Voltage Drop Question

[Mark Hennessy]

All very interesting, the question came about when I was asked to explain how to work out what cross sectional area conductor would be needed when adding auxiliary stuff to a vehicle, nothing specific was specified but I used a 100 watt load as an example for a nominal 12 v system, that would be a current draw of 8.3 amps, a 1 mm cross sectional area conductor (standard PVC auto cable) would just about cope with that but the voltage drop over say 3 metres or so would be around 0.45 volts, the general consensus for voltage drop in vehicles for auxiliary stuff seems to be a maximum drop of 3% of the supply voltage if possible so although a 1 mm cross sectional area conductor would probably just about pass from a safety aspect it wouldn't pass from a voltage drop aspect so I mentioned a general rule of thumb that I have used in the past, choose a conductor that has a cross sectional area that's double that of a conductor that has a current rating the same as the load current or near to it as possible.

The question then came up about working out the voltage drop more precisely for a couple of different supply voltages and could it be done without a test meter, nothing load specific apart from the fact it would be a filament bulb, the parameters used in my earlier posts were chosen at will to illustrate the problem.

Lawrence.

Ah, OK.

So, for ways to be very precise about it, assuming you are allowed to characterise the light bulb first, my opening paragraph suggests how to go about it.

In practice, companies will either follow established "rules of thumb", or will do a bit of R&D and derive their guidelines for generations of subsequent engineers to follow. Experience will reinforce or modify these rules of thumb. I suppose that in some cost-sensitive industries - like mass-market car production - there will be pressures to save every fraction of a penny, and the temptation to go to a thinner cable might need some justification?

In engineering, it is possible to calculate almost anything with the right input data. But generally, it's much easier to use more "approximate" methods, like Nick and I both did.

Academics might want answers with 3 digits of precision, determined via complex maths, but generally this is about making the poor student suffer the maths (which may or may not have any long-term value!).

Conversely, real-world engineers with a developed "sense" of engineering will understand that there is no point aiming for such precision in many circumstances. A classic example is a transistor circuit, where a student might calculate that you need a 3.782 MOhm bias resistor because of the given value of Hfe - that student doesn't yet realise just how shockingly variable Hfe actually is. Let alone the tolerances of the resistors and power supply and transistor Vbe.

I do have a 6.5 digit bench multimeter at home, but it only gets used for very specialist jobs - most of the time, 3 digits is plenty. When you have uV resolution, interpreting the readings takes some skill - forget about the notion of a stable reading like you get with the 6,000 count Flukes et al! Noise, thermal potentials, thermal drift - it's all there in the reading, and really messes with your mind!

3.14 is accurate, but not precise.

3.15018912432 is precise, but not accurate.

I'd suggest that measuring a car headlamp bulb in order to produce a curve like you posted earlier would be pretty tricky in itself. What sort of accuracy would it yield, even with the best test gear? How many samples would you need to measure to determine the tolerances of the filaments? From how many manufacturers?

Still, an interesting thought experiment...

Enough of my ramblings - all the best

Mark