On a DR panel datasheet, pixel pitch is the biggest number printed, and it's the one the standard looks at least. ISO 17636-2:2022 governs weld radiography with digital detectors, and it judges an image on two quantities measured on the image itself: basic spatial resolution, read from a duplex wire indicator, and normalized signal-to-noise ratio. The pixel feeds into both without matching either.
So buying a detector for digital industrial radiography, whether for an NDT lab or a fabrication and piping shop, is less straightforward than it looks. The same panel can pass class B on an 8 mm pipe and miss it on a 3 mm wall, or work well in a shielded bay and struggle after a year in the field. The sensible starting point is how the standard measures the image, working back from there to the datasheet lines.
The pixel on paper and the pixel you measure
Basic spatial resolution, SRb, is measured with the duplex wire indicator of ISO 19232-5:2018. The standard version carries thirteen pairs of high-density wires, from D1 at 0.80 mm to D13 at 0.050 mm, and in each pair the gap between the two wires equals their diameter. On the image you look for the first pair that no longer separates; that wire's diameter is the SRb, and total image unsharpness is twice that. ASTM E2002-22 uses the same indicator and states its equivalence with the ISO standard, so offers backed by US documentation can be compared on the same footing.
The measured value is almost always larger than the pixel pitch, and the reason is inside the panel. In an indirect-conversion detector, X-rays are absorbed by a scintillator that emits light, and that light spreads before it reaches the photodiodes, landing on neighboring pixels too. So the standard distinguishes detector SRb, measured with the indicator on the panel, from image SRb, measured in the actual exposure geometry.
SRb also enters the second requirement. Measured signal-to-noise ratio is normalized by multiplying it by 88.6 µm and dividing by SRb. The 88.6 µm figure is the side of the square equivalent to the 100 µm circular aperture once used to measure film graininess, as Ewert, Zscherpel and Bavendiek of BAM explain. The practical upshot is simple. A panel with bigger pixels collects more photons per pixel and shows a higher raw SNR, but normalization takes back most of that advantage, because the ratio is divided by a larger SRb.
How fine it has to be depends on penetrated thickness
ISO 17636-2 sets the maximum image SRb by penetrated thickness and technique class. In class B, for penetrated thicknesses between 12 and 40 mm, you need at least pair D10, that is 0.10 mm. Between 40 and 120 mm D9 at 0.13 mm is enough, while between 1.5 and 4 mm the requirement drops to 0.05 mm. According to the standard's drafters, who presented the 2022 changes at the 2023 DGZfP conference, these selection tables are unchanged from 2013. Ewert and Zscherpel sum them up with a rule of thumb: in class B the required SRb grows roughly with the cube root of thickness divided by 30.
That leads to a choice often made backwards. The panel should be sized for the thinnest wall that will come through the door, because that's where the resolution demand is strictest. A lab inspecting 3 mm tubes and 30 mm plate buys for the tubes. When the detector can't reach the required pair, the standard allows the second compensation principle. For each missing duplex pair you add one extra step on the single-wire indicator, for example D11 with W15 instead of D12 with W14. The 2022 edition allows up to three such compensations. Each one costs a higher signal-to-noise ratio, meaning longer exposures, and belongs in the written procedure before inspection starts.
Normalized SNR has its own table of minimums. In the 2013 version, as reported by Ewert, class B called for 150 below 50 kV, 120 between 50 and 150 kV and 100 between 150 and 250 kV. For iridium-192 on walls up to 50 mm the minimum was also 100. The 2022 edition reworked some application rules, including those for reduced tube voltages and for pipes inspected with flat panels. For a specification, check the values against the current text.
Magnification buys resolution and pays in penumbra
Moving the part away from the panel magnifies the image and divides detector unsharpness by the magnification factor. In exchange, geometric unsharpness grows with focal spot size times magnification minus one. ISO 17636-2:2013 combined the two in a formula: image unsharpness equals the square root of the sum of the squares of geometric unsharpness and twice the detector SRb, all divided by magnification. ASTM E2698 uses a cubic combination instead, and Zscherpel and Ewert have shown the two approximations lead to different optimum magnifications. In the 2022 edition the relationship between detector and geometric unsharpness moved to Annex F.
Two calculations with the 2013 formula, on our own assumptions, show how much the focal spot weighs. With a detector SRb of 0.10 mm and a 1 mm focal spot, the best image comes at contact, around 0.20 mm of unsharpness; magnifying by 1.2 makes it worse, about 0.24 mm. With the same detector and a 0.2 mm minifocus tube, a magnification of 2 brings unsharpness down to about 0.14 mm, and beyond that the gain disappears. The detector, then, can't be chosen apart from the tube. Focal spot physics is covered in our piece on X-ray tube focal spot and parameters, generator selection in the one on choosing an X-ray tube and generator.
Choosing a panel for your welds?
Send us wall thicknesses, source and required class: we'll tell you which resolution and which scintillator you need, with a no-obligation test on your own material and a tailored quote within 24 hours.
Email us Message us on WhatsAppPick the scintillator for the energy and the workplace
In most industrial panels the scintillator is thallium-doped cesium iodide (CsI) or gadolinium oxysulfide (GOS). CsI grows in thin columns that guide light toward the photodiodes like tiny optical fibers, so for the same thickness it spreads less. GOS is a packed powder in which light scatters more. There's a trade-off within the same family too. Zhao, Ristic and Rowlands compared two structured CsI layers in Medical Physics (2004): the high-resolution version gave about 50% higher sharpness and roughly 36% less light. A thicker scintillator stops more photons at high energy but widens the light spot. It's the same balance that separates powder and needle plates in computed radiography, described by Leblans and colleagues (2011).
Working energy also decides which tests apply. ASTM E2698-26 warns that some of its tests and requirements may not apply above 450 kV, and ASTM E2737-23 states it isn't intended for isotope sources. Then there's scatter: scintillator panels, like CR plates, are more sensitive than film to low-energy photons, and the 2007 BAM paper recommends lead or steel filters or screens to cut them. Anyone shooting iridium in the field therefore has two questions for the supplier: which characterization tests were run at that energy, and what filtration is planned.
The workplace completes the picture. The same BAM authors note that digital panels do best where temperature and humidity are stable, while CR plates also cope with mobile work in bad weather. Ingress protection, weight, cable or battery matter more in the field than in a bay, and should be compared over the real work cycle, including the repositioning a larger format saves on long seams.
Dynamic range and calibration: the noise a catalog doesn't show
A real joint contains different thicknesses: cap, heat-affected zone, base metal and, in double-wall pipe shots, even bigger steps. The panel has to stay clear of saturation on the thin side while holding the required signal-to-noise ratio on the thick one. Stated bit depth says little. More useful is the thickness range over which the detector keeps a given SNR, which ASTM E2597/E2597M-22 calls the specific material thickness range. Manufacturers measure it alongside SRb, efficiency, image lag and contrast sensitivity.
The noise left after exposure depends largely on calibration. Every panel is corrected with a dark image, which subtracts the offset signal, and one or more flat-field images, which even out the response of individual pixels. In its 2007 work BAM acquired calibration images with roughly ten times the integration used for measurement. Moving from one-point calibration to a four-point non-linear one raised the ceiling on normalized SNR from about 550 to 810. With careful calibration and long exposures, panels can show wall-thickness changes of about two parts per thousand. On an 8 mm test weld, the detector revealed micro-porosity that class C3 film did not show.
Dark signal depends on temperature, and when it creeps up over time it usually points to radiation damage, together with a growing count of bad pixels. ASTM E2597 sorts these into dead, over- and under-responding, noisy, non-linear and lagging pixels, and counts clusters of adjacent bad pixels separately. Software corrects them by interpolating from neighbors, and ISO 17636-2 accounts for the local loss of sharpness through its third compensation principle. So the offer should include the bad pixel map, the field calibration procedure and the recommended frequency.
Who checks the panel after acceptance
US standards split the responsibilities cleanly. ASTM E2597/E2597M-22 is the manufacturer's characterization, run on at least three units and explicitly not meant for users: it lets panels from different suppliers be compared on a common basis. ASTM E2737-23 is the user's practice, setting a baseline with a dedicated phantom and tracking its stability over time, and it states plainly that it is not an acceptance test. The examination itself follows ASTM E2698-26. Whoever signs reports needs to know which of these produced each number in the offer, a point picked up in our piece on who decides acceptance levels. When geometry or field work point to phosphor plates instead, the head-to-head comparison is in CR or DR in industrial radiography.
What to send us for an assessment
To recommend the right detector for radiographic inspection of your welds, we only need a few data points, but they have to be complete:
- material, joint type and minimum and maximum penetrated thickness, with diameter for pipes;
- technique standard and required class (A or B), plus the customer's acceptance standard;
- available source: maximum tube voltage and focal spot, or isotope;
- geometry: achievable source-to-object distance and room for the panel behind the joint;
- where the detector will work: shielded bay, line or field, with typical temperatures;
- volumes: exposures per shift and seam lengths;
- software and archive: reporting, DICONDE format, link to the quality system.
You can send them by email to info@dipitech.com or on WhatsApp, including drawings and reference radiographs.
The criterion to bring to the negotiating table fits in one line. The right detector reaches the SRb required by the thinnest wall you inspect, at the highest voltage you'll use, with no compensation beyond what the procedure states, and stays calibratable where it will actually work. If your joints and source are defined, the next step is a test on your own material, with a tailored quote within 24 hours and technical support in Italy. The panel is only one link in the chain, and our page on X-ray, CR and DR inspection systems shows the rest of it, from sources to software.
Frequently asked questions about DR detectors for weld inspection
What's the difference between pixel pitch and basic spatial resolution?
Pixel pitch is the center-to-center distance between two panel pixels. Basic spatial resolution (SRb) is half the unsharpness measured with the ISO 19232-5:2018 duplex wire indicator and includes light spread in the scintillator, so it's almost always larger than the pitch. ISO 17636-2:2022 judges SRb, and that's what to ask for in a quote.
What SRb does class B weld radiography require?
It depends on penetrated thickness. In the ISO 17636-2 tables, class B calls for pair D10, i.e. 0.10 mm, between 12 and 40 mm, D9 at 0.13 mm between 40 and 120 mm, and 0.05 mm between 1.5 and 4 mm. If the detector can't reach the required pair, the 2022 edition allows up to three compensations with the single-wire indicator, at the cost of longer exposures.
Is a CsI or a GOS scintillator better?
For fine detail, usually CsI, because its columnar structure guides light and spreads it less for the same thickness. Powder GOS spreads more. The choice should still be checked at your energy: the manufacturer's ASTM E2597/E2597M-22 characterization run close to the voltage or isotope you'll use is worth more than the material's name.
How often does a DR detector need calibrating?
Before use, and whenever voltage, filtration or temperature change noticeably, because dark signal depends on temperature and pixel response on energy. Long-term stability is documented with the ASTM E2737-23 baseline, which doesn't replace routine calibration and isn't an acceptance test.
Can a DR detector replace film on every weld?
In terms of standards, yes, if it reaches the SRb and normalized signal-to-noise ratio ISO 17636-2:2022 requires for that class and thickness. In practice it depends on geometry and location: where a rigid panel won't fit or the site is cold and wet, CR plates are often easier.
How much does a DR detector for weld inspection cost?
It depends on active area, SRb, scintillator, ruggedness, software and service, so a generic price would be misleading. DIPITECH prepares a tailored quote within 24 hours from wall thicknesses, joint type, source and required class, and can arrange a no-obligation test on your own material.