component values in circuit shown here?

Exactly and it can be in excess of 200 volts. Mr. Idiot has a lovely suggestion for testing if you don't believe that but it's rather painful as I discovered 40 years ago! Usually across the coil, convention says 86 POS (+) side and 85 NEG (-). In fact as long as you know how you're wiring it it doesn't matter. Diode cathode band to POS (+).

OK... last question: It looks like the diode goes across (in parallel) with the coil. Can't the surge still bounce back through the coil itself if it's big enough? Or does it follow the path of least resistance and just die in the diode? Almost seems to me that you'd be better off setting it up in series before the coil to just make sure the surge can't come back through (then again, I am very light on anything but simple concepts here :-) .

Actually I do it in series as you suggested. In fact I sometimes go "belt and braces" and do BOTH. But then I buy my diodes by the hundred.

The spike suppression diodes are always across the coil (with the band end towards +ve; if the wrong way the diode smokes as it shorts +12V to GND). The series (input) diode in this circuit is to stop the capacitor discharging thru other loads off the input supply. (Hence it can only discharge thru the relay coil (and any parallel resistor or path) at a known rate or "time constant". Though that series diode stops a +ve spike injection back into the source power/circuit, I'm not sure about stopping any -ve or GND surge. As I recall, it doesn't and there can be problems, but I'd want to review that. (But not now, its required 5 seconds of thought is just to taxing ATM.) The main spike that the coil produces is when power is removed. The coil tries to keep the current flowing, hence it produces a big voltage spike - just like an ignition coil. That spike is -ve, ie, its GND end is (say) 100-200V more +ve than its +12V end. But once that spike exceeds the diode's forward voltage drop ie ~0.6V, the diode conducts and "shorts out" that 50V or 200V whatever spike. The spike is actually making the -ve (GND or 0V) more +ve than the +12V supply. It's that reverse-bias that can breakdown many electronic components that should only be supplying or seeing a +ve voltage to/from the system. They may only be 50V PIV rated components (transistors, MOSFETs, 1N4001s, etc) and not expecting to see a 100V or 200V -ve spike. And I think the last para contains the solution to whether a series diode is adequate. Keep in mind that whether the +12V goes -ve or the GND goes +ve by 50-200V etc doesn't matter, it's all the same relative to any component across those supplies (aka supply or power rails) - though the component's other connections may have an impact on whether they are damaged.

Very informative gents, thanks. I'm a little surprised that more automotive relays don't have these built right in as I can see it being a problem these days with multiple computers in a vehicle. I think I'll be using them in any of the work I do from now on :-)

You are very astute. There are relays with inbuilt diodes - hence the convention that 86 is +ve and 85 is negative. However I prefer to add my own because: - why pay them to do it? - why have to worry about a polarity convention for an otherwise non-polar device? - some relay circuits do no have a polarity (ie, their +ve & -ve across the coil may swap); - how do you replace an internal blown diode? (ie, just add your own external $0.05 1N4004). Most of the time the spike quenching diode is not needed. And IMO, spike sensitive stuff should itself be protected - ie, have their own diode(s). In fact many ICs have that - most inputs and outputs have 2 reverse biased diodes - one to its +ve rail, the other to its -ve rail, hence clipping voltages to ~0.3V to ~0.6V above and below the supply rails. Incidentally, it seems that inbuilt-diode relays are now less common. Many instead have a parallel resistor (usually ~10kΩ ?) which is supposed to absorb some of the spike. (But does it?) I still prefer raw coils. I'll add diodes if desired. I won't bother with resistors. And any circuit I build will have its own protection. Relay diodes are usually then only to reduce electrical noise from the relay whether radiated or likely to be "seen" by dash bulbs or voltmeters and other sensing circuits etc. PS (the next day) - the testing that Howie alluded to (courtesy of Mr (Not!) Idiot) I assume is to simply touch the (-ve?) coil terminal/wire as it is de-energised. It's something I managed to avoid for decades until this year after my steering-wheel horn ring failed - I merely used an alligator clip to touch the horn relay's ground wire. It worked fine a couple of times, but later I must have been touching the relay GND as it de-energised. I fitted a dash-mounted push button that night! Maybe I should have tried a suppression diode across the coil, but the button was the proper interim solution. But thanks to Howie and Mr Idiot, I might consider this "new" method of testing the integrity of spike suppression relays. However I strongly suspect upon finding my first failed diode that I'll build some LED tester instead! FYI - I recently read that the current rating of such "across coil reverse biased diodes" should be double that of the coil. IE - with most car relays being 250mA or less, that means a 500mA diode. Hence the 1 Amp rated 1N400x series is ideal. The 1N4003 with 200V PIV rating should handle most spikes, but the now "standard" 400V 1N4004 adds extra voltage capability. So too the highest PIV in the series - the 1,000V 1N4007 - not that it's likely to be an advantage, but both the -04 & -07 are the common standardised and hence cheapest versions. (The others seem to have been dropped from production.)