Yes, its a tough call either way. FWIW, when I worked at Tektronix some 25 years ago they had an internal design interconnect handbook that recommended against IC sockets if at all possible. Their data showed the ICs were more reliable than the sockets. And if there were infantile reliability or spec problems, it was cheaper and better overall to prescreen/burn-in the ICs rather than sort them in production via sockets.
An exception to the no-socket recommendation is where firmware PROMs were likely to be upgraded. And in a rare instance, I've heard of memory tester boards that wouldn't work properly if the sockets weren't present (!). But outside of special cases like these, the ease of servicing advantage is outweighed by the socket reliability. Even a premium socket inserts an unnecessary interconnection. Save them for breadboarding. I haven't read anything along these lines since, but of course the industry has moved to SMT where soldering in the ICs is by far the norm.
Don
From: d.seiter@...
Reply-To: hp_agilent_equipment@...
To: hp_agilent_equipment@...
Subject: Re: [hp_agilent_equipment] HP 5342A processor board
Date: Mon, 18 Dec 2006 19:36:05 +0000
I would replace a supect socket with a good quality one rather than solder the IC in. If the IC is bad, it just means more work and possible damage to remove it again.
