RC car IR sensor
What you are seeing is simply that irradiance at your sensor from solar energy in the band passed by the epoxy "IR filter" of your sensor (roughly from 770 nm. to 1050 nm.) is much greater than the irradiance at about 880 nm. from your emitters. As I stated earlier, the world is very bright at near IR in daylight.
I was recently trying to measure displacement of an 880 nm. target (the Opto Diode OD-50L) at a distance of 30 to 50 feet on a sunny day using a 5" aperture reflecting telescope. Having difficulty getting a good signal, I viewed the target area with a Find-R-Scope IR viewer. While the OD-50L was clearly the brightest object in the view, the entire scene was very bright and highly visible. Passing a US$100, 880 nm., 10 nm. passband interference filter in front of the viewer objective made almost no difference in scene brightness! The "filter" on your BPW82 is EXTREMELY wide (280 nm.) and offers negligible effective filtering of ambient IR.
Don't forget, sunlight has a very broad spectrum, meaning that the amount of solar energy passing through a filter is (roughly) directly proportional to the bandwidth of the filter. Thus, if you use a filter with half the bandwidth of the 280 nm "filter" (really, just the natural transmittance of the epoxy encapsulant) of the BPW82, you will cut the solar energy immission in half, or get a 3 dB reduction of interfering solar energy. Not much benefit until you get down to a quite narrow filter bandwidth. Ideally, to get the best benefit from an IR filter, you need a filter passband that matches the emission bandwidth of your emitter. I had to resort to a HeNe laser, cube corner retroreflector target, and a 1 nm. passband filter to get a reasonable S/N ratio out of a silicon Position Sensing Diode in my telescope.
I think you will find that your emitters have a fairly broad beam angle. All the emitted IR that does not fall onto the sensor aperture is simply wasted by illuminating the landscape. You can get a reasonable idea of the emitted beam angle by observing the output from your sensor as you rotate the emitter around the vertical and horizontal axes at a fixed location in the field of view of the sensor.
The intensity of the IR from your emitter follows the inverse-square law. That is, at double the range, you get 1/4 the intensity due to spherical spreading. So, if you are having trouble with detection at a few mm. or cm. on your bench, you will be in deep stuff at a meter or two, even neglecting reflection losses. You can greatly improve this situation with optics on emitter and/or sensor, and with modulation/detection that allows detection of the desired emission buried deep in ambient optical noise. The latter is analogous to selective detection of a desired radio signal. (By the way, where did the "455 kHz" mentioned in your original post come from?)
Just shielding the light path and limiting the field of view of the sensor can improve your S/N ratio by reducing stray irradiance at the sensor from off-axis sources, but that does nothing to enhance the received signal level. Using a more directional source or external optics on the source can greatly improve the percentage of emitted IR that falls on the sensor.
Hope this helps.
awright