This is the second part of the article on buying a Digital Recorder. It addresses the needs of those wanting such a recorder for EVP work. Any recorder recommended here would also work as a general purpose recorder for interviews and investigation logging however it would be a bit of overkill considering the costs of these better machines. If you are intered more in the basics, Click Here for the requirements of logging recorders.
To begin with what I am not going to do is recommend specific makes and models. To do so would require I continually test and evaluate new models as they come out and to delete models from my recommended list as they are discontinued. I do not run a consumer agency here and don't have the resources to do constant testing. Instead what I will do is provide specifications that have been found to be acceptable in tests I have conducted. Any recorder which meets or exceeds these specifications should be acceptable for quality EVP work. I will say though I have tested and used two recorders to arrive at these standards. These recorders were the Fostex FR2 (list price $1,199) and a Samson H2 (list price $200) Both of these did (with certain qualifications) meet minimum standards. This is NOT to say these are the only ones which do, nor am I making any recommendation you purchase either of these machines. They are simply the recorders I used to determine the specifications given here. These two were selected because (A) The Fostex had multiple settings which allowed me to test it both in its high quality settings and to downgrade it which allowed me to test responses at less than optimum standards, and (B) the H2 was chosen because at the time of testing it was the cheapest recorder which met the minimum standards when used in its highest quality settings.
To summarize the standards required, there are five factors which must be met for any digital recorder being used for EVP work. These are:
- A digital recorder shall record using an uncompressed digital format. (PCM conversion, not the more common CELP method.) For the audio files, WAV or BWF are acceptable. Compressed WAV files or MP3 may not be used.
The recorder shall record in stereo. In addition the recorder will allow external microphones which can be placed 30 inches apart to allow for spatial analysis of the recording made.
- The sample rate shall be 96 KBPS or greater. (44 KBPS may be used however slight degradation may be noted. Not recommended but may be acceptable)
- The recorder shall use a 24 bit A to D conversion process. This is to assure good definition at low audio levels.
- The recorder shall employ adequate internal shielding to prevent external RF interference from affecting it during the record operations. This ensures that CB radio, shortwave, and other broadcasts do not cause false positives.
Any recorder meeting these specifications should be acceptable for EVP recording. The paragraphs below will outline the reasoning and importance for each of these requirements.
Uncompressed Digital Files
Compression is a technique used to shrink the size of digital files. In the case of audio, various techniques are used, some providing more compression than others. The problem is ANY compression may be too much! All compression techniques work by removing "unnecessary" parts of the audio data. Often it is not possible to tell what has been removed by simply listening to it. The problem is when it comes to EVP we don't understand what they are or how they were created. So based on that how can we know what can and cannot be safely removed from them? The answer is, we can't! So the obvious solution is don't remove anything. That way, if some future researcher discovers some yet unknown factor, some ambiance buried in amongst the noise, we can take our recordings and submit them for analysis to search for this factor. It would be a shame if we had what would otherwise be a good EVP, and we couldn't verify it simply because we allowed some compression algorithm to remove that key from it because it was "unnecessary". So the solution is simply Don't Compress Audio files.
There are two ways your audio may be compressed. The first is in the digitizing process within the recorder. Cheap recorders use what is known as CELP, (Code Excited Linear Prediction.) Without going intop the details, it is simply a way to create smaller audio files and extending record times. It cannot be used for EVP work. Instead your recorder should employ the PCM (Pulse Code Modulation) method. This is a non-lossy form of encoding. Some recorders will do both; It is the user's responsibility to select PCM when doing EVP work.
The second way a file may be compressed is when saving it. Since many recorders allow files to be saved in multiple formats it becomes simply a matter of choosing a non-lossy form. The solution is to use WAV, not the compressed MP3 format. Note some recorders may give you an option to reduce the size of your WAV file. If yours does, avoid it! That feature simply adds compression to the original WAV format and defeats the purpose of maintaining an uncompressed file format.
Record In Stereo
This one is almost a no-brainer. We have two ears, we hear in stereo, so why shouldn't we record the way we hear? Stereo recording has two important points. First, it provides redundancy. How often when taking photographs we've heard, "Take duplicate shots." It rules out the possibility of film or camera anomalies, bugs, and other one of a kind problems. With stereo recording you essentially have two recorders running. If your anomaly is picked up on both channels that means it has gone through two separate recording systems. By doing so you use two amplifiers and two microphones or sensors, Even the convrsion process can be a dual operation on some machines. It is unlikely a fluke on one channel will show up on the other as well.
But there is a second advantage. If you use the two microphone set up you can hear depth to your audio. You can gain a sense of direction. You don't just hear the sound, you may be able to determine where in the room it originated. And later, if you have the necessary hardware you can perform spatial analysis. That is a method where you can view the sound on an oscilloscope and determine from measurements how much the sound in one channel leads or lags the other. Since the speed of sound is a constant, and the difference can be measured, it becomes a matter of using mathematics to determine at what angle the sound had to originate with respect to the microphones to meet the known values. In other words if you have an EVP of Uncle Charley, now you have a way of putting the origin in his favorite chair!
Sample at 96 KBPS or Greater
We come now to the recorder itself, the internal design of it. The sample rate is the number samples made to the analog audio signal every second. If this rate is insufficient, conversion errors rise as the recorder attempts to fill in the gaps between samples. Plus as the Nyquist Point is approached the chance for distortion and false positives also increases dramatically. To understand the importance of this factor, you need to understand how a digital recorder works and what speech is comprised of.
Speech is made up of two major components, vocalizations and fricatives. Vocalizations are the sounds created by the larynx, fricatives are those sounds created by the positioning of the tongue and lips to create modifications to the sounds. For the purpose of this discussion we need only be concerned with the higher frequency components. These are the fricatives associated with the "S" and "T" sounds. The "S" is the highest frequency generated by speech; the "T" has the fastest rise time (that explosive start to the sound as the tongue releases that puff of air.) These two sounds have two important characteristics we need to note. They are about 3,000 Hz and they are non-sinosodial waveforms. It is from these that we must set the high frequency specification and sample rate on our recorders.
The Nyquist Point is established as the minimum number of samples needed to simulate a sine wave. Most agree that if the sample rate is seven times the highest frequency being sampled that is sufficient. In other words seven times the frequency of the highest fricative, or 7 X 3,000 = 21,000 BPS (21 KBPS). However that only applies to a sine wave; the "S" and "T" are non-sinosodial waves. For these what is known as a square wave must be employed. A square wave is a sine wave plus an infinite number of its harmonics. Since this is not practical another term, called a pseudo-square wave is used. This is the sine wave plus its first three harmonics.
So since we are concerned with the Nyquist Point of a Pseudo-square wave we can set our minimum sample rate as follows:
3,000 - (Frequency of "S")
X 7 - (Nyquist Point)
21,000 BPS (Rate for Sine wave)
X 3 - (3rd Harmonic)
63,000 BPS (Actual Sample Rate Needed )
Thus it is evident the minimum sample rate for a good digital recorder is 63,000 BPS or 63 KiloBytes Per Second (KBPS). This fits with testing conducted here where no noticeable distortion is present at 96 KBPS, and minimal degradation detected at 44 KBPS. (Just for comparison, most digital Voice Recorders sample at 8 to 16 KBPS, which is why I don't recommend using them for EVP work!)
One more thing may come into play here. Most recorders have some means of setting the quality versus record time. Always make sure the setting you choose maintains the minimum 96KBPS sample rate. Many recorders offer recording time ranging up to several hours. But to attain this they downgrade the sample rates! Even the most expensive professional grade recorders may offer these features. And they also downgrade quality when doing so. For EVP work it is your responsibility to choose a setting which maintains the quality required without regard to record time. Stick with the highest quality settings.
24 Bit A to D (Analog to Digital) Conversion
All sound is analog in nature. We hear in analog, and it is essential that a digital recorder process analog as it records its digital equivalent values. The problem is by its nature analog can have an infinite number of values, while digital must have definitive numeric values assigned to each sample made. Thus some rounding up or down must occur as the conversion takes place. However each bit added to the count will double the number of possible combinations, thus giving improved resolution to the audio quality. More bits means the recorder will not need to round off to as great of a degree to assign a specific digital value to a particular sample.
Consider, if we have a 16 bit A to D converter, there are 64,000 possible numeric values which can be represented. This may be suitable for non critical applications, but where data is involved and an analysis to be done, it is not sufficient. By going to the 24 bit system the number of combinations reaches 16 million. This allows for a much finer definition, and details can be refined much more accurately. The analogy is much like a digital camera. If you have a 1 megapixel camera you can take a decent picture. But some detail may be lacking. Compare that to a 10 megapixel camera and the sharpness and clarity will be apparent.
It becomes even more important when you consider that most EVPs are very low in volume. You are going to have to amplify them, sometimes hundreds of times, in order to bring them up where you can hear them. Just like the digital picture begins to pixelate when you zoom in, quality begins to deteriorate as you amplify the audio. And just like the digital picture, if you start with more detail you can amplify more before deterioration becomes a real problem. So it is advantageous to use the best resolution you can when you make a digital recording.
Internal Shielding is one area where you may encounter some difficulty determining whether or not your recorder is sufficient. Most manufacturers don't provide any specifications in that area. Ideally it will employ a single point ground system with a ferro-magnetic shield around critical electronic areas. RF bypass should also be used on all inputs and outputs.
You can test your recorder though. All you need to do is take it to any nearby radio station transmitter site. An AM broadcast station works best, preferably one of the high power clear channel ones. Get as near as possible to the transmitting antenna where signal level is highest. Don't plug any microphones in, but put the recorder in the record mode, turn the input gain all the way up, and position it in various ways while recording. Try all positions, then stop the recording. Play it back and listen for any evidence the recorder picked up any voices or transmissions from the station. If no sounds are heard, your recorder passed the test. The shielding is working properly.
These five areas are the most important when choosing a recorder for EVP work. Likely you will find that it is more expensive than the cheap voice recorders. But with all the shortcomings of the voice recorder it is imperative that the serious researcher get a recorder capable of serious research. If we want the paranormal field to be taken seriously, we must do serious research using serious equipment. Otherwise we can be ghostbusters, running around with our little voice recorders getting a lot of false positives and finding spirits behind every tree.