CRGO lamination core loss (W/kg) vs flux density: practical selection guide

1. Core loss vs flux density: what the curve really says

CRGO datasheets usually give you a couple of hard points:

• P1.5/50 or W15/50
• P1.7/50 or W17/50

Sometimes only one of them. The mill guarantees one induction point; everything else on the curve is “typical”.

The problem: you never run the core exactly at the guarantee point. You run at “whatever flux density the design and tolerances decide”, plus over-flux events, plus noise limits, plus your stacking reality.

So treat the P–B curve like this:

• Guarantee point: anchor for purchasing.
• Full P(B) curve: anchor for design.

A typical CRGO data set at 50 Hz, from a classic GOES brochure, looks like this for 0.23 and 0.27 mm grades (M-3 and M-4):

Three quiet points that engineers know, but purchasing sometimes doesn’t:

1. Loss jumps hard between 1.5 T and 1.7 T. For M-3, the table above shows ~52% increase (0.658 → 1.002 W/kg) just from that 0.2 T change.
2. Thickness hurts more at higher B. At 1.0 T, M-3 vs M-4 differ by ~0.055 W/kg. At 1.7 T, the gap is ~0.14 W/kg. Eddy terms are doing their job.
3. The “M-grade” label is only a window. Modern catalogues put typical 0.23 mm “M3” around 0.7–0.8 W/kg at 1.5 T and ~1.08–1.17 W/kg at 1.7 T, depending on mill and generation of steel.

So when somebody says “this is M3, 0.23 mm”, that’s not enough. You still need the curve, or at least two points on it.

2. Decide your flux density window first, then your grade

You can pick CRGO grade and lamination thickness in many ways. The most boring one works best:

Fix a realistic operating flux density band, then buy steel that behaves acceptably inside that band.

Approximate working windows at 50 Hz, for CRGO cores in oil, assuming ONAN/ONAF transformers and decent cooling margins:

• Distribution transformers (≤630 kVA)
  • B_work: 1.55–1.65 T
  • High-loss spec lines often use 1.7 T at rated tap; low-loss variants prefer 1.6 T.
• Medium power transformers (up to ~40 MVA)
  • B_work: 1.6–1.7 T
  • Push towards 1.7 T only when no-load loss penalties are mild and footprint matters; noise and over-flux limits bite here.
• Dry-type transformers
  • B_work: 1.5–1.6 T
  • Noise and partial discharge constraints usually pull flux down a notch; dry types are less forgiving to over-flux and local saturation.
• Reactors, specialty inductors at 50/60 Hz
  • Anything from 1.2–1.5 T, depending on ripple, dc bias and loss budget.

These are not rulebook values. They’re “numbers people quietly use” because they survive field experience and grid over-voltage habits.

Once you agree this window internally, grade selection gets much less noisy.

3. Translating a flux window into core loss expectations

Let’s use the M-3 / M-4 table as a simple model and assume your design sits at ~1.55 T in steady operation.

Engineers know loss vs B is not a perfect power law, but between 1.3–1.7 T it behaves “roughly” like:

P(B) ≈ P_ref · (B / B_ref)^n, with n somewhere around 1.6–2.0 depending on steel and frequency.

Now line up some scenarios at 50 Hz:

• M-3 at 1.5 T: P ≈ 0.66 W/kg (table)
• M-3 at 1.6 T: P ≈ 0.79 W/kg (table)
• M-3 at 1.7 T: P ≈ 1.00 W/kg (table)

For a 2,000 kg core, that’s:

• ~1.3 kW no-load at 1.5 T
• ~1.6 kW at 1.6 T
• ~2.0 kW at 1.7 T

Same steel, same lamination stacks. Only B moves.

So for a purchasing engineer, P1.5/50 and P1.7/50 are not just catalogue numbers – they are a quick way to sketch out exactly how much penalty you pay if the designer raises the flux 0.1 T to save copper.

4. When does it make sense to pay for Hi-B or domain-refined steel?

Most Hi-B or laser-scribed grades sit roughly one step “better” on the loss curve than conventional CRGO at the same thickness. Typical P1.7/50 values around 0.7–0.9 W/kg at 0.23–0.30 mm are common in modern catalogues.

That doesn’t automatically mean you should buy them.

Think in three quick passes:

1. Loss penalty lifetime cost
   • Use your utility or internal owning-cost model. Convert an extra 0.2–0.3 W/kg at your working flux into kWh over the guaranteed life.
   • Compare to the premium per kg of lamination stack for Hi-B.
2. Design push
   • If you are already at B_work ≥ 1.65 T and close to noise or temperature limits, cheaper conventional grades give you very little room.
   • Hi-B buys you either lower loss at the same B, or similar loss at slightly higher B (smaller core, less copper).
3. Spec stability
   • If your RFQ simply says “M3, 0.23 mm” without P1.5/50 or P1.7/50 numbers and test conditions, mills will offer whatever sits in their “M3-ish” drawer that month. That can be conventional CRGO one year and a mix with high-permeability variants the next.

In short: pay for Hi-B when you either:

• have a contractual penalty on no-load loss, or
• genuinely need the compactness/noise performance and have done the math.

Otherwise, a well-specified conventional CRGO (with explicit W/kg limits) plus a sensible B window is usually enough.

5. Lamination thickness: 0.23 vs 0.27 vs 0.30 mm in the real world

Plenty of blog posts already list thickness vs loss qualitatively. The usual story still holds: thinner strip, lower eddy losses, better at higher flux and frequency – and higher processing cost.

A practical way to think about it:

• 0.23 mm CRGO (often “M3”)
  • Good balance for distribution and many power transformers.
  • Typical P1.7/50 in real offers: roughly 1.0–1.2 W/kg.
• 0.27 mm CRGO (often “M4”)
  • Cheaper, easier processing, slightly higher loss, especially above 1.6 T.
  • Notice from the table earlier: the loss gap vs 0.23 mm widens as you go from 1.3 T → 1.7 T.
• 0.30 mm and 0.35 mm (“M5/M6”)
  • Attractive on price per kg.
  • A lot less attractive on no-load loss when pushed near 1.7 T, except for very cost-driven or retrofit projects.

So rather than “0.23 is premium, 0.27 is standard, 0.30 is budget”, phrase it this way:

“For a given B window and loss target, what thickness gives you the cheapest total package when you include copper, tank, and penalties on kWh?”

Many modern guides now explicitly show those trade-offs using total owning cost curves for distribution transformers.

6. Material data vs lamination stacks: correcting the fantasy

Data sheets are measured on carefully prepared strips. Your core is not a strip.

Three correction factors matter more than the rest:

6.1 Lamination / stacking factor

Spacemat’s GOES brochure shows typical lamination factors for CRGO around 95–97% at 50 psi, depending on thickness and coating.

That means:

• If your CAD model assumed “100% steel” stack height, you are already off by a few percent in effective cross-section.
• At fixed volts per turn, that translates into higher actual B than you think.
• Higher B shifts you further up the P(B) curve, making the real loss closer to model × lamination factor × “flux squeeze” penalty.

6.2 Building factor (core vs Epstein frame)

Nippon Steel’s ORIENTCORE HI-B data gives a neat comparison:

• Material core loss tested on strip: e.g. 1.48 W/kg at 1.7 T, 60 Hz.
• Built three-phase transformer core: around 1.72 W/kg at the same nominal point.
• Building factor ≈ 1.16.

Corner joints, flux rotation, local saturation in T-joints, air gaps at overlaps – they all add extra watts that never show in the bare strip test.

For conventional CRGO wound or stacked cores, building factors between about 1.1 and 1.3 are common depending on design and lap style.

6.3 Temperature

Counter-intuitive, but worth remembering: for GOES, core loss measured at 85 °C is often slightly lower than at 25 °C because resistivity increases with temperature and reduces eddy currents. Spacemat’s table shows W(85 °C)/W(25 °C) hovering around 0.95–0.98 across 1.0–1.7 T.

So if your spec quotes P1.7/50 “at 65 °C” and the datasheet quotes “at 20–25 °C”, losses will not scale in the obvious way. You still validate on the mill’s stated test condition.

7. How purchasing and engineering can specify CRGO lamination stacks together

Here’s a simple workflow that turns all the above into a defendable RFQ.

Step 1 – Freeze the design side inputs

From the transformer designer:

• B_work at rated tap (e.g. 1.60 T) and expected over-flux events (e.g. +10% for 1 minute).
• Target no-load loss at rated voltage and temperature.
• Core type (3-leg vs 5-leg, shell vs core), joint type (miter/step-lap), winding arrangement.

This lets you estimate:

• Required P(B_work) on strip,
• plus building factor,
• plus lamination factor.

Step 2 – Convert to material targets

For example, suppose:

• 50 Hz, ONAN distribution transformer.
• B_work ≈ 1.6 T, 0.23 mm CRGO, 2,000 kg core.
• You need core loss ≤ 1.7 kW at rated conditions.

Assume:

• Building factor ≈ 1.18 (stacked step-lap core).
• Lamination factor ≈ 96%.

Then the strip level target at 1.6 T becomes roughly:

Core loss per kg (strip) ≈ 1.7 kW / (2000 kg × 1.18) ≈ 0.72 W/kg at 1.6 T

From the M-3 table, 0.23 mm gives ~0.79 W/kg at 1.6 T, which is a bit higher. That tells you:

• Either tighten grade (closer to high-end M2/M3 or Hi-B),
• Or reduce B_work a little,
• Or accept higher no-load losses.

This is the kind of arithmetic that should show up in the design notes, not just in someone’s head.

Step 3 – Put it into RFQ language

Instead of “CRGO M3, 0.23 mm”, write something like:

CRGO lamination stacks, 0.23 mm, grade equivalent to M108-23 or better.

• P1.5/50 ≤ 0.70 W/kg, guaranteed as per IEC 60404-2 / JIS C 2550-1
• P1.7/50 ≤ 1.05 W/kg, same test conditions
• B50 ≥ 1.88 T (5000 A/m)
• Coating: C-5 equivalent, suitable for stress-relief annealing at 800 °C
• Lamination factor at 50 psi ≥ 96%

Numbers above are indicative, but this style of sentence is what keeps both sides honest.

Step 4 – Ask for the full P(B) curve

Don’t rely only on catalogue summary lines.

Request from your lamination supplier:

• Core loss vs B at 50 Hz over at least 1.3–1.7 T for the offered grade and thickness.
• Indicate whether numbers are “typical” or “guaranteed”.

If they cannot supply the curve, they at least should tell you which mill datasheet they are actually buying from.

8. Special cases where W/kg vs B gets tricky

A few situations where the nice Epstein-frame curve misleads you:

1. Complex joints and 5-leg cores
   • Local B in T-joints and yokes can run 10–20% higher than leg B. Losses there scale badly and dominate your noise complaints.
2. Mixed grades in one core
   • Some recent work mixes grades (e.g. Hi-B in legs, conventional in yokes) to balance cost and loss. Total P(B) then needs a weighted average, not a single W/kg.
3. Stress from punching and stacking
   • “As-sheared” vs “annealed” conditions affect the whole curve, not just the guarantee value. Differences of several tenths of a W/kg at 1.5 T are documented for stressed vs stress-relieved GOES.
4. DC bias and unbalanced loading
   • If your core sees dc offset or heavy harmonic content, Steinmetz-style models calibrated at sinusoidal B may under- or over-predict; exponents change with flux range and frequency.

When any of these show up, you either:

• test representative lamination stacks yourself, or
• insist on suitable model parameters from the steel supplier.

9. Quick checklist for CRGO lamination stack orders

For each new transformer design or major RFQ, make sure you can answer, on one page:

1. Flux density window
   • B_work at rated, B_max under worst over-flux event.
2. Target strip-level losses
   • P(B_work) per kg, implied from your allowed kW and assumed building factor.
3. Grade and thickness band
   • M-series label or IEC code, plus thickness (0.23 / 0.27 / 0.30 mm).
4. Guaranteed test points
   • P1.5/50 and/or P1.7/50 limits, test standard, sample condition (as-sheared vs annealed).
5. Lamination-specific details
   • Lamination factor requirement.
   • Burr height, flatness, coating type, annealing route.
6. Verification plan
   • How you will sample incoming coils or lamination stacks (frequency, lot size, test method).

If any of these boxes is empty, the core loss vs flux density curve will usually fill it in for you later, in the form of unexpected watts.

10. FAQ: CRGO lamination core loss vs flux density