First of all I would like to thank you all for such a great work.
I have been implementing the ESR meter, and checking Totoro's function I saw he is using the voltage supply as a parameter. Then I remembered about the hidden voltmeter to own supply voltage on 168 and 328 chips. I just added it on setup and it improves measurement quality. Actually, helps a lot regarding "well regulated quality supply" requirements ;)
Oh, regarding the coupling capacitor: as a bipolar capacitor I have 2 electros 330µF/35V in a back to back configuration. There is a lack of slope in the scope, which means that it is only measuring ESR. Actually, coupling cap ESR is not that important: 1Ohm ESR would only affect 1% over the 100 Ohm sense resistor. Capacitor value however is important, because a low capacitance value would bother the reading due to capacitor impedance. Regarding 4wire sensing: as the ground is shared, 4w cannot be used, but 3-wire sensing can and should be used because it is extremely easy to implement. Regarding Transistors: Transistor models are not important, they are operated as a switching device. However, they should be "fast" to switch. Currently I am using a 2SA733 and 2SC945 that I had in junkbox.
Oh, another thing, adding a decoupling capacitor between PNP transistor emitter and sense ground would be beneficial as well. Location on PCB should be as close as possible to those points.
Edit: Couple mods I think interesting: Playing with it, I just realized that if capacitors hold (or grow) any charge, readings get (easily) bothered. In the original schematic, there is a 10k capacitor to ground, I decreased it to 470 Ohm: well far to bother reading range and low enough to not grow bald while waiting to discharge 1000µF caps... The modification of the "commercial" schematic with BS170 does that, but I think there is no need for 50mA application. For LOW ESR Capacitors, original circuit may be a bit "short" in current terms. If increasing current to say 100mA, decoupling is probably needed for good measurement stability. According to my oscilloscope, measurement last 35µs, so I would use over 100µF...