Hangzhou City, Zhejiang Province,China-November 19, 2026
Whether a Filter Is Usable: Five Specs on the Datasheet Tell You Everything

The most frequently asked question when purchasing filters is “What is the peak transmittance?” This number is important, but it is just one point on the curve. What truly determines whether a filter is usable are several other positions on that curve.
A spectral curve is a continuous line with wavelength on the horizontal axis and transmittance on the vertical axis. Reading this line essentially means looking at its performance at five positions: where the peak is, how wide the peak is, how high the peak is, how low the out-of-band is suppressed, and how wide the suppressed range is. These five items correspond to five sets of parameters; missing any one will lead to a biased judgment.
The first position: where the peak is
The position of the peak is called the Center Wavelength (CWL), which is the wavelength corresponding to the highest transmittance within the passband. It must be aligned with the target’s characteristic wavelength; any deviation will reduce sensitivity.
How precise the alignment needs to be depends on the application. In gas detection, the absorption peak of carbon dioxide is at 4.26 μm, carbon monoxide at 4.67 μm, and methane at 3.31 μm. These peaks themselves have width, so as long as the center wavelength of the narrowband filter falls within the peak, absorption is valid. However, whether it falls at the center of the peak or on the shoulder makes a significant difference in sensitivity.

When looking at the center wavelength, you need to check two values: the nominal value and the tolerance. Nominal ±2 nm and nominal ±5 nm represent two different classes of products; the former requires a higher level of film thickness control precision in the coating process, and the cost goes up accordingly. During selection, you should not only ask “what can you achieve” but also “what is the consistency of this batch,” as the latter determines whether system calibration can be automated.
Also, pay attention to how the center wavelength is defined. Some manufacturers calculate it based on the peak vertex, while others use the midpoint of the two half-maximum points. For symmetrical passbands, the two basically coincide; for asymmetrical passbands, a difference of a few tenths of a nanometer to a few nanometers is normal. For narrowband products especially, the drawing must specify which definition is used; otherwise, the same filter might be read as different values during bidding comparisons.
The second position: how wide the peak is
The width of the peak is called Full Width at Half Maximum (FWHM), which is the wavelength interval between the two points where transmittance drops to half the peak value. It determines how accurately this filter “recognizes” the target.
The narrower the bandwidth, the smaller the spectral range that passes through, the more interfering gases can be excluded, and the better the selectivity. The cost is twofold: less energy enters the receiver, putting a strain on the signal-to-noise ratio; and once the center wavelength drifts due to temperature, the effective receiving band may deviate from the absorption peak.

The standard engineering practice is to make the bandwidth just cover the target absorption peak without cutting too much useful energy. A common value is 1% to 5% of the center wavelength, with most scenarios falling between 2% and 4%. This range is not arbitrary; it is jointly determined by the width of the gas absorption peak, the energy distribution of the light source, and the response of the receiver.
Absolute values must also be converted. 2% of 4.26 μm is 85 nm, and 2% of 8 μm is 160 nm. The same “2% bandwidth” corresponds to a physical width that differs by a factor of two in different bands, so you cannot just focus on the percentage when reading a datasheet.
The third position: how high the peak is
Peak transmittance is the highest point within the passband. It determines the absolute strength of the signal and affects detection distance and response speed.
This number is often used as a selling point, such as “95% peak transmittance.” But a high peak does not mean the overall performance is good, because it is single-point data. What is more worth looking at is the average transmittance within the passband, as well as the curve shape within the passband. A filter with a 95% peak but steep drops on both sides may actually provide less effective energy than one with a 90% peak and a flat passband.

For imaging applications, passband flatness directly affects color reproduction and grayscale uniformity; for thermometry applications, the absolute accuracy of transmittance is more important than the peak height. Temperature inversion relies on the received radiant power, and every fixed percentage deviation in transmittance translates into a systematic offset in the temperature inversion, rather than random noise.
This is also why the acceptance criteria for the same type of filter must be defined separately for imaging and thermometry. Imaging looks at resolution and contrast; thermometry looks at absolute accuracy and long-term stability of transmittance.
The fourth position: how low the out-of-band is suppressed
This is the most easily overlooked item, and also the one most prone to problems. How low the out-of-band transmittance is suppressed is called blocking depth, expressed in Optical Density (OD), with the relationship OD = -lg T.
OD3 corresponds to an out-of-band transmittance of 0.1%, OD4 to 0.01%, and OD5 to 0.001%. For every OD increase, out-of-band leakage is reduced by an order of magnitude, and the process difficulty goes up a step.

Why is this item critical? Because in real scenarios, a single gas almost never exists alone. Industrial sites simultaneously have moisture, carbon dioxide, and volatile organic compounds, and sunlight has energy across all bands. If the out-of-band is not deeply suppressed, energy from non-target bands will leak in, creating cross-interference that manifests as falsely high readings or a slowly creeping baseline.
There is a number that illustrates the magnitude difference. For the same out-of-band background, the residual interference of OD4 is two orders of magnitude lower than that of OD2. In scenarios where the background itself is already stronger than the target, this difference does not determine whether the reading is accurate, but whether the system can still function at all.
The fifth position: how wide the suppressed range is
The blocking range is the wavelength interval outside the passband that needs to be suppressed. It answers the question: “from where to where must it be blocked?”
Taking a narrowband filter for gas detection as an example, the blocking range must follow the response of the receiver and cannot use a generic value. The thermopile commonly used in NDIR responds from the visible region all the way to over ten micrometers in its absorption layer, so the entire response band must be suppressed.

This means the filter cannot just block a little on either side of the passband; it must suppress from the visible all the way to the far-infrared. If environmental thermal radiation from the 3 to 14 μm range leaks in, it will directly raise the baseline and drown out the tiny absorption signal at 4.26 μm.
Therefore, the blocking range is neither “the wider the better” nor “just give a number.” The correct approach is to base it on the actual response band of the receiver, plus the known interference sources in the system, and take the union of the two. For a system using a silicon photodiode from visible to near-infrared, suppressing up to 1.1 μm is sufficient; for a thermopile doing mid-infrared gas detection, it must be suppressed all the way to the far-infrared.
Vague blocking ranges are a common problem in bidding documents. Writing only “out-of-band blocking OD4” without specifying the range is equivalent to no definition at all. Whether it is 200 to 2500 nm or 400 to 1100 nm makes a huge difference in process difficulty and cost. This column on the datasheet must be filled with specific numbers.
There is also a segmented requirement. Some applications only care about specific bands not leaking in—such as avoiding a strong absorption band. In this case, the suppression effort can be concentrated on these few segments, which costs less than broadband suppression. This is custom design work that requires the application side to map out the interference sources clearly.
Looking at the five positions together
These five parameters are not independent; there are clear trade-offs between them.
Narrowing the bandwidth improves selectivity but reduces energy; deepening the blocking depth improves interference rejection but increases the number of layers and process difficulty.
The trade-off between peak transmittance and blocking depth is essentially a trade-off in coating stack complexity. Deep blocking requires more layers and greater total thickness, and every additional layer adds absorption and scattering loss. Therefore, once the blocking depth is given, the peak transmittance has an upper limit determined by the complexity of the coating stack. Trying to increase the peak often requires relaxing the blocking or bandwidth, rather than simply reducing layers. Reducing layers first sacrifices blocking and edge steepness.
These items form a typical four-corner constraint, and it is impossible to be optimal in all simultaneously.
Therefore, the correct order of selection is to first define the application constraints, and then determine the parameter combination. What gas is being measured determines the center wavelength; what interference exists on site determines the blocking depth and blocking range; the light source power and receiver sensitivity of the equipment determine how wide the bandwidth and peak transmittance can be. Conversely, picking parameters first and then finding an application usually leads to starting over during system debugging.

There is one more item to confirm separately: under what conditions these metrics were measured. A curve measured at room temperature and humidity is not the same as the curve after installation and operation at 60 degrees Celsius.
The rate at which the center wavelength shifts with temperature is strongly related to the material system and band, and a single number cannot be given generically. Common narrowband filters from near-infrared to visible are mostly on the order of 0.01 to 0.05 nm per degree Celsius; long-wave infrared filters using substrates with high refractive-index temperature coefficients, such as germanium, can reach 0.1 to 0.5 nm per degree Celsius.
A temperature difference of tens of degrees Celsius can accumulate to a few nanometers for the former and to over ten or even tens of nanometers for the latter. For a bandpass filter with a bandwidth of only tens of nanometers, this amount is enough to change the effective receiving band.
Therefore, specific values must be obtained from the supplier based on the actual coating stack and substrate by requesting the temperature drift curve; applying a generic order of magnitude will lead to calculation errors.
Returning to the curve itself. What you should truly have in hand during selection is not just a nominal curve, but three sets of data: the actual measured curve at room temperature, the temperature drift curve over the operating temperature range, and the center wavelength distribution over consecutive batches. The first two sets determine whether it will still be accurate after installation, and the third set determines whether the next batch will still match.
MULTI IR can provide these three sets of data for infrared filters, with a supported control level of center wavelength tolerance within ±2 nm, out-of-band suppression above OD4, and average transmittance in the infrared band above 91%. By looking at these numbers alongside the curves, you can estimate operating condition errors during the selection stage, rather than waiting until system debugging to find out.

The most worthwhile thing to spend time on during selection is not comparing the decimal places of peak transmittance, but clearly asking about the actual performance of the five positions under the current operating conditions, especially the curve shape after temperature changes. A filter that looks good on the datasheet but requires repeated compensation after installation into the system has a total cost far higher than one that was selected correctly from the start.
MULTI IR – National-level “Little Giant” Specialized and Sophisticated Enterprise. Over 10,000 types of infrared sensitive components in stock. Top 3 globally in comprehensive strength. Product portfolio covers infrared filters, optical coatings, infrared sensors, and other core categories. Lead drafter of the infrared filter industry standard.
Website: www.mirhz.com | Global Site: www.miroptech.com
About Us
Founded in 2007, Hangzhou MULTI IR Technology Co., Ltd. is an optoelectronic technology enterprise integrating R&D, production, and sales. Its products are widely applied in aerospace, medical care, AR/VR, display imaging, photography, and other fields, steadily holding the position of the world’s largest spot supplier of optical components.
Media Contact
Company Name: HANGZHOU MULTI IR TECHNOLOGY CO., LTD.
Contact Person: Media Relations
Email: Send Email
Country: China
Website: https://www.miroptech.com/
