Measurement errors – more on EVP and Ghost Hunting

So, Mickey Burrow recorded what he thinks is a ghostly voice. But before we examine more about "Electronic Voice Phenomena", or "EVP", I want to explain something about measurement errors, and why measurement errors are so important to ghost hunters.

Also, my apologies – I know I said I'd keep this short, but yet again I have something over 1000 words.


Let's start by making a measuring stick. We want to measure things in the real world, and we want some sort of accuracy, but maybe we're a little sloppy about it. So we'll use something handy. I happen to have a used pencil here, I have no idea how long it is – but I've been using it for a while and it still fits comfortably in my hand. Maybe it's half the length of a standard new pencil. We'll call this my "measuring stick".

As you might think, my measuring stick is handy for measuring the length of my desk. My desk is almost 17 measuring sticks in length, and only a little over 8 measuring sticks wide. This is accurate enough to figure out where to place my desk in my computer room. My computer room, incidentally, is about 28 by 33 measuring sticks in size.

Hey, this works pretty well! (I think the desk will fit riiiight over there.)

What about smaller things? Let's see, my laptop is a little over 2 and a half measuring sticks by 3 measuring sticks. Hm. That's kinda awkward, so I'll just round it to 3 by 3 measuring sticks. My coffee cup is 1 and 3/5ths tall, by 3/5's of a measuring stick wide. I'll just round that to two measuring sticks tall and one measuring stick wide. That's close enough, right?

Now I'll pull out a standard American Quarter, and measure it to be 1/5th of a measuring stick in diameter, and a little over a hundredth of a measuring stick in thickness. I'll round down the quarter to zero by zero measuring sticks in size.

Hey, according to my handy measuring stick, my quarter doesn't exist! How about that?

The problem that I've run into here is the problem with measurement errors. If I measure everything in terms of my measuring stick, and I can only use rounded values of my measuring stick, then there will always be a point where the measuring stick isn't very useful. The stick works just fine if I'm trying to figure out where to put my desk in my computer room. It works fairly well when I'm trying to figure out the proper placement of my laptop and coffee cup on top of that desk. But the measuring stick fails utterly at helping me figure out where I should place my favorite Quarter. According to my measuring stick, my quarter has a width and thickness of zero, which isn't very useful.


What does this mean to electronics?

I have a digital multimeter here. It's a “Fluke 77, series III” digital multimeter. According to it's specifications, it has a maximum resolution of 0.1mV, with an accuracy of 0.3%. I'm going to tell you now, that 0.3% accuracy is an optimistic specification. You can take my word for this, or you can read this PDF from Fluke.

Accuracy is is related to uncertainty and uncertainty is something that Scientists and Engineers can never eliminate. It's part of nature, and we just find ways to increase accuracy or live with inaccuracy. My little Fluke 77 multimeter is pretty accurate, but still it is slightly affected by temperature and by the basic nature of its input circuits. It is also affected by its limits of accuracy.

The accuracy of a standard voltmeter circuit isn't linear. This is a fancy way to say that the accuracy isn't the same across the entire range of measured voltages. Fluke gets around this by creating a sort of pseudo-linearity, by breaking up the volt meter range into 5 separate voltage measurement ranges. The most accurate of these is the 300 millivolt range, and we'll use that in the following example.

If you look at table 1-1 on page 1-6 of the PDF manual, you can see that in the 320mV range the accuracy is: (+/- 0.3% +1) That '+1' means that the meter is only able to resolve to one significant digit in this range. In other words, if I try to measure a voltage that is 310.94 millivolts the multimeter would think that it is somewhere between 309.9972 and 311.863 millivolts due to the +/-3% accuracy. But it would round it off to somewhere between 310 and 311.9 millivolts, because only one digit to the right of the decimal point is 'significant'.

Can you see that this is very much like my arbitrary measuring stick? Still, the measurement is still very useful at this point.

But remember what happened when I tried to use my measuring stick to measure a Quarter? We get the same sort of thing with the multimeter if we try to measure a signal that is somewhere between 0.01 and 0.05 millivolts.

The Fluke just can't do it. And what's worse, it won't round the measurement up – like we did with the coffee cup. Because of the nature of the circuit, we might see a reading of between 0.0 and 0.1 millivolts on the multimeter, but there would be no way to tell if it corresponded with the real world, or if it was just sort of “made up” internally due to the temperature and the inherent noise of the electronics. Even the age of the electronics could have an adverse and cumulative effect – which is why I have to get my multimeter calibrated on a regular basis to ensure it remains accurate!


But what does this mean?

At this point of an explanation, I usually notice that the eyes of my audience are glazing over. I'm sorry, I really tried to simplify it.

What this means is that my poor little multimeter is a poor tool for detecting voltages smaller than a certain point. At a small enough voltage range, the circuit becomes vulnerable to it's own internal instability (semiconductor noise) and it is vulnerable to changes in temperature, age of the electronics, and even to the electromagnetic interference of the environment. It does not “measure” these influences, it doesn't even correspond to these environmental changes in a one to one fashion. It merely becomes less accurate.

What this means is that any piece of digital electronics that samples the environment has a point beyond which it literally can NOT determine what is real, and what is not. Digital voice recorders are a great example. If you were to physically remove the microphone of a digital voice recorder, and then hit the “record” button, you would be recording the internal noise of the circuit, which may or may not correspond to the actual state of the circuit due to uncertainty. If you were to play that back, and amplify it, it is possible that your pattern-seeking brain could hear “voices” in the resulting random static.

This is in effect what ghost hunters do. They force their equipment to measure beyond the limits of what the equipment is capable of recording, and then claim what they hear is more than audio pareidolia, that it is in fact a ghost. Yet they can't prove it is anything more than random uncertainty, beyond the capability of their tools.

If you use a ghost hunter's 'measuring stick' you would believe that the world is jam-packed with Quarters. Maybe they're right, maybe they're wrong. There's really no way to tell. But ghost hunters build their careers on thinking that maybe this time they are really measuring something.