IR connection troubleshooting

Hello all

The place where I work, we have some processors communicating by means of IR. They work with a simple protocol, in that every time a byte is transferred, it is sent a number of times (typically 100+). The receiving processor then calculates the number of correct transfers. If 70% of the signals had the same value, the command is considered valid.
Most of the time this works, but sometimes for inexplicable reasons, we get communications errors.

The two IR-circuits are placed 20-40cm apart.
The processor is an ATMEGA162V running at 7.3728MHz (crystal).
The IR-circuit is based on a TFDU4100 with 165 Ohms in series with the IR-LED, speed is 115kbit/sec.
Power supply is 5V

The PCBs are placed in an industrial environment with a lot of noise. We have previously had problems with airborne noise as well as noise coming from the power supply which also supplies 3-phase motors without any filtering than some big caps in the PCB's internal power supply. The PCB has basic noise protection which has removed most other errors we had on this PCB.

Could some noise still cause poor IR communications?

Are there any "usual suspects" I should be aware of in connection with IR-communications?

I hope you experts can help me out. Thanks in advance.

/Malmkvist

Seems like a pretty noisy environment and noise getting into the IR signal is quite possible. Can you look at the received IR signal from the detector amp with an oscilloscope to see what it looks like?

115kb/s sounds like IRDA. I had trouble with a fluorescent lamp putting noise on the IRDA signal. The IR noise came from the ends of the lamp. I blocked about an inch on each end of the lamp and the problem went away.

The TFDU4100 is an IRDA part. Did you connect both resistors and caps as indicated on page 6 of the datasheet?
https://www.electro-tech-online.com/custompdfs/2009/01/tfd_4.pdf

Hello guys

For some reason I didn't get notified of your answers until today. I'll look at the schematics for the circuit tomorrow at work and answer your comments
(just so you didn't think I'd forgotten my thread)

crutschow: We did that, and signals coming from the IR-to-serial pin look normal, although often with the wrong code. When idle, its often seen that 0xFF and 0xFE are sent randomly. Although this is worrying, it doesn't have an immediate effect on the receiver, since none of those two signals have commands assigned to them.

duffy: Indeed it is IrDA - I forgot to mention this in my first thread.

blueroomelectronics: Yes almost so, except R2 is a 33R resistor. Do you think that might cause problems?

With a name like "baby face", I'de use a different tranceiver.

R2 is fine, but 165Ω is HUGE for R1. We use 7.5Ω on ours, you could go down to 14Ω on yours. With that other resistor, you are cutting down your transmit power considerably.