Explicit time step and mass scaling calculator

Work out the stable time step of a solid, shell or beam element, how many cycles your explicit run needs, and what mass scaling costs: the added mass for a target step, the largest step your mesh allows within an added-mass limit, and the DT2MS to type. You get the matching LS-DYNA *CONTROL_TIMESTEP, Radioss /DT/NODA/CST and Abaqus *FIXED MASS SCALING lines. Free, no sign-up, nothing leaves your browser.

Δt = TSSFAC × min(Le/c) · added mass m[(Δttarget/Δte)² − 1] · checked against an independent script and 145 LS-DYNA challenge reports · Updated 21 Nov 2026

1. One element

Size = edge length of a regular element (beam: its length). For a distorted element, enter the characteristic length your pre-processor reports and choose the hex (it uses the size as given).

Solver time step0.830 µsTSSFAC 0.9 × critical 0.922 µs
Cycles to end time1,20,487for 100 ms

    2. Your mesh: how far can you scale?

    Largest step within limit0.760 µsafter TSSFAC; 4.6× faster than unscaled
    Set DT2MS (kg-mm-ms)-8.440e-4with TSSFAC 0.9
    Your mesh: added mass against the time step0.1520%0.3275%0.50310%0.67815%0.85320%1.0325%solver time step before TSSFAC (µs)added mass (% of total)limit 2%
    added massyour limitlargest step within it

      3. Already ran trials? Hit the limit exactly

      |DT2MS| for the limit1.0288e-4from the runs at 1.00e-4 and 1.03e-4

        Cards use the largest step within your limit (planner, section 2). Set a target step in section 1 to use that instead.

        LS-DYNA *CONTROL_TIMESTEP

        $ time unit: ms (kg, mm, ms)
        $ step = TSSFAC x |DT2MS| = 7.59637E-4 ms
        *CONTROL_TIMESTEP
        $   DTINIT    TSSFAC      ISDO    TSLIMT     DT2MS      LCTM     ERODE     MS1ST
               0.0       0.9         0       0.0 -8.440E-4         0         0         0

        Radioss engine file

        # time unit: ms (kg, mm, ms)
        /DT/NODA/CST/0
        #            DT_sca              DT_min
                         0.9          7.59637E-4

        Abaqus/Explicit

        ** time unit: ms (kg, mm, ms); goes inside the *STEP, after *DYNAMIC, EXPLICIT
        *FIXED MASS SCALING, DT=7.59637E-4, TYPE=BELOW MIN

        How is the explicit time step calculated?

        An explicit solver (LS-DYNA, Radioss, Abaqus/Explicit) moves the model forward in small time steps. It stays stable only if a stress wave cannot cross an element in one step, the Courant condition. So every element gets its own limit, its characteristic length divided by the speed of sound in it, and the solver takes the smallest, times a safety factor:

        The sound speed depends on the element type, because a shell cannot bulge sideways the way a solid can and a beam only carries axial waves:

        For a 5 mm square steel shell (E = 210 GPa, ν = 0.3, ρ = 7,850 kg/m³):

        A 100 ms crash event at that step needs 120,487 cycles. The same 5 mm steel element as a hex solid steps at 0.833 µs (sound is faster in a solid, 6,001 m/s), as a regular tetrahedron at 0.680 µs (its length is the shortest altitude, 0.82 × the edge), as a triangle shell at 0.799 µs, and as a beam at 0.967 µs. With shell ISDO = 1 the square shell's length is the area over the diagonal, so 0.652 µs. Aluminium, lighter but softer, comes out almost the same as steel: 0.927 µs.

        5 mm steel elementLength usedWave speedCritical step
        Hex solid5.00 mm6,001 m/s0.833 µs
        Tetra solid4.08 mm6,001 m/s0.680 µs
        Quad shell, ISDO 05.00 mm5,422 m/s0.922 µs
        Quad shell, ISDO 13.54 mm5,422 m/s0.652 µs
        Triangle shell4.33 mm5,422 m/s0.799 µs
        Beam / truss5.00 mm5,172 m/s0.967 µs

        What does mass scaling do?

        The element step grows with the square root of density. So instead of re-meshing, you can make the few small elements heavier until their step reaches the one you want:

        To double a tiny element's step you multiply its mass by four, but if that element weighs a gram, the model barely notices. That is the whole trick: a mesh with a handful of small elements can often run several times faster for a fraction of a percent of extra mass. The catch is that the cost climbs steeply once the target reaches the step of the bulk of the mesh, because then thousands of elements need mass at once.

        Worked example: a 2 mm steel sheet with a few small elements

        Take an example shell mesh (made up for illustration, steel, 2 mm thick) of 9,400 elements at 5 mm, 300 at 3 mm around holes and fillets, 60 at 2 mm and 8 at 1 mm where two flanges meet. It weighs 3.736 kg. Unscaled, the eight 1 mm elements set the step: 0.1844 µs before TSSFAC.

        • Bring everything up to the 2 mm elements' step (0.3689 µs, twice as fast): added mass 0.010% of the model.
        • Up to the 3 mm elements' step (0.5533 µs, three times as fast): 0.15%.
        • The largest step within 2% added mass is 0.8440 µs, 4.58× the unscaled step, with all 368 small elements scaled. Within 1% it is 0.7018 µs; within 5%, 0.9328 µs, which now scales all 9,768 elements because it is past the 5 mm elements' own 0.9222 µs.
        Example 2: added mass against the time step for a 2 mm steel shell mesh with a few small elements0.1520%0.3275%0.50310%0.67815%0.85320%1.0325%solver time step before TSSFAC (µs)added mass (% of total)limit 2%
        added mass2% limitlargest step within 2%
        Added mass against the step for the example mesh. It rises slowly while only the 368 small elements take mass, then jumps once the step passes the 5 mm elements' own 0.922 µs. It is the sample in the planner above.

        TSSFAC and DT2MS in LS-DYNA

        Both live on *CONTROL_TIMESTEP. TSSFAC (default 0.9, which the manual says should stay below 1.0 for stability) scales the computed step. DT2MS turns on mass scaling: with a negative value, LS-DYNA adds mass only to elements whose own step is below |DT2MS|, just enough to bring them up to it, and then runs at

        Two things follow, and both trip people up:

        • Added mass depends on |DT2MS| alone, not on TSSFAC. Lowering TSSFAC at the same DT2MS gives the same added mass and a smaller step, so a slower run, never a cheaper one.
        • For a given step, keep TSSFAC high. Δt = 0.9 × 1.0 µs and Δt = 0.6 × 1.5 µs are the same step, but the second scales every element up to 1.5 µs and adds much more mass. Keep TSSFAC at its default and set DT2MS = −Δt / 0.9.

        A positive DT2MS is different: every element is scaled to that step, mass is added to small ones and removed from large ones. It is used for quasi-static runs, rarely in crash. In Radioss the equivalent of the negative form is /DT/NODA/CST in the engine file; in Abaqus/Explicit it is *FIXED MASS SCALING with TYPE=BELOW MIN.

        What 145 student reports on mass scaling show

        Skill-Lync's LS-DYNA course has a mass scaling challenge: a steel solid model in kg-m-s with a 600 s end time is given with DT2MS = −3.5E-05 and TSSFAC = 0.9; reduce the run time by changing DT2MS and TSSFAC without more than 8% added mass, then run it implicit and compare. We read the 145 public reports on Skill-Lync (127 titled "Week - 8 Mass Scaling", as of 21 Nov 2026). Because every student ran the same model, LS-DYNA printed the same added mass for the same DT2MS across reports, so their trials add up to one curve:

        Added mass reported by LS-DYNA against |DT2MS| in the Week 8 mass scaling challenge73.60%81.25%88.710%96.315%10420%11125%|DT2MS| (µs)added mass (% of total)limit 8%
        a trial in the reports8% limit|DT2MS| = 102.89 µs
        Added mass LS-DYNA reported against |DT2MS| (in µs; the model is in seconds, so 1.0289E-04 = 102.89 µs). Each point appears in 2 to 31 reports. From 100 to 110 µs the added mass goes from 6.4233% to 20.393%.
        • The limit sits at |DT2MS| = 1.0289E-04 (7.999%). Reports get there by trial and error, one run per digit (one report lists 16 explicit runs). The last 2.9% of step (100 to 102.89 µs) costs 1.58 points of added mass: the curve is past the knee.
        • Two runs are enough. Between two runs in which the same elements get mass, added mass is exactly a·T² − b:

        From the runs at 102.8 µs (7.9446%) and 103 µs (8.0661%):

        The same two runs predict 7.9992% at 102.89 µs and 8.0053% at 102.9 µs; LS-DYNA printed 7.999% and 8.005%. The runs at 100 and 103 µs give 102.88 µs, also right. Runs at 100 and 105 µs give 101.85 µs, too low, because between 103 and 105 µs a new batch of elements starts taking mass (the jump from 8.8759% to 10.748%). Bracket the limit closely, interpolate, then confirm with one run. Section 3 of the calculator does this.

        • Just over the line. 8 reports (two shared write-ups) settle on DT2MS = −1.029E-04, which LS-DYNA reports as 8.005%, over the 8% limit. The fourth digit matters at this slope.
        • TSSFAC below 0.9. 60 reports also try TSSFAC of 0.8, 0.7 or lower. The reports themselves show why that cannot help: at −1.03E-04 both 0.9 and 0.8 give 8.0661% added mass, while the step drops from 92.7 to 82.4 µs.
        • Estimated clock times are noisy. At least 56 reports compare runs by the clock time LS-DYNA estimates early in the run, which moves with machine load. One report lists 23 h 34 min for −1.0E-04 at TSSFAC 0.9 and 18 h 56 min at 0.8, although 0.8 takes 11% smaller steps for the same 6.42% mass. Compare cycle counts, or the final elapsed time on an idle machine.
        • Units. 11 reports call the end time 600 ms; the model is in kg-m-s (E = 2 × 1011, so pascals), so it is 600 s. At a step of 92.6 µs that is 6.48 million cycles, against up to 19.0 million at the starting 0.9 × 35 µs. The Abaqus manual calls explicit "computationally attractive for problems where the total dynamic response time that must be modeled is only a few orders of magnitude longer than the stability limit"; here it is almost seven. That is why the implicit runs in the same reports finish in minutes to an hour.

        What this calculator does not do

        • Distorted elements. It uses the length of a regular element of the size you give. A skewed or squashed element has a shorter length: enter the characteristic length your pre-processor or the d3hsp file reports.
        • Bulk viscosity and damping. LS-DYNA's solid formula includes the bulk viscosity term, which shortens the step slightly in elements being compressed fast. Ignored here (zero strain rate).
        • Contact, rigid bodies, springs and discrete elements. They have their own step limits; contact stiffness in particular can cut the step.
        • Selective mass scaling (LS-DYNA IMSCL, Abaqus variable mass scaling) adds mass only to high-frequency modes; this calculator covers the conventional kind.
        • Added mass during the run. The numbers here are for the undeformed mesh. As elements crush and get shorter, LS-DYNA keeps adding mass to hold the step (MS1ST = 0, the default), so the final added mass can be higher than at the start. Read it from glstat at the end of the run.
        • Radioss nodal step. /DT/NODA/CST works on nodal masses and stiffnesses, not elements, so its added mass differs a little from this element-based estimate. Radioss Starter prints its own table of target steps for 0.5% to 10% added mass.
        • Whether the result is still right. It tells you how much mass, not what it does. Check the added mass per part, kinetic and hourglass energy, and compare with a run at a smaller step.
        Explicit dynamics and crash

        Time step control, mass scaling and explicit vs implicit runs are a week of Skill-Lync's LS-DYNA course, and a standard question in CAE crash interviews.

        Frequently asked questions

        How is the time step calculated in LS-DYNA?

        For every element LS-DYNA divides a characteristic length by the speed of sound in the material, takes the smallest of these, and multiplies it by the scale factor TSSFAC (0.9 by default). For a 5 mm steel shell that is 0.9 × 0.005 / 5,422 m/s = 0.83 µs. The smallest or most distorted element sets the step for the whole model.

        What is TSSFAC in LS-DYNA?

        TSSFAC is the scale factor on the computed time step in *CONTROL_TIMESTEP: 0.9 by default (lower defaults apply with some contact settings and detonation). The manual asks for a value below 1.0 for stability. It keeps the step below the Courant limit for stability. Lowering it makes every cycle smaller and the run longer. With mass scaling on (DT2MS negative) the step becomes TSSFAC × |DT2MS|, so lowering TSSFAC alone shrinks the step without reducing added mass.

        What does a negative DT2MS mean in LS-DYNA?

        A negative DT2MS turns on mass scaling only where it is needed: LS-DYNA adds mass to elements whose own time step is below |DT2MS| so they reach it, and runs at TSSFAC × |DT2MS|. Elements that are already larger are untouched. To get a step Δt, set DT2MS = −Δt / TSSFAC. A positive DT2MS scales every element to the same step, adding or removing mass.

        How much added mass is acceptable in mass scaling?

        There is no universal number. The LS-DYNA manual allows negative DT2MS in transient runs "if the mass increases remain insignificant"; Altair's Radioss guide recommends keeping added mass below 5%, more for quasi-static runs; Abaqus suggests kinetic energy below about 10% of internal energy in quasi-static runs. Mass in a fast-moving part matters most, so check added mass per part (matsum) and compare with an unscaled run. Skill-Lync's LS-DYNA challenge allowed 8%.

        Why is my explicit time step so small?

        One small or badly shaped element sets the step for the whole model. LS-DYNA lists the 100 smallest element time steps in the d3hsp file; check which part and element they belong to. Fix the mesh there (remove slivers, collapse short edges, use a larger minimum size), or mass-scale those few elements. Stiff, light materials and rigid-to-deformable contacts also lower the step.

        Does mass scaling change the results?

        It can. Added mass adds inertia: a crash part with extra mass carries more momentum and kinetic energy, and natural frequencies drop by the square root of the mass ratio. Scaling a few tiny elements in a stiff region usually changes little. Check the added mass in moving parts, the energy balance (total energy, kinetic energy, hourglass) and compare key results with a run at a smaller step.

        What is the stable time increment in Abaqus/Explicit?

        Abaqus's conservative estimate is the minimum over all elements of the element length divided by the dilatational wave speed, the same idea as LS-DYNA; its manual gives the example of a 5 mm element at 5,000 m/s, about 1 µs. *FIXED MASS SCALING, DT=…, TYPE=BELOW MIN scales only the elements whose stable increment is below DT, like a negative DT2MS in LS-DYNA.

        Related on Skill-Lync

        Sources

        • LS-DYNA Keyword User's Manual Vol I R14, *CONTROL_TIMESTEP: card 1 DTINIT, TSSFAC, ISDO, TSLIMT, DT2MS, LCTM, ERODE, MS1ST; DT2MS "LT.0.0: TSSFAC × |DT2MS| is the minimum time step size permitted and mass scaling is done if and only if it is necessary to meet the Courant time step size criterion"; ISDO 0/1 shell lengths, 3-node shells by shortest altitude; glstat and matsum added mass.
        • LS-DYNA Theory Manual R14, chapter 25, Time Step Control: solid Le = Ve/Amax (tetrahedra: minimum altitude) with bulk viscosity term, shell and beam wave speeds, step = scale factor × minimum over elements. LS-DYNA support, Mass scaling.
        • Altair Radioss, Nodal time step control ("it is recommended to keep the amount of mass added to less than 5%"; MAS.ERR column) and /DT/NODA/Keyword3 (ΔTsca default 0.9, ΔTmin).
        • Abaqus 2025 documentation: Theory Guide, Explicit dynamic analysis (Δt = min(Le/cd), cd from λ + 2μ) and *FIXED MASS SCALING (DT, TYPE = BELOW MIN); Analysis User's Guide, Explicit dynamic analysis (n = T/Δt, kinetic energy "typically less than 10%" of internal energy in quasi-static runs).
        • Worked examples: computed by Skill-Lync with scripts/verify/explicit_timestep.py (element lengths from vertex coordinates, wave speeds from Lamé constants, the added-mass limit solved in closed form) and recomputed by this page's calculator code at every build. The example mesh is made up for illustration. Challenge figures: Skill-Lync's reading of 145 public student reports, 21 Nov 2026.
        Free counselling

        Make crash models that run, and run right.

        Time step, mass scaling and energy checks are what CAE crash interviews ask about. A Skill-Lync counsellor can map the course that fits the job you want.

        Talk to a CAE course counsellor

        Free. We call you back, usually the same day.

        By submitting you agree to our Terms and Privacy Policy.

        Talk to a career counsellor

        Talk to a career counsellor

        Free 20-minute call. We'll map the right course to where you want to work.

        By submitting you agree to our Terms and Privacy Policy.