DigiEntropy engineering blog

Plain-English engineering writing on refrigeration compressors, F-Gas Regulation, polynomial fitting (AHRI 540 / EN 12900), CO2 transcritical, and the multi-brand catalog economy.

• Sixteen Strangers, On Repeat: Making Persona Testing a Regression Test — We had a persona lab that ran once and impressed everyone. Turning it into something we run regularly meant freezing the test set, building a diff — and learning that three of five findings do not survive a real browser.
• Twelve Rounds of Persona Testing. Four Fixed the Wrong Thing. — We pointed 18 goal-driven persona agents at one new feature and kept going until they could use it. Here is every round, including the four we spent fixing a bug that lived in the measuring instrument.
• We Hired 18 Imaginary Users to Break Our Own Platform — Our tests all passed. They were written by people who already knew where to click. So we built 18 personas with goals instead of scripts, let them loose on the real site, and measured how it felt.
• Seven Frontier Models, 80 Engineering Tasks: We Re-Ran Our AI Benchmark — Our first benchmark got too easy — everyone passed. So we doubled the hard questions and added three more models. On the old questions the field is tied; on the new ones every model drops, and three real defects fell out.
• How to size a frequency inverter for a compressor — Which current sizes a VFD for a compressor — full-load, locked-rotor, or something else? The answer is the max operating current at the top frequency, checked against two criteria the compressor and drive makers spell out. Here is the sizing rule, sourced, with the tool that runs it.
• Inside Daisy's Runtime Upgrade: Better Roles, Recovery, and Models — We studied a cross-vendor prompt corpus, copied none of it, rebuilt Daisy's provider boundary, retired deepseek-chat safely, and measured the result across 49 real engineering scenarios.
• The balance point: how the System Simulator finds a real operating COP — Most selection tools impose the condensing temperature with a fixed approach guess. Our System Simulator solves for the point where the compressor and the actual condenser agree — so swapping a coil, or slowing an EC fan, moves the answer the way it would in the field.
• Why inverter compressors save energy at part load — the hidden physics — A variable-speed compressor and a fixed-speed one have almost the same COP at any single operating point. So where do the savings come from? Not the compressor — the coils. Here is the physics, with real polynomial data.
• Teaching Daisy Goal-Based Loops — Without Burning Tokens — We read Claude's agent-loops playbook, gave Daisy a deterministic goal check on its selection engines, and measured it: dead-end empty results became engine-backed answers, for zero extra tokens on healthy turns.
• Independent data for the refrigerant transition and cooling policy — A free, independent, vendor-neutral platform on real manufacturer data - for regulators, ozone and energy officers, and cooling programmes working on the refrigerant transition and cooling efficiency.
• Teaching refrigeration and HVAC with real market data, not textbook theory — A free, browser-based platform where students design and analyze with actual manufacturer performance data from 19 compressor brands — System Simulator, Lifelong Cost, Cold Room, Benchmark and more.
• We Let One AI Coach Ours — and Measured What Changed — How we used Claude Code to sharpen our dev workflow and teach our co-pilot Daisy to ground her answers and check her own work — measured before and after.
• Compliance Studio: five regulations, one verdict, one PDF — EN 378 charge limits, ASHRAE 15/34 concentration limits, the EU F-Gas GWP phase-down, EcoDesign efficiency, and lifetime TEWI — five rule sets in five documents. Compliance Studio runs all of them against one system and returns a single pass/warn/fail verdict, with a phase-out timeline and every citation.
• How to build the most efficient compressor: a first-principles guide to COP — Start from the one equation for COP — where mass flow cancels and efficiency turns out to be about matching, not size — then climb every lever from the Carnot ceiling down to the bearings, toward the most efficient compressor physics will allow.
• Inside Daisy: How We Measure and Improve Our AI Co-pilot — Our in-app AI engineer was useful but not 'wow'. So we built an eval to measure it across 33 questions, found four failure patterns, and fixed them — here's the whole story, in charts.
• Carbon Studio: lifetime CO2e for any compressor + refrigerant, on any grid — Refrigeration carbon has two halves — refrigerant that leaks and electricity over fifteen years. Carbon Studio computes both (TEWI and LCCP) for every option you want to compare, on the grid of any country, with a world map, scenario charts, and a one-click lower-GWP swap.
• Inside the dual-surface co-pilot: from one sentence to a page it drives — One plain-English request becomes a driven page plus a conversation of interactive cards. Here is every step of how — the inputs, the outputs, and the strategy behind each, with the alternatives we turned down.
• God Theory of Refrigerants: design a refrigerant, test it on a real compressor — A live blend lab for the classroom: mix your own refrigerant from pure molecules, get real CoolProp mixture thermodynamics, and predict how a real catalogue compressor would run on a fluid that doesn't exist yet.
• One catalogue number, any condition — rating water-cooled heat exchangers — Manufacturers publish a shell-and-tube condenser or evaporator as a single capacity at one refrigerant and one fixed temperature. Here is how we turn that one number into a physical model that rates the unit at your refrigerant, water temperature, flow and fouling — and why that is what makes an honest cross-brand comparison possible.
• How our AI co-pilot sizes a cold room — real engines, not a chatbot's guess — Watch our AI co-pilot size a cold room and shortlist compressors from one plain-English request — and see why the numbers come from real engines, not a language model's guess.
• From workflow packs to generative UI — rethinking the DigiEntropy AI assistant — We built our AI assistant on predefined, multi-step workflow packs. After a year, we asked whether that is still the best way to drive deep engineering tools with plain English — and prototyped a generative-UI answer that renders forms, results, and comparisons right inside the conversation. Here is the honest before-and-after.
• Pressure drop: the penalty selection software hides — Bitzer's selection software assumes zero pressure drop across the evaporator and condenser. A real loop has four pressures, not two — and once you put the piping back, the same compressor on R404A gives up about 17% capacity and 12% COP. Here is why, with the numbers.
• Operating envelopes, end to end: from a vendor PDF to a physics-generated polygon for every brand — An operating envelope is the TE/TC region where a compressor is qualified to run — and manufacturers publish it as a picture, not data. This walks the whole pipeline: the nine physical limits that shape the envelope, two ways to turn a vendor chart into a polygon, and how we now generate envelopes from physics for 15 brands (about 42,000 of them) when no chart exists.
• Water-cooled vs air-cooled condensers: same compressor, 28% more COP — Swap an air-cooled condenser for a water-cooled Bitzer K-type and the compressor's COP jumps about 28% — without changing the compressor. Here is the theory, and a System Simulator run that shows exactly why a lower condensing temperature does all the work.
• Condenser and evaporator type vs system COP — and how DigiEntropy makes the hidden losses visible — The compressor only sees four pressures. Condenser and evaporator type change those pressures through SST, SDT, line losses, superheat, and aux power — and a 1 K shift in lift moves COP by 2–4 %. Here is the engineering story, and a tour of the DigiEntropy pages that make every loss visible.
• The three axes of refrigerant goodness: theoretical COP, second-law efficiency, and volumetric capacity — Picking a refrigerant by COP alone gives you one of three answers that actually matter. This post unpacks the theoretical COP from a vapor-compression cycle, the Carnot-based second-law efficiency, and the volumetric refrigerating capacity that drives compressor displacement — and grounds each in actual polynomial COP from Bitzer compressors.
• EC fans in refrigeration: how electronically commutated motors cut energy and simplify control — Electronically commutated fans run cooler, quieter, and on a fraction of the electricity of conventional AC motors — and they give refrigeration controllers a continuous speed dial instead of a coarse on/off switch. Here is how they work, how they are controlled, and what the engineering numbers actually look like.
• Cross-Reference — finding an equivalent compressor when lead time is tight — Compressor lead times in 2026 swing from 4 weeks to 6 months depending on brand, motor type, and supply-chain noise. Cross-Reference finds same-capacity, same-refrigerant equivalents across all 13 brands so the project doesn't stall on the spec.
• From single compressor to parallel rack — when to add the next compressor — A single-compressor system is simpler + cheaper. A parallel rack (multiple compressors on one suction header) handles bigger capacity + better part-load efficiency + redundancy. The crossover point depends on duty profile and reliability needs.
• The Chinese-brand wave — Snowkey, Carlyle SE, Invotech — Chinese refrigeration brands have moved from low-spec OEM production to engineering-grade catalog presence in the EU market. Snowkey, Carlyle Southeast, Invotech (and others) now ship compressors that compete on spec + pricing with established European brands.
• Copeland Discus class — when Discus beats screw — In the 30-150 kW band, the natural competitor isn't always screw — Copeland's Discus class semi-hermetic reciprocating compressors win on cost + serviceability + part-load COP in exactly the band where screw is often the default.
• Dorin CO2 — the Italian transcritical specialist — Officine Mario Dorin (Florence, Italy) builds reciprocating compressors that almost exclusively target transcritical and subcritical R744 applications. The CD-series is the EU reference for compact CO2 commercial refrigeration. A focused brand worth knowing.
• BOCK HG/HGX — German semi-hermetic recip for industrial duty — BOCK's HG and HGX families are the German engineering standard for industrial semi-hermetic reciprocating compressors. R744-qualified ECOLINE-equivalent variants, ammonia industrial, transport refrigeration — BOCK covers ranges where Bitzer doesn't quite fit.
• Hanbell RC2 — the 15-coefficient screw quirk — Hanbell's RC2 screw compressors publish polynomials using a 15-coefficient 4th-order bivariate form, not the standard 10-coefficient EN 12900 / AHRI 540 cubic. Why they did it, what it costs in interoperability, and what it buys in fit accuracy.
• The Bitzer Octagon family explained — HE, HES, FE, GE, JE, EES, CES — Bitzer's Octagon family is the EU reference for semi-hermetic reciprocating compressors. The letter classes (HE, HES, FE, GE, JE, EES, CES) encode bore + stroke + cylinder count. Knowing the system lets you pick the right model from a part number alone.
• F-Gas Quota 2026 — what the new allocation rules mean — Regulation (EU) 2024/573 changed the HFC quota allocation rules in non-obvious ways. The producer/importer registry is now tighter, newcomer quotas shrank, and quota trading is more transparent. What's the practical impact on prices and availability.
• Multi-stage CO2 cascade for ultra-low temperature — Single-stage R744 hits its envelope below -45 °C. Multi-stage cascade architectures push ULT to -90 °C with R744 as the low-temp stage and R744 or R290 as the high-temp stage. The pharma + biotech ULT market is rapidly converging on cascade CO2.
• Discus vs Scroll vs Reciprocating — picking the right compression technology — The four dominant compression technologies (semi-hermetic recip, Discus, scroll, screw) overlap heavily in mid-capacity bands. The right choice depends on duty cycle, refrigerant, capacity control needs, and serviceability. A decision tree.
• Industrial ammonia (R717) — still the king of cold storage — Ammonia has been the industrial cold-storage refrigerant for a century. F-Gas Regulation 2024/573 has pushed even more capacity back onto R717. The compressor ecosystem is mature; the safety playbook is well-understood; the COP is unmatched at industrial scale.
• Heat reclaim from R744 transcritical — the killer feature — Transcritical CO2 systems reject heat at 60-90 °C — perfect for store heating, DHW, or floor heating. A reclaim-equipped pack covers 80-100 % of winter space heating in temperate climates. Often the dominant economic argument vs R454C centralised.
• Refrigerant pricing 2026 — R404A scarcity meets HFO ramp — Virgin R404A prices in the EU have tripled since 2023. Reclaimed is the only legal new supply but tight. Meanwhile HFO blends (R454C, R455A) and naturals (R744, R290) are ramping with very different supply dynamics. A pricing snapshot.
• Inverter vs fixed-speed compressors — when the premium pays off — An inverter compressor costs 30-60 % more than the fixed-speed equivalent. The premium pays back in three patterns of duty: highly-variable load, long off-design hours, and tight discharge-temperature control. Picking right needs the operation simulation, not the nameplate.
• R32 vs R454B for residential AC — picking your A2L — Residential air conditioning has moved to A2L mildly-flammable refrigerants. R32 dominates today's stock; R454B is the OEM-favoured successor. The decision drivers are GWP, COP at high ambient, manufacturer ecosystem, and the regulatory window.
• F-Gas and natural-refrigerant subsidies in Europe — country-by-country 2026 — Germany BAFA, France CEE, Italy Conto Termico, UK BUS, Spain MOVES. The national subsidy landscape that covers natural-refrigerant transitions in 2026, what each pays, and how to qualify.
• Retrofit vs replace — the F-Gas-era compressor decision framework — Retrofitting an R404A compressor to R448A is cheaper but defers the problem. Replacing with R744 is capital-heavy but future-proof. A framework for picking the right path per system age, duty, and capital plan.
• F-Gas Regulation 2026 — what changes for refrigeration compressors — EU F-Gas Regulation (EU) 2024/573 applied from March 2024. The HFC quota does not move on 1 January 2026 — but two bans land, and the quota nearly halves in 2027. Here is what it means for compressor selection, retrofits, and the refrigerants you can still buy.
• Envelope collapse: why a whole compressor series shares one operating limit — You might expect every compressor + refrigerant pair to carry its own operating envelope. It doesn't. Across five brands' digitised vendor data, 20,621 model-and-refrigerant combinations collapse to 1,459 distinct envelope polygons — each shape reused 6 to 27 times. The envelope is fixed by the platform, the refrigerant, and the application variant, and is independent of displacement. We show the pattern, the physics behind it, and the one place it breaks.
• How to pick an R404A replacement in five minutes — DigiEntropy's F-Gas Retrofit tool walkthrough — R404A is on borrowed time. Picking the right drop-in or near-drop-in replacement used to require spreadsheets, datasheet hunting, and guesswork on PFAS exposure. Here's the five-minute workflow with the DigiEntropy F-Gas Retrofit tool.
• AVP Quotation — condensing-unit pricing across brands — AVP-format quotations (Aria Verde Pacchetto — the Italian standard for commercial condensing units) used to mean three Excel templates and a calculator. DigiEntropy's AVP Quotation page produces them for any condensing unit in the corpus.
• PH chart fundamentals — reading the vapour-compression cycle — The pressure-enthalpy diagram is the engineer's microscope for refrigeration. Compressor work, evaporator duty, condenser duty, superheat, subcooling — all geometric on the right plot. The PH Chart page lets you draw any refrigerant cycle in seconds.
• R744 (CO2) vs R290 (propane) — picking the right natural refrigerant for commercial refrigeration — Both R744 and R290 survive every regulatory threat on the horizon — F-Gas, PFAS, the lot. Both are commercially mature. So which one for your project? A frank comparison from someone who's watched the catalog data accumulate.
• Lifelong cost — what a compressor really costs over 15 years — A compressor's sticker price is the smallest of four numbers. Lifelong Cost prices all of them — initial, installation, operational and downtime — from real catalogue, tariff and city data, so the cheapest tag doesn't win by default.
• Compliance + TEWI in one click — F-Gas direct emissions + indirect kWh emissions + GWP + safety class + EN 378 charge limit, all in one regulatory verdict and one PDF. The Compliance page is the slide auditors and sustainability committees actually look at.
• Line components — the EXV, solenoid, sight glass, drier picker — Line components rarely get attention in compressor-focused tools. Wrong size or wrong refrigerant compatibility quietly kills systems in the field. DigiEntropy's Line Components page centralises them across Castel, Danfoss, Sanhua, and others.
• Unit cooler and condenser selection for cold rooms — the catalog pair — Cold-room sizing is three components, not one. The compressor is the headline; the unit cooler (evaporator) and condenser set the SST/SDT the compressor actually sees. Pick them wrong and the compressor's nameplate doesn't match reality.
• Supermarket transcritical R744 pack sizing — start to spec — Transcritical CO2 has dominated new European supermarket refrigeration since ~2018. Pack sizing has its own conventions — booster vs flooded, parallel compression, ejectors, gas cooler approach. The Supermarket Refrigeration page walks through it.
• R448A vs R449A vs R454C — comparing the three R404A drop-in refrigerants — When you retrofit an R404A system, three blends dominate the shortlist: R448A, R449A, and R454C. They overlap on paper. They diverge once you look at PFAS exposure, capacity match at LT, and 2026/2030 phase-out exposure.
• The PFAS restriction proposal — what it could mean for HFO refrigerants — ECHA's proposed universal PFAS restriction under REACH could phase out HFOs and HFC/HFO blends starting in 2026-2028. If it lands as drafted, natural refrigerants become the only future-proof choice for long-life refrigeration systems.
• System Simulator — drag-and-drop refrigeration loop design — Compressor + condenser + evaporator + expansion valve + pipes — the full refrigeration loop simulated against vendor polynomials and a closed-loop solver. System Simulator is where you go after Cold Room when the loop topology is the unknown.
• Heat-pump compressor selection in five minutes — Air-to-water heat pumps need a compressor sized for HEATING capacity at the design ambient, not cooling at MT. Most selection tools default to cooling. DigiEntropy's Heat Pump page flips the entire workflow to heating mode so the selection actually matches the duty.
• Condensing temperature and ambient conditions: the biggest variable nobody talks about — Most refrigeration systems are designed for a peak summer ambient. But the same system will run at very different condensing temperatures across the year, and each degree of condensing temperature costs or saves measurable electricity. Understanding how ambient conditions propagate through the system allows engineers to optimise for annual energy consumption, not just peak load.
• Two-stage compression and flash-gas economy: why low-temperature systems need extra help — Single-stage compression works well down to around −30 °C evaporating temperature. Below that, the pressure ratio climbs to the point where volumetric efficiency collapses, discharge temperatures become dangerous, and COP is poor. Two-stage systems with flash-gas economisers recover much of that efficiency.
• Isentropic efficiency: how real compressors deviate from the ideal — The isentropic process is a theoretical ideal — no friction, no heat loss, perfectly reversible. Real compressors fall short of it. Isentropic efficiency quantifies that gap and explains why two compressors with identical displacement can consume very different amounts of electricity for the same cooling output.
• Refrigerating effect and mass flow: the engine behind cooling capacity — Cooling capacity is the product of two quantities: how much refrigerant flows per second, and how much heat each kilogram absorbs in the evaporator. Understanding both reveals why the same compressor can deliver radically different capacities depending on refrigerant, operating conditions, and pipe sizing.
• Superheat and subcooling: two small numbers that change everything — Every refrigeration system runs with a superheat at the evaporator outlet and a subcooling at the condenser outlet. These two deltas look small — typically 5–15 K each — but they shift capacity, discharge temperature, and COP in ways that compound across the whole system.
• Pressure drop and line sizing: the hidden performance killer in refrigeration circuits — Pipe pressure drop in a refrigeration circuit is not just a hydraulic nuisance — it directly reduces the effective evaporating temperature and degrades COP. This post explains the physics, quantifies the penalty, and shows the practical line-sizing rules that keep pressure drop within acceptable limits.
• Heat exchanger sizing: from LMTD theory to catalog selection — Condenser and evaporator sizing rests on a small set of heat transfer equations that every engineer should be able to apply by hand before trusting a selection tool. This post works through the LMTD method, explains approach temperature as the central design variable, and shows how the theory maps to real selection results.
• Refrigerant selection: matching fluid properties to application requirements — Choosing the right refrigerant is not just a regulatory box-ticking exercise — the fluid's thermodynamic properties directly determine the system's operating pressures, efficiency, compressor displacement, and long-term serviceability. This guide explains the key properties and how they map to application requirements.
• Coefficient of Performance: what drives efficiency in a vapour-compression system — COP is the single number that summarises a refrigeration system's thermodynamic efficiency. This post unpacks its definition, explains what the Carnot limit tells us about the maximum possible COP, and shows how real-world deviations from that ideal show up in measured data and simulation results.
• The pressure-enthalpy diagram: reading the heart of a refrigeration cycle — The p-h diagram encodes every thermodynamic trade-off in a single chart. This guide explains how to read it, what each process line means for real hardware, and how a running simulation maps its calculated operating point directly onto the diagram.
• Real parts, real cycle — how System Simulator stays accurate across brands — A 12 kW R448A medium-temp cold room, built end-to-end in System Simulator with a Bitzer compressor, Güntner heat exchangers, Castel valves, Danfoss drier and EXV, and ASTM B280 copper. The same physics module drives every component's pressure drop, so the converged cycle agrees with the vendor curves you would read from each manufacturer's own selector.
• Cross-brand compressor benchmark in one click — Bitzer vs Frascold vs Copeland on the same chart — Picking between a Bitzer 4PCS-10.2 and a Frascold S30-153Y used to mean two PDF datasheets, a calculator, and an hour. With the DigiEntropy Benchmark tool it's two clicks. Here's how it works and why cross-brand benchmarking finally got easy.
• Cold room sizing in five minutes — load, system, simulation in one tool — Cold room sizing has always been a multi-tool workflow: a load spreadsheet, then a compressor selector, then a simulation. The DigiEntropy Cold Room page collapses it into one — and you can go from blank to specified system in about five minutes.
• Screw rotor universality: within-rotor and across-rotor η_v collapse — The kinematic mass-flow law extends to twin-screw compressors with the rotor swept volume in place of N_cyl × V_cyl. We show that the η_v(Π) curves of every motor variant in a single Bitzer CSH rotor family overlap to within ~3 % CoV, and that the curves across distinct CSH rotors collapse to within ~4 % CoV — tighter than the equivalent reciprocating result.
• The universal predictor: three parameters from geometry to capacity — A closed-form capacity calculator built on the kinematic equation and the universal η_v(Π) curve reproduces vendor EN 12900 polynomials to a mean absolute error of ~4 % across multiple Bitzer reciprocating families on R134a and R404A — using only three universal parameters per refrigerant.
• From PDF catalog to polynomial fit in 60 seconds — the PolyFitter tool — Got a compressor datasheet that publishes a capacity table but no polynomial coefficients? Paste the numbers into PolyFitter and get a clean AHRI 540 / EN 12900 fit in seconds — with residual, 3D surface, and contour plots to validate the fit visually.
• Geometry collapse: one volumetric efficiency curve fits the entire Octagon family — We extend the η_v(Π) analysis to all reciprocating geometries in the Bitzer Octagon family. A single universal clearance-volume law with three parameters (ε, n, β_bore) fits the empirical volumetric efficiency to a mean absolute error of ~3 % on R134a. The dominant residual is explained by cylinder bore.
• Cylinder-count symmetry: η_v(Π) is invariant to N_cyl in semi-hermetic reciprocating compressors — The kinematic mass-flow law treats cylinder count as a simple multiplier on swept volume. We verify the prediction by computing the empirical volumetric efficiency η_v(Π) for several Bitzer HE-family models on R404A and observe that the curves overlap within ~2 % across cylinder counts.
• Frequency collapse: mass flow is linear in shaft speed for reciprocating compressors — An empirical re-derivation of the kinematic mass-flow law from vendor polynomial coefficients. For a reciprocating compressor, ṁ scales linearly with rotational speed N to within ~0.2 % across 30–70 Hz inverter operation. We show the result for one Bitzer Octagon model and note that BOCK and Hanbell inverter ranges share the same form.
• The Compressor Galaxy — 12,000+ refrigeration compressors plotted on a single chart — 13 brands. 12,000+ compressor models. Every qualified refrigerant. All on one interactive chart you can pan, zoom, and filter. The Compressor Galaxy is the discovery layer of DigiEntropy — and it's how engineers find candidates they didn't know existed.