Log acquisition, processing and presentation

Quick refresher

A curve in a file is the result of a chain: sensor → digitisation → communication channel → tie to depth → calibrations and corrections → recording. We keep the raw data, so that we can repeat the processing.
The sampling interval must be several times smaller than the vertical resolution of the tool. The noise of nuclear methods comes from counting statistics (the tool counts random pulses, so a short count scatters more). At half the speed the noise is smaller by a factor of $\sqrt2$ . Averaging neighbouring samples also reduces the noise, but it widens the averaging window of the tool.
We check the master calibration (made in the workshop) at the well, before and after recording. The repeat section must agree with the main log within the statistical scatter. A systematic disagreement is a sign of drift (a slow change in the response of the tool) or a calibration error, not of geology.
We match runs by markers on the gamma ray log (GR), which is recorded in every tool string. The shift usually changes with depth, so we set it at several markers and interpolate between them (estimate the values in between).
The heading is part of the data. It gives the depth reference, the bit size, the mud type, $R_m$ , $R_{mf}$ and $R_{mc}$ (the resistivities of the mud, filtrate and mudcake) with the temperature at which they were measured, the bottom-hole temperature and the time since circulation stopped, the tool string, and the processing.
We show resistivities on a logarithmic scale (classically four decades, 0.2–2000 Ω·m; a decade is a factor of ten). We show density and neutron porosity on compatible scales (scales on which the two curves coincide in a reference rock). A depth scale of 1:200 is for detailed study of beds, and 1:500 is for correlation.

Key relationships

\sigma_2 = \sigma_1\sqrt{v_2/v_1}, \qquad \sigma_k = \sigma_1/\sqrt{k}

statistical noise at a different logging speed and after averaging $k$ samples

L_{\text{eq}} \approx \sqrt{L^2 + (k\,\Delta z)^2}

equivalent window of a tool with window $L$ after averaging $k$ samples at interval $\Delta z$

s(z) = s_1 + \dfrac{z - z_1}{z_2 - z_1}\,(s_2 - s_1)

shift of a run between markers at depths $z_1$ and $z_2$

R = R_{\min}\cdot 10^{\,N x}

reading on a logarithmic grid of $N$ decades; $x$ is the fraction of the width from the left edge

z_s = z_{\text{b}}(t) - d_s, \qquad \delta z = v\,\delta t

depth of an LWD sensor from the time of the sample, and the depth error when the clocks disagree

Common pitfalls

Taking a curve from a file without its heading: the mud, the temperature, the matrix of the neutron curve (the rock type that its porosity scale assumes) and the corrections already applied are unknown.
Blaming geology for the disagreement between the main log and the repeat section without checking the calibrations.
Shifting a run by one constant amount from a single marker.
Reading a backup curve on the main scale, or a logarithmic scale as a linear one.
Mistaking the crossover of the neutron and density curves (the curves overlap) in sandstone on limestone scales for gas.
Using real-time LWD data for quantitative interpretation when the memory data are available.

After this lesson you can

describe the acquisition chain (sensor, downhole electronics, communication channel, surface system), and explain where the data are processed and how the LAS and DLIS file formats differ
explain how the sampling interval, logging speed, statistical noise and vertical resolution are linked, and choose between slower logging and averaging
check a recording against the calibrations and the repeat section, and tell a random disagreement from a systematic one
match runs by depth using markers (sharp, easy-to-recognise features) on the gamma ray log when the shift changes with depth
find the mud and temperature data in the heading, estimate the formation temperature, and convert the resistivities of the mud, filtrate and mudcake to that temperature
read tracks, logarithmic and compatible scales, backup curves and depth scales; tell real-time logging-while-drilling (LWD) data from memory data; and check a recording against a checklist

From the sensor to the file

A logging tool does not measure porosity or the resistivity of the formation. It measures the response of a sensor: the number of gamma rays that reach a detector, the voltage (potential difference) between electrodes, the electromotive force (voltage) induced in a receiver coil, and the arrival time of an elastic wave (a sound wave in the rock). To become a curve on a log, this response passes through several links. At each link information can be lost or an error can enter.

Links of the acquisition chain

1
Sensor
Detectors, electrodes, coils or acoustic receivers turn the response of the rock into an electrical signal. Their sensitivity slowly drifts (changes) with time, so we calibrate the tools regularly (see the section on calibrations below).
2
Downhole electronics
The electronics amplify the signal and digitise it (analogue-to-digital conversion: turning a voltage into numbers). They count pulses in energy windows (ranges of particle energy). They also calculate intermediate quantities, for example density from two detectors with a correction for the mudcake. The more processing the tool does, the less data it has to send. But less of the original information remains.
3
Communication channel
Over a logging cable, the digital telemetry of modern systems sends hundreds of kilobits per second or more. In LWD (logging while drilling), mud-pulse telemetry (pressure pulses sent through the mud) gives a few bits per second, and rarely tens (details). The rate of the channel decides what we can send at once and what has to stay in the tool memory.
4
Surface system
The surface system receives the data and takes depth from the measuring wheel and the cable marks. It shifts the readings of each sensor by its position in the tool string. It applies calibrations and environmental corrections (corrections for the effect of the hole and the mud). It calculates derived curves and shows everything on the screen during recording.
5
Recording and transmission
The surface system writes the raw and the processed data to files and prints the field log. The files go to the client (the company that ordered the logging) by satellite or land link, often while the work is still going on.

The capacity of the channel largely decides where the data are processed (example below). What the channel cannot send in time has to be compressed or processed in the tool itself. The logging unit does the calibrations, the environmental corrections, and a quick evaluation of the formations for decisions at the rig. Joint processing of all the methods, together with core, tests and data from neighbouring wells, is done later, in an interpretation centre (Log interpretation workflow). The less processing is irreversibly “built into” the tool, the better. Raw samples that are kept alongside the processed curves let us recalculate the curves years later, with new corrections and models.

Worked exampleHow much data a run gives, and what the channel can send

Given (assumed data)

Recorded interval: 1453.0–2703.0 m, sampling interval 0.1524 m (6 in)
Scalar curves (one value at each depth): 25 curves, 4 bytes per sample each
Waveforms of the sonic tool (recorded sound waves): 8 receivers, 400 values each, 2 bytes per value
Logging speed: 540 m/h
LWD mud-pulse channel for comparison: 8 bit/s

Solution

The interval $2703 - 1453 = 1250$ m at a sampling interval of 0.1524 m is 8202.1 steps, and the number of samples along depth is 8203.
Scalar curves: $25\cdot 4 = 100$ bytes per sample, and $8203\cdot 100$ bytes $\approx 0.82$ MB per run.
Waveforms: $8\cdot 400\cdot 2 = 6400$ bytes per sample, 64 times more than all the scalar curves together; per run $\approx 52.5$ MB.
At 540 m/h the tool covers $540/3600 = 0.15$ m per second, that is $0.15/0.1524 = 0.984$ samples per second. The scalar curves need 0.79 kbit/s, and the waveforms need 50.4 kbit/s. Wireline telemetry can send both.
Over a mud-pulse channel of 8 bit/s one set of waveforms would take $6400\cdot 8/8 = 6400$ s, about 1.8 h for one depth.

Answer: A run gives about 0.82 MB of scalar curves and 52.5 MB of waveforms. Wireline telemetry sends them during recording; a mud-pulse channel cannot.

What it means. That is why in LWD the tool itself processes the waveforms, only the interval transit times (the travel time of sound over a set distance) go to the surface, and the full waveforms are written to memory. The same calculation explains what the file format needs. A text table with one depth per row is poorly suited to hundreds of values at each depth, so arrays need a format like DLIS (below). The data in this example are assumed for illustration.

The result of recording is a field log (a log printed at the well) and digital files. The industry uses several formats for exchanging log data.

Log data formats

Format: LAS (Log ASCII Standard)
What it is: a text file: heading sections and a table of values, usually one depth per row; versions 1.2 (1989), 2.0 (1992) and 3.0 (1999)
Where it is used: exchange of curves between programs and organisations; the most widespread exchange format

Format: DLIS (Digital Log Interchange Standard)
What it is: a binary, self-describing format (stored as bytes, and the file carries its own description), first issued by the American Petroleum Institute (API) in 1991 as recommended practice RP 66
Where it is used: field data of service companies and the archive: it stores curves with different sampling intervals, waveforms and images

Format: LIS (Log Information Standard)
What it is: an earlier binary format, in use since the late 1970s
Where it is used: old archives

Format: WITSML (Wellsite Information Transfer Standard Markup Language)
What it is: an industry standard for transferring data from the rig, based on the XML markup language
Where it is used: LWD and drilling parameters sent to the office during drilling

Format	What it is	Where it is used
LAS (Log ASCII Standard)	a text file: heading sections and a table of values, usually one depth per row; versions 1.2 (1989), 2.0 (1992) and 3.0 (1999)	exchange of curves between programs and organisations; the most widespread exchange format
DLIS (Digital Log Interchange Standard)	a binary, self-describing format (stored as bytes, and the file carries its own description), first issued by the American Petroleum Institute (API) in 1991 as recommended practice RP 66	field data of service companies and the archive: it stores curves with different sampling intervals, waveforms and images
LIS (Log Information Standard)	an earlier binary format, in use since the late 1970s	old archives
WITSML (Wellsite Information Transfer Standard Markup Language)	an industry standard for transferring data from the rig, based on the XML markup language	LWD and drilling parameters sent to the office during drilling

The LAS format was introduced by the Canadian Well Logging Society (CWLS); version 2.0 remains the most widespread, while version 3.0 has a more complex structure and fewer users.

Sections of a LAS version 2.0 file

Section: ~V
What it contains: the format version and the line-wrap mode (WRAP: with YES, the values of one depth occupy several lines); always first

Section: ~W
What it contains: the well: name, field, date; the start and stop depths and the step (STRT, STOP, STEP), the null code (NULL)

Section: ~C
What it contains: the curves: mnemonic (short name), units and description, in the order of the columns of the table

Section: ~P
What it contains: parameters: mud resistivities, bottom-hole temperature, bit size and so on; optional

Section: ~O
What it contains: free text; optional

Section: ~A
What it contains: the table of values; the first column is depth (or time); always last

Section	What it contains
~V	the format version and the line-wrap mode (WRAP: with YES, the values of one depth occupy several lines); always first
~W	the well: name, field, date; the start and stop depths and the step (STRT, STOP, STEP), the null code (NULL)
~C	the curves: mnemonic (short name), units and description, in the order of the columns of the table
~P	parameters: mud resistivities, bottom-hole temperature, bit size and so on; optional
~O	free text; optional
~A	the table of values; the first column is depth (or time); always last

Missing values are written as a conventional number, most often −999.25. A program that reads this number as a measurement spoils any calculation. Read the units — metres or feet, g/cm³ or kg/m³, µs/m or µs/ft — from the ~C section; do not guess them.

DLIS is built differently. It is a binary “container” (stored as bytes, not as text). It keeps together with the curves everything that we need to understand them: a description of the well and the tools, processing parameters, calibrations, several sets of curves with different depth sampling (called frames), and arrays of values at each depth. The client usually gets both the DLIS with the field data and a LAS with the main curves. When DLIS is converted to LAS, the arrays and some of the parameters are lost. So we keep the original DLIS, and the description in the LAS states what processing has already been done.

Sampling interval, speed and resolution

People often mix up three quantities. The sampling interval is the distance along depth between adjacent samples. The vertical resolution is the smallest bed thickness that the tool shows with its true value. The design of the tool sets it (details). The logging speed is the speed at which the tool moves along the hole. It decides how much time the tool has for each sample.

We choose the sampling interval from the resolution. There must be at least two samples per resolution length, and preferably three or four. Otherwise a thin bed can fall between samples. The standard sampling interval of wireline tools is 0.1524 m (6 in) for a resolution of about 0.6 m. For high-resolution tools it is a few centimetres, and for electrical imagers (tools that make a picture of the borehole wall) it is millimetres. In metric systems the interval is usually a round fraction of a metre, for example 0.1 m. More frequent sampling does not improve the resolution by itself. It only describes in more detail the curve that the tool has already smoothed.

The physics of each method limits the speed. Nuclear tools count random pulses. The faster the tool moves, the fewer pulses it counts per sample, and the greater the statistical noise (counting statistics). A nuclear magnetic resonance tool needs time to polarise the protons, that is, to line up hydrogen nuclei with a magnetic field (recording mode). A sonic tool needs time to record waveforms. A tool string moves at the speed of the slowest tool in it, so slow methods are sometimes recorded in a separate run.

Statistical noise at a different logging speed and after averaging

\sigma_2 = \sigma_1\sqrt{\frac{v_2}{v_1}}, \qquad \sigma_k = \frac{\sigma_1}{\sqrt{k}}

where:

$\sigma_1$ standard deviation of a sample (the typical scatter of one reading) at logging speed $v_1$ , curve units
$\sigma_2$ the same at speed $v_2$ , curve units
$v_1,\ v_2$ logging speeds, m/h (any, the same for both)
$\sigma_k$ standard deviation after averaging $k$ adjacent samples at speed $v_1$ , curve units
$k$ number of samples averaged, —

Units: noise in the units of the curve (g/cm³, API, or porosity units: one porosity unit is 0.01, or 1%, of porosity); speeds in the same units

Holds when

Counting statistics set the noise: the number of pulses per sample is proportional to $\Delta z/v$ . We ignore other interference.
Adjacent samples are independent and the sampling interval $\Delta z$ is the same.
The bed is uniform over the averaged samples.

Both results follow from the counting formula $\sigma_N/N = 1/\sqrt N$ (the relative scatter of $N$ counted pulses). At half the speed the noise is smaller by a factor of $\sqrt2 \approx 1.41$ . Averaging three samples reduces it by a factor of $\sqrt3 \approx 1.73$ without losing time. The price of averaging is resolution.

Equivalent window of a tool after averaging

L_{\text{eq}} \approx \sqrt{L^2 + w^2}, \qquad w = k\,\Delta z

where:

$L$ averaging window of the tool itself, about equal to its vertical resolution, m
$w$ filter length: $k$ samples at interval $\Delta z$ , m
$L_{\text{eq}}$ length of a single uniform window equivalent to the tool with the filter, m

Units: m

Holds when

The tool and the filter are uniform windows, as in the thin-bed model (vertical resolution).
We replace the two windows by one window with the same second moment (a measure of spread): when two averagings are applied one after the other, their variances (squared spreads) add.
Compared with the exact calculation of two windows, the approximation slightly exaggerates the loss of contrast (in the example below, 0.149 against 0.152).

The tool and the filter smooth the curve one after the other, and their “blurs” add up like the two short sides of a right triangle. While the filter is clearly shorter than the tool window, the resolution hardly suffers. A filter of the same length widens the window by a factor of about $\sqrt2$ .

Worked exampleSlower logging, or averaging?

Given (assumed data)

Neutron porosity noise: 1.2 porosity units per sample at 600 m/h (one standard deviation)
Sampling interval and tool window $L$: 0.1524 m and 0.45 m
Task: 400 m of a carbonate sequence; a porous bed 0.5 m thick with a porosity of 0.18 among tight rocks with a porosity of 0.04
Options: logging at 300 m/h, or averaging three samples at 600 m/h

Solution

At half the speed, 300 m/h: $\sigma = 1.2\sqrt{300/600} = 1.2/\sqrt2 = 0.85$ porosity units. Logging 400 m takes 80 min instead of 40.
Averaging three samples at 600 m/h: $\sigma = 1.2/\sqrt3 = 0.69$ units. The filter length is $w = 3\cdot 0.1524 = 0.457$ m.
Equivalent window: $L_{\text{eq}} = \sqrt{0.45^2 + 0.457^2} = 0.64$ m, which is already greater than the thickness of the bed.
By the uniform-window model, the reading in the middle of the bed is $\dfrac{0.5}{0.64}\cdot 0.18 + \left(1 - \dfrac{0.5}{0.64}\right)\cdot 0.04 = 0.149$ . The exact calculation of two successive windows gives 0.152.
Without the filter, the tool window (0.45 m) is smaller than the thickness of the bed. In the middle of the bed the tool then reads the true 0.18, at either speed.

Answer: Logging at half the speed reduces the noise to 0.85 units without loss of resolution, but it takes twice as long. Averaging reduces the noise to 0.69 units without loss of time, but the porosity of a half-metre bed then reads as 0.149–0.152 instead of 0.18.

What it means. The choice depends on the section. In thick uniform beds averaging costs almost nothing. In a thinly bedded carbonate sequence it underestimates the porosity of exactly those thin beds that the logging is run to find. Here slower logging is better, although an hour of rig time is expensive. Digital filters are also applied during processing, so the heading and the description of the file state what averaging has already been done. The data in this example are assumed for illustration.

Calibrations, checks and the repeat section

The sensitivity of sensors changes with time. Detectors and sources age, the electronics drift, and the temperature and pressure in the well affect the readings. So we link the tool reading to the measured quantity by calibration, and we check it regularly. In Russian-language practice, the link between the readings and the measured quantity that is set at the base against reference standards is also called graduation (the “graduation characteristic” of the tool). A check with portable devices at the well, before and after the measurements, is called field calibration.

Master calibration: The service company does it from time to time at its base, against standards: blocks of known density for the density tool, a tank of water or a calibrator for the neutron tool, rings of known diameter for the caliper, and a test loop for the induction sonde. At the same time the company sets up the portable calibrators that go to the well. The gamma ray log is often calibrated at the well itself, with a portable radioactive standard.
Before-survey check: A check at the well with portable calibrators. We compare the result with the master calibration.
After-survey check: The same check after the tool is pulled out. The difference between “before” and “after” shows how much the tool drifted during the job. The values and tolerances (allowed limits) are printed at the end of the log and written to the file.
Check in known environments: We compare the readings in casing, in evaporites (rocks such as salt and anhydrite) and in clean water-bearing beds with values that we can predict in advance (table below).

What tools should read in known environments

Environment: Casing
Reference values: caliper: the inside diameter of the casing; sonic log with no cement behind the casing: 187 µs/m (57 µs/ft)
What is checked: the caliper calibration and the interval transit time scale

Environment: Anhydrite
Reference values: density 2.98 g/cm³, interval transit time 164 µs/m (50 µs/ft), neutron porosity close to zero
What is checked: the density, sonic and neutron logs

Environment: Rock salt
Reference values: apparent density 2.03 g/cm³, 220 µs/m (67 µs/ft), neutron slightly below zero; with fresh mud, washouts form because the salt dissolves
What is checked: the same methods and the condition of the hole

Environment: Clean water-bearing bed
Reference values: the density and neutron curves coincide on compatible scales; the apparent water resistivity is close to the formation value
What is checked: consistency of the porosity and resistivity methods

Environment: Thick uniform shales
Reference values: resistivity curves of different depths of investigation almost coincide: there is no invasion
What is checked: the calibration of the resistivity sondes

Environment	Reference values	What is checked
Casing	caliper: the inside diameter of the casing; sonic log with no cement behind the casing: 187 µs/m (57 µs/ft)	the caliper calibration and the interval transit time scale
Anhydrite	density 2.98 g/cm³, interval transit time 164 µs/m (50 µs/ft), neutron porosity close to zero	the density, sonic and neutron logs
Rock salt	apparent density 2.03 g/cm³, 220 µs/m (67 µs/ft), neutron slightly below zero; with fresh mud, washouts form because the salt dissolves	the same methods and the condition of the hole
Clean water-bearing bed	the density and neutron curves coincide on compatible scales; the apparent water resistivity is close to the formation value	consistency of the porosity and resistivity methods
Thick uniform shales	resistivity curves of different depths of investigation almost coincide: there is no invasion	the calibration of the resistivity sondes

The values for evaporites and casing are the same as in the topics Density log and photoelectric factor (Pe) and Sonic log. Zero neutron porosity is only a guide, good to within a few porosity units: the readings in salt and anhydrite depend on the tool.

A repeat section (in Russian-language practice also called a repeat, or control, measurement) is an interval that the same tool records a second time. It is usually several tens of metres long. It is more often recorded before the main log, in the interval of interest. We do not compare the two records by eye. We compare the differences between the readings at the same depths. The mean difference shows a systematic offset, and the spread of the differences shows repeatability.

Checking the repeat section from the differences

\bar\Delta = \frac{1}{N}\sum_{i=1}^{N}\Delta_i, \qquad s_\Delta = \sqrt{\frac{\sum_i (\Delta_i - \bar\Delta)^2}{N - 1}}, \qquad s_\Delta \approx \sigma\sqrt2

where:

$\Delta_i$ difference “repeat minus main log” at the $i$ -th depth, curve units
$N$ number of depths compared, —
$\bar\Delta$ mean difference, curve units
$s_\Delta$ standard deviation of the differences, curve units
$\sigma$ statistical error of one sample at the given logging speed (from the tool specification), curve units

Units: units of the curve being checked

Holds when

Both recordings are depth-matched. We choose the depths in uniform parts of beds, not at boundaries.
The errors of the two recordings are independent and equal: the difference then has a standard deviation of $\sigma\sqrt2$ .
The standard error of the mean difference is $s_\Delta/\sqrt N$ . A mean that is several times larger than this error is a systematic offset.

A spread of the differences close to $\sigma\sqrt2$ means that the recordings differ only by noise. A systematic offset points to tool drift or a calibration or depth error. We look for the cause in the calibrations before and after recording.

Worked exampleA repeat section in a water-bearing sandstone

Given (assumed data)

Density differences at eight depths: +0.006; −0.004; +0.009; −0.010; +0.002; +0.007; −0.006; +0.004 g/cm³
Density error per sample: $\sigma = 0.006$ g/cm³
Neutron porosity differences at the same depths: +0.031; +0.006; +0.024; +0.012; +0.035; +0.009; +0.020; +0.011
Neutron porosity error per sample: $\sigma = 0.010$ (one porosity unit)

Solution

Density: the sum of the differences is 0.008 g/cm³, and the mean is $\bar\Delta = 0.001$ g/cm³. The sum of the squared deviations from the mean is 0.00033, so $s_\Delta = \sqrt{0.00033/7} = 0.0069$ g/cm³.
The spread expected from noise alone is $0.006\sqrt2 = 0.0085$ g/cm³, and the observed spread is no greater. The standard error of the mean is $0.0069/\sqrt8 = 0.0024$ , and the mean is only 0.42 times that error. The recordings agree.
Neutron: the mean difference is $0.148/8 = 0.0185$ and the spread is $s_\Delta = 0.0107$ . This is within what noise alone would give, 0.0141, so the random part of the disagreement is normal.
The standard error of the mean is $0.0107/\sqrt8 = 0.0038$ . The mean is 5 times larger. This is a systematic offset of 1.85 porosity units, not chance.

Answer: The density tool repeats itself within the statistical scatter. The neutron curve of the repeat section is systematically higher than the main one, by 1.85 porosity units.

What it means. In a water-bearing sandstone the hydrogen index (the hydrogen content of the fluid) of the filtrate and of the formation water is almost the same, so invasion does not explain such a difference. The next step is to look at the calibrations before and after recording. If they differ by the same amount, the tool has drifted, and we correct the main log from the calibrations or record it again. We make a repeat section for exactly this reason: to catch such cases while the tool is still at the well. The data in this example are assumed for illustration.

Depth: tying in a run and matching runs

Depth on a log is the distance along the hole from the depth reference, usually the rotary table or the rig floor. The heading states the height of this reference above the ground or above sea level. Within a run, the measuring wheel and the magnetic marks on the cable give the depth, and the readings of the sensors are shifted by their position in the tool string (details). The topic on wireline work covers cable stretch and its errors (depth errors).

We tie each run to the hole at points whose depth is known independently: the shoe of the previous casing string (the bottom end of the casing, clearly visible on the caliper log and on the resistivities) and the bottom of the hole. The difference between the driller's depth and the logger's depth at these points is the first check. In deep wells it can reach a metre or more, and it usually grows with depth.

Then we bring the records of different runs to one reference depth. This is usually the depth of the run by which we pick the beds (mark their tops and bases), most often the main resistivity run. The tool for depth matching is the GR. It is recorded in every tool string, and the other sensors of the string are already tied to it by their offsets. On the reference GR and on the GR to be matched, we look for markers: thin, sharp beds that cannot be confused with their neighbours, such as carbonate beds, highly radioactive shales, coals and sharp boundaries. The shift of a run at a marker is the difference between the marker's depth on that run and its depth on the reference run. The shift is rarely constant: cable stretch depends on the tension, that is, on the weight of the tool string and on depth. So we take markers over the whole interval and interpolate the shift between them.

Shift of a run between two markers

s(z) = s_1 + \frac{z - z_1}{z_2 - z_1}\,(s_2 - s_1), \qquad z_{\text{ref}} = z - s(z)

where:

$z$ depth of the reading on the run being matched, m
$z_1,\ z_2$ depths of the markers above and below, on the same run, m
$s_1,\ s_2$ shifts at the markers: depth on the run being matched minus depth on the reference run, m
$z_{\text{ref}}$ depth of the reading in the depth system of the reference run, m

Units: m

Holds when

Between markers the shift changes smoothly (cable stretch, thermal elongation). Interpolation does not correct jerks of the tool at ledges in the hole.
Above the uppermost marker and below the lowest one we take the shift as constant, equal to the shift at the nearest marker.

This is linear interpolation (a straight-line estimate between two known values). We shift each reading by the amount that matches its position between the markers. We shift all the curves of a run together with its GR.

Markers on the GR of runs 1 and 2 of the assumed well

Marker: M1
What it is: thin carbonate bed
Run 1, m: 2571.40
Run 2, m: 2571.95

Marker: M2
What it is: highly radioactive shale bed
Run 1, m: 2603.00
Run 2, m: 2603.75

Marker: M3
What it is: thin carbonate bed
Run 1, m: 2644.60
Run 2, m: 2645.55

Marker	What it is	Run 1, m	Run 2, m
M1	thin carbonate bed	2571.40	2571.95
M2	highly radioactive shale bed	2603.00	2603.75
M3	thin carbonate bed	2644.60	2645.55

Run 1 (resistivities) is the reference. The beds of the well are analysed in the lesson Log interpretation workflow.

CalculatedGR of two runs of the assumed well in the interval 2596–2612 m before and after depth matching. Before matching, run 2 (dashed curve) lags behind the reference run 1 by 0.75 m: the peak of the highly radioactive shales (marker M2) and the top of the sandstone appear that much deeper. After we shift the run by the markers, the curves coincide within the noise.

Worked exampleMatching a density run whose shift grows with depth

Given (assumed data)

Markers on the GR: table above: M1, M2, M3
Density reading: recorded by run 2 at a depth of 2615.80 m
Reference run: run 1

Solution

Shifts at the markers (run 2 minus run 1): M1 $2571.95 - 2571.40 = 0.55$ m, M2 0.75 m, M3 0.95 m. The shift grows with depth: run 2 is “stretched” relative to run 1.
The reading lies between M2 (2603.75 m) and M3 (2645.55 m) in run-2 depths. The shift is $0.75 + \dfrac{2615.80 - 2603.75}{2645.55 - 2603.75}\cdot(0.95 - 0.75) = 0.808$ m.
The depth of the reading in the run-1 system is $2615.80 - 0.808 = 2614.99$ m.
With the mean shift over the three markers (0.75 m) the result would be 2615.05 m, an error of 0.06 m. With no shift at all the error is 0.808 m.

Answer: The run-2 reading belongs to a depth of 2614.99 m on run 1. The shift at this depth is 0.808 m.

What it means. Eight tenths of a metre seem a trifle. But in this well a tight bed C, 1.2 m thick, lies above bed D. Without depth matching, the density of the porous sandstone B would “move” onto the upper 0.8 m of bed C, and the density of bed C would move onto the upper 0.8 m of bed D. A point-by-point calculation would then show a false reservoir in the tight bed, with high porosity and high resistivity, and a non-reservoir at the top of D. The mean shift here is wrong by only 0.06 m. But in a long interval where the shift keeps growing, the error of a constant shift builds up. The data in this example are assumed for illustration.

The same technique ties other data together. We bring core to the log depth by the gamma activity measured on the recovered core (Core analysis). We shift LWD data, which are tied to the driller's depth, to the wireline log by the GR, or the other way round (driller's and logger's depth). The main thing is to choose one reference depth and bring everything to it.

The log heading

The log heading opens every log. In a digital file it is stored in the parameters. It is not decoration but part of the data: without it we can neither correct nor interpret the curves. At the end of the log, part of the heading is repeated, and the calibrations and the repeat section are printed.

What is in the heading and why

Block: Well
What it contains: company, field, well number, coordinates
Why: to identify the data unambiguously

Block: Elevations
What it contains: what the depths are measured from (rotary table, rig floor) and its height above ground or sea level
Why: to convert depths to elevations above sea level and to compare wells

Block: Run
What it contains: date, run number, total depth by the driller and by the logger, recorded interval, casing shoe
Why: to check the depths

Block: Hole and mud
What it contains: bit size; mud type, density, viscosity, fluid loss (how much filtrate the mud gives up in a standard test), source of the sample; $R_m$ , $R_{mf}$ , $R_{mc}$ with the temperature of measurement
Why: borehole corrections, calculations involving the filtrate

Block: Temperature
What it contains: maximum bottom-hole temperature, the time circulation stopped and the time of measurement
Why: formation temperature

Block: Tools and processing
What it contains: tool string configuration, sensor offsets, centralisers (devices that keep the tool in the middle of the hole), the matrix of the neutron curve, assumed densities, filters
Why: to understand what has already been done to the curves

Block: Remarks
What it contains: hole problems, mud additives (barite, potassium chloride), deviations from the programme
Why: to decide which corrections and exclusions are needed

Block	What it contains	Why
Well	company, field, well number, coordinates	to identify the data unambiguously
Elevations	what the depths are measured from (rotary table, rig floor) and its height above ground or sea level	to convert depths to elevations above sea level and to compare wells
Run	date, run number, total depth by the driller and by the logger, recorded interval, casing shoe	to check the depths
Hole and mud	bit size; mud type, density, viscosity, fluid loss (how much filtrate the mud gives up in a standard test), source of the sample; $R_m$ , $R_{mf}$ , $R_{mc}$ with the temperature of measurement	borehole corrections, calculations involving the filtrate
Temperature	maximum bottom-hole temperature, the time circulation stopped and the time of measurement	formation temperature
Tools and processing	tool string configuration, sensor offsets, centralisers (devices that keep the tool in the middle of the hole), the matrix of the neutron curve, assumed densities, filters	to understand what has already been done to the curves
Remarks	hole problems, mud additives (barite, potassium chloride), deviations from the programme	to decide which corrections and exclusions are needed

Extract from the heading of the assumed well

Heading line: Depth reference
Value: rotary table, 6.5 m above ground; vertical well

Heading line: Total depth: driller / logger
Value: 2705.0 / 2705.8 m

Heading line: Casing shoe, 244.5 mm: driller / logger
Value: 1452.0 / 1452.4 m

Heading line: Bit
Value: 215.9 mm (8½ in)

Heading line: Mud
Value: water-based, clay–polymer, 1.18 g/cm³

Heading line: $R_m$ / $R_{mf}$ / $R_{mc}$
Value: 0.62 / 0.48 / 0.85 Ω·m at 19 °C

Heading line: $R_m$ at bottom-hole temperature (run 2)
Value: 0.241 Ω·m

Heading line: Run 1
Value: induction log, microresistivity tool, GR; bottom-hole temperature 79.4 °C, 6.5 h after circulation stopped

Heading line: Run 2
Value: density and neutron logs, caliper, GR; bottom-hole temperature 82.6 °C, 14.0 h after circulation stopped

Heading line	Value
Depth reference	rotary table, 6.5 m above ground; vertical well
Total depth: driller / logger	2705.0 / 2705.8 m
Casing shoe, 244.5 mm: driller / logger	1452.0 / 1452.4 m
Bit	215.9 mm (8½ in)
Mud	water-based, clay–polymer, 1.18 g/cm³
$R_m$ / $R_{mf}$ / $R_{mc}$	0.62 / 0.48 / 0.85 Ω·m at 19 °C
$R_m$ at bottom-hole temperature (run 2)	0.241 Ω·m
Run 1	induction log, microresistivity tool, GR; bottom-hole temperature 79.4 °C, 6.5 h after circulation stopped
Run 2	density and neutron logs, caliper, GR; bottom-hole temperature 82.6 °C, 14.0 h after circulation stopped

The same data are used by the quick-look interpretation of the well in the lesson Log interpretation workflow.

The resistivities in the heading are measured at the surface on a sample of the mud. The filtrate and the mudcake are obtained in a filter press (a small device that squeezes filtrate out of the mud). For calculations we bring the resistivities to formation temperature by the Arps formula. If $R_{mf}$ was not measured, we estimate it from $R_m$ . The usual order $R_{mf} < R_m < R_{mc}$ serves as a check on the record (details). We estimate the formation temperature from the bottom-hole temperature. If there are several measurements, we correct for the cooling of the bottom of the hole by circulation (circulation correction).

Worked exampleFormation temperature and mud resistivities from the heading

Given (assumed data)

Heading: table above
Mean annual surface temperature: 11 °C
Bed of interest: at a depth of 2618.0 m from the rotary table

Solution

Depths are counted from the rotary table, while the surface temperature refers to ground level. The bottom of the hole by the logger is $2705.8 - 6.5 = 2699.3$ m below ground, and the bed is $2618.0 - 6.5 = 2611.5$ m below ground.
Temperature gradient (the rise of temperature with depth) from the later run 2: $G_T = (82.6 - 11)/2699.3 = 0.02653$ °C/m, that is 2.65 °C per 100 m.
Temperature of the bed: $T = 11 + 0.02653\cdot 2611.5 = 80.3$ °C.
Resistivities at this temperature (measured at 19 °C, Arps formula): $R_{mf} = 0.48\cdot\dfrac{19 + 21.5}{80.3 + 21.5} = 0.191$ Ω·m; in the same way $R_m = 0.247$ Ω·m and $R_{mc} = 0.338$ Ω·m.
Check of the line “ $R_m$ at bottom-hole temperature”: $0.62\cdot 40.5/(82.6 + 21.5) = 0.241$ Ω·m, which agrees with the heading.
Run 1 (79.4 °C, 6.5 h after circulation stopped) would give 77.2 °C and $R_{mf} = 0.197$ Ω·m, which is 3% more.

Answer: The formation temperature is at least 80.3 °C. At this temperature

R_{mf} \approx 0.191

R_m \approx 0.247

and

R_{mc} \approx 0.338

Ω·m.

What it means. Both bottom-hole temperatures are too low, because circulation has cooled the bottom of the hole. The later measurement is closer to the formation temperature. So the formation temperature is a lower-bound estimate (the true value is at least this), and the resistivities at that temperature are upper-bound estimates (the true values are at most these). Two measurements are too few to check the Horner line (a straight-line plot that extrapolates the temperature to its undisturbed value). If there are temperature measurements from formation tests in the area, we compare with them. If the line “ $R_m$ at bottom-hole temperature” agrees with the recalculation, that is a simple check for typing errors in the heading. The quick-look interpretation (a fast first-pass interpretation) of the same well in the lesson Log interpretation workflow uses this 80.3 °C and $R_{mf}$ = 0.191 Ω·m. The data in this example are assumed for illustration.

Tracks, scales, depth scales and shading

A standard of the American Petroleum Institute (API) set the layout of a paper log, and software still reproduces it. A sheet 8.25 in (21 cm) wide is divided into three tracks of 2.5 in (63.5 mm) each. A narrow depth track of 0.75 in is left between the first and the second. The first track has a linear grid for the GR, the spontaneous potential (SP) and the caliper log. The second and third tracks are laid out to suit the task. They can share one logarithmic grid of four decades (1.25 in per decade), or share one linear grid, or be set up separately: resistivities on the logarithmic second track, and the porosity methods on the linear third. In Russian-language practice a composite plot of several methods with a common depth track is called a planshet. On a screen there can be any number of tracks, but the logic of the classic layout is preserved (how to read a log).

A linear scale suits quantities that change within a narrow range: GR, density, neutron porosity and interval transit time. Resistivity needs a logarithmic scale. In one section it varies from tenths to thousands of ohm-metres, and only a logarithmic grid shows all the beds with the same relative accuracy. On this grid the divisions inside a decade are not equal steps; each one is a multiplier.

Reading on a logarithmic grid

R = R_{\min}\cdot 10^{\,N x}, \qquad x = \frac{\lg(R/R_{\min})}{N}

where:

$R$ value of the curve, Ω·m
$R_{\min}$ value at the left edge of the grid, Ω·m
$N$ number of decades across the width of the grid, —
$x$ distance from the left edge as a fraction of the grid width, —

Units: resistivities in the same units; $\lg$ is the decimal logarithm (logarithm to base 10)

Holds when

The grid is logarithmic across the whole width; on a split grid the formula applies only to the logarithmic track.

Each decade occupies the same width and multiplies the value by ten. The classic grid of 0.2–2000 Ω·m has $N = 4$ across the width of two tracks.

We choose compatible scales so that two curves coincide in a reference rock. Density and neutron porosity share one track. For limestone the density scale is 1.95–2.95 g/cm³ and the neutron porosity scale is 0.45…−0.15 in limestone units. For sandstone the scales are 1.90–2.90 g/cm³ and 0.45…−0.15 in sandstone units. On both scales porosity increases to the left. Any separation of the curves in a clean water-bearing bed then points to shale, gas or a different matrix (a different rock type) (details). For the same reason, interval transit time is most often plotted increasing to the left.

A backup curve is a spare copy of a curve in the same track, on a shifted range. It appears when the main curve goes off the edge of the scale. For example, the GR scale is 0–150 API and the backup scale is 150–300 API. The backup is drawn with a different line type. If we read a backup curve on the main scale, we are wrong by a whole scale range.

Depth scales

Scale: 1:200
1 cm (1 in) of the log: 2 m
Purpose: detailed study of productive beds

Scale: 1:500
1 cm (1 in) of the log: 5 m
Purpose: general surveys over the whole hole, correlation

Scale: 1:1000
1 cm (1 in) of the log: 10 m
Purpose: overview of the hole

Scale: 1:2000, 1:5000
1 cm (1 in) of the log: 20 m, 50 m
Purpose: composite sections for geologists

Scale: 1:240
1 cm (1 in) of the log: 20 ft (5 in per 100 ft)
Purpose: detailed scale in feet

Scale: 1:600, 1:1200
1 cm (1 in) of the log: 50 and 100 ft (2 and 1 in per 100 ft)
Purpose: correlation scales in feet

Scale	1 cm (1 in) of the log	Purpose
1:200	2 m	detailed study of productive beds
1:500	5 m	general surveys over the whole hole, correlation
1:1000	10 m	overview of the hole
1:2000, 1:5000	20 m, 50 m	composite sections for geologists
1:240	20 ft (5 in per 100 ft)	detailed scale in feet
1:600, 1:1200	50 and 100 ft (2 and 1 in per 100 ft)	correlation scales in feet

A scale is the ratio of the length on paper to the length of the hole: 100 ft is 1200 in, so 5 in per 100 ft is 1:240 and 1 in per 100 ft is 1:1200. On a screen we can choose any scale, but printed logs and reports are still made at these scales.

Shading makes a log readable at a glance. We shade the sandstones, where the GR is below the cut-off value, yellow. We shade the separation of the neutron and density curves in shales grey, and the crossover in gas-bearing beds red or yellow. The area between resistivity curves of different depths of investigation is a sign of invasion. The colours are not standardised, so the legend of scales and shading is a required part of a log.

CalculatedAn assumed well in a format close to the classic one, 2596–2636 m. Track 1: GR 0–150 API with a backup curve of 150–300 API (it appears opposite the highly radioactive bed at 2603 m) and shading where the GR is below 63 API. Track 2: resistivities on a logarithmic grid of 0.2–2000 Ω·m. Track 3: density and neutron porosity on scales compatible with sandstone; the separation of the curves is shaded grey.

Worked exampleThree readings that are easy to get wrong

Given (assumed data)

Resistivity: the curve is at 0.244 of the width from the left edge of a 0.2–2000 Ω·m grid (four decades)
Porosity: clean water-bearing limestone and sandstone with a porosity of 0.10, on limestone-compatible scales
GR: the 150–300 API backup curve is at 0.12 of the track width

Solution

Logarithmic grid: $R = 0.2\cdot 10^{4\cdot 0.244} = 1.89$ Ω·m. If the same point is wrongly read on a linear scale of 0–20 Ω·m, the result is $0.244\cdot 20 = 4.88$ Ω·m, which is 2.6 times more.
Limestone: $\rho_b = 2.71 - 0.10\cdot 1.71 = 2.539$ g/cm³, at $(2.539 - 1.95)/1.00 = 0.589$ of the width; the neutron curve is at $(0.45 - 0.10)/0.60 = 0.583$ of the width: the curves coincide.
Sandstone: $\rho_b = 2.65 - 0.10\cdot 1.65 = 2.485$ g/cm³, at 0.535 of the width. A neutron tool in limestone units reads about 0.057 in it (the model of the lesson Porosity from the neutron log), at 0.655 of the width.
The neutron curve is to the right of the density curve by 0.120 of the width, that is by 7.6 mm on a track 63.5 mm wide.
GR backup curve: $150 + 0.12\cdot 150 = 168$ API; on the main scale the same point would be read as 18 API, clean sandstone instead of shale.

Answer: The resistivity is 1.89 Ω·m. In sandstone on limestone scales the curves separate by 7.6 mm “as if from gas”, although there is no gas. The backup curve gives 168 API.

What it means. The crossover of the curves in clean sandstone on limestone scales is a matrix effect, not gas. So in sandstone sections we use scales compatible with sandstone, and we confirm gas by the resistivity and by recalculating the porosity (Gas, shale, carbonate and pressure effects on porosity logs). The size of the shift for a real tool depends on its characteristics; here we take it from the idealised model of the lesson. The data in this example are assumed for illustration.

LWD data: real time and memory

An LWD tool writes data to its memory by its own clock, at a set time interval, not at a set depth interval. The tool does not know the depth. The surface system measures the depth from the position of the travelling block (the moving block that hangs the drill string) and the length of pipe run in, and it records the bit depth against time. After the assembly is pulled out, we merge the memory data with this record. For each sample we find the bit depth at its time and subtract the distance from the bit to the sensor.

Depth of an LWD sensor from the time of the sample

z_s = z_{\text{b}}(t) - d_s, \qquad \delta z = v\,\delta t

where:

$z_s$ depth of the sensor at the time of the sample, m
$z_{\text{b}}(t)$ depth of the bit at time $t$ from the surface record, m
$d_s$ distance from the bit to the sensor, m
$\delta z$ depth error when the clocks of the tool and the surface system disagree, m
$v$ rate of penetration at that time, m/h
$\delta t$ clock disagreement, h

Units: m, h

Holds when

We know the bit depth from the surface record. Its own errors (pipe stretch, heave at sea: the up-and-down motion of the rig) are not included here.
The second formula holds as long as the rate of penetration is constant within the clock disagreement.

The depth of memory data is only as accurate as the clocks. The error grows with the rate of penetration, so we compare the clocks of the tool and of the surface system before running in and after pulling out.

Worked exampleDepth and time of LWD data

Given (assumed data)

Surface record: bit at 1912.0 m at 10:00 and at 1918.0 m at 10:24, drilling at a steady rate
Density sensor sample from memory: stamped 10:16 by the tool clock
Distances from the bit: density sensor 21.6 m, resistivity sensor 8.5 m
Variant: the tool clock was 3 min fast

Solution

Rate of penetration: $(1918.0 - 1912.0)/24\cdot 60 = 15$ m/h.
Bit depth at 10:16: $1912.0 + 6.0\cdot 16/24 = 1916.0$ m; the sensor is 21.6 m higher: $1916.0 - 21.6 = 1894.4$ m.
If the tool clock was 3 min fast, the sample was taken at 10:13 by the surface clock. The bit was at 1915.25 m and the sensor at 1893.65 m. The error of 0.75 m equals the rate of penetration multiplied by the clock error: $15\cdot 3/60 = 0.75$ m.
Time since penetration: the bit passed the point at 1894.4 m $21.6/15 = 1.44$ h, that is 86 min, before the density measurement. The resistivity sensor saw the same point 34 min after penetration.

Answer: The sample belongs to a depth of 1894.4 m. A clock error of 3 min would shift it by 0.75 m. Density and resistivity at this point were measured 86 and 34 min after penetration.

What it means. The different times since penetration explain why sensors of the same assembly see different amounts of invasion. This quantity is recorded as a separate curve. Here the drilling is steady. With stops and connections (adding a length of pipe), the time since penetration grows in jumps, and we calculate it from the depth–time record. We compare the memory data with the real-time data, which are tied to depth at the surface at once. A systematic shift between them reveals a clock error. The data in this example are assumed for illustration.

Only part of the data goes over the mud-pulse channel in real time: the main curves, at a coarse sampling interval (why). The surface system ties them to depth at the moment of reception. While the pumps are off, for example when pipe is being added, the mud-pulse channel is silent and gaps appear in the real-time data. During drilling the data are sent to the office over a network, usually in the WITSML standard. After the assembly is pulled out, the memory data replace the real-time data in all quantitative calculations.

Real-time and memory LWD data

Feature: When available
Real time: during drilling, with a delay for the distance to the sensor
Memory: after the assembly is pulled

Feature: Depth sampling interval
Real time: coarse: depends on the drilling speed and the channel
Memory: fine: set by the sampling period

Feature: Tie to depth
Real time: by the bit depth at the moment of reception
Memory: by time, by merging with the depth–time record

Feature: Content
Real time: the main curves, sometimes compressed
Memory: all curves, arrays and service data

Feature: Purpose
Real time: steering the well, operational decisions
Memory: interpretation and estimation of hydrocarbons in place

Feature	Real time	Memory
When available	during drilling, with a delay for the distance to the sensor	after the assembly is pulled
Depth sampling interval	coarse: depends on the drilling speed and the channel	fine: set by the sampling period
Tie to depth	by the bit depth at the moment of reception	by time, by merging with the depth–time record
Content	the main curves, sometimes compressed	all curves, arrays and service data
Purpose	steering the well, operational decisions	interpretation and estimation of hydrocarbons in place

A repeat LWD log recorded while tripping (running the drill string in or out) or reaming (working the bit back through the hole to clean it) shows how the zone near the borehole has changed over hours and days. We compare it with the log recorded while drilling, in the same way as the repeat section of a wireline tool.

A checklist before interpretation

Checking the data is the first step of interpretation, not a formality. It is convenient to work through a list. We write down each defect that we find, and we mark on the log the interval that it relates to.

Checklist

1
Heading
Check that it is complete. Check that the mud, the resistivities with the temperature of measurement, the bit size, the run number and the recorded interval agree with each other. Look for signs of copying from the previous run.
2
Depth
Check the depth reference, the casing shoe and the bottom of the hole by the logger against the driller, and the overlap with the previous run. Make sure that all runs are depth-matched by the GR to one reference.
3
Calibrations
Check that the calibrations before and after recording are within tolerance and close to the master calibration, and that the “before–after” drift is small.
4
Repeat section
Check that there is one, that it agrees with the main log within the statistical scatter, and that there is no systematic offset.
5
Speed and tension
Check that the logging speed is no higher than the slowest tool allows. Look for sticking events or jerks on the tension curve, and for depth “steps”. There should be none.
6
Quality indicators of the methods
Compare the caliper log with the bit size. Check the correction $\Delta\rho$ of the density tool (guidelines) and cycle skipping on the sonic log (the receiver misses a wave cycle) (signs). Check also the corrections of the neutron log (details) and of the resistivity sondes (details).
7
Known rocks
Check that the readings in casing, evaporites, thick shales and clean water-bearing beds are as expected.
8
Processing
Find out which environmental corrections and filters have been applied, in which matrix the neutron curve is recorded, and which densities have been assumed.
9
Digital file
Check the units, the null code, the step and direction of depths, the mnemonics and the splicing of runs (joining them into one curve). Make sure that the LAS agrees with the DLIS.
10
LWD data
Use memory data instead of real-time data. Compare the clocks. Find the time since penetration for each sensor. Compare with wireline logging where it exists.

Terms and symbols

Acquisition chain: Sensor → downhole electronics → communication channel → surface system → files and log.
Sampling interval: The distance along depth between samples. It must be several times smaller than the vertical resolution.
Logging speed: The speed at which the tool moves. For nuclear methods it sets the statistical noise.
Calibration: Linking the tool reading to the measured quantity by comparison with standards. It is done at the base, and at the well before and after recording.
Repeat section: An interval recorded twice with the same tool, to check repeatability (glossary).
Marker: A thin, sharp bed by which the records of different runs are matched.
Depth matching: Bringing the records of runs, LWD and core to one reference depth (glossary).
Log heading: The conditions of recording: elevations, run, hole, mud, temperature, tools and processing.
LAS and DLIS: LAS is a text format for exchanging curves. DLIS is a binary, self-describing format for field data.
Compatible scales: Scales of two curves on which the curves coincide in a reference rock.
Backup curve: A copy of a curve on a shifted scale range, for stretches where the main curve goes off the edge.

Common mistakes and misconceptions

Wrong

“A curve in a file is the measurement itself.”

In fact

It is the result of a processing chain: calibrations, corrections, filters and conversion to a matrix. The heading says exactly what has been done. We keep the raw data so that the processing can be repeated.

Wrong

“The repeat section did not agree, so the formation has changed.”

In fact

First check the calibrations before and after recording, and the depth matching. The formation changes only where invasion is taking place: in permeable beds, and for tools that are sensitive to the flushed zone.

Wrong

“It is enough to shift the runs by a single marker.”

In fact

The shift usually changes with depth. We take markers over the whole interval and interpolate the shift between them.

Wrong

“The bottom-hole temperature from the heading is the formation temperature.”

In fact

It is the temperature at the bottom of the hole after circulation has cooled it, so it is a lower-bound estimate. We get the formation temperature by correcting several measurements, or from test data.

Wrong

“LAS stores everything the tool recorded.”

In fact

LAS is a table of curves with a single depth index. Arrays, calibrations and many parameters remain only in the DLIS, so we keep the original DLIS.

In depth: advanced detail

Automatic depth matching: cross-correlation

Depth-matching programs do what the eye does, but they use all the points at once. Take a window of the reference curve a few metres long. The program tries different trial shifts of the second curve and calculates a correlation coefficient (a number that measures how well two curves match) for each. The shift with the greatest coefficient is the shift in this window. The program slides the window along the hole and so gets the shift as a function of depth.

Correlation coefficient of two curves at a trial shift

r(\tau) = \frac{\sum_i (x_i - \bar x)\,\big(y_i(\tau) - \bar y\big)}{\sqrt{\sum_i (x_i - \bar x)^2\,\sum_i \big(y_i(\tau) - \bar y\big)^2}}

where:

$x_i$ reference curve (GR of the reference run) at the depths $z_i$ of the window, API
$y_i(\tau)$ curve to be matched, read at the depths $z_i + \tau$ , API
$\bar x,\ \bar y$ mean values in the window, API
$\tau$ trial shift, m
$r$ correlation coefficient, from −1 to 1, —

Units: curves in the same units; shift in metres

Holds when

The window contains distinct beds. On a uniform stretch $r$ hardly depends on the shift, and the maximum cannot be trusted.
The shift is constant within the window.
The curves were recorded by tools of similar resolution.

The correlation coefficient does not depend on the level and range of the curves, so the matching works even when two GR tools differ slightly in calibration.

CalculatedCorrelation coefficient of the GR of runs 1 and 2 of the assumed well in the window 2596–2612 m as a function of the trial shift. The maximum is at a shift of 0.75 m, as given by marker M2 in this window. Even with no shift the correlation is high (0.85). The large packages of shales and sandstones create it, and only the thin beds make the maximum stand out.

A person always checks an automatic result. A cyclic section, with alternating similar sandstones and shales, gives several almost equal maxima, and the program may pick the “neighbouring” bed. A window that is too short picks up noise. A window that is too long averages out the change of the shift. Tools of different resolution show thin beds differently, so their correlation maximum is lower, although the shift is determined correctly.

Connections to other topics

Open-hole logging: why, how and when — how wireline logging and LWD are carried out, vertical resolution and the classic log.
Wireline operations — cable stretch and depth errors on wireline.
Formation water and formation temperature — converting resistivities to formation temperature and the circulation correction of temperature.
Determining formation water resistivity — the resistivities of the mud, filtrate and cake from the heading in $R_w$ calculations.
Gamma ray log and natural gamma ray spectrometry — the GR as the common depth “ruler”, and counting statistics.
Density log and photoelectric factor (Pe) — signs of unreliable density and calibration against reference blocks.
Neutron log — limestone units and compatible scales.
Core analysis — tying core to the log depth.
Well correlation — comparing the sections of different wells, as distinct from matching the runs of one well.
Log interpretation workflow — what is done with the checked data: the order of interpretation.

Check yourself

1
A tool with a vertical resolution of about 0.6 m was recorded with a sampling interval of 0.5 m instead of 0.15 m. What will happen?
1. A.Thin beds may fall between samples, and the curve will lose details that the tool saw
2. B.The noise will decrease threefold
3. C.The resolution will improve: there are fewer samples, so less noise
4. D.Nothing: the resolution is set by the design of the tool
Show the answer
Correct answer: A. Thin beds may fall between samples, and the curve will lose details that the tool saw
There would be only slightly more than one sample per resolution length instead of four. The tool still sees a bed 0.6 m thick, but the samples may fall on its boundaries, and the peak will disappear. More frequent sampling does not improve the resolution. But fewer than two or three samples per resolution length lose information.
2
Neutron porosity at 900 m/h has a noise of 1.5 porosity units per sample (sampling interval 0.1524 m, tool window 0.4 m). What will the noise be at 450 m/h, and when four samples are averaged at 900 m/h? What is the price in the second case?
Show the answer
At 450 m/h: $1.5/\sqrt2 \approx 1.06$ units. When four samples are averaged: $1.5/\sqrt4 = 0.75$ units, but a filter $4\cdot 0.1524 \approx 0.61$ m long widens the window to $\sqrt{0.4^2 + 0.61^2} \approx 0.73$ m, almost twofold. The tool will then show beds thinner than 0.7 m with too little contrast.
3
The repeat section of a density tool is higher than the main log by 0.02 g/cm³ on average in all beds, including shales, and the spread of the differences is small. What is the most likely cause?
1. A.A change in the formation due to filtrate invasion
2. B.Gas in the flushed zone
3. C.Counting statistics
4. D.A systematic error: tool drift or a calibration error; the calibrations before and after recording must be checked
Show the answer
Correct answer: D. A systematic error: tool drift or a calibration error; the calibrations before and after recording must be checked
An identical offset in all beds, including impermeable shales, cannot be a consequence of invasion, and statistics would give a spread without an offset. It is a sign of a systematic error. We look for it in the calibrations before and after recording.
4
From the markers, run 2 is deeper than the reference run 1 by 0.4 m at 2100 m and by 0.9 m at 2300 m (depths on run 2). To what depth of run 1 does the run-2 reading at 2240 m belong?
Show the answer
The shift at 2240 m is $0.4 + \dfrac{2240 - 2100}{2300 - 2100}\cdot(0.9 - 0.4) = 0.4 + 0.35 = 0.75$ m. The depth in the run-1 system is $2240 - 0.75 = 2239.25$ m. A constant shift taken from one of the markers would be wrong by 0.35 or 0.15 m.
5
The heading gives, at one temperature, $R_m = 0.85$ Ω·m, $R_{mf} = 1.10$ Ω·m and $R_{mc} = 0.95$ Ω·m. What raises doubts?
1. A.The cake should be more conductive than the mud
2. B.The filtrate has a higher resistivity than the mud: normally $R_{mf} < R_m < R_{mc}$ , so the record must be rechecked
3. C.The mud resistivities are not needed for interpretation
4. D.Nothing: all the values are about 1 Ω·m
Show the answer
Correct answer: B. The filtrate has a higher resistivity than the mud: normally $R_{mf} < R_m < R_{mc}$ , so the record must be rechecked
The solid part of the mud hardly conducts current. The filtrate, from which the solids have been removed, conducts better than the mud. The mudcake, where the solids are concentrated, conducts worse. So normally $R_{mf} < R_m < R_{mc}$ . The reverse order is a reason to check the record (columns mixed up, different temperatures of measurement) before calculating with these values.
6
On the field log of a sandstone section, the density and neutron curves on limestone scales diverge in clean water-bearing sandstones: the neutron curve is a few millimetres to the right. What is it?
1. A.Shale
2. B.A malfunction of the neutron tool
3. C.The matrix effect of sandstone on limestone-compatible scales
4. D.Gas
Show the answer
Correct answer: C. The matrix effect of sandstone on limestone-compatible scales
A neutron tool calibrated in limestone reads less than the true porosity in sandstone, while density porosity on a limestone scale reads more. The result is a crossover of the curves “as if from gas” in a bed where there is no gas. We confirm gas by the resistivity and by recalculating porosity to the matrix of the bed. In sandstone sections, scales compatible with sandstone are more convenient.

Updated: 16 October 2026