Why Your IR Lamps are Probably Ruining Your Glass
If you’ve ever dealt with batch glass annealing, you know the frustration of “batch drift.” It’s a nightmare. You run one load and it comes out flawless. Then, for no apparent reason, the next load is a mess—internal stress, weird optical distortions, the works. Most engineers start by obsessing over the timer or checking the furnace insulation. But usually, the real problem is hiding in plain sight: your infrared (IR) lamps aren’t actually putting out the same amount of heat.
The problem with “close enough”
Here is the thing about IR lamps. When you wire up a bank of them, you just assume every tube is doing the same job. But cheap lamps have a tolerance of maybe 5% to 10%. That doesn’t sound like much, but in a batch furnace, a 5% dip in heat creates a cold spot. Glass is picky. If one section doesn’t hit that annealing point at the exact same time as the rest, you’re left with residual stress. That’s why we obsess over high-consistency lamps. We try to get that variance down to ±2%. It keeps the thermal profile flat across the entire furnace. No surprises.
Picking the right gear (without burning it out)
Choosing a lamp based on your glass thickness is a bit of a balancing act. Shortwave IR is great because it penetrates deep into the glass. Medium-wave is your go-to for surface absorption. We usually use quartz envelopes with specific coatings to get the emission spectrum just right. But be careful. You can’t just crank up the wattage to “fix” a cold spot. If you over-drive a lamp, it’ll burn out way too fast. And if you cram a high-density 2kW lamp into a tight space, you better make sure your cooling fans are up to the task, or you’ll end up melting your sockets.
Stop the guessing game
When your lamps are consistent, your ramp-up and soak times actually become predictable. You stop guessing. The glass reaches the soak temperature at the same time, regardless of where it’s sitting on the rack. It means you can finally stop messing with your PID controllers every time you start a new batch. It just works.