How to Choose kV and Filter for Your Micro-CT Experiment

Most CT (computed tomography) artifacts can be traced back to selecting the wrong X-ray energy. The question I’m asked most often is simple: “Which kV and filter are best for my sample?” I understand why people ask this; it’s one of the first things I consider when collecting CT data. Together, kV and filter choice determine X-ray penetration, image contrast, and artifact levels. The choice can make the difference between high-quality and poor-quality CT data.

The good news is that you can use some basic guidelines and tools to determine the right combination. In this article, I share a practical workflow for selecting the right kV and filter combination, show some examples, and introduce quality metrics commonly used to evaluate X-ray CT data. Hopefully, you will find some tips to apply when you plan your next X-ray CT experiments.

How to choose kV and filter for your micro-CT experiment

1. Why is kV and filter choice important?
2. How to choose kV and filter for your sample

Step 1: Estimate the optimal X-ray energy for your sample.

Example: Single-material sample (Quartz core)

Step 2: Consider contrast for multi-material samples

Why a low-Z vs low-Z materials combination is challenging

Example: Water vs. polyethylene

Why low-Z vs high-Z materials can also be challenging

Step 3: Map energy choice to instrument parameters (kV and filter)

Understanding kV (X-ray tube voltage)

How filters modify the X-ray spectrum

Step 4: Defining image quality metrics before you scan.

Quantitative metrics

Qualitative metrics

Example: Effect of artifacts

Example: Segmentation-friendly data

Example: Comparing data collected at different kV

Step 5: Practical starting points by sample type

3. Troubleshooting: What if your CT scanner can’t reach ideal conditions?
4. Quick workflow summary

Takeaways

1. Why is kV and filter choice important?

Why is it so important? As we discussed in our introduction webinar, X-ray CT is an X-ray absorption imaging technique, and the transmitted intensity through a sample follows Beer-Lambert’s law:

I=I₀e^-μ_lt

where I₀ is the intensity of the incident X-ray beam, I is the intensity of X-rays transmitted through the sample, and t is the sample thickness. The linear attenuation coefficient, μ_l, is proportional to density, ρ, and the mass attenuation coefficient, μ_m, as follows:

μ_l=μ_mρ

In practice, atomic number also influences how much a sample will attenuate based on the plot below (Toda, H., 2021. “X-Ray CT: Hardware and Software Techniques.” Springer Singapore, Singapore.). 

Thus, X-ray attenuation increases as density increases, and materials with higher atomic number tend to attenuate more strongly at X-ray energies typically used for X-ray CT experiments. 

So, what does all this mean? Basically, if the X-ray energy is too high for a given sample density and thickness combination, the sample doesn’t absorb the X-rays enough, and you don’t achieve good contrast. If the X-ray energy is too low, low-energy photons are preferentially absorbed as the beam passes through the sample. This shifts the effective energy spectrum toward higher mean energy, a process known as beam hardening, which can introduce cupping and streak artifacts in reconstructed images. This means that you need to choose the X-ray energy to match the sample properties to achieve:

• optimal penetration and transmission of X-rays for a sample

• sufficient contrast between materials within the sample

• artifact reduction

So, how do we decide which combination makes the most sense in practice? Luckily, we have physics and many resources available to help us narrow the scope of ideal kV and filter choice.

2. How to choose kV and filter for your sample

Step 1: Estimate the optimal X-ray energy for your sample.

Image quality in CT data can be evaluated in several ways, including metrics such as signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR). These rely on the detected intensity of the transmitted X-ray signal. Transmission, T, is easy to calculate; it’s the ratio of the transmitted intensity (I) and the incident beam intensity (I₀). Using the Beer-Lambert law, we can express this directly as e^-μ_lt, giving,

T=I/I₀ =e^-μ_lt.

So, what transmission should we aim for? CT standards generally recommend a minimum transmission of about 10–20%, with 14% often cited as a benchmark. These values appear in documents such as EN 16016-2:2011 and EN ISO 15708-2:2025. (“EN 16016-1:2011 - Non destructive testing - Radiation methods - Computed tomography - Part 2: Principle, equipment and samples.” “EN ISO 15708-2:2025 - Non-destructive testing - Radiation methods for computed tomography - Part 2: Principles, equipment and samples (ISO 15708-2:2025).”) 

A group of researchers ran tests to determine the optimum transmission that worked across different sample types and found something important: there is no single “best” transmission that works for every material or geometry. What did hold true is that keeping minimum transmission in the 10–20% range ensures the beam can penetrate the sample well enough to produce usable data. (Reiter, M., Kasperl, S., Kuhn, C., Erler, M., Weiß, D., Heinzl, C., Gusenbauer, C., Kastner, J., 2012. “Evaluation of transmission based image quality optimisation for X-ray computed.” e-Journal of Nondestructive Testing 17.) In other words, the guideline is a solid starting point for selecting X-ray energy, but you’ll still need to fine-tune it based on your sample’s composition and shape.

Since we usually know what our sample is made of, we can use the NIST X-ray Attenuation Database to quickly estimate how it absorbs X-rays at different energies. We could also simply place the sample in the CT system, turn on the X-rays, and measure the transmission directly. But it’s helpful to see how NIST’s online tool lets us predict that ahead of time. Let’s take a look at that process. (Hubbell, J.H. and Seltzer, S.M. (2004), Tables of X-Ray Mass Attenuation Coefficients and Mass Energy-Absorption Coefficients (version 1.4). [Online] Available: http://physics.nist.gov/xaamdi [2025 11 05]. National Institute of Standards and Technology, Gaithersburg, MD.)

Example: Single-material sample (Quartz core)

In this example, we’ll use a cylindrical quartz rock core as our theoretical sample. Quartz is silica, or silicon dioxide (SiO₂), with a density of roughly 2.65 g/cm³. 

The NIST tool provides the mass attenuation coefficient, µm/ρ, at different X-ray energies which we can plot for SiO₂, as follows.

To compute transmission, T, we first convert this to the linear attenuation coefficient by multiplying by the density of quartz (2.65 g/cm³). From there, transmission follows directly from 

T=e^-μ_lt.

For our example, let’s imagine three quartz cores with diameters of 10 mm, 20 mm, and 50 mm. Now we can look at how the transmission varies with X-ray energy for each of these sample sizes.

From the plot, a few things stand out.

• The smallest core, at 10 mm, reaches the minimum transmission target at 30 – 35 keV.

• The 20 mm core needs a higher range, about 42–52 keV.

• The thickest core, at 50 mm, doesn’t reach minimum transmission until around 95–200 keV, which may exceed the energy limits of some X-ray CT systems. 

From this calculation, you can tell that 42-52 keV is a good place to start if you were to scan a 20 mm thick solid quartz.

Keep in mind that this calculation assumes a fully solid quartz core. Real rock samples usually have pores, cracks, and other heterogeneities that change how they absorb X-rays. So, treat this as a baseline or a starting point. You’ll still want to measure transmission on your CT system to get an accurate value for the real sample.

Step 2: Consider contrast for multi-material samples 

Real samples are rarely made of a single solid material. Once multiple materials are involved, considering one transmission rate alone isn’t enough. You also need to think about contrast, the grayscale differences that let you distinguish one material from another.

Why a low-Z vs low-Z materials combination is challenging

Some combinations, such as water and polyethylene, are difficult to image because both materials have similar densities and are composed of low-Z elements. Their attenuation values track closely, giving you only subtle grayscale differences to work with.

A practical way to understand the subtle attenuation differences is to plot the linear attenuation coefficient for the materials of interest. The NIST database (and tools like MuCalc) make this easy. (Hanna, R.D., Ketcham, R.A., 2017. “X-ray computed tomography of planetary materials: A primer and review of recent studies.” Geochemistry 77, 547–572. https://doi.org/10.1016/j.chemer.2017.01.006) 

Example: Water vs. polyethylene

Let’s look at a simple low-Z vs. low-Z case: a polyethylene container (ρ=0.88 g/cm³) partially filled with water (ρ=1.0 g/cm³). The linear absorption coefficients of these two materials, calculated using the NIST database, show a larger separation at 15 keV than at 25 keV, indicating greater contrast at the lower energy. This demonstrates how low-energy X-rays enhance grayscale separation between low-X materials.

Why low-Z vs high-Z materials can also be challenging

It might seem then intuitive that mixing a low-Z material with a high-Z one should produce strong contrast. And in many cases, it does, but not without complications. Low-Z vs. high-Z combinations bring their own challenges:

• Beam hardening: The high-Z material strongly absorbs low-energy photons, distorting the effective X-ray energy spectrum and causing artifacts.

• Photon starvation: The high-Z regions may attenuate low-energy X-rays so strongly that almost no photons reach the detector, again creating artifacts and streaking.

• Energy trade-offs: Increasing kV reduces artifacts and improves penetration, but it can also reduce contrast in the low-Z material, especially when the attenuation curves flatten out at higher energies.

So, what approach does one take when their sample contains both low-Z and high-Z materials? Generally, I start with test scans with filters and higher kV to reduce artifacts as much as possible, but may tune down the X-ray energy so that the signal from the low-Z material is still visible.

Step 3: Map energy choice to instrument parameters (kV and filter)

Up to this point, we’ve focused on X-ray photon energy in terms of keV. In practice, what we actually control on a microCT system are tube voltage (kV), tube current (μA), and the X-ray filter. These three parameters shape the final X-ray spectrum your sample sees.

Understanding kV (X-ray tube voltage)

Most laboratory X-ray CT systems use a tungsten (W) X-ray source because its high atomic number improves X-ray production efficiency and its high melting point enables operation at higher power levels.

When high-energy electrons strike the target, they decelerate and generate Bremsstrahlung radiation. Because Bremsstrahlung radiation forms a continuous energy distribution, any given kV setting produces a broad spectrum of photon energies rather than a single energy. Increasing the tube voltage shifts this spectrum toward higher energies and increases the maximum photon energy to a value equal to the applied kV. For example, a 130 kV system can produce photons up to 130 keV. The plot below shows X-ray spectra calculated for a tungsten (W) target at several tube voltages. In these calculations, tube current was held constant and no added filtration was applied.

CT systems are typically described by their maximum tube voltage because this value defines the upper limit of material thickness and density that the system can penetrate. A higher maximum kV allows imaging of denser or thicker samples, while lower-kV systems are better suited to low-density materials requiring higher contrast at lower energies.

For a tungsten target without added filtration, the mean photon energy is typically on the order of one-third of the applied kV, as shown below.

How filters modify the X-ray spectrum

After you choose a kV, you can further “tune” the X-ray spectrum by placing a thin sheet of metal between the X-ray source and the sample. The plot below shows X-ray spectra for a W target, calculated using SpekCalc. (Poludniowski, G., Landry, G., DeBlois, F., Evans, P.M., Verhaegen, F., 2009. “SpekCalc: a program to calculate photon spectra from tungsten anode x-ray tubes.” Phys. Med. Biol. 54, N433–N438. https://doi.org/10.1088/0031-9155/54/19/N01) Notice how different filter materials shift the X-ray spectra and increase the mean X-ray energy.  

X-ray filters remove low-energy photons by absorbing them, effectively ‘hardening’ the beam. This shifts the spectrum toward higher mean energy before it enters the sample.

The downside of using X-ray filters is that they reduce the incident X-ray beam intensity. The plot below, generated using MuCalc, shows the transmission rate for several common CT filters as a function of X-ray energy. Both the filter material and its thickness influence how much of the beam gets through. As a result, you may need to compensate by increasing the scan time.

Because different kV and filter settings can produce similar mean X-ray energies, you’ll still need to test a few combinations to see which one gives the best results. The optimal choice is the one that delivers the best data quality for your specific sample.

A side note about X-ray current:

You might be wondering, “What about the X-ray current? Don’t I need to set that too?”

Yes. The X-ray current (μA) controls the number of electrons striking the target and therefore the total X-ray photon output, as shown below. 

Increasing the current improves photon counting statistics for your CT data but also increases the heat load on the target. To dissipate that heat, many microfocus sources increase the effective focal spot size. A larger focal spot introduces geometric unsharpness and reduces spatial resolution. Selecting current therefore requires balancing photon flux against focal spot size and the spatial resolution required for the CT measurement. 

Most systems specify how the focal spot grows with current, usually shown in plots of the horizontal and vertical spot dimensions, as shown below. Because of this, selecting the current requires balancing maximizing X-ray intensity with maintaining a small X-ray spot size for sharpness and the achievable spatial resolution needed for your sample and imaging goals.

Step 4: Defining image quality metrics before you scan.

Once you have a sense of which scan conditions might work for your sample, it’s tempting to jump straight into collecting test data. Before doing that, it’s worth deciding how you'll judge image quality. Choosing your metrics early will make your test scans far more efficient and meaningful.

Why define metrics first? Different CT experiments demand different metrics for judging image quality. These might include ease of segmentation, SNR, CNR, artifacts, and other criteria. By identifying your evaluation criteria before scanning, you ensure you collect data that truly aligns with your experimental goal.

Quantitative metrics

Commonly used quantitative metrics are SNR and CNR. The SNR is defined as the ratio between the mean intensity (I_object) for an object or region of interest (ROI) and the standard deviation (σ_object) of intensity.  (Rodríguez-Sánchez, Á., Thompson, A., Körner, L., Brierley, N., Leach, R., 2020. “Review of the influence of noise in X-ray computed tomography measurement uncertainty.” Precision Engineering 66, 382–391. https://doi.org/10.1016/j.precisioneng.2020.08.004)

SNR=Signal/Noise=I_object/σ_object

SNR quantifies the stability of measured intensity within an ROI. A higher SNR means that random intensity fluctuations are small relative to the signal, making voxel values more statistically reliable. Although SNR does not directly measure attenuation, it reflects the precision of gray values within an ROI under a given set of scan and reconstruction conditions. In practice, higher SNR indicates lower relative noise in the reconstructed volume, which improves the reliability of downstream tasks such as threshold selection, segmentation, and dimensional measurement.

The CNR is defined as 

CNR=Contrast/Noise=| I_object-I_background |/σ_object

The CNR quantifies how distinguishable two materials are relative to image noise.  Higher CNR indicates that the difference in gray values between materials exceeds the background noise level, improving segmentation and boundary detection reliability.

Qualitative metrics

Qualitative criteria, such as ease of segmentation, artifact severity, and edge sharpness, are also important to consider. Each of these affects data analysis in different ways. 

For example, ease of segmentation will influence the difficulty level and time required to separate phases in your CT data. An ideal situation is one where you can use automated or threshold-based segmentation workflows.

Additionally, artifact severity will affect how well you can interpret details in the image. That might matter most in cases where your goal for CT data is failure analysis or the detection of small features. Artifacts also affect dimensional measurements and feature extraction, such as porosity/inclusion analysis.

Let’s take a look at examples of how these metrics are used.

Example: Effect of artifacts

In this example, we’ll look at some data collected for a 3D printed polymer sample, used in a blog article describing artifacts and how to minimize them. This sample was mounted in the CT instrument so that its edges were parallel to the incident X-ray beam. In the parallel orientation, severe scattering artifacts are present, making it difficult to segment the phases and to determine sample surfaces cleanly. A second scan was collected after tilting the sample. In the tilted orientation, streaking artifacts are greatly minimized. This example shows how artifact reduction or correction is paramount to data quality because the artifacts conceal surfaces and hamper dimensional analysis. 

Example: Segmentation-friendly data

Below is a 2D cross-section from a rock core scan. The rock grains form a clear, high-intensity peak in the histogram. Because the grayscale separation is strong, Otsu binarization easily segments the grain material from the background. In this case, segmentation takes only seconds, and you can quickly move to the next step in your CT data analysis, such as porosity calculation. 

Example: Comparing data collected at different kV

To illustrate how quantitative metrics shift with energy, let’s revisit the water-filled polyethylene sample we discussed earlier. This sample is a polyethylene (p) tube containing water (w) and air (a). 

Scans were collected at 50 kV, 90 kV, and 130 kV, each using 8 W of total tube power to avoid any current-related changes in focal spot size. Using the table of mean X-ray energy above, we see these kV settings correlate to 15.3 keV, 28.8 keV, and 42.6 keV, respectively. Remember when we evaluated plots of linear attenuation coefficients, our analysis predicted better contrast for data collected at low-energy X-rays. To investigate this hypothesis, ROIs were drawn for each phase to compute mean intensities and standard deviations.

From the SNR calculations, we see that the highest SNR values occur for data collected at 130 kV (SNRw: 21.2, SNRp: 20.6, SNRa: 14.5). This improvement makes sense because we know that higher X-ray energy increases transmission through the sample, which improves detected photon counting statistics and reduces relative noise.

The CNR calculations tell a different story, as shown below.

The best CNR occurs at 50 kV, matching the prediction based on the attenuation curves we looked at previously. Lower kV increases the grayscale separation between water and plastic.

For this specific multi-material case, if the goal is material segmentation or phase distinction, a lower kV setting (e.g., 50 kV) would be preferred despite the lower SNR. However, if the priority is reduced noise or improved X-ray transmission through thicker sections, a higher kV setting may be justified. The optimal setting depends on whether contrast or SNR performance is more critical for your analysis task.

The takeaway message is that no universal metric defines “good” CT data for every sample. High SNR improves the statistical stability of gray values, while high CNR improves material separability. In many cases, minimizing artifacts such as beam hardening or streaking may be even more important than maximizing either SNR or CNR. Clarifying your evaluation criteria before scanning ensures that acquisition parameters are optimized for the measurements you actually intend to perform.

The table below provides a practical guide for aligning acquisition strategy with analysis objectives.