Frank@sri-unix (07/13/82)
So far, I've discovered that these chips do not like to have their recv buffer read while the xmit buffer is still draining. The fix to that one is simple (if your chip exhibits this problem of garbled input - not all do): just check both bits (03H) when checking input status. That's all well and good. Now for the strange part: when using a program like MODEM to send out a file, the last byte of the block, which normally should be the checksum byte, is the NAK received up front to kick off the process! Looks like the xmit drain problem in reverse, doesn't it? Again, not all of the chips have this problem - and the problem doesn't appear to be run specific either. Last I heard (rumor), Signetics has withdrawn the 2661. I have not heard about the 2651. The 2651 is used in the Thomas Engineering QLC-100, four-port card. The 2661 is used in the Action DPC-180 and DPC-100 cards. If anyone uses either of these chips and has run across the second problem, and discovered a solution, I would like to hear about it immediately! Thanks, Frank