Choosing The Proper Digital Recorder for EVP Work
by Jim Brown for NPS
First, to be clear I am not recommending any particular make or model. To do so would require that I have actually tested it in my lab under controlled conditions. Since my name is not Donald Trump, I can’t afford to buy one of everything and conduct the level of testing needed to make such a determination of which is “Best”.
What I can do is post a list and explain the specifications your recorder must meet to conduct such research and minimize false positives. You can also use this list to determine the proper setting for your recorder; even the most expensive recorders can be downgraded by their settings to be rendered useless. No matter what recorder you use it is your obligation to use settings which meet or exceed these specifications.
So let’s begin.
- Your recorder shall record in stereo (Two or more independent tracks). This is done for two purposes. First since we have two ears it allows us to form a sense of direction for the source. That same audio can also be used with an oscilloscope and spectrum analyzer should you want to conduct more intense spatial analysis of the area. (I won’t go into all that here, suffice it to say spatial analysis requires two channels sourced from mics a known distance apart.) The second reason is redundancy. Two mics using two separate amplifiers are unlikely to pick up the same external interference should it occur. Thus stereo recordings provide their own reference against each other.
- Your recorder shall use a 24 bit wide A to D converter. Many voice recorders only use 16. The number of bits represent the number of possible audio levels available for conversion. This is the resolution of your audio and directly affects the amount of conversion errors. (All conversion has some.) An analogy to this is better understood in photography. All factors equal, a 15 megapixel camera can produce a much sharper picture than a 4 megepixel camera. Thus the possibility of canversion errors and pareidolia is reduced by the higher bit count.
- Your recorder shall sample at 96 KBPS or higher. The sample rate is the number of conversions made each second. I won’t go into Nyquist Points and how the conversion errors increase as the signal approaches the frequency of the signal. Suffice it to say that to obtain a good conversion with minimal aliasing the sample rate must be at least 3 times the highest frequency being converted. Since the “S” sounds in speech require up to 4 kHz, one might suspect that a sample rate of 12 kBPS would be sufficient. But that would be wrong.
The Nyquist Point applies to sinosodial waveforms. Speech is irregular and contains harmonics needed to provide proper inflections. If we were to look at how the various frequencies interact on a scope it can be shown that we must provide the second, third, and fourth harmonic to properly represent the signal. Thus our “S” sound is not 4 kHz, but 12 kHz! So the minimum sample rate has just been bumped up to 36 KBPS.
I added the extra sampling simply because we don’t know what EVP is. We can’t definitively say there is no EVP audio above the “S” sound. So extending it to 96 KBPS actually allows good converion with minimal error up to 6 Khz, just in case something else is present at the higher audio frequency.
- Your Recorder shall use a non-lossy conversion protocol. One of the problems facing digital recording is file size. As a result to keep size down a system called Code-Excited Linear Prediction (CELP) was developed. While I won’t get into the complex algorithms involved, in summary it looks at each sample then makes a prediction what the next SHOULD be. Thus some samples can be skipped to minimize memory usage, the algorithm fills in for the ones ommitted. But since we don’t know what represents an EVP, how can we determine what can be safely removed from one?
The solution is to use a protocol known as Pulse-Code Modulation. (PCM) This is a much simpler method; it simply samples the audio and saves each conversion. Thus there are no conversion errors introduced, but at the expense of requiring much more memory capacity and large files. Thus to satisy this requirement your recorder shall use PCM, not CELP for conversion.
- Your Recorder shall use a non-lossy file format. Related to conversion is file handling. Losses and errors are introduced if a file is compressed. Many recorders use the MP3 format to keep file size down. This is unacceptable. Rather the proper format is WAV. This simply saves the digital equivalent of the audio without compression. So your EVP recordings must be in uncompressed WAV format. (Note: Beware, some variations of WAV exist which do involve compression algorithms. These are also unacceptable.)
All of the above specifications can be obtained from manufacturer’s spec sheets for any machine you may be considering. There is a sixth requirement that is generally more difficult to obtain but is very important. Usually (Rule of thumb here) the more the recorder costs the more likely it is to meet this requirement.
- Your recorder shall incorporate adequate shileding against stray EM Fields. Failure here produces “EVP” caused by radio broadcasts or other outside interference. Tests to verify proper shielding can be done in a lab where a strong RF field is generated around the recorder while recording. The recorder should pick up no interference from the RF. You can also do a quick and dirty version of this test yourself if you live near a strong AM radio station. Simply visit the transmitter site and record silence. If the radio station can be heard the recorder fails the test. (Not definitive but it does rule out some of the more obvious failures before they get used to create more false EVP.)
Next: EVP Session methods.