Given the sudden change in direction this investigation has taken, I thought it best to summarize exactly where we are. Today we will examine what type of signal we are looking for, the equipment needed to produce the signal, the equipment needed to hunt for the signal and how to narrow down the possibilities.
Unlike previous articles we are going to get very specific on the requirements and provide people, with the appropriate equipment, a chance to not only locate the signal, but defend against it.
The hunt begins in earnest.
With the new understanding of plasma around the axon opening voltage-gated channels to trigger action potentials, we can now speculate on the type of hardware required to track, interface and control neural firing patterns.
It has become obvious that we are seeking a high frequency phased-array with electronic steering. This could imply both ground-based and satellite-based systems. For those unfamiliar with phased arrays you can learn more here:
To achieve the discrimination required to trigger individual neurons means that the plasma around axons respond mainly to particular frequencies. No doubt this relates to the plasma frequency, but I have yet to discover the exact mechanism. Although, my gut reaction is that the system exploits resonant frequencies and may even have a method of determining that frequency. My cursory examination of plasma antenna theory demonstrates that this information can be revealed remotely by a signal.
For amateur hams, or even engineers, attempting to locate this signal I would focus on channels that appear to be occupied by radars or complex data streams. Standard radio equipment will not be of much use here, the resolution is too low to discern the structure of the signal.
Also, given that we are looking for a steerable phased array, it means that the wavefront where the most energy is concentrated may be out of phase depending on your location. This means that the signal strength will be distributed in time, so some dynamic reconstruction may be required to analyze the signal if you are not located at the focal point. There is also the possibility that more than one array is being used at the same time on a single target. This may prevent hotspots from developing.
So, let's give you the technical specifications of a radio system that will provide the required resolution.
High Resolution Software Defined Radio
For those who have experience of software defined radio, you will be aware that most high-end systems can manage around 1.6Mhz bandwidth across the range 10KHz - 32MHz with around 80MSps on a 16 bit ADC. This is a fantastic specification on a general purpose receiver, but you would be unable to analyze the signal to any great detail.
The reason is the way FFT works. The greater resolution you have in the time domain, the less you have in the frequency domain. We can look at some of the equations here:
bins = sample input length / 2
bin size = (samplerate / 2) / bins (or Sampling rate / sample input length)
time taken = (1 / samplerate) * sample input length
If we provide some practical examples based upon the specifications of the high-end radio given above we can see the limitations:
binsize = 3200000 / 524288 = 6.1Hz
time taken = (1 / 3200000) * 524288 = 16.4ms
If a signal is modulated faster than 6.1 times per second, it would appear as a continuous line on the spectrum. Also, if two more signals are started within a 16.4ms time frame, they would appear to start at the same time. All signals that lie within 6.1Hz of each other would be seen as a single signal.
If we attempt to increase our frequency resolution, our time resolution becomes worse:
binsize = 3200000 / 2000000 = 1.6Hz
time taken = (1 / 3200000) * 2000000 = 0.625ms
From this you think there is a direct relationship between the time taken and bandwidth of an FFT bin. We can demonstrate this is not the case by examining a lower sample rate.
binsize = 96000 / 524288 = 0.183Hz
time taken = (1 / 96000) * 524288 = 5.461s
To see the modulation in the type of signals we have been discussing, we need to strike a balance between frequency and time resolution. This can be very difficult.
Firstly, we need to define what exactly we are looking for. This will set the specifications of the receiver required to analyze the structure of the signal. As I mentioned above, we are looking for a steerable phased array. The signal itself will be a densely packed structure of high frequency radio waves modulated in the sub-3KHz band. This should cover most, if not all, of the potential operating frequencies that the plasma around the axon can be driven by.
Thus, we should see pulses with varying timings layered together in a narrow frequency band. In short, low frequency pulse width modulation (PWM) on a high frequency carrier. No doubt there are similarities to GSM and the application of Gaussian filters to initiate action potentials. For that reason I am not going to rule out a network of fake cell towers as a potential source. That said, given that their transmissions would not conform to GSM standards, or any known standard, they would stick out to a professional with the right gear.
Further, I do not expect the signals to be confined to a single band of contiguous frequencies. That said, I fully expect the signal to be located in bands allocated to the US and UK military. Given that this has most certainly been figured out by Russia and China, cross-referencing the overlaps in allocation of bands to the military may reduce the search further. Initially though, I would focus on the US allocations particularly between 35Mhz and 1Ghz.
From these requirements we can see that we need a time resolution of around 0.3ms but we are guessing at the bin size. In this case, we would also need a frequency resolution as fine as possible, microhertz or better.
That presents a problem. The only way to get a frequency resolution below 1Hz is for the sample input length to exceed the sample rate. We can see this in the last equation we performed. In doing so, we are guaranteed that our time resolution will be in the order of seconds, not the fraction of milliseconds that we require. Zero padding and overlaps are of no help here.
Thus it should be clear that FFT is the wrong tool for the job and, as such, commercial SDRs and spectrograms linked to radios are pointless for signal analysis. So, if you ever see a "SIGINT" system with an FFT, have a quiet giggle.
It should be obvious now why amateur hams have never reported such a system. They can't see it on SDR and on a standard radio it would sound like a complex digital mode transmission.
So, how do we detect it?
FFT provides what frequency components exist in a signal over a defined period. We require a transform that provides fine grained resolution of both time and frequency. We have options such as the short time Fourier transform, FWT, Wigner distributions, etc.
STFT has resolution issues, as does FWT. So, we can ignore these methods as they will not reveal the signal structure. Given the similarities with GSM, the best approach would be to modify the method used to sniff GSM packets in the air. I found a very good article on this by David A. Hall from National Instruments:
Unlike GSM I do not expect a defined carrier, at least not one that represents the axons of an entire person. It is quite possible that axons close in frequency terms will create a "carrier like" signal on a vector signal analyzer. This should, in theory, betray the location of the signals. It may be much harder to track down signals that are narrower than the resolution of the vector signal analyzer. At this point, it may be useful to examine the basics of vector signal analysis. I found another good resource from Agilent on this:
As we can see from this document, all the real work is performed in the digital domain. With that information, we now know that we can turn any decent SDR into a vector spectrum analyzer just by modifying the software. That said, it can be computationally quite heavy and parallel processing expands capability. Again, there is a resolution issue in narrowband signals such as 100 μHz but it is typically faster than other methods. So, brief signals spread out could potentially hide, however, for complex integration with the brain frequencies would need to be sustained revealing carriers.
The next stage is to pass this information to a Gabor Spectrogram to reveal a break down of the low frequency pulse modulation. How effective this will be in terms of frequency separation is unknown, but the time resolution should be adequate enough to determine pulse rates, assuming the frequencies do not overlap in the image causing longer or continuous lines. You may need to apply some bandpass filters to isolate the signal.
If you are having difficulties with this part, or you suspect multiple layers, then try the Time Domain IQ visualization as described by the National Instruments article. I don't know how revealing this will be, but it is one of the last steps towards eventual decoding of the signal. I use the term decoding, rather than demodulation, as the signal just triggers neurons and contains no embedded data. That said, it may contain elements similar to any phase-shift keying as the signal would need to track through neuron clusters to provide fine control. Thus, the signal would be a form of continuous-phased frequency shift keying.
As a side note, this brings up the role of Collins Radio Company (now Rockwell Collins) in early tests. Although, their involvement is not strictly necessary as the patent has been available since 1958. That said, they would have held the patent until 1975 and field tests were well underway by then.
Whilst people play with this idea and make recordings of suspect signals, I will look into better methods and appropriate software solutions that can be used.
It must be stressed at this point that what we will be receiving is neural patterns from Mr Computer. I don't expect that anyone will receive neural signals from people. These neural signals will be a mixture of conversations, verbal commands, images, sounds, ideas, conclusions, emotional stimuli and motor controls. Whilst we may not be able to decode them immediately, that will come in time as we learn the meaning of neural codes. One good indicator that you have stumbled upon the correct signals, is the patterns will be relatively similar in terms of structure and given the nature of a phased array will have different signal strength and reception timings. Depending on the footprint of the wavefront, we may find these signals spread across a band of frequencies, or the same frequencies reused but within a defined spacial constraint. That is, the same frequencies are used for everyone but the signal only develops the correct strength around the target.
One good thing to come out of this is the fact that we know the signal can be blocked by a Faraday's Cage. At this stage, we don't know if a second signal reveals the neural patterns, or if it is direct detection of emissions. A phased array has a limited power output and a defined signal strength must be maintained at the target to induce neurons into firing. Whilst it may be feasible to increase the power to overcome the attenuation by a single Faraday cage, if a significant proportion on the system do this, it becomes unfeasible to maintain the power requirements. I will come back to this issue of shielding later and attempt to develop a low-cost solution that should suit most people's needs.