I wouldn't expect you to remanufacture your present circuit, ... but ... in the interest of maintaining a "single-point" ground node, I'd put the regulators close to pin 20 of the MCU, and situate the digital logic components close to that as well. The reason for this is that the single most important circuit feature with respect to controlling noise, is the single-point ground, with low-impedance paths to all connected pins. A low impedance path from the regulator to Vcc is not a bad idea either, but the ground node can make or break a system, noise-wise.

I my opinion, and it's not much more than that, the active portion of the circuit should have been arranged in a smaller area. I see long wires in several places. Capacitor leads between Vcc and GND should be kept short, i.e. 1-2 mm. That's particularly important for the small bypass cap's. On a perf-board such as this one, they should be immediately adjacent to the Vcc or GND pin of the IC they're bypassing. Keeping the circuit small keeps the Vcc and GND node small, which makes bypass more effective and mitigates the risk of contamination from external sources.

Just to make sure you're not fooling yourself, I'd recommend you move the 'scope-probe cable around so you can see which signals change radically as you move the cable, and which do not. You can have standing waves and other "artifacts" appearing on your 'scope display, which are not real. Probing Vcc in different locations in the circuit will tell you some things. If you have a two-channel 'scope, measuring power-to-ground noise by using the 'scope in differential mode will tell you some things, too. If you're fortunate enough to have a four-channel 'scope, it will be informative to see what portions of the power-ground noise appear on the output pins.

Your circuit is small and simple. Nevertheless, it is possible to create some peculiar forms of reflected noise, among others, despite the fact the primary switching noise generator is the MCU. Were I in your situation, I'd look at the power and GND leads between the MCU and the two other IC's. You may be surprised at the magnitude of the noise on those leads. Small, low-value cap's within 1 hole in the perfboard, will be close enough to help considerably.

It looks as though you've used AWG #22 wire to route power and GND. I'd recommend you consider, as much for your own education as for the benefit of the circuit, changing to #16 or larger. It's extreme, but it will show you some things. Its also easier to attach ground straps to #16 than to #22.

BTW, be sure to read the spec's for the ZIF socket, particularly pin-to-pin capacitance, and consider how they affect things.

good luck!

RE