: rack airflow = supply drawn + exhaust pulled back, every m³/s the fans move reaches an aisle mouth, return − supply = air heat ÷ ṁ·cp. Above the plan, the liquid path (IT → cold plates → CDUs → TCS → FWS → coolers) and the air path (rack air → cold aisles → CRAHs → supply → return → CHW → chillers) show each link’s state under the fault; hover a rack, CRAH, CDU or aisle for its numbers. Faults: CDU feeder, one CDU, CRAH losses, facility water, chiller trip, chilled water — with the first-order transient (C·dT/dt = Q − f·Q·(T − Tsink)/(T0 − Tsink)) checked against its analytic solution. A rack that loses cold-plate flow throttles in seconds, so the default shows IT shed; "if they kept running" is the worst case. Racks above 27 °C are outlined and above 32 °C hatched, so colour is not the only channel; the map is keyboard and touch operable. It is a thermal approximation for training, not CFD.
04 ASHRAE — psychrometrics and envelopes
Every humidity quantity is derived from dry-bulb, dew point and site pressure. Nothing is read off a printed chart.
| Class | Dry-bulb | Low humidity | High humidity | Max DB derating above 900 m |
|---|---|---|---|---|
| Recommended (A1–A4) | 18–27 °C | −9 °C DP | 15 °C DP and 70 % RH | 1 °C / 300 m above 1,800 m |
| A1 | 15–32 °C | −12 °C DP and 8 % RH | 17 °C DP and 80 % RH | 1 °C / 300 m |
| A2 | 10–35 °C | −12 °C DP and 8 % RH | 21 °C DP and 80 % RH | 1 °C / 300 m |
| A3 | 5–40 °C | −12 °C DP and 8 % RH | 24 °C DP and 85 % RH | 1 °C / 175 m |
| A4 | 5–45 °C | −12 °C DP and 8 % RH | 24 °C DP and 90 % RH | 1 °C / 125 m |
| H1 | 5–25 °C (rec. 18–22 °C) | −12 °C DP and 8 % RH | 17 °C DP and 80 % RH | not modelled |
Binding limit. For each class the tool computes the signed margin to all six boundaries (dry-bulb high/low, dew point high/low, RH high/low) and reports the smallest — the limit the condition would break first. Above 3,050 m every class is outside the envelope.
Liquid classes are set by the maximum facility-water supply temperature: W17, W27, W32, W40, W45, W+ (above 45 °C). The free-cooling check compares the required supply with what a dry cooler delivers at the design dry-bulb: leaving water = ambient + approach; a negative margin means trim cooling for those hours.
Condensation. Room dew point from dry-bulb and RH; the coolant supply must exceed it by the chosen control margin. The tool also inverts the problem: the highest room RH a given supply tolerates is es(supply − margin) / es(DB).
Sensor log. Each reading is classified against the chosen class; time in each region is weighted by the interval to the next reading. Rate of change is the largest swing between any two readings inside a 60-minute and a 15-minute window, compared with 5 °C/h (tape media present) or 20 °C/h and 5 °C in 15 min, and 5 % RH per hour. A coarse log can miss a fast step — log at 5 minutes or faster to prove the 15-minute limit.
05 Tier — maintainability and fault tolerance by test
The Tier Standard: Topology defines Tiers by objective. The simulator tests the objective directly on DH01’s topology, generated from the shared plant: the plant sets N (DH01’s share of generators, transformers and UPS modules, its CDUs and CRAHs, the central chiller plant and liquid coolers), the Tier sets the spares and the paths. Every removable element is taken out one at a time, in two modes — one unit of each identical group (they are symmetric), and every bus, UPS output board and pipe loop whole, because a component is a component however many units hang on it.
| Tier | Power path modelled | Cooling path modelled | Expected result |
|---|---|---|---|
| I | Utility or N generators → board → N UPS → PDU → single-corded load | N plant → N pumps → one loop → N CRAH | neither |
| II | N+1 generators and UPS, single path | N+1 plant, pumps, CRAH; one loop | neither (the path is not redundant) |
| III | N+1 generator plant → board A (active) / B (alternate) → UPS → PDU → dual-corded load | N+1 plant and pumps → loop A (active) / B (valved alternate) → N+1 CRAH | maintainable, not fault tolerant |
| IV | Two independent active systems | Two independent active systems | both |
The gate asserts these four outcomes for both power and cooling, so an edit that breaks a Tier fails the build. Utility appears only for Tier I/II: for III/IV the Standard treats on-site generation as the primary source. Compartmentalisation and continuous cooling (Tier IV) are not modelled.
06 TIA — cabling, spaces, bonding
Fiber budget. Loss is built from TIA-568.3-D maximum component values and compared with the IEEE 802.3 channel insertion-loss limit and reach of the application.
| Application | Fiber | Reach | Channel IL |
|---|---|---|---|
| 10GBASE-SR | OM3 / OM4 | 300 / 400 m | 2.6 / 2.9 dB |
| 25GBASE-SR | OM3 / OM4 | 70 / 100 m | 1.8 / 1.9 dB |
| 40GBASE-SR4 | OM3 / OM4 | 100 / 150 m | 1.9 / 1.5 dB |
| 100GBASE-SR4 | OM3 / OM4 | 70 / 100 m | 1.8 / 1.9 dB |
| 400GBASE-DR4 | OS2 | 500 m | 3.0 dB |
| 10GBASE-LR / 100GBASE-LR4 | OS2 | 10 km | 6.2 / 6.3 dB |
Copper channel. 90 m permanent link plus 10 m of cords (Category 8: 24 m + 6 m, 30 m channel). 10GBASE-T runs 55 m on Category 6 and 100 m on 6A.
Port count. Ports = racks × ports per rack × (1 + spare), rounded up; panels = ports ÷ ports per 1U; MPO trunks = duplex ports ÷ (fibers per MPO ÷ 2). Every horizontal link ends on a panel at both ends, so the HDA carries the sum of all racks.
| TBB / BC run | < 4 m | 4–6 | 6–8 | 8–10 | 10–13 | 13–16 | 16–20 | > 20 m |
|---|---|---|---|---|---|---|---|---|
| Minimum size | 6 AWG | 4 AWG | 3 AWG | 2 AWG | 1 AWG | 1/0 AWG | 2/0 AWG | 3/0 AWG |
TIA-607-B/C sizes the backbone bonding conductor at 2 kcmil per linear foot and caps it at 3/0 AWG; the larger sizes some references list come from the superseded J-STD-607-A. Local electrical code may require more.
07 NFPA — agent, detection, sequence
Agent quantity uses the NFPA 2001 total-flooding equations at the minimum design temperature; the concentration the same quantity produces at the maximum room temperature is checked against the agent's NOAEL, which decides whether the space may be occupied.
| Agent | k1 | k2 | NOAEL | Discharge |
|---|---|---|---|---|
| HFC-227ea | 0.1269 | 0.000513 | 9.0 % | 10 s |
| FK-5-1-12 | 0.0664 | 0.0002741 | 10.0 % | 10 s |
| IG-541 | 0.65799 | 0.00239 | 43 % | 60 s Class B · 120 s Class A/C |
| IG-100 | 0.7997 | 0.00293 | 43 % | 60 s Class B · 120 s Class A/C |
Design concentration defaults are editable starting points: the real value is the listed system's minimum extinguishing concentration × 1.2 (Class A and C) or 1.3 (Class B). Container counts use the listed container size, also an input.
Room size and air changes default to DH01 from the shared plant (length, width and height from the hall geometry; air changes from CRAH airflow ÷ hall volume), each overridable. Detection. High air movement reduces the area each spot smoke detector may cover (NFPA 72 high-airflow table: 60 ACH → 125 ft², 30 → 250, 20 → 375, 15 → 500, 12 → 625, 10 → 750, 8.6 → 875, ≤ 7.5 → 900). Between rows the higher-airflow row is used. Detectors are laid on a square grid of that area; cross-zoned release doubles the count, each zone meeting spacing on its own. Aspirating sampling holes are treated as spot detectors for spacing.
Release sequence. A state machine — normal, alert, fire 1, pre-discharge, abort held, discharge — driven event by event. Delays, which outputs trip at which stage and whether an abort switch is fitted are site programming approved by the AHJ, not fixed by NFPA 2001.
08 ISO — KPIs, Sankey, baseline
KPIs use 12-month energies. PUE is reported per ISO/IEC 30134-2 category, because the metering point moves losses between IT and overhead. By default the twelve-month energies at M0–M3, the water and the renewable and reused energy come from the shared plant, so nothing is entered; Metered data reopens the inputs, prefilled with the model, and reports each meter against it.
09 Diagrams and symbols
Every block diagram is drawn by the RZ diagram engine (js/rz-diagram.js): nodes sized from their own text, orthogonal connectors routed around boxes, labels placed by search, legend derived from what was drawn. Equipment glyphs are the IEC 60617 / ISA-5.1 symbols DCMOC uses (assets/engineering/, emitted as js/rz-eng-symbols.js), plus twelve web-only symbols for detectors, sprinklers, agent containers, EPO, meters, patch panels, transmitters and valves.
Connectors carrying load move; each dash pattern advances by exactly one of its own periods per cycle, so the loop never snaps back, and reduced-motion settings stop it. Data charts (psychrometric, loss profile, Sankey, CUSUM, PUE bars) are computed from their scales, not hand-placed.
10 Verification and limits
tools/test-ltc-workbench.mjs (142 assertions) and tools/test-ltc-plant.mjs (the plant: energy balance on 202 plants, 100 MW hand checks, metering order, transients against the analytic solution, DH01 behaviour, recomputed practitioner claims), both in the ship gate, pins each module's arithmetic to hand-checkable values, requires every Tier topology — power and cooling — to meet its own objective, and holds each page's frame: script order, diagram tokens in both themes, cache tokens on every asset, and the fail-closed access gate.