Awesome, thanks for the replys guys. I find all this very interesting and I've definitely learnt something here.
Just for the hell of it I added the 470 ohm resistor and it definitely turns off much quicker.
Hmm, though not too quickly. Ah, let me see, that's running around 100mA load isn't it, so it should have 5-10mA base drive to saturate reliably; but this is only giving it about half so it's kinda marginal. Halving the resistors should get it about as fast as it can go, which still looks like it should be faster (under 100ns should be perfectly reasonable for this transistor) but maybe there's a lot of emitter path inductance or something also limiting it.
Or maybe I'm not appreciating how much capacitance the zeners have, that could be the limiting factor. But it doesn't look like it; for that to be true, base voltage would be a lot lower by the time the collector is really swinging, and it's not.
Or, assuming the transistor really is what you think it is; maybe it's an eBay special...
Alternately, putting a 47-100pF in parallel with the 1k, so that for a brief instant, the MCU pin drives the base HARD. Also not recommended with sloppy layout, but handy for faster drive or pulsed circuits, while using less power.
I also noticed the collector voltage rised much quicker as a result. Am I correct in saying that the peak voltage reached if I didn't have the zeners fitted would also be much higher?
With automotive ignition systems, my understanding is that one of the advantages of modern semiconductor switched ignition coils over old mechanical breaker points, is they can switch off the coil primary much faster which results in more energy being induced and a more powerful spark from the secondary. Is this correct and is this essentially what I'm seeing here?
Maybe, maybe not. I mean, you've indeed measured that it is the case, that little ringing on top of the zener clamp is higher now -- though only very slightly so. (That BTW is the collector, and probably probe, capacitance resonating with the zeners and return loop inductance. It again seems like an awful lot, which makes me wonder what the transistor really is.)
Thing is, ignition coils have quite a lot of capacitance, and the core inductance may not all be available in an instant, either. The latter is very relevant to solenoids, relays and other things of electromechanical nature that they don't care about operating terribly quickly.
The effect of capacitance, is simply that it takes time for the voltage to swing up. If you put say a 1nF cap from C to GND, you'll see it rise much slower (the ringing will be slower as well). An ignition coil certainly doesn't have that much on its primary (well, one from the olden days of "points" might have a fractional uF cap, actually!), but the secondary might have a few tens or hundreds of pF, and with a winding ratio of some hundreds or thousands, the impedance ratio is 10k-1M (the square of the voltage ratio) and so the equivalent measured at the primary can be quite substantial!
Speed is limited by whatever dominates, so with slow drive, the transistor may limit, but above some point there's rapidly diminishing returns.
Which, here, you're pretty well in that case -- most of the inductor's energy is going into the zeners (hence the relatively wide flat-topped pulse, or, well, it'd be flat if you could measure across the zeners themselves, without interference from the ringing in the ground loop). If the waveform took like 10us to rise, you can imagine a lot of that pulse would end up gobbled by the transistor as it's taking its sweet time. (To illustrate this, try increasing Miller effect: put say 100R in series with 100pF, and wire that from base to collector. Play around with values and see; mind the resistance shouldn't be too low as it'll probably oscillate without any.)
As it happens, automotive coils are often driven by IGBTs, and not just any kind, but types that are customized for low Vce(sat) at low Vge(on) (since "cold cranking" might only have 5-6V available for the ECU), and slow switching (including internal series gate resistor) to reduce EMI emissions. The HV risetime might be 10s of microseconds, even 100s (compared with a pulse width of ~100s us during the spark) so it doesn't need to be anything fancy.
As for the core, the effect is that, when made of solid metal (pole piece, armature, etc.), it takes time for the magnetic field to "soak" into the bulk. This happens because current is induced in the metal, opposing the magnetic field. The current in any given layer, decays over time (due to the metal's resistance; it does not decay for superconductors (Meissner effect)), and so the field is able to "soak" in. When the metal is divided up (laminated strips or powder), the magnetic field can "bathe" everything simultaneously, raising the frequency response.
"Soak" is a fairly appropriate term, as the flow is a diffusion process -- this manifests in the circuit as a characteristic having equal parts resistance and inductance (45 degree phase), proportional to sqrt(F) over some range.
Here's an impedance plot of a nanocrystalline core (plotted over the original datasheet background for tweaking purposes), notice it's inductive at low frequencies (Z ~ F), diffusive at mid frequencies (half slope), then capacitive and finally complex at high frequencies (Z ~ 1/F, with peaks and valleys).
Nanocrystalline cores are quite fancy materials (they're partly metallic glass, what the hell, right?), but they're still just metal, a very thin ribbon wound into a tight spiral. The thickness of that ribbon sets the cutoff frequency, above which diffusion takes over (apparently ~10kHz).
It's not a huge asymptote (~20-400k), and is easier to read on other materials or components, at least when they provide such a plot. But this example was handy. (Diffusion is also relevant to chemical systems, like batteries and ionic capacitors (supercaps). Which is why a full charge can take so damn long on a battery, the charge voltage is a step and the charge current is the step response -- with a long "drooling" tail due to diffusion.)
Here's the equivalent circuit of that impedance:
The impedance-step* response from such a load, is limited by its own impedance. Essentially, the magnetic field trapped inside the material, takes time to escape, in the process depositing some of its energy in internal materials. The ideal case of a pure inductance, gives an infinite flyback pulse (assuming turn-off is also instant), which is nonphysical. If we add node capacitances, core losses and so on, we find the waveform is nicely finite and continuous, and we have a useful baseline to compare switch speed to, for example.
The impedance-step response of this particular component, will still be pretty sharp (~microseconds), but if you simply imagine the curve displaced to the left a few decades, you've got iron-core power transformers (laminated iron, versus the thin (~0.001"?) strip of this particular core), or even further for solid metal.
*There's not really a name for this, so I'm making one up. It would be a plain old "step response" if it were a linear system (varying the input voltage, holding impedances constant), but the thing is we're varying the switch impedance, and that's what makes these circuits so interesting at times. A piecewise-continuous model (t < 0, switch on; t > 0, switch off; etc.) is easy enough to start with, though reality is never piecewise so we may use that as just the starting point, and expand on it as needed to develop the results we expect (like the risetime measured here, which to model, will require adding capacitors, or a reasonable model of the switch, etc.).
Tim