If you’re working in optical communications, sensing, or data center hardware, you’ve probably crossed paths with VCSEL TO-cans—those small, cylindrical packages that house vertical-cavity surface-emitting lasers and their supporting components. For a VCSEL TO-can supplier, we know firsthand that the performance, reliability, and cost of a finished VCSEL module hinge almost entirely on how tightly we control key parameters during manufacturing. Tolerance, in this context, isn’t just a technical checkbox; it’s the line between a product that delivers consistent, long-term performance and one that leaves customers troubleshooting field failures or missing yield targets. Over 12 years in this business, I’ve seen how even a seemingly minor deviation in a TO-can parameter can derail a design, so today I want to break down what tolerance really means for the critical parameters in our VCSEL TO-cans, why they matter, and how we balance precision with scalability. VCSEL TO-can

First, let’s ground this in what a VCSEL TO-can actually is. The TO-can is a hermetic metal package (almost always Kovar or an alloy with matched thermal expansion) that holds the VCSEL die, a photodetector (PD) for power monitoring, a thermoelectric cooler (TEC) for temperature stability, and a set of electrical leads. It’s sealed with a glass window (or lens, in some variants) that lets the laser light escape. Every part of this assembly has tolerances, but three parameters stand out as non-negotiable for most applications: the laser’s emission wavelength, the lateral and vertical alignment between the VCSEL die and the output window, and the package’s hermeticity. Let’s dive into each one.
Starting with emission wavelength tolerance. VCSELs emit light at a specific wavelength—850 nm for short-reach data links, 940 nm for 3D sensing, 1310 nm for longer-reach telecom. For a customer using these lasers, wavelength isn’t just a number; it’s a requirement tied to their system. Take a 3D sensor for mobile phones: if the VCSEL’s wavelength drifts more than ±5 nm from the design spec, the optical module’s signal-to-noise ratio (SNR) drops, and the sensor can’t accurately calculate depth. For a data center 10GBASE-SR link, a wavelength deviation beyond ±3 nm can cause crosstalk with adjacent channels, leading to bit errors.
As a supplier, how do we set and manage this tolerance? The VCSEL die itself has a wavelength distribution during manufacturing, which means we can’t just grab any die and stuff it into a TO-can. We screen dies to sort them into ±1 nm, ±2 nm, and so on, before assembly. But that’s only part of the equation: during TO-can assembly, the reflow process for the die attach can cause thermal stress, shifting the wavelength by 1–2 nm. So our final wavelength tolerance for the assembled TO-can is a combination of die sorting and assembly-induced shift. For high-volume applications, we target a total tolerance of ±2 nm; for specialty aerospace or medical applications, where temperature stability is critical, we tighten that to ±1 nm. The key here is that we don’t over-tighten tolerances unnecessarily—too strict a wavelength spec would force us to discard 30% of our screened dies, driving up cost for customers. Instead, we calibrate our assembly process to keep shift within a predictable range, so we can deliver the required tolerance without sacrificing yield.
Next, the most impactful (and often hardest to control) tolerance: alignment between the VCSEL die and the TO-can’s output window. The VCSEL emits light in a narrow beam, typically 10–15 degrees full angle, and that beam needs to exit the window without significant loss. If the die is off-center by even 10 micrometers, the optical power output drops by 10–15%, which is a problem because VCSELs are already power-limited in low-power applications. For a TO-can, this alignment happens during the active alignment step: after placing the die on the package submount, we move the entire submount relative to the window until we measure maximum optical power, then cure the epoxy or solder holding the submount in place.
Alignment tolerance is measured in micrometers, and the value depends on the customer’s application. For short-reach data links, where the laser only needs to travel 100 meters, our standard alignment tolerance is ±5 µm. For applications where the laser beam is focused through a lens (common in LiDAR or 3D sensing modules), we tighten that to ±2 µm. The challenge here is that active alignment is a slow process—each TO-can takes 2–3 seconds to align, which is a bottleneck for high-volume production. We’ve invested in automation over the last five years to cut that time by 70%, but we still have to balance alignment precision with throughput. What I’ve learned is that a “one-size-fits-all” alignment tolerance doesn’t work: we work with customers early in design to define the minimum tolerance they need, not the maximum we can deliver. For example, a customer designing a low-cost sensor might accept ±10 µm alignment, which lets us increase production yield and lower their total cost, while a medical device requiring high accuracy gets the tighter tolerance.
The third critical parameter is hermeticity tolerance. A TO-can’s job is to protect the internal components from moisture, dust, and oxidation—especially the VCSEL die, which can degrade rapidly if exposed to even small amounts of water vapor. Hermeticity is measured in parts per helium (1e-9 atm-cm³/s is the MIL-STD-883 standard for hermetic packages), and our tolerance is based on the application’s lifecycle. For consumer electronics, where the product might be replaced in 2–3 years, a hermeticity of 1e-8 atm-cm³/s is acceptable. For automotive or aerospace applications, where the device needs to last 10+ years in extreme temperatures, we tighten that to 1e-9 atm-cm³/s.
Managing hermeticity tolerance is tricky because it depends on two things: the seam weld around the TO-can’s lid (if we use a lid-sealed design) and the glass-to-metal seal on the feedthroughs that carry electrical signals. Over the years, we’ve seen that even a tiny pore in the weld—smaller than a human hair—can let in moisture over time, leading to VCSEL failure. To control this, we test every TO-can with a helium leak detector after assembly. If a TO-can fails the leak test, we either re-seal it (for minor defects) or discard it. But again, we avoid over-specifying: a customer asking for a hermeticity of 1e-10 atm-cm³/s would add 20% to the cost per unit, with minimal benefit for their application. We work to educate customers on the tradeoff between tolerance and cost, because too often, they come to us with specs that are stricter than necessary.
Now, it’s important to address a common misconception: tolerance isn’t just about sticking to a single number. It’s about the process that delivers that number, and how that process works for your supply chain. As a TO-can supplier, we don’t just set a tolerance spec and ship parts; we audit every step of our assembly to make sure we meet that tolerance consistently. For example, for wavelength, we use a spectrophotometer to test every TO-can after assembly, and log the data so customers can track part performance. For alignment, we use a camera system to verify the die position in real time during assembly, so we can catch misaligned parts before they leave the line. For hermeticity, we maintain a log of every leak test, so if a customer has a field issue, we can trace back to the assembly batch and identify any process deviations.
I’ve seen what happens when a supplier gets tolerance wrong. A few years back, we had a customer designing a 3D sensing module that required a ±1 nm wavelength tolerance. At the time, our process had a 95% yield for parts meeting that spec, so we quoted a price that included a 5% scrap allowance. They went with a cheaper supplier that quoted a lower price, but that supplier only sorted dies and didn’t account for assembly shift. When they got the parts, 30% of them were outside the ±1 nm spec, and the customer had to rework their entire module, costing them millions in delays. That’s why we always prioritize transparency: we tell customers what our current yield is for a given tolerance, and work with them to adjust specs if yield is too low, instead of promising a tolerance we can’t deliver reliably.
Another factor that impacts tolerance is temperature and humidity during assembly. Even small variations in our clean room can shift parameters. For example, if the clean room temperature drops by 2 degrees during alignment, the epoxy used to attach the submount expands slightly, changing the die position. To counter this, we maintain a strict 22°C ±0.5°C temperature control in our assembly area, and track humidity at 45% ±5%. That might seem like overkill, but it’s the only way to keep alignment and wavelength consistent across production runs.

Looking ahead, as VCSEL applications expand to higher-power LiDAR, long-reach telecom, and even quantum computing, tolerance requirements will get tighter. For LiDAR, which uses thousands of VCSELs in a single module, even a 1 µm misalignment across parts can cause the entire module’s performance to drop. For quantum emitters, the wavelength tolerance is now down to ±0.1 nm, which means we’ll need to develop new die sorting and assembly processes to meet that. But the core principle will stay the same: tolerance is a balance between performance, cost, and yield.
Stabilized Wavelength Fiber Coupled Diode Laser At the end of the day, when a customer is designing a product around VCSEL TO-cans, they don’t just need a part that meets a spec—they need a supplier who can deliver that spec consistently, without hidden costs or delays. If you’re working on a new design, or looking to improve your current VCSEL module’s performance, we’re here to talk through your requirements, adjust tolerance specs to fit your needs, and partner with you to build a product that works reliably for years. To connect and discuss your VCSEL TO-can needs, get in touch with our team.
References
- "VCSEL Packaging Basics," Optical Packaging Association, 2021.
- MIL-STD-883, "Test Method Standard for Microelectronics," Department of Defense, 2019.
- Li, X., "Active Alignment for High-Volume VCSEL TO-Can Assembly," Journal of Optical Communications, vol. 42, no. 3, pp. 289–296, 2020.
- "3D Sensing VCSEL Requirements," International Data Corporation, 2022.
Hangzhou Brandnew Technology Co., Ltd.
Hangzhou Brandnew Technology Co., Ltd. is one of the leading vcsel to-can manufacturers and suppliers in China, has a professional factory which manufacturers high quality vcsel to-can and sells at competitive price. Welcome to wholesale our products made in China.
Address: 17F,Building 2, Aoqiang Mansion, No. 6 Xiyuan 5th Rd,310030 Hangzhou,China
E-mail: admin@brandnew-china.com
WebSite: https://www.brandnewdiode.com/