The question then arises, what is the best way to wire EMGs to 18V?
-Should the user be able to toggle between 9V and 18V at will? (i.e. is there any benefit to this)
In the long term, I don't really think so — maybe temporarily as a good way to quickly (and thus effectively) compare 9V to 18V.
Paralleling the batteries won't magically double the current flow through the circuit. What it does do is double the the amount of current that can be supplied to the circuit for a given voltage drop occurring due to the batteries internal resistance (i.e. halve the voltage drop for a given amount of current draw), but this should be more or less negligible in either case (a few ohms multiplied by a few milliamps is in the tens of millivolts at most).
Series is the only way to significantly increase headroom. The idea is that the output of any (straightforward) active circuity is limited to be within the power supply rails minus some margin, with any larger signal being clipped. For example using a single 9V battery, the output voltage (Vo) will at best be bounded by
0 ≤ Vo ≤ 9. But, like I said there is some margin on that too, a typical op-amp might only provide output up/down to around 1V away from the power rails, so around
1 ≤ Vo ≤ 7, a maximum peak-to-peak voltage of 7V. Whereas, doubling the supply voltage to 18V gives roughly
1 ≤ Vo ≤ 17, or V
PP = 16V — a little over twice the maximum output as with the 9V supply. As such, voltage spikes that get clipped when running on 9V should either have significantly reduced clipping or avoid clipping at all when running on 18V.
Where this can make a significant difference is in some kinds of fuzz/distortion pedals, where vast swathes of the signal are intentionally clipped against the power rails (though, in discrete transistor designs, the voltage change likely also throws the biasing somewhat askew). With an EMG preamp, where the gain is much less, only the very loudest initial transients of a pluck/strum should come close to clipping, even at 9V.
antigua's tests of an EMG 85 in the previously linked thread showed maximum peak-to-peak voltage of around 4V and no sign of clipping.
As
sumgai has said the older battery will continue to discharge the newer until they reach some equilibrium point, however I can't see either battery being particularly happy undertaking that journey. The internal resistance is generally quite low (a few ohms), potentially allowing a lot of current (hundreds of mA) between the batteries and causing a lot of heat. But aside from that, you'd also effectively be attempting to recharge the older battery, which in the case of single-use / non-rechargeable batteries is a bad idea.
But, if the batteries start out in equilibrium, the two should theoretically discharge evenly avoiding the above problems — whilst giving twice the capacity, thus twice the runtime, of a single 9V.
Conversely, whilst a series connection sounds like it should also give extended runtime: i.e. an almost dead 9V battery giving only 5V will result much reduced headroom, two in series would give 10V, still more than a single fresh battery. In reality, at this point the voltage will already be quickly dropping of a cliff — that, in combination with the fact the circuit running at a higher voltage will draw more current, probably makes the 18V runtime equal (or even less than) that of a single 9V.
The possible improvement in sound with 18v may be an issue with 'slew rate' which measures how well the amp can follow a fast change in input voltage such as the transients at the start of a note. A higher supply voltage will raise the slew rate of the same amp.
Hmmm, interesting.
The commonly held belief is that EMG used to use the now obsolete LM4250 (see, for example,
Electrosmash's analysis of an EMG 81). With a brief glance at the middle-left graph on page 6 of
the LM4250 datasheet, it appears that the supply voltage affects the slew rate only a little: a factor of less than one and a half, for an order of magnitude difference in V
CC, so we might as well treat that as constant for a mere doubling of the voltage. However, it's important to not miss that's a plot against I
SET which does have what is basically a linear relationship to the supply voltage (i.e. if we ignore the variability of the drop from Q9 in the op-amp, see the schematic on page 9 of the datasheet).
Using
this SPICE model I get slew rates of about 170mV/μs at 9V and 360mV/μs at 18V which seem reasonable based upon the datasheet graphs. Adding a little context to that, we can use the relation
f ≤ Slew Rate / (π VPP) to determine the maximum frequency of a sinusoid at a given voltage level that will be immune to slew rate limiting. That is: for a sinusoid of 7V
PP (again, about the maximum we might reasonably assume we could get from a 9V supply), we get
f ≤ 7.7kHz with a 9V supply, but
f ≤ 16kHz with an 18V supply. However, the initial transients may of course have a steeper gradient than a sinusoid.
If slew rate really was a huge issue, a better solution would've been a different op-amp. For example a TL061 has a typical slew rate of 3500mV/μs, though that does come at a cost of an increased current consumption of around 175—200μA for 9—18V. By comparison, with the given 1Meg R
SET resistor, the LM4250 draws around 80μA at 9V and 180μA at 18V — though it's also worth mentioning the up-to-date specs for an EMG 81 (presumably now with a non-obsolete op-amp) list it as having 400μA of current draw at 9V, and so all of this may no longer be relevant.