Here is the general approach you should take.
Any connectors and contact points to the outside world should have protection. The outer perimeter of the board should have a guard ring running around it with protection on it. This is because people tend to handle the board by the edges. This guard ring should be the ground for the rest of the system with a another ring concentric just inside.
All other internal nodes probably can't be protected, in fact bare boards are never considered to be ESD proof for all possible scenarios.
You can encapsulate the board to provide a little more protection, depending upon the encapsulation it can be quite good. 32 um of parylene would really help but is expensive (this is a mil standard).
But ultimately, the whole board should be inside of an enclosure to prevent contact and discharge.
Almost every charger and regulator out there nowadays has an absolute max of 6V. If you have a USB rated TVS and some capacitance on the 5V bus, you probably won't have problems there. You can use a high voltage cap if you want, and put it close to the barrel jack.
For buttons, use a series resistor (as large as you can accept without causing problems with your pullup or pulldown) and a shunt cap. Put the resistor near the button and put the cap near the IC where the signal goes. You probably don't need TVS on the buttons, but it is a good idea to put it in the layout (close to the IC) in case it is needed. A wide variety of 0402 sized ESD diodes are available. If you have a large series resistance between the ESD gun and the IC, the ESD diode will not have to dissipate much energy. The capacitor will help prevent steep dV/dt.
Most importantly, you need to test. You can't solve ESD analytically. You can give it your best shot, but then you need to test to verify.
The two main problems you get from ESD are:
- IC or transistor inputs damaged (fix with capacitance or ESD diodes
or TVS)
- poorly terminated inputs toggle due to pulse coupling and cause some undesired behavior such as reboot/reset/poweroff.
Rarely does ESD damage anything other than an IC input.
Also make sure you consider case 2. Any reset inputs or interrupt inputs need to be scrutinized. An internal 200k (or whatever) pullup may not be enough. You may want to add either a cap, or a stronger pullup/pulldown.
Finally, the best defense against ESD, when possible, is to keep the gun away from the components by recessing them within a plastic housing. That is mainly a mechanical design issue, though. If the gun tip cannot get near the components, there will be no discharge in the first place.
Best Answer
You would potentially save area, but I can think of a big reason why you wouldn't want to put a TVS with a spark gap underneath it. If you put solder anywhere near the spark gap you run the risk of shorting it out. That reason alone would be good enough for me to run them side by side in parallel instead of a TVS with a spark gap underneath. For me the saving of PCB area would not be worth the risk of shorting out the signal.
If your talking surge (like IEC61000-4-5) then you'd better size your TVS to handle that surge because it will get blown out if your not careful. Air breaks down at ~3kV/mm with a 0.2mm gap that's roughly 600V. The TVS is going to see the current before the spark gap below 600V unless the TVS breakdown is higher than 600V. If this is a low impedance surge current, the TVS has the possibility of blowing out.
I would do one or the other (TVS OR Spark Gap) depending on what you want to protect and protect against. Either that or I would put some kind of current limiter between the TVS and spark gap with the spark gap facing the input of the surge.
There are also other devices that are better at protection than a spark gap depending on the design