Post History

At first glance, your design looks reasonable.  I assume D1 and D2 are really supposed to be on opposite data lines, and them both being on the A line is a typo?

One thing that sticks out to me is the value of the clamping diodes.  Why so high?  What's your expected common mode range?  A large common mode range may be OK for receiving, but will cause trouble when sending anyway.  RS-485 signals are supposed to be from 0 to 5 volts.  Why not clamp with 6 V or so Zeners?

I just looked at the MAX485 datasheet, and it shows the absolute maximum <i>Driver input voltage</i> as only 500 mV beyond power at each end.  This means it's not meant to handle a large common mode range anyway.  I really don't get the 33 V clamps now.  5 V Zeners would make a lot more sense.

I vaguely remember a long time ago having issues with Maxim RS-232 parts.  We ended up going with the TI version, which was pin-compatible.  I remember the TI datasheet specifically pointing out the part's protection.  Of course avoiding Maxim back then was necessary anyways for small volume purchasers.  They were notorious for unavailable parts months on end unless you wanted to buy a few million.

I can't imagine why anyone would spec the unprotected part.  What advantage does it provide?  Pin protection circuitry isn't just for operation, although that's important for signals that go off-board like your RS-485.  Pin protection also helps during production and general part handling.  Proper ESD precautions are <i>supposed to</i> be followed, but stuff happens.  One possibility is partial damage during production followed later by actual failure.

Your theory about ringing seems at least plausible.  The ferrite chokes L2 and L3 are rated for 512 &Omega; at 100 MHz.  That comes out to 815 nH, which is quite plausible for a small ferrite bead or chip inductor.  815 nH together with 10 nF has a resonance frequency of 1.8 MHz.  That's well within the range of harmonics that could easily be on the line.  That together with only 500 mV over and under voltage allowed on each line does sound like a bad match.

Put another way, this design is relying on the part clamping the signals to safe levels, but the chosen part specifically <i>doesn't</i> do the clamping.  That makes no sense.