SDR Spirit Box

[Incognitus]

Over a year ago I started working on a home made spirit box with a Raspberry Pi and a RTLSDR dongle. This thread will be to track how testing is going, perhaps I'll post some audio samples, and eventually a link to Github will be provided to download the actual software.

I had a few Pi's laying around and had done some simple stuff, like attaching a camera and sensors for temp, humidity, motion, etc. Lots of fun. I set one up with a SDR (software defined radio) dongle and an upconverter and listened to HAM radio for awhile. Lots of whack jobs on 80 meter.

I'm also interested in all these "ghost hunting" gadgets you see on shows, mostly from a technical/nerd standpoint. How were their toys built, and in what way are they potentially "backdoored" to fake results. EMF detectors are easy to set off at will, for example. It's incredibly simple and cheap to make a short range FM transmitter (eg,

Please, Log in or Register to view URLs content!

) to fake voices on a spirit box. Someday I might get something "traditional" like a SB7, but honestly I'm no electrical engineering and I have no way of knowing if the SB7 or similar devices contain any chips or circuitry that encourage (or insert) fake results.

Before anyone sues me, I'm not saying the SB7 IS backdoored in some way to produce results. I'm saying I personally would have no way of knowing. Building my own with a Raspberry Pi and a software defined radio dongle means I know how each piece of it works and know there's no fakery involved.

Someone else has recently debuted a SDR based spirit box. I'm not sure if it's supposed to be public or not, so I won't mention anything identifiable. Theirs is completely different, and I've been working on mine for over a year. Theirs works via a plugin for SDR# which hops frequencies with no squelch, and they have it hooked up to a satellite dish, so they get SO many channels. TYhere's some pros and cons to their method. On the pro side, it's very visual because when they're live-streaming they can have the spectrum waterfall on camera. On the con side, they need a fairly powerful laptop to run SDR# in that way and the dish is so good that they get maybe too many channels, so it can be a steady stream of non-stop talking. They need resources for the GUI, for Windows itself. Due to the size of the dish and the need for a laptop, you can't just carry it around with you.

Mine, which I started long before I had any knowledge other people were working on something similar, has completely different goals. I wanted it to be portable so you could easily walk around with it. That meant a reasonable size antenna, ability to use a small battery for power and ease of use (no booting into Windows, running SDR# and running the plugin, no having to SSH in to start anything).

So I've come up with a fairly simple system:

• Raspberry Pi (currently 3b, but could be a 4) -
  Please, Log in or Register to view URLs content!
• Battery - At least 10,000mAh, with the correct plug (pi 4 is USB-C, the 3 isn't) -
  Please, Log in or Register to view URLs content!
• Antenna -
  Please, Log in or Register to view URLs content!
• SDR - I used a NooELEC Nano -
  Please, Log in or Register to view URLs content!
• Cable to go from SDR to antenna -
  Please, Log in or Register to view URLs content!
• Bluetooth wireless noise canceling ear buds - I'll be using over the ear noise canceling headphones eventually for comfort. Can also pair to a speaker, but I prefer the Estes method which I'll describe below. I have something like these (but not these)
  Please, Log in or Register to view URLs content!

Install is fairly easy. The Pi 3b I'm using right now has the last version of Raspbian on it still. I tested on a Pi 4 (8gb) running the 64bit version of Raspios and it worked just as well. Maybe better, actually, since it's a 64bit OS. At any rate, either will work and I'll likely be switching the 3b over to Raspios tonight. Using the minimal install, with no GUI to save on space and RAM.

I'll go over some of the software and configuration next.

 

[Incognitus]

The first ghost box was created by Frank Sumption, who died in 2014, and was really to talk to aliens.

I'm going to document the setup using Raspios since that'll be the OS I use going forward. You can use 32bit or 64bit. You should use a sd card at least 32GB.

To start, grab the Raspberry Pi Imager from

Please, Log in or Register to view URLs content!

. If you want to use 32bit, you can install with the imager as is (it includes the 32bit software).

If you want to use 64bit, download the image from

Please, Log in or Register to view URLs content!

and click on Custom in the Imager software and choose the image you download. You preferably want the Lite images, because the GUI will use more disk, RAM and power and really isn't needed for anything.

Either way, you'll want to setup a network connection (wired is fine, you don't need network while you're using it) and enable SSH. I suppose you could also plug in a keyboard, mouse and monitor, but you'll still need a network connection to run updates and install software.

I won't go over all the aspects of OS install because there are a lot of resources out there already.

Make sure to run updates and if necessary run raspi-config and expand the filesystem (I didn't have to do this on the pi4, so maybe it automatically uses 100% of the sd card now).

Once you've done one or two of these installs, it takes maybe a total of 10 minutes. Very easy.

 

[Incognitus]

The biggest pain in all of this was getting bluetooth working. Plugging in headphones would work fine, but I wanted to have the least amount of wires involved as possible.

Bluetooth is the most painful thing to get setup on the Pi's, especially if you're going to use ALSA tools (such as aplay) to output audio to a bluetooth device. What you'll see a lot of times is you'll pair, trust and connect your device, but the pi will automatically disconnect like 1 second later. Really really annoying. Honestly, if this post helps anyone beating their head against a wall trying to get bluetooth working on a Pi, I'll consider all this work worthwhile.

The trick is to install and start pulseaudio before pairing and connecting your bluetooth device. You also need the pi-bluetooth package:

Please, Log in or Register to view codes content!

Reboot, SSH in and as the pi user start pulseaudio, and use bluetoothctl to pair, trust and connect your device:

Please, Log in or Register to view codes content!

Obviously replace 00:00:00:00:00:00 with the MAC address of the device you're pairing. You'll have to put your bluetooth device into pairing mode. The MJAW thing I have goes into pairing mode automatically for a minute if it doesn't connect to anything it's already paired to, but my bluetooth speaker has a button to hold down to put it into pairing mode. You should see it pop up in bluetoothctl when the Pi sees it available for pairing. It'll have a green NEW in front of it.

As long as pulseaudio --start is run on boot (via cron or some other method), from this point on you only need to power on the pi and power on the bluetooth device and it'll connect. Nothing manual is needed.

I currently have a script that runs on boot that starts pulseaudio and then watches the /var/log/messages file for the bluetooth buds to connect and then auto starts the ghost box software.

 

[Incognitus]

For the script I mentioned in the previous post, in /etc/crontab I have:

Please, Log in or Register to view codes content!

The contents of that script is currently the following:

Please, Log in or Register to view codes content!

This is my script for Raspios. It had to be slightly different for Raspbian because it logs bluetooth connected devices differently. It needs to be written better. It'll get triggered when you shut your bluetooth device off as well.

At any rate, it waits 30 seconds, starts pulseaudio, waits 5 seconds and uses tail to watch the /var/log/messages log for the line saying my ear buds are connected. When it sees that, it auto starts rtl_gb (the actual software), pipes it through sox (a general audio toolbox, you can use low/high pass filters, noise reduction, etc, through sox) and finally plays the piped audio via aplay.

So when in actual use, you plug the battery into the Pi, wait 1-2 minutes, then turn your bluetooth devices on and 10 seconds or so later the ghost box automatically starts. Power off the bluetooth device and the rtl_gb software is stopped on the Pi. So after setup, you don't really need to SSH in unless you either want to make changes or something to that effect.

The rtl_gb command can be piped through tee, which will allow you to save a local copy of the feed while also outputting the audio to bluetooth. Changing that line to the following would save the audio in a file on the Pi named file.wav:

Please, Log in or Register to view codes content!

 

[Incognitus]

Now on to the software itself. I'm not posting the source for rtl_gb yet because it's a mess and I'm still working on it. Ultimately, it is a modification of the rtl_fm tool from the rtl-sdr software.

You can see the install process in the following Quick Start guide, the GETTING STARTED ON LINUX section:

Please, Log in or Register to view URLs content!

. I'd recommend doing the manual install. It's very easy and you'll eventually need to compile rtl_gb so you'll need the source installed anyways.

At this point, plug in your SDR, connect your antenna, and you can use the normal tools, such as rtl_tcp (which will allow you to point SDR# running on another PC at the PI's SDR feed), rtl_fm to listen to specific stations or scan, rtl_power to do frequency surveys, etc.

Using rtl_power with heatmap.py from

Please, Log in or Register to view URLs content!

will let you create PNG files of scanned frequency ranges making it easy to see where the strongest signals are.

For example, here's a heatmap from the Pi3 running Raspbian:

https://imgur.com/B3dqMfa

And here's a heatmap from the Pi4 running Raspios:

https://imgur.com/iGyKBHd

So a couple things here... The same SDR Dongle and antenna are being used on both. The Pi4 has USB3 ports, but the Pi3 only has USB2. The Pi4 has 8GB of RAM, and the Pi3 only has 1Gb. And lastly the Pi4 is running 64bit Raspios and the Pi3 is running 32bit Raspbian.

The pi3 has a ton more noise. Its image is overall brighter. There's more of a contrast in the Pi4 heatmap between actual FM stations and just static. The Pi4 has some weird moving interference (the lines that aren't completely vertical). The actual stations themselves seem stronger in the Pi3 heatmap, but maybe that is because the noise level is overall higher. I dunno, I still know very little about radio itself.

Anyways, both show radio stations, so there's bound to be at least some speech going on. The Pi3 will have more static, which you may or may not consider a good thing.

The rtl_gb program supports everything rtl_fm does. In addition, it lets you set the squelch at 0 (so that you will hear static while hopping frequency), a configurable delay between frequency hop and frequency hop mode.

Please, Log in or Register to view codes content!

Now... I can't say that the hop delay ends up being accurate. I think it ends up being hop delay + however long it takes rtl_gb to run through 1 iteration of its main loop. I think accurate enough for our purposes. I'm not trying to will any science awards, and honestly 99% of the credit goes to the author of rtl_fm, Kyle Keen (

Please, Log in or Register to view URLs content!

). I am working on coding something from scratch, but I'll likely release rtl_gb first because it works well.

I like to use it with hop delay set at -1 for random and for mode set to random as well. Hoping sequentially almost guarantees you'll hear full phrases from strong stations because they'll be broadcasting over a large frequency range.

As I've probably mentioned, we live in what's basically a cabin in the woods on a mountain, so we normally don't get too many FM stations. Hearing full phrases when random frequency hopping is creepy AF. And it happens ALL THE TIME, which seems to defy logic. I regularly get direct answers to questions.

This will eventually be released on Github with instructions on how to grab and compile it. At that point, you'll have a completely portable, SDR based ghostbox that you built yourself and you know has no fakery built into it.

One note about why I'm using FM...These SDR dongles are actually made for digital TV. Except for the newer SDRs, you need an upconverter to even tune in AM radio frequencies. An upconverter is a separate device that modifies the frequency by adding 125Mhz to it. So 0Mhz becomes 125Mhz so that it's in a range supported by the SDR and in your SDR software there is usually an offset you can enable so it automatically subtracts 125Mhz.

The upconverter requires power, and normally you'd use a USB cable to the Pi. So now I have something that needs a case, where right now it's easy to carry around with some velcro around it (pictures soon). And it's going to pull more power and there's more heat. There is a nano version of Ham It Up and I might give it a try some day. Would be nice to have AM as an option.

Newer SDR dongles have a direct sample mode which theoretically allows you to tune to AM frequencies without an upconverter. It's not very good and needs a really good antenna, which again makes it less portable.

There are a lot of other frequency ranges you can use for both AM and FM. You could tell rtl_gb to hop through the entire range the SDR supports (28Mhz-1900Mhz or something like that). There are other things like local fire departments that use AM and FM, but most of that range is pretty much empty.

 

[Incognitus]

Estes Method

I first saw this watching Hellier. If you like creepy stuff, I highly recommend. The Estes method was created in 2016 by Karl Pfeiffer, Connor Randall and Michelle Tate, and it's been used in the TV show Kindred Spirits and by Greg and Dana Newkirk in Hellier.

With a normal ghost box session, you're using a radio with a built in speaker or you're outputting to a bluetooth speaker so everyone can listen, ask questions, interpret what they're hearing, etc.

With the Estes method, a single person listens to the ghost box blindfolded and preferably using noise canceling, over the ear headphones. The idea here is the person listening to the ghost box has no way of being prompted. They don't know when something is said or a question was asked, or what was asked. Want to REALLY freak someone out? Make sure the person listening can't detect you moving, quietly move to another room and ask a question. Even if the responses are coincidental, it's still such a thrill when it happens.

I really like that disconnect. Obviously, the person listening still knows the topic, and likely knows what type of questions will be asked. It would be really interested to see if you could have someone do the Estes method after taken some place blindfolded, with no previous knowledge of the history of the place or any details.

We can get into the likely psychological, and maybe subconscious, forces at work here, but at the end of the day it's creepy and it's a fun thing to do on a Saturday night during a pandemic.

 

[SkullTraill]

I'm not at all interested in ghost hunting, but if you've seen my posts on hacker culture and democratized networks, you'll know that I'm always interested in home technology/communication devices. Very interested in seeing your progress! Thanks for jounraling on here, and keep it up!

 

[SkullTraill]

Estes Method

I first saw this watching Hellier. If you like creepy stuff, I highly recommend. The Estes method was created in 2016 by Karl Pfeiffer, Connor Randall and Michelle Tate, and it's been used in the TV show Kindred Spirits and by Greg and Dana Newkirk in Hellier.

With a normal ghost box session, you're using a radio with a built in speaker or you're outputting to a bluetooth speaker so everyone can listen, ask questions, interpret what they're hearing, etc.

With the Estes method, a single person listens to the ghost box blindfolded and preferably using noise canceling, over the ear headphones. The idea here is the person listening to the ghost box has no way of being prompted. They don't know when something is said or a question was asked, or what was asked. Want to REALLY freak someone out? Make sure the person listening can't detect you moving, quietly move to another room and ask a question. Even if the responses are coincidental, it's still such a thrill when it happens.

I really like that disconnect. Obviously, the person listening still knows the topic, and likely knows what type of questions will be asked. It would be really interested to see if you could have someone do the Estes method after taken some place blindfolded, with no previous knowledge of the history of the place or any details.

We can get into the likely psychological, and maybe subconscious, forces at work here, but at the end of the day it's creepy and it's a fun thing to do on a Saturday night during a pandemic.

God damn how do people do this stuff without shitting themselves. I can't even bear being blindfolded and left in a room alone, forget about adding a eerie/spooky radio receiver and hearing fucking weird shit... fuck me I'd die.

 

[Incognitus]

God damn how do people do this stuff without shitting themselves. I can't even bear being blindfolded and left in a room alone, forget about adding a eerie/spooky radio receiver and hearing fucking weird shit... fuck me I'd die.

I'll admit right now, nothing really scares me. Horror movies don't do anything to me. However, there are times when using the spirit box I get full body goosebumps because I'm so freaked out. Next post is going to be an overview of the testing we did yesterday and next steps.

 

[Incognitus]

We did two 15 minute sessions yesterday. Jaide was the listener and I was the questioner. I did record both sessions and I may see about pulling out some of the exchanges and posting them. Not sure yet. I don't like the sound of my own voice heh...

The first session was with the Pi4 running Raspios. Since I don't have a battery pack yet for the Pi4, it was plugged into power and was sitting on a tray table in front of Jaide (about a foot away). The antenna was fully extended.

The second session was with the Pi3 running Raspbian running off a battery pack. This should be relevant because power supplies usually introduce a TON of radio interference. So I expected this to be clearer. The antenna was fully extended and it way sitting on the same tray table in the same way as the Pi4.

For both, Jaide was using noise canceling ear buds (she can't hear anything other than the spirit box) and was wearing a blind fold. This was a little odd for us because we normally do this at night. This time we did it during the day. There was a 5-10 minute period between sessions.

Both sessions used rtl_gb with a random hop delay, random frequency each hop in the 88Mhz-108Mhz range (standard FM). It's hard to tell in the screenshots in one of the previous posts because of all the interference, but there's very few actual stations I can tune to in that range (not all of those vertical bars are stations, much of it is interference). Also the random hops are very random. It's exceedingly rare for it to hop to a frequency near the one it's currently on (reducing the possibility of getting full phrases from 1 hop). Some of the sentences Jaide heard were a different voice for each word (NOT the same station in other words). There were complete sentences.

So both sessions had a fair amount of back and forth chatter, but the Pi4 was better as far as clarity goes.

The Pi4 session was creepy as hell. First of all, Jaide kept hearing her name. Actually 3 versions of her name, multiple times. Her first name isn't totally unheard of, but it's not a super common name either, and it said her full name, her short name AND her nickname. There were also a LOT of "CALL US" "HELP US". There were repeated mentions of the number 66. At one point, there was a long period where it said complete sentences such as "I myself am over fire, hurry, save me". When I asked, who are you? The answer came back pretty much immediately: "BITCHES!". Well ok. Conversation seemed to die out when we got closer to the 15 minute mark. Throughout this, I was asking (very quietly so Jaide couldn't even feel the vibrations of me talking) "Who are you?" "What is your name?" Etc...

The Pi3 session seemed to pick up on the same conversational thread, but both to me and Jaide it was less clear. I think it's because there's just so much more interference when running on the Pi3, which is confusing to me. The Pi4's are known to output more interference, and since it was using a cheap power plug, I expected the Pi4 to be less useful, but the opposite happened. There was still a fair amount of "SAVE US" "CALL US" and that type of thing. This time I asked "Who are you?" and it said "WHORE!". Both those responses were directly after I asked, as if in response. Keep in mind, Jaide cannot see or hear anything and has no idea when I'm talking, when I'm just sitting and waiting, or even if I got up and walked into another room.

Not really sure what to make of "bitches" and "whore". Or 66. 66 seemed to be important. I asked "66 what?" and I didn't get a clear response.

So now the question is, why do they function different? It may just be the difference in RAM (8GB in the Pi4, 1GB in the pi3), or different OS (64bit in Pi4 vs 32bit in Pi3), or any other number of reasons. It could be that the Pi4 supports USB3, but the Pi3 only supports USB2. Today I'm installing Raspios 64bit on the Pi3 and will do some frequency analysis and testing and see if it makes any difference. I'd like to use the Pi3, just with the same quality, if it's possible, but honestly it may just be the faster "everything" on the Pi4 that makes it work better.

One issue I have to solve, if I move forward with the Pi4 is that the Pi4 gets MUCH hotter. It's not particularly comfortable to hold in your hand. It'll still be 100% easily portable with a battery pack, but might require not holding it during the session. The Pi3 gets warm, but not too bad. You can hold it in your hand (bonus: makes it easy to hold the antenna, using your body as an antenna) without any discomfort. For the Pi4 I will probably need to design a 3d printed case, possibly with room for a small fan (if I can find a fan that doesn't add a ton of interference).

All in all, it works. It works so well sometimes it's scary. More to come.

 

[SkullTraill]

We did two 15 minute sessions yesterday. Jaide was the listener and I was the questioner. I did record both sessions and I may see about pulling out some of the exchanges and posting them. Not sure yet. I don't like the sound of my own voice heh...

The first session was with the Pi4 running Raspios. Since I don't have a battery pack yet for the Pi4, it was plugged into power and was sitting on a tray table in front of Jaide (about a foot away). The antenna was fully extended.

The second session was with the Pi3 running Raspbian running off a battery pack. This should be relevant because power supplies usually introduce a TON of radio interference. So I expected this to be clearer. The antenna was fully extended and it way sitting on the same tray table in the same way as the Pi4.

For both, Jaide was using noise canceling ear buds (she can't hear anything other than the spirit box) and was wearing a blind fold. This was a little odd for us because we normally do this at night. This time we did it during the day. There was a 5-10 minute period between sessions.

Both sessions used rtl_gb with a random hop delay, random frequency each hop in the 88Mhz-108Mhz range (standard FM). It's hard to tell in the screenshots in one of the previous posts because of all the interference, but there's very few actual stations I can tune to in that range (not all of those vertical bars are stations, much of it is interference). Also the random hops are very random. It's exceedingly rare for it to hop to a frequency near the one it's currently on (reducing the possibility of getting full phrases from 1 hop). Some of the sentences Jaide heard were a different voice for each word (NOT the same station in other words). There were complete sentences.

So both sessions had a fair amount of back and forth chatter, but the Pi4 was better as far as clarity goes.

The Pi4 session was creepy as hell. First of all, Jaide kept hearing her name. Actually 3 versions of her name, multiple times. Her first name isn't totally unheard of, but it's not a super common name either, and it said her full name, her short name AND her nickname. There were also a LOT of "CALL US" "HELP US". There were repeated mentions of the number 66. At one point, there was a long period where it said complete sentences such as "I myself am over fire, hurry, save me". When I asked, who are you? The answer came back pretty much immediately: "BITCHES!". Well ok. Conversation seemed to die out when we got closer to the 15 minute mark. Throughout this, I was asking (very quietly so Jaide couldn't even feel the vibrations of me talking) "Who are you?" "What is your name?" Etc...

The Pi3 session seemed to pick up on the same conversational thread, but both to me and Jaide it was less clear. I think it's because there's just so much more interference when running on the Pi3, which is confusing to me. The Pi4's are known to output more interference, and since it was using a cheap power plug, I expected the Pi4 to be less useful, but the opposite happened. There was still a fair amount of "SAVE US" "CALL US" and that type of thing. This time I asked "Who are you?" and it said "WHORE!". Both those responses were directly after I asked, as if in response. Keep in mind, Jaide cannot see or hear anything and has no idea when I'm talking, when I'm just sitting and waiting, or even if I got up and walked into another room.

Not really sure what to make of "bitches" and "whore". Or 66. 66 seemed to be important. I asked "66 what?" and I didn't get a clear response.

So now the question is, why do they function different? It may just be the difference in RAM (8GB in the Pi4, 1GB in the pi3), or different OS (64bit in Pi4 vs 32bit in Pi3), or any other number of reasons. It could be that the Pi4 supports USB3, but the Pi3 only supports USB2. Today I'm installing Raspios 64bit on the Pi3 and will do some frequency analysis and testing and see if it makes any difference. I'd like to use the Pi3, just with the same quality, if it's possible, but honestly it may just be the faster "everything" on the Pi4 that makes it work better.

One issue I have to solve, if I move forward with the Pi4 is that the Pi4 gets MUCH hotter. It's not particularly comfortable to hold in your hand. It'll still be 100% easily portable with a battery pack, but might require not holding it during the session. The Pi3 gets warm, but not too bad. You can hold it in your hand (bonus: makes it easy to hold the antenna, using your body as an antenna) without any discomfort. For the Pi4 I will probably need to design a 3d printed case, possibly with room for a small fan (if I can find a fan that doesn't add a ton of interference).

All in all, it works. It works so well sometimes it's scary. More to come.

That is so fucking creepy. I wonder why it's so hard to extract an answer for the question "who are you?"

Maybe the scariest part is they don't even know.

I'm just generalizing not only for your particular setup, but for all anecdotes I've heard regarding "spirit communication"

I'm a bit sceptical about the random hop delays to different frequencies... Why do that? Surely that implies complete variation and a spirit cannot be expected to hop frequencies with you?

 

[Incognitus]

I'm a bit sceptical about the random hop delays to different frequencies... Why do that? Surely that implies complete variation and a spirit cannot be expected to hop frequencies with you?

The random hop delay was my attempt at introducing more randomness to the whole process. I've used it at 2ms delay and that works real well. Using a long hop delay on each freq change sits on one channel too long. It's kinda weird, but It seems to work equally as well regardless of if you use a random delay per hop or a set delay like 2ms.

The random hop frequency is to avoid hitting strong radio stations multiple times in 1 hop sweep, thereby staying on that station for potentially a second or more. That will almost always pull complete phrases from a single radio station, which I'm trying to avoid. It can go up and down sequentially also, but it always sounds more like a conversation when using random frequencies.

For my purposes, I'm trying to introduce randomness as much as possible. You make a point that the randomness would require the spirit to hop frequencies, and from a completely logical standpoint, I'd agree. You could go even further, if it's "ghosts", and they're from hundreds of years ago, how are they going to use this at all? If it's entities from another dimension, with completely different technology, same problem really.

Despite all that, from my own tests, the more randomness I inject into the process, the more sensory disconnect we use between the listener and questioner, the better the conversation ends up being. It's like the randomness is part of the process which is being used to communicate.

I dunno. Honestly, I started this just as a fun raspberry pi project. I didn't expect it to work, from a technical standpoint. Then when it did work technically, I didn't expect anything except static and random sounds and words that made no sense, but I sometimes get extended answers to questions (never the ones I really want answers to, like "who are you?"). I am a skeptic at the end of the day, which is why I didn't want to buy a device to use for fear of fakery and I'm a bit shocked at how well it works. Not sure what to really make of it, but definitely requires more experimentation.

I'm thinking of doing something really crazy next session. In another thread I posted a link to an audio file that was created by importing the sigil of King Paimon. I cleaned the audio up significantly and I think I'm going to slow it down, enhance the bass, and play it in the background during the next session. If Jaide will go for it.

 

[Incognitus]

So a lot has happened here. Honestly, I may need to just start a spirit box 2.0 thread and start from scratch. The Bluetooth info above is missing some info and Skull said he would help me replace that (thanks Skull), but I’ve made a lot of changes.

First things first, I’ve decided on a name. It will be called PiGB (Raspberry Pi SDR Ghostbox). Yay for names.

The way it worked previously is there was a script started on boot that watches a log file and would automatically run rtl_gb when it sees a headset connect. This was my initial attempt at making it function on its own without having to bring a laptop. This method doesn’t allow any settings changes.

So yesterday I went another route. I installed RaspAp which turns the Pi into a Wi-Fi AP (rather than a client). This way, you can just whip out your phone, connect to the PiGB Wi-Fi hotspot and access a web page I built that allows you to change settings (hop delay, hop mode, whether it also writes to a file, can set a time limit) and has buttons to start and stop the gb software. I like this much better. Everyone always has a phone on them and it’s extremely easy to use. I’m really happy with this. The web page needs to be prettied up, but this makes it a 100% portable, carryable, manageable device without needed a laptop or an existing Wi-Fi connection. Even better, if you plug an Ethernet cable into the Pi, the AP will provide internet access. You can install a SSH client on your phone and ssh into the Pi directly from your phone if something needs to be tweaked.

I did a lot of testing using rtl_power with and without the Pi Wi-Fi in AP mode and I don’t see any interference in the frequency ranges that would be typically used for a ghostbox. That made me happy because I want this to be easily configurable without a computer or laptop.

I spent a bunch of time upgrading the Pi3 I have to Raspios 64bit so I could compare the pi3 (usb2.0, slower cpu and 1gb ram) with the pi4 (usb3.0, faster cpu quad core, 8gb ram) using the same software versions and config. I’ve been concentrating on the Pi3 because thats the one I currently has a battery for, it doesn’t take as much power and runs slightly cooler.

I’ll do a session report in the next post, but suffice to say, the Pi4 (maybe unsurprisingly) functions a lot better. There seems to be much more (and more accurate) communication. To a scary extent. More to come.

 

[Yazata]

66

I have no idea who said it where, but I remember listening to a podcast years ago where the guest said 66 is the number of the fallen Angels.
Crowley gives 66 as the Mystical number if the qlipoth.

 

[Incognitus]

I have no idea who said it where, but I remember listening to a podcast years ago where the guest said 66 is the number of the fallen Angels.
Crowley gives 66 as the Mystical number if the qlipoth

I'm going to have to look into this more. Thanks! The other odd thing is that Jaide had been using a deck of faery oracle cards, and there just happens to be 66 of those. Either way, very odd.

Making some great progress. I have a very basic web interface. This works real nice. Plug the Pi into a battery, wait a minute, and then you can attach to the wifi hotspot on the Pi itself. From there you can pull up a web interface on your phone to control the software.

Here's what that looks like now. Like I said, very basic HTML and PHP just to get functionality without having to SSH in. You CAN still SSH in once attached to the PiGB's hotspot. I scripted most of the web interface on an iPhone X using Termius. Worked great.

https://imgur.com/9O5dFAV

So at this point, it's completely manageable without a laptop or another Internet connection. Setup is much simpler from when I started. Very happy with it.

Also fixed some resampling issues. Sox was changing the play speed somehow, which made things tend to sound deeper in pitch and more bassy. Some sort of oddity with Sox on Pi4 RaspiOS 64bit I think. I also had issues with saving the ghost box output to file. THAT was saving it at TWICE the speed. Shesh. Both fixed now. Wav filename is now date/time recording starting. This is recording output from the ghostbox, not anything through a microphone (at least not yet).

One thing I need to do is grab a newer copy of rtl_fm and make the ghost box mods to that. Or code something from scratch. Not sure yet.

 

[Jarhyn]

So, I am a professional embedded/firmware/systems engineer, and am currently involved in a number of signal processing projects surrounding a very sensitive EM sensor.

I'll level insofar as I'm not entirely sure what your goals are in terms of I/O and capture.

Have you considered collecting entropy from your input and using that entropy to drive frequency hop randomizations?

It seems your goal is to allow as much information as possible, particularly CMBR, to drive the output of (outside agent?) phrase generation?

The more you drive entropy in from the ccollector...

 

[Jarhyn]

At any rate, if you want, I can build anything you want as long as you're willing to run the frontend coding aspects. I would think important features to code would be as referenced allowing the thing to collect entropy from a mic to start with, running a rolling XOR operation on the most recent piece of data before spinning on (maybe as you speak your question, even), and then start running in the audio data on top of the xor stack for selecting the next frequency.

This gives you highly chaotic (at first) and then true random (once it taps into the radio receiver).

Oh, and a rolling buffer you can activate to record back-logged stuff that you otherwise would have missed because it already passed

 

[Incognitus]

I'm not saying what you're talking about wouldn't be an interesting project, but isn't really what I'm doing here.

I'll level insofar as I'm not entirely sure what your goals are in terms of I/O and capture.

This is running Linux, therefore librtlsdr already exists, and sox can be used for saving (and audio filtering if desired). It would be silly to code from scratch an audio toolbox when such an awesome one already exists. The wheel already exists.

Have you considered collecting entropy from your input and using that entropy to drive frequency hop randomizations?

No, that's a different type of device. One exists, I forget the name of it, and there's no public information about it. It may just be TV (bullshit) magic. The concept is interesting tho. It supposedly uses environmental variables for word and phrase generation. That isn't related to what I'm doing here.

Spirit boxes typically have no "input" in that way. Questions are asked verbally.

It seems your goal is to allow as much information as possible, particularly CMBR, to drive the output of (outside agent?) phrase generation?

SDRs were originally digital TV receivers. What digital TV receivers are you aware of that covers the 70-217Ghz range?

Most radio noise (static) is natural electromagnetic atmospheric noise, which includes manmade RF interference. You likely get more static from the devices in your house, and the receiver itself if it's low quality.

I am not generating phrases. In fact, I'm doing my best NOT to intentionally generate anything. The input is AM or FM radio (or really, any frequency and modulation the SDR dongle and software supports). The "output" is WAV audio.

At any rate, if you want, I can build anything you want as long as you're willing to run the frontend coding aspects.

One of my stated goals was to build something I knew had no fakery and I said I do not have enough electrical knowledge to take apart a commercial radio spirit box and verify it's not doing some sort of intentional phrase generating (in other words, faked). I'm not sure what you would build. This is a Raspberry Pi project, as stated. Either way, I don't know you nor trust you, so the problem would be exactly the same as just buying someone else's $80 modified radio.

I would think important features to code would be as referenced allowing the thing to collect entropy from a mic to start with, running a rolling XOR operation on the most recent piece of data before spinning on (maybe as you speak your question, even), and then start running in the audio data on top of the xor stack for selecting the next frequency.

This gives you highly chaotic (at first) and then true random (once it taps into the radio receiver).

Oh, and a rolling buffer you can activate to record back-logged stuff that you otherwise would have missed because it already passed

I think, perhaps, the misunderstanding here is that you might think I want to have a microphone that somehow drives some sort of generator. That is not what a spirit/ghost box is or why I mentioned it. The possibility of a mic would be so that the Pi can record the questioner (see below), and then mux that audio track on top of the raw track coming from the radio (that the listener is hearing). That way you would have a single audio file with the question and answer sync'd up. That was the only purpose for potentially adding a mic, and I may not even consider it as that's more processing wearing down the battery.

A spirit/ghost box is essentially just a radio that scans through FM or AM radio frequencies very fast, staying on each frequency just a matter of milliseconds. Most sweep those ranges sequentially ascending. You can usually set the hop delay (the amount of time the radio plays the current frequency, even if it's static, before moving on), with 100 milliseconds being common. You usually can't change anything else.

This is just a pet project to build something like that, just with a Raspberry Pi running Linux and a RTLSDR dongle. Because it's written in C and I'm writing it myself, I can give the ability to tweak settings you can't do on something like a SB7. You can hop random frequencies (instead of just sequential ascending), plus configure things such as sample/bandwidth rate, resample rate (typically 48khz for me), use audio filters via sox (high/low pass filters, loudness normalization, etc), save to a WAV file (which a lot of radio ghost boxes can't do at all), and define what frequency ranges to hop through. Feel like the universe is going to talk to you somewhere between 400-500Mhz today? Sure, you can set that range with what I built, but you can't do that on a commercial box.

In reality, this should return gibberish or static all the time. If you set the hop delay to something very small, like 1ms, you won't even pull a full word from a strong FM station, so it's a bit weird when complete phases come out.

To further add to the randomness, there are 2 people involved, the questioner and the listener. This is called the Estes method. The listener, the person listening to the radio hops, sits blindfolded with noise cancelling headphones. The listener can't hear what questions are asked, when they are asked or if the questioner is even still in the same room. The questioner asks questions verbally. The listener says what they hear.

Bullshit? A type of audio pareidolia? Sure, maybe. I dunno. Appears to work, but even if it is complete BS it's a fun little project. I like building things. Either way, it's a combination of technology and supernatural/occult, and that interests me.

 

[Jarhyn]

I'm not saying what you're talking about wouldn't be an interesting project, but isn't really what I'm doing here.



This is running Linux, therefore librtlsdr already exists, and sox can be used for saving (and audio filtering if desired). It would be silly to code from scratch an audio toolbox when such an awesome one already exists. The wheel already exists.


No, that's a different type of device. One exists, I forget the name of it, and there's no public information about it. It may just be TV (bullshit) magic. The concept is interesting tho. It supposedly uses environmental variables for word and phrase generation. That isn't related to what I'm doing here.

Spirit boxes typically have no "input" in that way. Questions are asked verbally.



SDRs were originally digital TV receivers. What digital TV receivers are you aware of that covers the 70-217Ghz range?

Most radio noise (static) is natural electromagnetic atmospheric noise, which includes manmade RF interference. You likely get more static from the devices in your house, and the receiver itself if it's low quality.

I am not generating phrases. In fact, I'm doing my best NOT to intentionally generate anything. The input is AM or FM radio (or really, any frequency and modulation the SDR dongle and software supports). The "output" is WAV audio.


One of my stated goals was to build something I knew had no fakery and I said I do not have enough electrical knowledge to take apart a commercial radio spirit box and verify it's not doing some sort of intentional phrase generating (in other words, faked). I'm not sure what you would build. This is a Raspberry Pi project, as stated. Either way, I don't know you nor trust you, so the problem would be exactly the same as just buying someone else's $80 modified radio.




I think, perhaps, the misunderstanding here is that you might think I want to have a microphone that somehow drives some sort of generator. That is not what a spirit/ghost box is or why I mentioned it. The possibility of a mic would be so that the Pi can record the questioner (see below), and then mux that audio track on top of the raw track coming from the radio (that the listener is hearing). That way you would have a single audio file with the question and answer sync'd up. That was the only purpose for potentially adding a mic, and I may not even consider it as that's more processing wearing down the battery.

A spirit/ghost box is essentially just a radio that scans through FM or AM radio frequencies very fast, staying on each frequency just a matter of milliseconds. Most sweep those ranges sequentially ascending. You can usually set the hop delay (the amount of time the radio plays the current frequency, even if it's static, before moving on), with 100 milliseconds being common. You usually can't change anything else.

This is just a pet project to build something like that, just with a Raspberry Pi running Linux and a RTLSDR dongle. Because it's written in C and I'm writing it myself, I can give the ability to tweak settings you can't do on something like a SB7. You can hop random frequencies (instead of just sequential ascending), plus configure things such as sample/bandwidth rate, resample rate (typically 48khz for me), use audio filters via sox (high/low pass filters, loudness normalization, etc), save to a WAV file (which a lot of radio ghost boxes can't do at all), and define what frequency ranges to hop through. Feel like the universe is going to talk to you somewhere between 400-500Mhz today? Sure, you can set that range with what I built, but you can't do that on a commercial box.

In reality, this should return gibberish or static all the time. If you set the hop delay to something very small, like 1ms, you won't even pull a full word from a strong FM station, so it's a bit weird when complete phases come out.

To further add to the randomness, there are 2 people involved, the questioner and the listener. This is called the Estes method. The listener, the person listening to the radio hops, sits blindfolded with noise cancelling headphones. The listener can't hear what questions are asked, when they are asked or if the questioner is even still in the same room. The questioner asks questions verbally. The listener says what they hear.

Bullshit? A type of audio pareidolia? Sure, maybe. I dunno. Appears to work, but even if it is complete BS it's a fun little project. I like building things. Either way, it's a combination of technology and supernatural/occult, and that interests me.

Oh, I'm not trying to cast aspersions on what you are doing.

By input, I'm merely talking about cryptographically improving the randomness of your RNG, and then using additional data to randomize the RNG.

Radio static is commonly caused by a few sources, as you mention this includes "human sources", but most of it is the cosmic microwave background radiation left over from the big bang, pulsars singing, black holes merging, and other such chaos.

As your application uses randomizations to frequency hop, it's rather trivial to sample a mic to get initial "entropy", and then sample the input from the radio itself to improve that "entropy".

This would improve the fidelity of any "spookiness" as relates to the quality of the "randomness" of your frequency hop algo.

It just ALSO means that you have a fun input for seeding your RNG that directly incorporated the question.

 

[Incognitus]

@Jarhyn - You inspired me to try something. Not quite along the lines of what you were talking about, but that led to this idea.

Sox (the audio toolkit) has a noise reduction filter. To use it, you feed it some noise and it creates a profile, then you can apply that profile to audio and theoretically it removes any noise similar to what was profiled. I've messed with this some in the past, because it can create some neat audio effects.

Here's a 1 minute WAV recorded and saved on PiGB:

https://soundcloud.com/user-830161545%2F2022-02-02-01-57-12a%2Fs-MRNVTkZXgcw
I created the sound profile off of the above file:

sox 2022-02-02_01\:57\:12.wav -n noiseprof noise.prof

(Gotta escape the colons). This created a noise profile based on the original recording. Then I applied noise reduction to that same original recording, using the sound profile also created off the same file.

sox 2022-02-02_01\:57\:12.wav ./noisered.wav noisered noise.prof 0.21

And this is what I get:

https://soundcloud.com/user-830161545%2Fnoisered%2Fs-hCp1cqDEwY6
That is creepy as hell! With a little more processing, it would be even cleaner. At any rate, I thought it was neat to both create the noise profile and use it to filter the same file.