Nyquest, etc. Questions

[jeff_p58]

05/28/2016 at 8:01 pm

OK, I am a new Prime member and have been playing around with writing and creating and scripts. I have been using Audacity 2.1.2 and had some issues with the Nyquist effect. I researched the forums and learned a lot and in the end came up with the following notes. I would love to get feedback on my notes and my questions below:

1. Develop Script See separate documentation on how best to phrase and structure subliminal script files

2. Debug Script Using a Text2Speech program like ‘TextAloud” play your script and look for places where pauses might be inserted or where the text reader has pronunciation problems. Make sure that the spoken text flows the way you want it to

3. Speak to File After you are happy with your script: Speak the script to a .wav file – wav files are in a lossless audio format and work best for editing inside Audacity

4. Start Audacity You can create a new project or add to an existing project – we assume here you are creating a new project

5. Import Drag & Drop you .wav audio file into Audacity

6. Set Project Rate Double check that it reads at least 44,000 in the lower left corner – the higher the rate the higher quality of overall project but the larger the files

7. Select All In a new project; Just hit Control Key +A and it the whole thing should turn a deep blue to indicate it has been selected

8. Set Volume [Effect>>Amplify] menu This adjusts the volume of the recording in decibels-Dream Girls recommends reducing the amplitude/volume by entering -44.2 db in first box

9. Resample Just to be safe hit Ctrl-A again to make sure all of track is selected – then on [Tracks>>Resample] menu, type in 192000 without commas, or select it from the drop down box. Make sure there are three 0’s if you typo it with a missing 0, it won’t be re-sampled at a high enough frequency for the Nyquist to do its job

10. Shift Frequency Just to be safe hit Ctrl-A again to make sure all of track is selected – On the [Effect>>Nyquist Prompt] menu selection enter the following script:
(highpass8
(lowpass8
(mult
(lowpass8 (highpass8 s 300) 7000)
(hzosc 45000)
(hzosc 59500))
24000)
15000)
Before you hit [OK] on later versions of Audacity, make sure you haves selected the ‘Use Legacy (Version 3) Syntax’ checkbox

11. Add Background If this is not to be a silent recording, you can Import and Set Volume for a new parallel track of sounds like rain or ocean waves. Be careful when you are adjusting the volume you are only affecting the new track(s)

12. Extend Select all of the subliminal track not including any background tracks and [Effects>>Repeat] menu
• You want to repeat the subliminal message to be just short of the background track – for example a 5 minute subliminal would be repeated 4 more times to come just shy of the 27 minute rain background track from Dream Girls.
• Extend to 10 minutes or so if you are using a playlist and there is no background noise track; that way your playlist doesn’t jump from topic to topic so quickly.
• If you only play one silent file, then this is unnecessary.

13. Export [File>>Export] to an MP3 file

Question 1[/b]
Step 10 – shift frequency – I interpret the Nyquist script inside out:
(highpass8
(lowpass8
(mult
(lowpass8 (highpass8 s 300) 7000)
(hzosc 45000)
(hzosc 59500)
)
24000)
15000
)
1. Filter all frequencies above 7000 and blow 300 out of source (s)
2. Multiply result by a 45 kHz sinewave and a 59.5 kHz sinewave – this results in a 14.5 kHz carrier wave modulated by the results of step 1
3. Filter out all frequencies above 24 kHz and below 15 kHz

OK my questions is – if we are creating a modulated 14.5 kHz carrier wave, and we are then filtering all frequencies below 15 kHz – how does this work?

Question 2
Several folks in the forum discuss testing or QA’ing the results here by decoding it. I have not seen any discussion yet on how to do this or what to look for.

Question 3
In step 8 when adjusting volume. I am unsure about why there is a 50 db range and how to judge setting the new volume levels for a frequency adjusted audio file

Any assistance on either of these questions would be greatly appreciated.

[Fizbin]

05/29/2016 at 4:25 am

Hi, and welcome to the forums!

I’ll give some brief answers here, but feel free to ask for more technical detail if needed.

1. What the last filter does is remove the carrier itself (14,5000hz), as well as the “lower sideband” from the encoded signal. This ensures minimal stuff to interfere in the true audio range, while still providing the actual information that can be decoded in the near-ultrasonic space. This is essentially what old “CB Radios” did back in the 1970’s to produce “SSB” or “Single Sideband” signals.

2. The main thing you are looking for with your QA is that the decoded file is still audible and intelligible. There are quirks in digital signal processing that can, if the frequencies are precisely aligned, actually nullify everything and result in true silence.

3. Level setting is actually one of those places where this process is as much art as science, and depends almost as much on your playback device as your source file. Essentially, set the levels of your unencoded file such that when played back under the conditions (player, speakers, playback volume) you expect to use, the volume is reasonable at your intended recipient’s listening position.

[jeff_p58]

05/29/2016 at 6:35 am

Thanks for the quick reply Fizbin,

1. I sort of understand that is what that last set of filters is supposed to do. But I do not follow how filtering everything not int the 15-24 kHz range does that.

2. OK, we should decode, play and make sure that the file says what we think it should say – how exactly do we decode it?

3. I think my earlier issues with Nyquist (when I was not checking to use legacy syntax) was blocking my understanding of volume levels – now that the frequency shift is working (I Think – LOL) it seems to be OK

[Fizbin]

05/29/2016 at 3:55 pm

OK, Amplitude Modulation (AM) theory 101 🙂

Let’s say you have a typical analog audio signal, with a frequency range of between 0 and infinity. “Perfect” human hearing can discern frequencies between 20hz and 20,000hz. Our recording devices, therefore, are optimized for that tonal range. In practice, this range is not linear, and sensitivity to the extremes dims with age, so even within the perfect range, a lot of signal is just so much excess baggage.

For example, AM radio has a frequency cap of around 5khz, and standard analog telephony has a frequency cap of just 3.5khz. Both are intelligible, but also well below human maximums.

Now, to broadcast this signal, you need to modulate a carrier wave. In theory, this carrier can be anything, but in practice it should be at least twice the highest frequency of content you wish to broadcast. If your carrier is a perfect sine wave, it takes up zero “bandwidth” – literally the “width” of a “band” of frequencies centered on the carrier itself. If it is only the carrier, there is no band.

Once you start modulating the carrier, it starts taking up more space (aka bandwidth) in the frequency spectrum. Specifically, it takes up twice the space of the highest frequency signal used to modulate it. This is because just as a wave goes above and below zero amplitude, modulating a wave makes it go both above and below its core frequency by the amount of the modulating frequency. These are called the upper (above) and lower (below) sidebands of the carrier. The sidebands (normally) are exact mirrors of each other, so mathematically, all you really need to re-create the modulating signal is to know the carrier frequency and ONE of the sidebands.

In reality, depending on the demodulation technique, you don’t even need that. The crystalline hairs of the inner ear are one such carrier-insensitive demodulator. Thus, the final Nyquist filter removes the lower sideband and the carrier: 14.5k + 300 is about 15,000. 14.5k + 7k is about 22k, plus a little overhead because the filter is a slope, not a step-function.

[jeff_p58]

05/30/2016 at 8:06 am

thanks for the explanation. am going to have to let this rattle around in my brain for a bit to see what takes. LOL

[Tichi]

05/31/2016 at 2:13 am

I am not sure if this answers your original Q2 you can use this Nyquist block which Fizbin supplied to decode your ‘encoded’ mp3 back to normal.

To reverse the Nyquist filter and hear the file as ‘normal’.

Depending on your original encoding carrier, the eBook uses 14500 so just copy paste and apply to your file to decode it back.

(setq carrier 17500)
(lowpass8 (mult 2 s (hzosc
carrier)) 8000)

-or if you followed the eBook instructions-

(setq carrier 14500)
(lowpass8 (mult 2 s (hzosc
carrier)) 8000)

You may consider changing the amps to -30 instead of -44 or see what works for your playback device by creating a test file with some spoken text which is innocent (testing 123, etc) to see how loud it actually plays back. Follow the usual instructions but don’t apply the Nyquist filter. This is important to prevent hearing loss in case it’s too loud especially if listening with headphones on.

Also another check is to analyse the file in audacity and plot the frequency. You should see a graph with distinctive peaks.

Checking Spectral Analysis:
On a completed file, perform a spectral analysis (Audacity – Analyze – Plot Spectrum). You should see a peak graph around the 13.5K – 17K mark.

[randy_srs]

09/03/2016 at 6:54 pm

so my question is around being able to verify silent / subliminal tracks .

using a anyizar from apple app . i don’t see anything running in the 16000k range . so i am wondering how to verify the coverage range of a speaker and placement for best results .

[randy_srs]

09/03/2016 at 10:40 pm

another question 🙂

been doing research on decoding and verifying what the encoded file sounds like .
(setq carrier 16500)
(highpass8
(lowpass8
(mult
(lowpass8 (highpass8 s 300) 7000)
(hzosc 45000)
(hzosc 59500))
24000)
14000)

when i decode it with
(setq carrier 16500)
(lowpass8 (mult 2 s (hzosc
carrier)) 8000)

sounds like crap (computerized not anything like the amy recording at all )
is this correct ? just trying to verify the sub is correctly

[Fizbin]

09/04/2016 at 4:49 am

You’re not decoding at the same carrier as you are encoding. The “carrier” statement in the encoding stage isn’t actually doing the work – the 45000 and 59500 are. The delta between them is 14500, which is the net carrier signal. Use that instead of 16500 in your decode phase and you should get a clear output.

[randy_srs]

09/04/2016 at 11:19 am

Thank you fizbin that corrected it awesome.
Now for the next newbie question

Should a iPhone spectrum analyzer be able to pickup the frequency
Just looking at verifying the signals is there on the recordings

I know this is a lame questions

[Tap]

09/04/2016 at 11:29 pm

I think you should keep the same nyquist code as in the pdf file and don’t mess with it. If you want the squeaking louder, simply change the amp to -30 or a lower negative number, instead of -44.2.

[randy_srs]

09/05/2016 at 10:10 am

Hey Tap ,

another newbie question , what I am wondering is how to verify the scripts are actually playing on a device . other then wife responding to the scripts .
or how to see how far the speaker Carrier travels in a given room .

thanks for the info.

[Fizbin]

09/06/2016 at 8:14 pm

If you can decode it, the signal is there. A spectrum analyzer should detect it, as long as the pickup device has a suitable frequency response. I’m not sure if the mic on an iPhone does.

Generally, I suggest setting the playback device volume with a “control” song so that it can be heard at a reasonable volume at the expected listening distance, that should keep your subliminal from being so loud it damages hearing, but actually “detectible” by whatever actual mechanism picks them up (the inner-ear’s crystalline fibers, by current thinking).

[Tichi]

09/13/2016 at 2:27 am

You could create a “test” file with some innocent text in textaloud like “this is a test file, can you hear me?”.

Do all the steps to create the silent, except for the Nyquist prompt part (scrambling the message). Then play it on your device as a test. You can then play with speaker positiong and volume, etc.

Yes I see now not a volume question… In theory recording it then trying to decode it should work provided the device can adequately capture the sound. Must try that myself.

[randy_srs]

10/07/2016 at 12:59 am

Hey guys

So stupid question but I did a test

Did 2 recording
1. Using the the exact same wav file created
A silent recoding using the above steps

2. Using the SRS X1 recorder yes it’s 100 bucks
But here is the question

Test one is truly silent only here white noise. And tests from the spectrum analyzer from my phone to the other device pick upon audio in the 16khz range

Test 2 using the srsx1 though has a slight squeeknas you all call it hits mid volume in the 16 kHz range

So my question is does that mean that one recoding is being heard better than the other

Thoughts and comments would be great

[Fizbin]

10/07/2016 at 4:51 am

Every playback device is different. Some might have a better frequency response curve at the high end. Some that have noise reduction and/or other forms of internal signal processing may actually partially decode the message at higher (or in some cases even moderate) volumes. This can also vary somewhat by file, depending on the source. So, it is always a good idea to test your files on the actual playback device before exposure to your subject.

How that impacts reception in the subject is another matter. Just as with devices, each person’s hearing is different. What is “silent” to one may be received as “squeaks” to another. Age and past exposure to loud noises are significant factors here.

That said, it is probably safe to say that a device whose noise output is “differentiated” in a pattern consistent with the content of the recording may be more effective than one which is simply “pure” white noise. When you used your meter, did you detect any patterning in the device 1 output (regardless of whether you could hear it)?

If both showed good differentiation, then both should be effective. You may need to reduce the volume on the “squeaky” device, though, to ensure that the squeaks are not heard by your subject.

An acid test would be to record the output of your devices, and run THAT through the decoder algorithm to see if the decoded signal is intelligible. If both give you back your original sound at equivalent quality, then you’re golden with either device. Otherwise, I would lean toward whichever device gave you the better decoded results.