Can you elaborate on the type? I thought schottky diodes were a type of junction diode.
Yes... there's always a junction somewhere.... but more importantly, it's a Si-Si, P-N rectifying junction (with metallic junctions elsewhere to eventually connect to normal wires). Whereas schottky are usually made with PtSi, or something weird like that, that makes a rectifying (rather than metallic) Schottky junction.
When I look for components, the primary categories seem to be general, rectifier, PIN, schottky, TVS, varactor, and zener. Is the type you're talking about one of those?
Schottky and varactor are the special cases in this list: schottky are chemically and electrically different. Varactors are junction diodes, but with a characteristic doping profile to achieve the characteristic (and specified) capacitance, and probably special layouts to improve Q factor.
The rest are all PN junction diodes, designed for, and controlled for, various properties.
Zeners are optimized for reverse (true (low voltage) Zener, or (high voltage) avalanche) breakdown under continuous operation; TVSs are optimized for peak current handling. (Both may work well in the opposite role, you just don't have the manufacturer's assurance that they'll work reliably under those conditions.)
General purpose rectifiers range from simple diffused PN diodes, to good old PIN diodes at high voltage (the 'I' region being necessary to achieve the high breakdown voltage). They can indeed be used as PIN diodes for RF purposes, but probably won't be optimal (or specified) in series resistance or recovery time. High speed rectifiers are optimized through doping profiles, patterning and carrier lifetime controls (dopants like gold or platinum, or electron irradiation); forward voltage drop and reverse breakdown voltage trade off against recovery speed.
All rectifiers are avalanche capable, but are rarely specified for operation that way; in the old days, junctions were made with poor process control and purity, and often suffered from fatal breakdown (hence, textbooks up until the 70s or so always recommended using a resistor and capacitor in parallel with each diode, when connecting strings in series). You might even get away with, say, using a 1N4007 as an ~1100V zener diode, but I wouldn't count on it for a real design.
So, from that space of diodes, you can pick whatever's best for the problem at hand. A too-big diode exhibits way more leakage and capacitance than a small-signal circuit should have to deal with. Always use components best proportioned for the task: just because you can use a humongous power transistor for controlling milliamperes, doesn't mean you should -- and there are probably very good reasons against it that you haven't considered.
Can you clarify this point?
For this circuit, the load on the 5V test source is minimal - within the mA range. Having a forward current capability of 3A is more than an order of magnitude of headroom. I took the critical parameter to be the breakdown voltage. I needed to protect against 300V, so I chose a diode rated to 600V.
Why is this not suitable?
With a 300k source resistance, you're counting on >>30Mohm equivalent to get anywhere near even 7 bits of ADC reading, let alone 10+ bits typical of microcontrollers' internals, or even better external units. Or from 5V, leakage well under 100nA.
The MURS360 is 3.0uA max at room temperature, so you're already screwed on that -- by a factor of >30. The typical case might be better, but considering the typ. and max. at 150C aren't very far apart, the typ. at room temperature probably isn't much better (1.5uA?).
Now, that is at 600V. And you could extrapolate the graph (very unusual to see a graph of maximum allowed values -- very good of On Semi to provide that!) down to, say, 3nA at a couple volts, which is now a factor of <30 better than needed, not bad. But if it heats up at all, you're screwed again. So it's still very dubious, and if you want more bits of accuracy, you don't have a choice.
What's the maximum current it needs to handle, anyhow? 300V / 300k = 1mA. So it's 3.5 orders of magnitude oversized, not counting ESD.
A typical solution would be a BAT54S, but this does 2uA leakage at room temperature already. Like I said, schottky isn't the way to go here. BAV99 is rated for 25nA max at 20V, 25C -- now we're talking. And even at elevated temperature, the typical isn't above 100nA until 70C or so.
By the way, leakage currents do cancel, to some extent, but this depends on matching between the two diodes, which is not guaranteed. Very likely, one is down by half of the other, so you still end up with +/- half the total leakage. It also varies nonlinearly with voltage, which will cause a much more insidious error to your ADC measurement than a simple straight-line constant current error would!
An even smaller diode would be beneficial, so you could get guaranteed leakage down in the single digit nA. But now you start to run afoul of possible ESD problems, depending on how and where your circuit actually connects -- if the high-voltage end of the 300k is available to the outside world, then as much as a 10kV spark (assuming the resistor doesn't break down in the process -- an axial resistor might actually be okay, but an SMT, doubtful) could deliver no more than 33mA, nary a scratch. But if the ADC end is exposed to the world, you better watch out, because, come winter time, even a casual perusal of fingers on PCB can develop thousands of volts, and the sparks, tens of amperes peak. Just by operating current, you might not even need external diodes at all, you could probably get away with the diodes inside the ADC chip -- but that's not good design practice, and definitely not good practice when it comes to ESD.
Tim