Validation and differences from the desktop version¶
Clause 10.2 of the Italian building code NTC 2018 asks the designer to assess the reliability of the calculation code they use. That judgement is yours; our job is to give you the material to form it: fields of use and limits, theoretical basis, test cases that are fully worked and repeatable.
This page is the summary. It is a validation, carried out by those who develop the program and verifiable by anyone: no third-party body is involved.
If you come from the desktop program and want to know why a number has changed, go to Differences from the desktop program.
Fields of use and limits¶
Downhole NX processes the down-hole seismic test: a source at the surface, at a distance d from the borehole axis, and a receiver lowered to successive depths. From the first-arrival times of P and S waves it computes:
- the source–receiver distance and the corrected times for the obliquity of the path;
- with the direct method, for each seismic layer defined on the travel-time curve: Vp, Vs, unit weight, Poisson's ratio, G, Ed, E, Ev;
- with the interval method, the same parameters between two consecutive receivers and their averages per seismic layer;
- the equivalent velocity Vs,eq between an initial depth and a depth H (Vs,30 with H = 30 m).
Limits
- The result is only as good as the arrival times. The app can propose them, but a proposal is not a reading: they must always be looked at again on the traces, starting with those the app itself marks as doubtful. The same holds for the proposed seismic layers: they are the subdivision that best follows the times, not a geological interpretation.
- It assumes straight ray paths between source and receiver. With strong velocity contrasts near the surface the real path is refracted and the corrected times of the first receivers are approximate; the effect grows with d and dies out with depth.
- With a single receiver the interval-method velocities are pseudo-interval velocities: the two times come from two different shots and the trigger error enters the difference in full.
- The automatic unit weight comes from an empirical correlation: it is an estimate. If you have measured values, assign them in the table.
- The moduli are dynamic, small-strain moduli: they are not the moduli for a settlement calculation.
- It does not assign the subsoil category.
- It applies no correction for borehole deviation or for the elevation of the source relative to ground level.
- It reads SEG-2 and text traces; it does not merge P and S files with different numbers of channels.
Theoretical basis¶
The formulas, with the references, are on the Theory page: corrected times, direct method, interval method, elastic parameters, unit-weight correlation, equivalent velocity, automatic proposals, rounding.
Test cases¶
Eight cases, repeated automatically at every version of the program. The first five have a closed-form reference that anyone can redo by hand; the sixth is the comparison with the desktop program; the seventh covers anomalous inputs; the eighth the automatic proposals.
| Case | Subject | Reference | Outcome |
|---|---|---|---|
| 1 | Homogeneous medium, Vp = 1 500 and Vs = 300 m/s, d = 3 m, 20 receivers | exact times of the straight path | interval Vs 299.98 ÷ 300.02 (≤ 0.01 %); Vp ≤ 0.03 %; Vs,eq = 300.0 |
| 2 | Two layers: 0–8 m at 600 / 200 m/s, 8–20 m at 1 800 / 450 m/s | closed form | Vs 200.0 and 450.0; Vp 600.0 and 1 800 ± 3; Vs,eq = 300.0 |
| 3 | Elastic parameters with γ = 19.6133 kN/m³, Vs = 200, Vp = 400 m/s | hand calculation | ν = 0.3333; G = 80.00; Ed = 320.00; E = Ev = 213.33 MPa |
| 4 | Unit weight from the correlation | hand calculation | 18.405 kN/m³ (Vs = 200 m/s, z = 5 m); 23.536 (Vs = 1 000 m/s) |
| 5 | Vs,eq of a known column: 0–10 m at 200, 10–30 m at 400 m/s | hand calculation | 300.00 between 0 and 30 m; 333.33 between 5 and 30 m; 200.00 between 0 and 10 m |
| 6 | Example of the desktop program | desktop program | see below |
| 7 | Anomalous inputs | expected behaviour | see below |
| 8 | Automatic proposals of times and seismic layers | operator's readings, known profiles | see below |
The deviations in cases 1 and 2 are the effect of rounding: corrected times to 4 decimals of a millisecond, times at the bases to 2. On a 1 m interval at 1 500 m/s they amount to ± 0.2 m/s.
Consistency check (case 5). In the direct method the sum of the travel times through the layers is the corrected S-wave time at H: Vs,eq cannot depend on where the layers are placed. On the example it comes out at 242.86 m/s with one layer, with five layers and in the interval method, with or without seismic layers.
Case 6 — the desktop example¶
This is the project you open with File → Open the example: 30 receivers every metre from 1 to 30 m, d = 2.6 m, automatic unit weight, five seismic layers with the bases saved by the desktop program (5.04 – 10.30 – 17.78 – 25.32 – 30.11 m).
Corrected times and interval method match the desktop program to the last printed digit, on all 30 rows. For instance, first interval (0–1 m): Vp = 348.21 and Vs = 163.09 m/s, γ = 20.71 kN/m³, ν = 0.3595, G = 56.17, Ed = 256.06, E = 152.73, Ev = 181.17 MPa.
Direct method. With the conventions of the desktop program put back into the calculation, Downhole NX returns exactly the desktop numbers: this shows that the deviations below are the declared corrections, and nothing else.
| Layer | Base [m] | Vp [m/s] | Vs [m/s] | γ NX | γ desktop | G NX [MPa] | G desktop |
|---|---|---|---|---|---|---|---|
| 1 | 5.04 | 367.35 | 182.74 | 18.77 | 18.04 | 63.90 | 61.42 |
| 2 | 10.30 | 482.57 | 209.31 | 18.19 | 18.54 | 81.28 | 82.81 |
| 3 | 17.78 | 1 108.15 | 202.00 | 17.56 | 18.07 | 73.04 | 75.20 |
| 4 | 25.32 | 1 702.03 | 337.21 | 19.27 | 20.05 | 223.39 | 232.49 |
| 5 | 30.00 | 1 181.82 | 409.45 | 19.84 | 21.17 | 339.12 | 379.20 |
In the desktop program layer 5 has its base at 30.11 m, Vp = 1 209.60 and Vs = 419.07 m/s. Vs,30 = 242.86 m/s against 243.37 for the desktop program (−0.2 %).
There are two deviations, both intended:
- Last layer. The desktop program puts the base of the last layer at the slant distance of the last receiver (30.11 m = √(30² + 2.6²)) instead of at its depth (30 m). With the same time the layer comes out thicker, and its velocity is overestimated by 2.3 %.
- Unit weight. The desktop program feeds the correlation with the thickness of the layer instead of the depth. Downhole NX uses the mean depth. The unit weights change by −6 % to +4 %, and with them G, Ed, E, Ev, which depend on them linearly. Velocities and ν do not change.
Case 7 — anomalous inputs¶
| Input | Behaviour |
|---|---|
| Duplicate, out-of-order or non-positive depth | No calculation; the error points to the row |
| S-only test (Tp = 0 everywhere) | P columns “--”; Vs, G and Vs,eq computed |
| A missing time at one receiver | The two intervals touching it stay “--” for that wave; times at the bases are interpolated on the measured receivers |
| Decreasing corrected time | Interval “--”, warning with the receiver number |
| Vp ≤ Vs, or Vp/Vs ≤ √2 | ν and E “--” (and Ev if negative) |
| Decreasing time at the base between two bases (direct method) | The layer takes the parameters of the previous one and is marked “*” |
Case 8 — the automatic proposals¶
The proposals do not enter the calculation: what is measured here is how close they come to what a person chooses.
Arrival times, on a real survey. 18 depths from 1 to 18 m, three-channel SEG-2 records sampled at 64 µs. The comparison is with the times the operator picked by hand.
| Wave | Deviation from the manual pick | Times marked doubtful |
|---|---|---|
| P | median 0.3 ms, maximum 1.4 ms | 4 out of 18 |
| S | median 1.4 ms, maximum 5.3 ms | 10 out of 18, including all those above 4.5 ms |
On P waves the proposal matches the operator's reading within the uncertainty of the reading itself; the doubtful times are those that do not increase with respect to the previous depth, by half a millisecond or less.
On S waves the deviation is a few milliseconds, on times of 40–110 ms, and more than half of the proposals are marked doubtful. At the first two depths the shear wave is buried under the P-wave train; between 6 and 11 m the wavelet changes shape, and there the operator's times are irregular too (74.5 – 75.8 – 77.5 – 84.9 ms). At 18 m the deviation is 4.2 ms and the time is not marked: a time that is not doubtful is not right for that reason. This is why the S proposals are checked one by one.
Arrival times, on synthetic traces (damped waves with a known arrival, on noise): P within 0.6 ms; S within +3 / −6 ms. The early pick comes from the zero-phase filter, which widens the wave train on both sides.
Seismic layers, on profiles with known layers (exact times of three and of five layers, also with 0.5 ms of noise on the times): the bases are recovered exactly, within one metre with the noise, with no extra layers.
Seismic layers, on the example of case 6: the proposal is 6 – 12 – 18 – 27 – 30 m, with a misfit of 0.57 ms; the interpreter had chosen 5.04 – 10.30 – 17.78 – 25.32 – 30 m. The proposed bases lie on the measurement depths, one every metre: the difference is within two receivers. Vs,30 is the same, 242.86 m/s, as it must be (case 5).
Repeating the cases¶
Case 6 is repeated by opening the example and reading the tables of the Processing tab; on the same example Propose seismic layers gives the subdivision of case 8. Cases 1–5 can be redone by hand with the formulas in Theory: enter the measurements in the Measurements tab and compare.
Differences from the desktop program¶
Downhole NX does not reproduce some behaviours of the Downhole desktop program, because they are calculation defects. Anyone comparing an old report with a new one finds here the reason for each deviation.
| Desktop program | Downhole NX | Why | Effect |
|---|---|---|---|
| Base of the last seismic layer at the slant distance SR | At the depth of the last receiver | SR is a distance along the ray, not a depth | Velocity of the last layer and Vs,eq (example: −2.3 % and −0.2 %) |
| Direct-method γ from the correlation applied to the thickness of the layer | From the mean depth of the layer | The correlation is a function of depth; the desktop program itself uses depth in the interval method | γ and moduli of the direct method (example: γ from −6 % to +4 %) |
| Interval-method seismic layer whose base does not coincide with a receiver: every layer gets the mean of the whole profile | Base snapped to the nearest receiver | A layer must average its own intervals, not all of them | Averages per layer of the interval method |
| Interval-method Vs,eq from the arithmetic mean Vs of the seismic layers | On the measurement intervals | The arithmetic mean of velocities is not the velocity of the layer: it overestimates Vs,eq and makes it depend on where the boundaries are | Interval-method Vs,eq lower, on the safe side, and equal to that of the direct method (example: 265.81 → 242.86 m/s, −9 %) |
| ν of the previous row when Vp ≤ Vs; E = 2·G without ν; negative Ev printed | “--” | A value that cannot be computed is not a test result | Anomalous rows |
| Layers that do not reach H: H divided by the time of the existing layers | Last layer extended to H, and declared | The formula needs stretches covering the whole thickness between z0 and H | Vs,eq of tests shorter than H |
| Numerator of Vs,eq truncated to the integer | H − z0 | 25.5 m is not 25 m | Vs,eq when H − z0 is not an integer |
| Missing time treated as zero | Interval “--” | A time that was not measured is not a zero time | Tests with incomplete receivers |
What does not change. Corrected times, values per measurement interval (except the anomalous rows of the table) and Vp, Vs, ν of the direct-method seismic layers other than the last one match the desktop program.
Choices of the desktop program that are kept. These are choices of method, not defects:
- the layer velocity of the direct method is the chord between the times at the two bases, not a least-squares line;
- the first interval runs from ground level, at time zero, to the first receiver: with a shallow first receiver and a large d it is the value most sensitive to the straight-path assumption;
- a layer with a decreasing time at its base copies the previous one;
- the unit-weight correlation and the intermediate rounding.
Comparing an old report
Open the .dwhx with File → Open… and compare the corrected times and the
interval-method table first: they must match. Then look for the differences where the
table above predicts them: last layer, γ and moduli of the direct method, Vs,eq.
Version¶
This page describes Downhole NX from version 2026.1.286 (October 2026), which introduced the automatic proposals; cases 1–7 hold from 2026.1.285. The version of the app is in the status bar at the bottom and in the ? button, Version tab.
Found an error on this page? Let us know or open a Pull Request.