I want to protect the PCB-mounted components from vibrations. What kind of glue/substance should I use? What are the most prone to failure components due to vibrations?
For long term survivability, anything that needs more mechanical strength than its solder connections provide should use 'proper' mechanical restraints such as brackets, mounting clamps etc.
However, it is common in many consumer products for larger mechanical items to be "held in place" or mechanically supported by an adhesive. Properly done this is a legitimate manufacturing method, even though sometimes frowned on by "proper" designers.
Asian solution: The most common method in Asian products is a white glue that sets to a hard ceramic white finish. It has good initial bonding capabilities and strength, is fast setting and fails miserably within a period of from months to a year or so.
Western solution: The most common western method is hot-glue or hotmelt glue. It has good initial bonding capabilities, reasonable strength, sets within the time taken to cool (tens of seconds to minutes depending on amount and thermal conduction) and fails miserably within a period of from months to a year or so. People still use it and say how good it is.
A superb solution: An excellent solution is neutral cure silicone rubber. It has good bonding capabilities once set - these vary with materials but primers or special variants are available for difficult materials. It has reasonable strength but low modulus (somewhat stretchy) so resists vibration and impact forces superbly. Setting time can be minutes to many hours - usually longer than shorter. Once set it will typically last for 20++ years. The long setting times are not an issue as long as the adhesive is not used to retain the objects in position but to add long term vibration resistance. Most silicone rubbers will not flow during setting so setting time is unimportant except that the surface will be tacky at first. "Skin over" occurs typically in 10 to 20 minutes. Setting in achieved by action of atmospheric moisture so low humidity causes longer setting times. Full setting rate depends on water thickness with rates of about 3mm depth per 24 hours being typical.
An excellent combination where mechanical retention or location using the adhesive is required is to use a "dab" of hotmelt glue to hold things in place initially and then use silicone rubber as the long term binding and antivibration agent.
As well as their excellent low modulus and long life, SR's make excellent waterproofing and sealing agents. However they are not wter vapor sealants - ater vapor will pass through SR with time - liquid water wont. So they "breathe" - which may be good or bad depending on application.
There are numerous other adhesives which my work for you. Cyanoacrylate glues set fast (super glue, elephant glue), risk marring surfaces in some applications and will debond with time on some surfaces. Latex and contact glues are low modulus and can be useful. Solvents tend to be used which may have adverse effects. Others exist, but I've found silicone rubber to be the best overall.
- Silicone Rubber versus Epoxy Resin:
I didn't mention epoxy resin (ER), and Mike has recommended it in preference to silicone rubber (SR). Epoxy resins are certainly highly useful. They are available with "pot lives (usable time before setting) of under 1 minute to hours, and time to full curing of an hour or so to a day or so. Elevated temperatures increase curing rate and reduce pot life. ER's are usually '2 pot" - two different components are mixed just prior to use. SR's are also available in 2 part formulations, usually with rapid setting times (minutes to 10's of minutes - can be longer by design).
A factor which I cited as an advantage, Mike has cited as a disadvantage. Both of us are correct (and both wrong? :-) ) and it is worth understanding why.
The factor is "modulus". Modulus (more properly "Elastic Modulus") is the ratio of stress to strain - the amount of deformation experienced per applied force. There are various measures of modulus, mainly relating to directional aspects of the forces and deformations. Some more details here in this Wikipedia article on elastic modulus..
Silicone Rubbers are "low modulus" - while the actual values can vary widely by design, those normally encountered will be "substantially "more elastic" than typical epoxy resins. Mike and I are agreed on this :-). How useful this is is a matter for discussion.
An epoxy resin will generally act in the same way as a clamp or bracket or mechanical mounting system. It rigidly attaches the component to a fixture or PCB etc. Typical silicone rubbers are by no means "jelly like" - they set to a stiff rubber - typically as flexible as a less than stiff block eraser - solid but not concrete like. The hardness of a rubber formulation is not intuitively related to modulus* but the two are related enough to be useful. An appropriate scale for hardness for rubbers is the "Shore A" hardness scale. Car tyres would be about Shore A 70 - a rather stiff rubber. A typical rubber band would be Shore A 25-30 - rather stretchy. Silicone Rubbers are available that set to either of these extremes, but most would be intermediate, with a tendency towards the stiff side in most cases.
Now consider, if you have ever encountered a vibration damper, earthquake protector, shockproof mounting, auto engine mounting, ... - what mechanical characteristics do they have? What sort of materials would you expect them to be made of?
M Alin asked about what sort of things most needed to be glued.
Essentially, objects that are liable to move under impact or vibration and to sustain damage if they do.
(* A formula that approximately relates Shore A to Young's modulus is given below. Unlikely to be useful in this context.)
Small SMDs probably don't need mechanical fixing, since they hardly exert a force on the soldering thanks to their low weight, often milligrams. (An 0402 thin film chip resistor weighs exactly 1mg.)
For bigger components, like large electrolytics, I wouldn't use a glue which remains elastic. This kind of glue still allows the part to vibrate, though with a lower amplitude. Even at these lower amplitudes over a long time they may cause solderings to break due to creep.
Back in my days in the aerospace industry, small components that were mounted 'substantially' flat against the PCB (ie through-hole axial resistors, diodes, DIP ICs etc) required no special treatment, likewise SMD components. Larger components such as radial capacitors and components mounted proud of the PCB (transistors etc) were fixed in place with epoxy resin adhesive. Silicone rubber (RTV) was only used to secure preset pot screws as it is pliable and doesn't afford much protection against vibrational stress.
PCBs were fixed to chassis using anti-vibration mounts (nitrile-rubber etc). Conformal coatings probably provided some slight additional protection too.
generally speaking most devices have enough surface tension on the solder to keep the components in place even at highish temperatures and large vibrations. Most at risk would be the heavy and large components. if you have any large SMD electrolytic capacitors i would make them thru-hole, along with any connectors.
The only use of glue in SMD work that i can think of is using it to keep components from falling off during oven re-flow.