help-with-radio

[untold needle]

Error message with that code @granite spear @half plover

[granite spear]

I'm not familiar enough with numpy to know exactly what's going on, but it looks like you're reading the data in from a text file, but never actually converting it to a complex number. It's still a string?

[untold needle]

I just added a line to convert it to a complex number

f = np.array([complex(x) for x in f])

just now

this shrinks the error down @half plover @granite spear to just:

Traceback (most recent call last):
  File "C:\Users\aryan\PycharmProjects\Fast Fourier Transform\main.py", line 21, in <module>
    f = np.array([complex(x) for x in f])
  File "C:\Users\aryan\PycharmProjects\Fast Fourier Transform\main.py", line 21, in <listcomp>
    f = np.array([complex(x) for x in f])
ValueError: complex() arg is a malformed string

[half plover]

complex() I believe requires two inputs, first the real part, then the imaginary

[untold needle]

I see

[half plover]

So

a=1.03
b=0.8
c=complex(a,b)

As an example

[untold needle]

but I have the data in the form -0.0117647058823529-0.019607843137254943j

[remote pond]

complex() should parse complex number strings in that form

[untold needle]

throwing errors

[half plover]

try print(numpy.real(num))

[primal warren]

Seems fine here ```

complex('-0.0117647058823529-0.019607843137254943j')
(-0.0117647058823529-0.019607843137254943j)

[untold needle]

then why is it giving me errors

wait

its because its not in string form originally

but it is

surely

[primal warren]

It looks like you're making arrays of strings, then handing them to complex() two at a time?

[untold needle]

am I?

how so

import pyfftw
import multiprocessing
import numpy as np
import matplotlib.pyplot as plt

plt.rcParams['figure.figsize'] = [30, 21]

plt.rcParams.update({'font.size' : 10})

#Create a signal with two/multiple frequencies and some noise

dt = 4.16666667e-7 #Time div


with open('time.txt', 'r') as a:
    t = [line.strip() for line in a]

with open('voltages.txt', 'r') as a:
    f = [line.strip() for line in a]

f = np.array([complex(x) for x in f])

#FFT computation

n = len(t)
print('works')
fhat = pyfftw.interfaces.numpy_fft.fft(f, n)
print('works')
PSD = fhat * np.conj(fhat) / n
print('works')
freq = (1/(dt*n)) * np.arange(n)
print('works')
L = np.arange(1, np.floor(n/2), dtype = 'int')

indices = PSD > 40
PSDclean = PSD * indices
fhat = indices * fhat
ffilt = pyfftw.interfaces.numpy_fft.ifft(fhat) #Filtered signal

#Plot

fig, axs = plt.subplots(3, 1)


plt.sca(axs[0])
plt.plot(freq[L], PSD[L], color = 'c', linewidth = 2, label = 'Noisy')
plt.plot(freq[L], PSDclean[L], color = 'k', linewidth = 1.5, label = 'Filtered')
plt.xlim(freq[L[0]], freq[L[-1]])
plt.xlabel('Frequency(Hz)', fontsize=12)
plt.ylabel('Power(PSD)', fontsize=12)
plt.legend()

plt.show()

if you look at the complex line

f = np.array([complex(x) for x in f])

@primal warren it is just converting them to complex numbers

[half plover]

So it can’t be a string I’m finding

[untold needle]

im getting the string malformed error though

[young cove]

pyrtlsdr output .npy files natively, which you can read back in as complex arrays, so can you avoid the conversion to text?

[untold needle]

I have the .npy files

but I would like to fix this error

[remote pond]

what exactly does one of the problem input strings look like?

[primal warren]

Probably there's something in your file that's throwing it. Either print each value before converting it, or catch the exception and print the resulting string to see which one it gags on (I'm guessing a blank line)

[half plover]

If I have quotes around the number it returns the whole string

[untold needle]

its a txt file

here it is

if you look at my code, I unpack it earlier line by line

[young cove]

you could use numpy.loadtxt() to read it directly, setting dtype=complex,

https://numpy.org/doc/stable/reference/generated/numpy.loadtxt.html

instead of reading it line by line

[untold needle]

hmm

[young cove]

works perfectly for me:

>>> import numpy as np
>>> a = np.loadtxt("Downloads/message.txt", dtype=complex)

the values print match perfectly

[untold needle]

well erm

@young cove that was beautiful

[young cove]

array([-0.02745098-0.01176471j,  0.00392157+0.01176471j,
        0.00392157-0.03529412j, -0.01176471+0.00392157j,
        0.01176471+0.00392157j, -0.00392157-0.01960784j,
       -0.00392157+0.01960784j,  0.00392157-0.01960784j,

...

[untold needle]

it worked perfectly

[young cove]

numpy has all kinds of tools for data I/O, we should use them

[untold needle]

yeah no that definitely works

@young cove thanks a lot

numpy is actually great

[young cove]

numpy is fantastic, and is designed to be easy to use

[primal warren]

I remember the bad old days when building numpy would take a couple of days, but these days I just install it with pip and life is good.

[remote pond]

you have two blank lines at the end of your file, and complex() doesn't like empty strings as input

[young cove]

is pyfftw a separate package? There are FFT's in scipy (which builds on numpy)

[untold needle]

yes it is but pyfftw is basically just a faster version, it still builds on numpy

the syntax is just pyfftw.interfaces.(numpy or scipy)

[young cove]

got it

[untold needle]

this is weird though @young cove its running fine and all, I think I've calibrated the axis wrong

a 1MHz signal on a 92.2MHz received frequency?

Must be some dodgy axis calibration

[young cove]

the fft is going to make some assumptions about the center frequency, doest it have that as an optional arg?

[untold needle]

probably?

I will have a look

[untold needle]

from what I'm trying, it seems as though if I change the time divisions(dt) it changes the axis center frequency, I'm not sure how to tell the fft to set a center frequency

maybe fftshift()

[young cove]

i looked at the doc, and I see no way. The FFT is assuming a band starting a 0, I guess, and it's up to you to scale it

[untold needle]

I think maybe fftshift

not sure thoug

[young cove]

https://numpy.org/doc/stable/reference/generated/numpy.fft.fftshift.html

do you need the speed of pyfftw right now? You could use the standard numpy/scipy routines, and have more examples avaialble. Later you can optimize with a faster FFT

[untold needle]

not really

[young cove]

i would say stick with the integrated toolset for now

[untold needle]

ok

[young cove]

an FFT itself is kinda unitless

[untold needle]

this is true

Its about axis calibration though

[young cove]

the signal the pyrtlsdr records is downcoverted to a baseband. The data has no knowledge it was at 92-94MHz (or whatever the bandwidth you are using is). So the result is going to be 0-2MHz, etc. You can just add an offset.

if I understand correclty

[untold needle]

That's what chat gpt told me but I didn't listen to it.

Do I just add an offset of the centered frequency, so 92.2e6?

[young cove]

chatgpt didn't say that; some smart person wrote something that chatgpt paraphrased. it's better to dfind that smart person

is it center or left edge?

[untold needle]

Center

Center is 92.2 in pyrtlsdr

In the pi

[young cove]

and you gave that center to gqrx?

[untold needle]

Well yes

I gave it the parameters we discussed yesterday

[young cove]

So you want to make an adjustment so that it looks the same in your numpy code as it does in the gqrx plot

[untold needle]

Yes

[young cove]

and does the graph make sense, based on the signal the rtl-sdr was recording?

[untold needle]

Which graph?

The fft?

[young cove]

the gqrx graph and your fft output graph

[untold needle]

The gqrx one does

I haven't matched gqrx vs fft actually but looking at local radio stations, fft does seem right

[young cove]

they both have a known peak somewhere, yeah, so you can just say, ok that's the offset I need

center or left, whichever works

and your signal was from a beacon at a known frequency, so that's the test

[untold needle]

Oh so just brute force it fine

Yeah I see

Just Try different offsets

That seems perfectly reasonable

Thanks.

Thanks for fixing the complex error too btw

[young cove]

yeah, it's not really brute force, it's just that the system is at a frequency, but the signal is shifted when received, and then you offset it it back up so it matches the scanned frequency band

[untold needle]

You have really helped with everything in the last few days

[untold needle]

Yes

I will try tomorrow, as I have class I should probably sleep considering that it is 23:20

Thanks a lot for your assistance.

[young cove]

i am not an expert by any means, but I use SDR's for recreation. Also I helped a group shift from Matlab to numpy a few jobs ago.

[untold needle]

Oh fine, so you are an expert

🤣🤣🤣

[young cove]

you're doing great, you have to apply the rather theoretical stuff you learned in class to this practical problem. (I am glad you folks didn't build a receiver from scratch)

but I have never actually used FFT for my own devious calculation purposes

[untold needle]

Oh we are still doing that but we have an unlimited time frame now

Unless we decide to return the components

[young cove]

what happened to the competition deadline?

[untold needle]

Oh we are using this rtl sdr for it

Hardware submission by 12th March

Including raspberry pi and code.

[young cove]

got it

[untold needle]

The building receiver from scratch is like a side hobby now just for fun, won't contribute towards competition although we might talk about our pains and struggles during the end presentation

We will likely mention adafruit/external assistance

[young cove]

is this all theoretical or is it going to fly?

[untold needle]

Oh it's flying

It's going in a balloon

30km into thr atmosphere

[young cove]

well, looking forward to hearing about

it

[untold needle]

Definitely, I'll make an announcement about it in this channel here at least

[young cove]

good night and good luck!

[untold needle]

Thats why I also need to automate the pi to run the code when it boots up automatically

[untold needle]

Thanks a lot

[untold needle]

@primal warren Do you have some arduino code for the AD9850 function generator, we are getting dodgy results with our code and oscilliscope.

we have a module

we are getting a flat line

rather than a sine wave

it is extremely dogy

[primal warren]

Have you tried the example code? How did you wire it?

[untold needle]

Well the red LED turned on

It was done on a breadboard

[weak niche]

very dodgy

we rewired it twice i think

[primal warren]

I used a breadboard too, worked fine (however this was an AD9833, I'm unsure what an AD9850 is)

[untold needle]

just different specs

we need to hookup the yagi to the function generator

[primal warren]

So, what do you mean by "extremely dodgy"? That seems different from "we are getting a flat line". Did you connect the FSYNC wire?

When you say "we are getting a flat line", what is your oscilloscope deflection factor? The signal from these boards isn't a high voltage one. I get about 600mV peak to peak from mine.

If you're on something like 5 volts per division, it would look like a flat line.

[untold needle]

Hmmm Ill check with my team as I was coding the fourier at the time.

You could try with @tropic fractal who is part of my team but he's not particularly active

ill check it out though

[tropic fractal]

This was the code we were using today, it's just example code provided on the GitHub page for the library we're using (https://github.com/F4GOJ/AD9850):

const int W_CLK_PIN = 13;
const int FQ_UD_PIN = 8;
const int DATA_PIN = 10;
const int RESET_PIN = 9;

double freq = 1500000;
double trimFreq = 124999500;

int phase = 0;

void setup(){
  DDS.begin(W_CLK_PIN, FQ_UD_PIN, DATA_PIN, RESET_PIN);
  DDS.calibrate(trimFreq);
  DDS.setfreq(freq, phase);
}

void loop(){
  
}```

We made sure to connect the Arduino's pins as described on the library's page and other tutorials, and the connection to the arduino seems to be working. As for the "dodgy" results, the sine wave output was a constant voltage, usually around 30mV, though this could change after restarting the program without any obvious explanation (it tended to vary between 20 and 40mV, so nothing like the 600mV peak to peak that you've been getting @primal warren ). Our oscilloscope has been known to act up, so it could just been an issue with that. Thanks for any help 🙂

[untold needle]

@young cove @granite spear Sorry to bother you guys once again, it appears that my fourier is returning bogus results, here is the code:

import pyfftw
import multiprocessing
import numpy as np
import matplotlib.pyplot as plt

plt.rcParams['figure.figsize'] = [30, 21]

plt.rcParams.update({'font.size' : 10})

#Create a signal with two/multiple frequencies and some noise

dt = 4.16666667e-7 #Time div
center_freq = 92.2e6
n_samples = 24000

with open('time.txt', 'w') as f:
    for i in range(n_samples):
        f.write(f'{dt:.20f}\n')
        dt += 0.000000416666667


with open('time.txt', 'r') as a:
    t = [line.strip() for line in a]

''''with open('voltages.txt', 'r') as a:
    f = [line.strip() for line in a]'''

f = np.loadtxt('voltages.txt', dtype="complex")

#FFT computation

n = len(t)
fhat = pyfftw.interfaces.numpy_fft.fft(f, n)
PSD = fhat * np.conj(fhat) / n
freq = (1/(dt*n)) * np.arange(n)
freq = freq + center_freq
L = np.arange(1, np.floor(n/2), dtype = 'int')

indices = PSD > 40
PSDclean = PSD * indices
fhat = indices * fhat

#Plot

plt.plot(freq[L], PSD[L], color = 'c', linewidth = 2, label = 'Noisy')
plt.plot(freq[L], PSDclean[L], color = 'k', linewidth = 1.5, label = 'Filtered')
plt.xlim(freq[L[0]], freq[L[-1]])
plt.xlabel('Frequency(Hz)', fontsize=12)
plt.ylabel('Power(PSD)', fontsize=12)
plt.legend()

plt.show()

Now I will show you my graph vs GQRX graph for the same values

GQRX graph

my graph, it is clear in gqrx there is no spike at 92.2MHz and it appears that my fourier is only scanning a bad of 100Hz?!?!?! which is really weird

thanks guys @young cove @granite spear

[young cove]

why n = len(t) ? Is n number of samples? In that case why not len(f)?

[untold needle]

I mean that also works

Unfortunately that isnt the error

[young cove]

are they the same?

[untold needle]

yes

they are both 24000

[young cove]

so, in the sample data, are you expecting a spike at 92.2MHz? I don't know what generated the sample data/

[untold needle]

I mean the gqrx one is correct

I am not expecting a spike ther

[young cove]

ok, so it's not sample reception data

[untold needle]

I want my fourier to show the same data as gqrx but mine is dodgy

[untold needle]

no

its something with the fourier or something

[young cove]

how did you calculate dt? I'm surprised it's a numerical constant. I thought you could calculate it

[untold needle]

its not

its just 1/sample rate

surely

[young cove]

I think you could print out the freq values, etc. as you compute them to see if they look reasonable

[untold needle]

hmmm ok

[young cove]

if it's just 1/sample_rate, then do that calculation in the code, then you are assured it is not off by a decimal place, or one too many zeros, etc.

[untold needle]

ok sure

[young cove]

I'm not saying it's wrong, but the fewer mysterious constants there are, the better

[untold needle]

Wait I just wrote a new fourier program with scipy, it is giving me promising results

hmmmmmmmm

its different to gqrx but I think it is still right

@young cove this seems promising but still different to gqrx

import numpy as np
import scipy.fft
from scipy.fft import fft
import matplotlib.pyplot as plt

dt = 0.000000416666667
n_samples = 24000

with open('time.txt', 'w') as f:
    for i in range(n_samples):
        f.write(f'{dt:.20f}\n')
        dt += 0.000000416666667

# Load the voltages and times from file
voltages = np.loadtxt('voltages.txt', dtype=np.complex128)
times = np.loadtxt('time.txt')

# Define the sampling rate and center frequency
sampling_rate = 1 / (times[1] - times[0])
center_freq = 92.2e6
bandwidth = 2e6

# Define the frequency range to compute the FFT over
freqs = np.linspace(center_freq - bandwidth / 2, center_freq + bandwidth / 2, len(voltages))

# Compute the FFT of the voltages
fft_voltages = scipy.fft.fft(voltages)

# Save the results to a file
with open('fft_results.txt', 'w') as f:
    for i in range(len(freqs)):
        f.write('{} {}\n'.format(freqs[i], abs(fft_voltages[i])))

# Plot the results
plt.plot(freqs, abs(fft_voltages))
plt.xlabel('Frequency (Hz)')
plt.ylabel('Magnitude')
plt.show()

[young cove]

why write time.txt and then read it back in? Why not just use a linspace to generate times?

times isn't even used except you compute the sampling rate from the first two values

but the sampling rate is a given (what was it, 2 MHz? I forget)

[untold needle]

2.4MHz

I mean thats just triviality, good practice I guess

@young cove I do believe this fourier is completely correct, I honestly dont understand why GQRX thinks I am wrong

maybe my fourier/gqrx fourier picked up on something that the other didnt?

[young cove]

you could bd supplying data to gqrx in some wrong way

[untold needle]

because rn in the UK, bbc radio 3 is playing and it operates at 92MHz

[untold needle]

I think I have done it correctly with all the right parameters

[young cove]

I think you should start with some sample data that has known stuff in it

[untold needle]

oh so like model a frequency

of like 92MHz

using 2pi*Omega

I see

thats a good shout

[young cove]

yes, or real data that you get from gqrx by connecting it to the rtl-sdr, with known carriers. e.g. ok in this 2MHz part of the FM band, I can hear these stations, so I'd expect to see these carriers in gqrx (spread over the modulation width). Then save that data from gqrx and put it through your FFT

or sample the AM broadcast band. There you'll have fix carriers with sidebands, maybe easier to spot

[untold needle]

how do you save data from gqrx as a file?

[young cove]

i don't know, maybe not possible, but maybe not necessary. You can see those stations in real-time on gqrx. Then you record with pyrtlsdr the same band, and expect to see similar data ("there's Radio 4, there's ...", etc.)

[untold needle]

oh fine ok

[young cove]

the point is just to get some data that you know what it should look like, and then you can vet the processing steps. Or use a signal generator to generate a single carrier in some band, record it with pyrtlsdr (if it's just nearby, it will certainly pick it up), and look for a spike at that freq. Or generate artificial data, as you suggested

[untold needle]

yes good ideas

[untold needle]

As you predicted @young cove I took a live reading with the rtl sdr dongle and got 91MHz strong signal, 91.3MHz strong signal and 90.8MHz some signal whereas the GQRX when I imported a file, was completely messed up

[young cove]

ok, so our import attempts are no good, have you run the data through your fft programs yet ?

i thought the gqrx display looked kinda junky

[untold needle]

well yes, only one of the spikes is 'correct'

It correctly identifies a spike at 90.3MHz but misses the spikes at 91Mhz and 91.3MHz

[young cove]

so maybe your starting frequency is wrong? Like, is it center frequency or lowest frequency? Maybe the spike is from some other place in the FM band

[untold needle]

Im running live gqrx right now

[young cove]

that is 90.3 is not really the signal at 90.3

[untold needle]

The spikes are clearly at 90.3, 91 and 91.3MHz

[untold needle]

I am not understanding

[young cove]

on the live gqrx?

[untold needle]

yes

right now

on the live gqrx

maybe it needs more samples?

[young cove]

ok, so when you record the same band with pyrtlsdr, are you saying that you only see one carrier? YOu should record for several seconds

[untold needle]

several seconds!!!

right now I am doing millisecond

several milliseconds

[young cove]

but since it's FM, you should see _something" all the time, the carrier is always there, but it's modulated and moves around

well how about 0.5 seconds

[untold needle]

that is 1.2 million samples

that is a large text file indeed

[half plover]

For reference, spec analyzer of FM frequencies around me

[untold needle]

thats what I get on gqrx

and my fourier

but gqrx and fourier dont match

[young cove]

don't use the text file. save the binary data with .save(), an then read it in with .load(), instead of .loadtxt()

it will be a lot smaller

no reason to use text

[untold needle]

correctly identifies this FM signal

my fourier

well I have to get the data over from pi ---> laptop

so even using .raw files is large

I'll go up by 1 order of magnitude

[young cove]

can't you do an scp or something?

[untold needle]

scipy?

[young cove]

no, scp network copy program

[untold needle]

no idea

[young cove]

how are you copying it over now?

is the Pi on wifi now?

[untold needle]

copying text file into discord

then copying discord to notepad

then notepad to pycharm

very pedantic

extremely

[young cove]

you can just upload the binary data file to google drive or something

[untold needle]

in theory yes

[young cove]

scp mylocalfile.npy username@remotehost:somename.npy

what is the desktop computer you are using?

is it windows or Linux?

[untold needle]

windows

[young cove]

scp requires ssh server to be enabled on the receiving side

are you running a desktop on the Pi or just command line?

[untold needle]

discord method is superior

14MB

oh my

[untold needle]

desktop

[young cove]

so use Google drive or similar instead, then you'll have an archive of all this, no?

instead of random discord files

and you can upload binary files, they will be much smaller

and faster to upload and download

one sample is 16 bytes in binary and about 55 bytes in ascii

[untold needle]

mhmm

[young cove]

i mean, you can do what you want, I'm just saying how I would do it.

[untold needle]

yeah I'll try that

[young cove]

the peak in the FFT plot above is centered on 9.025 MHz, not 90.3 MHz ??

[untold needle]

this is true

but it spreads over

to 90.2 and 90.3

is it an issue

[young cove]

yeah but off by a factor of 10 in frequency also

not sure you are recording the same thing

the x axis label seems wrong

[untold needle]

factor of 10

no the x axis label is right

910^7 = 90(10^6) = 90MHz

[young cove]

ah ok

is the 90.3MHz station much stronger than the others?

[untold needle]

just plotted the data with 240000 samples rather than 24000 so one order of magnitude higher

it produced the same results

it likes the 90.25MHz

this is very annoying

@young cove I have no idea what the error could be

as my program is insisting on the 90.25MHz

[young cove]

maybe try other band slices to see if you see other FM stations at expected places. I'm still a bit suspicious. Or, expand the y asxi scale so you can see lower peaks more easily. madbodger suggested log scale

[untold needle]

it is extremely suspicious

[young cove]

or just let the 90.3 go off scale and widen 0-100

[untold needle]

how do I use a log scale

[young cove]

i dunno, it's in matplotlib for sure

[untold needle]

ok

got it

[young cove]

your spectrume analyzer photo clearly shows one station as much stronger

[untold needle]

yes it does

ill plot a log scale

woah woah woah

@young cove the log scale did something devious to it

[young cove]

the x axis is different, how come?

[untold needle]

I just increased bandwith

still the same

89.5MHz to 92.5Mhz

3MHz bandwith

[young cove]

is there a station at 89.x?

[untold needle]

let me ask GQRX

@young cove the plot looks very messy now though

[young cove]

lots of high variable noise

reduce the scale maybe

[untold needle]

89.6Mhz

my plot shows 89.8x

very wrong

I am not sure why

it is so wrong

maybe the offset is wrong @young cove ?

[young cove]

I don't know, you could recheck your code.

[untold needle]

I dont think anything is wrong though

[young cove]

does the spectrum analyzer show 90.3 smack dab on 90.3? My impression is that the reported frequencies might be rounded?

in the US, the frequencies are correct, but you might have different spacings. (like the AM band is 9 kHz in UK, and 10 kHZ in the US)

don't know about the FM band

[untold needle]

hmm I cant seem to figure out what the issue is

[young cove]

... but gqrx shows it smack dab on 90.3, right?

[untold needle]

well now its gone

the radio station must have stopped broadcasting

at the time ye

yes

[young cove]

i would say check and simplify your code

[untold needle]

now the 2 main stations broadcasting are 89.6Mhz and 91MHz

[young cove]

you could generate an artificial signal as we discussed. Generate a 90 MHz sinewave or something, or even square wave

if you generate a 1MHz square wave, you can check the harmonics

[untold needle]

if I generated a sin wave I know my fourier works

I've tried it many times

[young cove]

try AM band, carriers don't move

[untold needle]

good cal

[remote pond]

yeah broadcast FM if you're sampling for only a few milliseconds, you're probably picking up the instantaneous audio-frequency deviation of the carrier, not the carrier itself

[untold needle]

I tried like 0.1s too

Ill try this AM band

101-103Mhz band

[young cove]

i mean MW, 500-1600 kHz. AM in MHz is something else, like aircraft, not likely to hear much

[untold needle]

oh like very low

ok

[young cove]

i have to stop for the evening soon

[untold needle]

ok same here

[young cove]

check that gqrx can hear stuff there

[untold needle]

sure

just doing that now

@young cove 24MHz lowest limit

[remote pond]

there might be a WWV signal at 25MHz? haven't checked recently

[untold needle]

just seems like a bunch of noise to be honest

[remote pond]

100MHz is still broadcast FM in the US

[young cove]

tyr without log

also TV stations, but don't know where they are in UK

[remote pond]

oh, they're in the UK? hm, there might be a Finnish time service broadcast at 25MHz too

[untold needle]

This fourier is so mind boggling

@primal warren @granite spear do you guys have any examples of code to plot the PSD of a file or wave as mine is just not working.

[grizzled vapor]

I do have some jank fft code if that helps. It's for an EEG plotter thing

I'll send it later

Maybe you can play the data back slower as audio and feed it into an fft you know works correctly

Then you just need to multiply each frequency by the reciprocal of the slowing down factor I think

And you would have audible verification of the things the fft shows

[grizzled vapor]

Takes a FIFO buffer and runs FFT on it

[untold needle]

The data is an array of complex values in the form a+xj

I'll try it though

[untold needle]

oh my god

@young cove it worked

After 5 hours of trying with the same file and changing the code, I get 2 noticeable peaks at 91 and 91.3MHz which is what I recorded with the gqrx live data

[young cove]

Great! … I am wondering if there is a Nyquist issue here as well. For a bandwidth b, Nyquist says you need to sample at at least 2b to recover the signal and avoid aliasing. But I think you said you were sampling 2 MHz (and wider) at 2.4 MHz Should be at least 4 MHz sampling frequency for 2 MHz bandwidth.

[grizzled vapor]

You don't need phase value. So you don't need complex numbers

You can just use rfft

[untold needle]

Probably, a bandwidth of 1MHz is acceptable though.

[glass jungle]

seems like you have a lot more experience with this than me

I want to use my keyboard to type messages as a kind of point to messenger. is this possible or am i just stuck with using the buttons to send pre-baked messages?

[primal warren]

Yeah, that should be fairly straightforward.

[sage spire]

hey guys, what the range of this radio? https://www.adafruit.com/product/3178

[half plover]

Page said the tested with quarter wave wires at a range of 2km

Direct line of site though

[sage spire]

oh ok thanks!

[normal drift]

The range is highly dependent on the antenna and environment. Trees, hills, buildings, windows have a huge impact. If you need more than 1km you may have to pay attention to these things.

[young cove]

Range has been discussed multiple times in the forums, with some people doing systematic tests:
https://forums.adafruit.com/search.php?keywords=lora+range&terms=all&author=&sc=1&sf=titleonly&sr=topics&sk=t&sd=d&st=0&ch=300&t=0&submit=Search

[sage spire]

Thanks!

[frozen monolith]

sky is the limit. as long as mg is less than 256 characters, it can be anything you want.

[glass jungle]

i asked the wrong question, sry.

do you know of example code that i could look at that takes input from the user's terminal, displays it on the OLED, and then sends it after pressing enter?

i imagine writing the interpreter that will break up a message into 256 character chunks if it goes over the limit wouldn't be too hard would it?

[noble thorn]

I have an ESP32-S3 running CircPy 8.x and I want to connect to wifi on boot.py but then use HttpServer in code.py. Is this possible? HttpServer docs show connecting to wifi in code.py and during soft reloads my ESP32 has to reconnect to wifi all over again.

Right now I connect to wifi on boot, the little TFT screen on the ESP32 confirms it connected.
Then my code.py runs and the top of the TFT says "No IP" and wifi.radio.ipv4_address == None

[normal drift]

I'd recommend ask this in #help-with-circuitpython . There have been recent changes in how the esp32s2 wifi is initialized that may well help you.

[noble thorn]

Thank you @normal drift

[sage spire]

hey guys can this radio acept 5v? https://www.adafruit.com/product/3072

[granite spear]

Yes, it has an onboard regulator to convert down to the 3.3V the chip uses.

[sage spire]

perfect, thanks!

[untold needle]

@granite spear @primal warren So the project is somewhat functional but we have one problem. They won't allow us to put rechargeable lithium ion batteries on thr balloon so is there anyway to power the pi without a UPS hat like for example using some lithium energizer batteries and stripping a micro USB or even some AA batteries. Many Thanks.

[granite spear]

Yep, that's totally possible. You'll probably need some sort of voltage regulator to produce a stable 5V. We used the Energizer lithiums on balloons before, as they're reasonably tolerant of low temperatures.

Do you know how much current the whole system draws? That will set the specs of your needed regulator and battery pack.

[grizzled vapor]

are you allowed to use a power bank?

that is the easiest solution

but also low temps

could do some stuff to it

EdKeyes's advice is good

[untold needle]

well its only the receiver(the pi 3b) so I would estimate a max of 1.8-2A

1.8A is the standard lipo battery hat for the pi 3b

but then again we cant have rechargeable batteres

[granite spear]

The AA ultimate lithiums are rated to 2.5A continuous current, so that should work fairly well. A 4-pack of those would get you around 6V which you could regulate down to 5. That would last probably a few hours depending on what the peak versus average current of the Pi is.

Here's an example buck converter you could use with it: https://www.adafruit.com/product/1385

[untold needle]

doesnt this hookup to the PI's GPIO pins?

but wait

@granite spear could we strip the ends off and shove it into a microusb that has also been stripped?

[granite spear]

I'm not familiar with the Pi pinout off the top of my head... Does it have an unregulated voltage input or only takes USB?

[untold needle]

well microusb definitely preferred but you can hook it up to the GPIO pins

though without regulation that is very risky

[granite spear]

So you'd likely want a regulator like the one linked above. And then you could wire it either directly or through a USB connector.

[untold needle]

exactly

but how do you wire it through a USB connector is my questin

question

just strip the wires

[granite spear]

Yeah, the easiest thing would be to just cut a USB cable apart. Typically the power and ground will be red and black wires inside it.

[untold needle]

exactly

and then just solder together

im guessing

[granite spear]

Yup.

[untold needle]

yeah doesnt seem hard at all

so just connect red and red and then black and black Im guessing

and how do we know which one is ground in this thing and supply

[granite spear]

Yes, that should be correct. Not 100% guaranteed since the colors inside a cable are not actually standardized, but pretty common. Black should be ground.

[untold needle]

yeah usually

do you hookup likes together

like ground --> ground and supply --> supply

[granite spear]

Yes, that's correct.

[untold needle]

yep

then the last step is just to get it to run the python script on bootup

[granite spear]

(And a good general rule. It's pretty rare to have to connect power directly to ground.)

[untold needle]

and there are tutorials for that

[untold needle]

thats true

[untold needle]

Do you think 4 of those lithium energizers will be fine or should we go for 6 instead

https://www.amazon.co.uk/HALJIA-Plastic-Battery-Holder-Storage/dp/B07MSDQHZV/ref=asc_df_B07MSDQHZV/?tag=googshopuk-21&linkCode=df0&hvadid=430825534028&hvpos=&hvnetw=g&hvrand=17822128022478886900&hvpone=&hvptwo=&hvqmt=&hvdev=m&hvdvcmdl=&hvlocint=&hvlocphy=9045970&hvtargid=pla-737420469011&psc=1&tag=&ref=&adgrpid=104042186550&hvpone=&hvptwo=&hvadid=430825534028&hvpos=&hvnetw=g&hvrand=17822128022478886900&hvqmt=&hvdev=m&hvdvcmdl=&hvlocint=&hvlocphy=9045970&hvtargid=pla-737420469011

This seems good

[untold needle]

ok that problem is fixed now

@granite spear @primal warren @young cove Sorry to bother you guys, for some reason, I am trying to run the python script on startup but it simply isnt working. I have tried using both crontab and rc.local but neither of them appear to work at the moment.

I am trying systemd right now

[untold needle]

ok I see the error it is throwing

Traceback (most recent call last):
File 'home/pi/Radio.py', line 13 in <module>
np.save('samples.npy', samples, allow_picle = True, fix_imports = True)
File '<_array_function__internals>', line 5, in save
File '/usr/lib/python3/dist-packages/numpy/lib/npyio.py', line 524, in save
file_ctx = open(file, 'wb')
PermissionError : [Errno 13] Permission denied: 'samples.npy'

Mar 14 12:41:49 raspberrypi python3[1339]: Found Rafael Micro R820T tuner
Mar 14 12:41:50 raspberrypi python3[1339]: [R82XX] PLL not locked!
Mar 14 12:41:50 raspberrypi python3[1339]: Traceback (most recent call last):
Mar 14 12:41:50 raspberrypi python3[1339]: File "/home/pi/Radio.py", line 13, in <module>
Mar 14 12:41:50 raspberrypi python3[1339]: np.save('samples.npy', samples, allow_pickle = True, fix_imports = True)
Mar 14 12:41:50 raspberrypi python3[1339]: File "<array_function internals>", line 5, in save
Mar 14 12:41:50 raspberrypi python3[1339]: File "/usr/lib/python3/dist-packages/numpy/lib/npyio.py", line 524, in save
Mar 14 12:41:50 raspberrypi python3[1339]: file_ctx = open(file, "wb")
Mar 14 12:41:50 raspberrypi python3[1339]: PermissionError: [Errno 13] Permission denied: 'samples.npy'
Mar 14 12:41:50 raspberrypi python3[1339]: Reattached kernel driver
Mar 14 12:41:51 raspberrypi systemd[1]: Radio.service: Main process exited, code=exited, status=1/FAILURE
Mar 14 12:41:51 raspberrypi systemd[1]: Radio.service: Failed with result 'exit-code'.
Mar 14 12:41:51 raspberrypi systemd[1]: Radio.service: Consumed 3.449s CPU time.
Mar 14 12:41:51 raspberrypi systemd[1]: Radio.service: Scheduled restart job, restart counter is at 7.
Mar 14 12:41:51 raspberrypi systemd[1]: Stopped Radio Service.
Mar 14 12:41:51 raspberrypi systemd[1]: Radio.service: Consumed 3.449s CPU time.
Mar 14 12:41:51 raspberrypi systemd[1]: Started Radio Service.

[young cove]

rc.local or systemd run as root. You can run as the the pi user instead. (Though as root it should just run.)

[untold needle]

exactly, it should run

[young cove]

https://forums.raspberrypi.com/viewtopic.php?t=314455

[untold needle]

I'm getting the above error though

Its something to do with it not being able to access the samples.npy file

[young cove]

https://forums.raspberrypi.com/viewtopic.php?t=314455

[untold needle]

@young cove it should be working and the above error message shows that it was trying to work but its something to do with the samples.npy file needing access or something

[young cove]

the home directory for root is not /home/pi

so it may be looking in the wrong place

more direct pointer to the guide above: https://github.com/thagrol/Guides/blob/main/boot.pdf

[untold needle]

well it wasnt actually /home/pi

it was /home/aryan

[young cove]

could you upload the script, using the + sign?

[untold needle]

aryan is the name of the account

with root access

[young cove]

you can run as a user other than root. See section 4.2.4.3 in the guide

[untold needle]

@young cove well the issue is with saving it to the file

the program is running

but it cant save to samples.npy file

[young cove]

in rc.local or whatever put in sudo -u aryan -i name/of/the/script. I think you want to & it or it will hold up the boot process.

[untold needle]

some permissions error or something

[untold needle]

sudo -u aryan -i Radio.py &

[young cove]

then it will run as aryan and you don't have to worry about what happens when it runs as root. If you run it too early perhaps the filesystem hasn't been mounted r/w yet

[untold needle]

but whats the error with the samples.npy

[young cove]

is that what you are doing now?

[untold needle]

@young cove Thats what I will add now

oh wait

[young cove]

reference it with a full path: /home/aryan/Radio.py

[untold needle]

I made the rc.local file executable

ok

I made it executable though

sudo chmod +x /etc/rc.local

so I cant write to it now

@young cove what should I do in systemd

to stop the samples.npy issue

[young cove]

better: sudo -u aryan -i python3 Radio.py

why did you make it executable?

[untold needle]

thats what a guide told me

I would rather do it in systemd as crontab and rc.local failed

[young cove]

if you arent using /etc/rc.local remote it

[untold needle]

sure, so remove /etc/rc.local and delete the crontab extra lines I added

what should I do in systemd then

[young cove]

I would say don't use systemd -- use plain crontab instead. Remove what you added to systemd. See the guide above, section 4.3: Using cron

[untold needle]

ok fine

but crontab is failing for me

it isnt working

[young cove]

crontab -e
# add this line when editing crontab
@reboot sudo -u aryan -i python3 Radio.py &

not sure if the & is needed here or not. I'll try to find out

[untold needle]

oh ok fine

[young cove]

oh wait, that's not necessary, holdon , don't do anything yet

[untold needle]

thing is @young cove I cant see the errors crontab is throwing but with systemd I know what is going wrong(in this instance it cant save to the samples.npy file) but in crontab I dont know what is going wron

[untold needle]

ok

do you want to know what I currently have in my crontab file @young cove

[young cove]

remove what is in your crontab file, remove /etc/rc.local

undo what you did with systemd. Get it back to nothing

I haven't done this in about 20 years, gotta study it a bit

[untold needle]

ok

[untold needle]

how does one remove /etc/rc.local, ok I disabled the systemd thing

https://www.youtube.com/watch?v=Gl9HS7-H0mI

I used this guy

for crontab

[young cove]

ok, don't remove it. I can't watch videos for this. I read.

[untold needle]

I ran my file in the terminal and it worked perfectly

[young cove]

don't trust the videos

[untold needle]

Ok fair enough

[young cove]

takes too long to watch a video

[untold needle]

Thats true

[young cove]

what is in rc.local now?

[untold needle]

dont know as its executable now

but when I execute the /etc/rc.local file it works

[young cove]

sudo cat /etc/rc.local

[untold needle]

but it doesnt work on bootup

[young cove]

sudo chmod -x /etc/rc.local will undo the executable

wow there is a lot of bad advcie out there

do this:

[untold needle]

yes there is so much bad advice everywhere

[young cove]

sudo cat /etc/rc.local what does it say?

[untold needle]

it says exactly what I added into the file

its the standard file

but then before the exit0 line

[young cove]

was it empty before or did it not exist?

please just show me what's in the file

[untold needle]

no it always existed

ok sure

[young cove]

I don't have an RPi to boot up right now

[untold needle]

sending now

discord loading on rpi

#!/bin/sh -e

rc.local

This script is executed at the end of each multiuser runlevel.

Make sure that the script will "exit 0" on success or any other

value on error.

In order to enable or disable this script just change the execution

bits.

By default this script does nothing.

Print the IP address

_IP=$(hostname -I) || true
if [ "$_IP" ]; then
printf "My IP address is %s\n" "$_IP"
fi

python3 /home/aryan/Radio.py &
exit 0

@tight sail

@young cove

[young cove]

change that to sudo -u aryan -i python3 /home/aryan/Radio.py &

that will run Radio.py as you instead of as root

[untold needle]

ok

[young cove]

did you add the "My IP address" stuff, or was it in there?

[untold needle]

already in there

@young cove ok I did that

[young cove]

ok, reboot

[untold needle]

alright

ok I did a sudo reboot

now waiting

[young cove]

what is your deadline on this? >March 12 now

[untold needle]

oh they gave us an extension until the actual launch date

so until monday I guess

we had to fill in a waiver for an extension

it was very annoying

but we are still tight on time

[young cove]

can you get another team member who is a sysadmin type person?

[untold needle]

I will be in belgium by next tuesday

[untold needle]

none of them are sysadmin type people

[young cove]

in the whole university?

[untold needle]

@young cove ok so I just checked the last time the samples.npy file was written to and it says 12:04 and currently it is 13:25 for me

this means that it didnt work

[young cove]

does it print "My IP address is ..." to the syslog?

[untold needle]

how do I check the syslog

[young cove]

look in /var/log/syslog

[untold needle]

ok

[young cove]

i have to get set up here to try this myself. I'll be back in a bit

[untold needle]

oh wait it was running systemd

should I disable the systemd program

[young cove]

don't disable all of systemd. Just disable what you added.

how did you add Radio.py to systemd?

[untold needle]

https://www.thedigitalpictureframe.com/ultimate-guide-systemd-autostart-scripts-raspberry-pi/

this is how

[untold needle]

yep did that

Ill try rebooting again

I also removed the stuff I added to crontab

ok Ill try again

[untold needle]

@young cove ok nothing in syslog

not even IP address

earlier systemd was showing something

but now i disabled my systemd service

nothing is happening

[young cove]

i am trying a simple test, will be a few more minutes

[untold needle]

did it work

[young cove]

not yet, but I'm getting closer

the problem is getting the output monitored

still working, i'll be back

[untold needle]

what are you trying?

running an arbitrary program on pi startup?

[young cove]

rc.local, it's working, the issue is making sure it keeps running

[untold needle]

the closest I have gotten is with systemd

[untold needle]

its working?

[young cove]

halfway

[untold needle]

use an ampersand(&)

at the end of the line

[young cove]

of course, that's not the issue right now

just give me more time

[untold needle]

oh ok sure

@young cove my systemd was working but the issue was that it didnt have permission or something to save to the samples.npy file

[young cove]

systemd is large and confusing and a mess

[untold needle]

fair enough

my rc.local and crontab have shown nothing in the syslog either though

and have no evidence of working

[young cove]

right, that's what I'm trying to fix. using shell redirection into a file properly. It's partly working, but there's no output in the file, as if the script has stopped running. debugging that

[untold needle]

ok

[untold needle]

I see, my rc.local and crontab just fail and I have no idea why

@young cove maybe add an echo to see if its working or not

[young cove]

ok, problem was that python was buffering the output, so it appeared there wasn't any. Here's what I think will work for you in rc.local:

sudo -u aryan -i /bin/bash -c 'python3 -u Radio.py >run.log 2>&1 &'

I'm using /bin/bash so I can redirect the output. The -u causes python3 not to buffer the output, so you can see it line by line in run.log. You can change the name of run.log, of course.

this is tested on an RPi

[untold needle]

ok sure

@young cove have a look at this forum too

https://raspberrypi.stackexchange.com/questions/107933/cannot-run-2-apps-with-rc-local-but-it-works-from-command-line

it says rc.local is deprecated

[young cove]

i don't care if it's deprecated, it works

[untold needle]

alright Ill try it

I also dont care as long as it works

[young cove]

you can try it with systemd later, but if you get this working, you know you have something that works

in the long run, yes, you should use something else, but you have a short time here

[untold needle]

yep

I agree

[young cove]

get it working, and then comment it out with # (so you can put it back), and try similar from systemd. But you have to work out how to run as aryan from systemd as well. crontab might be better to run as yourself.

[untold needle]

I am rebooting now I hope this works

[young cove]

systemd is "after my time", I haven't had to use it for anything

[brave willow]

You'd put User=aryan in the [Service] section of the systemd .service file and it'll do it

[untold needle]

should I make it executable or not?

[young cove]

rc.local does not need to be executable

[untold needle]

yeah the systemd was working but I was getting some sort of permission error

ok I'm rebooting now

[young cove]

thanks - if you have further suggestions later, that would be great

[untold needle]

Mar 14 12:41:49 raspberrypi python3[1339]: Found Rafael Micro R820T tuner
Mar 14 12:41:50 raspberrypi python3[1339]: [R82XX] PLL not locked!
Mar 14 12:41:50 raspberrypi python3[1339]: Traceback (most recent call last):
Mar 14 12:41:50 raspberrypi python3[1339]: File "/home/aryan/Radio.py", line 13, in <module>
Mar 14 12:41:50 raspberrypi python3[1339]: np.save('samples.npy', samples, allow_pickle = True, fix_imports = True)
Mar 14 12:41:50 raspberrypi python3[1339]: File "<array_function internals>", line 5, in save
Mar 14 12:41:50 raspberrypi python3[1339]: File "/usr/lib/python3/dist-packages/numpy/lib/npyio.py", line 524, in save
Mar 14 12:41:50 raspberrypi python3[1339]: file_ctx = open(file, "wb")
Mar 14 12:41:50 raspberrypi python3[1339]: PermissionError: [Errno 13] Permission denied: 'samples.npy'
Mar 14 12:41:50 raspberrypi python3[1339]: Reattached kernel driver
Mar 14 12:41:51 raspberrypi systemd[1]: Radio.service: Main process exited, code=exited, status=1/FAILURE
Mar 14 12:41:51 raspberrypi systemd[1]: Radio.service: Failed with result 'exit-code'.
Mar 14 12:41:51 raspberrypi systemd[1]: Radio.service: Consumed 3.449s CPU time.
Mar 14 12:41:51 raspberrypi systemd[1]: Radio.service: Scheduled restart job, restart counter is at 7.
Mar 14 12:41:51 raspberrypi systemd[1]: Stopped Radio Service.
Mar 14 12:41:51 raspberrypi systemd[1]: Radio.service: Consumed 3.449s CPU time.
Mar 14 12:41:51 raspberrypi systemd[1]: Started Radio Service.

[young cove]

i have been using unix for 45 years, things change, and I don't have the same needs at the moment

[untold needle]

this was my systemd error

[untold needle]

ok I rebooted

lets see the last time the file was written to

@young cove nope

failed again

[young cove]

there should be a log in run.log

[untold needle]

last time samples.npy was written to was 13:43 and it is 14:31 for me

[untold needle]

the syslog?

[young cove]

no, I gave you a command to put the output from Radio.py into run.log

[untold needle]

yes

how do I check run.log

[young cove]

is there a run.log?

just more it

[untold needle]

more?

[young cove]

or less it or cat it

it's just a text file

[untold needle]

oh fine

ok

i did sudo cat run.log

@young cove cat: run.log: No such file or directory

[young cove]

show me the line you added to /etc/rc.local

[untold needle]

ok

[young cove]

cat /etc/rc.local

[untold needle]

yep

[young cove]

@brave willow if you are available to continue, I have to do other things

[untold needle]

sudo -u aryan -i /bin/bash -c 'python3 -u Radio.py >run.log 2>&1 &'

[young cove]

that looks ok

[brave willow]

That's going to try to write run.log in whatever directory you might be in when that script starts

[young cove]

right, but -i -u should get it into aryayn's homedir

[brave willow]

-i is an interactive shell, is it not?

[young cove]

it forces running .bash_profile, etc. that's what I want here, so the $PATH will be right

[brave willow]

Might be worth just putting an absolute path for the run.log to see if that works. I'm not sure you want an interactive shell from rc.local

[young cove]

we could also do sudo -D

it worked in a local test for me

here is how I am testing:
in /etc/rc.local:

sudo -u pi -i /bin/bash -c 'python3 -u test.py >test.output 2>&1 &'

test.py is:

import time

while True:
    print(time.time())
    time.sleep(5)

I reboot, and ps shows python3 running, and I see output in /home/pi/test.output

@untold needle you could try a simple test.py like the above instead of Radio.py for now

[untold needle]

ok

[untold needle]

there is an output file?

I feel like with systemd I was very close

it was just some sort of permissions error

[young cove]

>test.output is writing the outputfile.

[untold needle]

This startup pgoram is confusing me so much

program

[young cove]

@brave willow if you can suggest how to do this with systemd, that woudl be great

[brave willow]

@untold needle post your .service file, and double-check the output path for the output file

[young cove]

must run as aryan, in aryan's homedir, and be spawned, because it runs forever

thank you

[untold needle]

sure

[young cove]

@untold needle so comment it out in /etc/rc.local before you reboot

[untold needle]

yep

[Unit]

Description=Radio Service
After=multi-user.target

[Service]

Type=idle
User=aryan
ExecStart=/usr/bin/python3 /home/aryan/Radio.py
Restart=always

[Install]
WantedBy=multi-user.target

@brave willow

do you want the log/error message as well

[brave willow]

Should user be aryan instead?

I think you need to specify WorkingDirectory= and put your home dir in there as well

[untold needle]

my bad

[brave willow]

You can use WorkingDirectory=~to make it auto go to your home dir

[untold needle]

@brave willow ok just edited that

where do you add the WorkingDirectory part?

[brave willow]

In the [Service] section

[untold needle]

ok

[untold needle]

wait so specify my homedir

so /home/aryan

[brave willow]

So if you are using User=aryan, putting WorkingDirectory=~ will make it run out of aryan's home directory automatically

[untold needle]

now I just need to find the curly thing after the equals sign on my keyboard

[brave willow]

Oh, it's a tilde, on US/ANSI keyboards it's next to the 1 key

[untold needle]

my keyboard doesnt have it

should I just specify the homedir

[brave willow]

That should work for testing

[untold needle]

testing?

[untold needle]

what testing

[brave willow]

Oh, I meant you could go back and change it to ~ later 🙂

[untold needle]

oh ok

[untold needle]

thanks, is there anyway I can give my service root access or something

as I think that is the problem potentially

[brave willow]

You can run it as root if you want

[untold needle]

how do I do that?

[brave willow]

I can't remember if you use User=root or not- might have to google on that one

Or if omitting it makes it run as root automatically

[untold needle]

I think omitting makes it run as root automatically

[brave willow]

So then you would need WorkingDirectory=/home/aryan instead of the ~ in that case

[untold needle]

yep

[untold needle]

yes I see

wait a second

@brave willow I think it might have worked

it says the last time samples.npy was written to was 14:58 which is the time right now for me

It might have actually worked

this is a shocking revelation

[brave willow]

Woohoo!

[untold needle]

I am skeptical as with all things that include command line tools

@brave willow I am 80% sure it is working now

but radio.service isnt stopping running

it is writing to samples.npy literally every minute

the last time samples.npy was written to literally just now

It is writing to it every 10-20 seconds @brave willow thats weird

[brave willow]

I think someone mentioned the writes are being buffered?

[untold needle]

maybe

its spam writing to it

this perhaps proves that systemd is superior to crontab and everything else

oh it works

@brave willow it works perfectly

it actually writes to the file

but it writes to it like every 5-10 seconds

which probably isnt ideal

is there anyway to add a delay @brave willow

probably

[brave willow]

Maybe, I'm not sure where that buffering is happening

Usually it defaults to that so you don't murder your I/O

[untold needle]

aha there is

@brave willow RestartSec=3600

you add this line

in seconds

so if I make that 60

it writes every 60 seconds

you add that under service I think

[brave willow]

Oh, you want the script itself to get auto-restarted every 60 seconds

?

[young cove]

may be easier to make Radio.py be an infinite loop and put in a time.sleep() rather than restart it

[untold needle]

this is a very good point actually

[brave willow]

Sorry, I assumed that was the case already

[untold needle]

nope

[brave willow]

Yeah I'd do that for sure

[untold needle]

I dont know what happened actually

it just randomly started working

thats really weird

maybe it was the working directory thing

this definitely confirms systemd supremacy

[brave willow]

It's a beast but is generally pretty good at what it does

[untold needle]

at least now I'm back in my comfort zone with just having to change the python code

I'm not so familiar with sysadmin and command line tools

plus I've been ill for the last few days including today so I've missed some class and doing work with a raging headache isnt particularly easy

[young cove]

ty @brave willow I was much more fmailiar with sysv init

[brave willow]

Certainly nothing wrong with sysv init- at least not until different distros started doing it differently/using different directories to make it more complicated

[young cove]

yes, it got out of hand. Not sure systemd was the best solution, amazing about the tension it created

[untold needle]

tension?

[young cove]

people declared they would never use systemd. They started new distributions rather than using it (devuan, etc.)

[untold needle]

I was getting very stressed about it, I knew I was close with systemd to be fair, I was only missing one line it seems

[untold needle]

there is so much bad advice out there

[brave willow]

systemd is pretty heavy-handed but I feel like it's in a good place now with what it provides- all the auto restart and logging features are really nice

[untold needle]

logging features are very good

[young cove]

it is very non-unix-y, which bothered people

[untold needle]

systemd>crontab>rc.local

in my opinion

from what I have tried

crontab and rc.local just flat out failed for me

[brave willow]

I mean, each has its use- crontab is good for periodic jobs for sure. I've used rc.local even fairly recently, but having to reboot every time to test changes is a real drag. Plus yeah, it is being deprecated, though I imagine it'll continue to work for the foreseeable future

[spiral canyon]

Anybody know what sort of rf transmitter chips would be in kinetic switches? (would they be special, or just normal ones?)
https://youtu.be/9Pw7U0XFgUM?t=312

[granite spear]

I think they'd generally be fairly normal sub-GHz transmitters, though likely to be on the simple side instead of having a complicated protocol like Bluetooth... just power up, spit out a few ID bits on a carrier, then power down again. The receiver would be continually listening since it has power all the time.

[untold needle]

Today we built a voltage regulator of 5V to power the pi. It only provides like 1-1.5A though

hopefully thats enough for pi+dongle

it should be

we need to add a heat sink though

[granite spear]

It probably won't matter for the short flight, but note that convection-based heat sinks become less effective at high altitudes because the air isn't as dense.

[untold needle]

Thats true

Its not that short

It's like 3 hours

[spiral canyon]

Not sure how high you're going, but I felt inspired to make this.

[grizzled vapor]

You almost definitely want a buck converter

[untold needle]

teacher said voltage regulator is fine, and tbf it is working(even though I would prefer a buck converter), only issue is heat

[grizzled vapor]

Heat means inefficiency, which means less runtime. Also, that reg would be throwing out probably 4.5W with a 1.5A load. 4.5W can get that reg really hot

Dissipation in watts is calculated by (voltage drop across regulator*amps drawn from reg)

[untold needle]

yeah of course I calculated that

P=IV

however

"

@grizzled vapor with some calculations I thinks its likely to be more like 800mA-1A so like 3.2-4W which is still a lot

I'm sure we have heat sinks that can account for that much at least

pretty high; 30km upwards

[grizzled vapor]

You would need a beefy heatsink to keep that from overheating

At that point it may more of become a slug of metal with enough thermal capacity to prevent overheating during the flight

There are cheap buck converter modules on amazon

[primal warren]

Or reliable ones from Pololu

[untold needle]

what do you think of a heat sink?

[grizzled vapor]

Also it's cold up at 30km

[untold needle]

How beefy a heat sink are we talking @grizzled vapor

[untold needle]

so surely thats good isnt it?

[grizzled vapor]

Maybe

A lot of components are rated for -40 to 125 degrees Celsius

But vreg will still get really hot

Maybe you can stick an old laptop cooler on it

But that's overkill when you could just use a buck

[untold needle]

I asked him to get a buck but my professor/teacher has some aversion to it

also if we get a buck it has to arrive literally in 2 days

[grizzled vapor]

You might be able to parallel linear regulators

Then each one gets less hot

[untold needle]

https://thepihut.com/products/ubec-dc-dc-step-down-buck-converter-5v-3a-output

This?

we can feed it 6V@2A or 1.5A using 4 lithium energizers in a battery pack

[grizzled vapor]

That will work

[untold needle]

I think me or my friend will probably buy it ourselves to save time rather than emailing the professor/teacher to buy it for us, plus he doesnt like it for some reason

[primal warren]

Buck regulators can interfere with radio receivers, which may be why your professor wants to avoid them. However, a little separation, copper tape shielding, and ferrites can keep the interference from your receiver

[untold needle]

copper tape shielding? And where/how do we add ferrites. And if it is possible could we use this current LM7805 Voltage regulator with a heatsink

[primal warren]

You'd have to deal with the heat dissipation and reduced battery life with a linear regulator like that, as people have explained

[untold needle]

Well the buck converter certainly seemed better until you said it interferes with radio. How do you keep the interferece away like you mentioned. Where do we add the ferrites/copper tape shielding and what is separation

[granite spear]

Heh heh, maybe you'll pick up the buck frequency in your FFT, so you know that part works. 😉

[grizzled vapor]

Ferrite beads should go in between the buck and the Pi on the power line. Maybe add some decoupling caps after ferrites for more filtering

For copper tape/aluminum foil, just connect it to ground and wrap it around the buck

I think that's right

[untold needle]

thanks, @primal warren can you explain how to connect the ferrite beads and if I should add the tape to the converter or the outside of the dongle

[wary arrow]

What is gain of a receiver antenna?

[untold needle]

depends on which one

[untold needle]

hwo bad is the interference? Also do we need copper tape shielding in the buck or on the dongle, the dongle we have is pretty high quality so is already in a metal casing. Also I am interested in how we add these ferrites, and since we might not have those, is there anything else we can do as well.

[wary arrow]

I mean in general, just trying to understand

[granite spear]

It's usually a measurement of how much better or worse it is than a simple dipole receiver. An antenna with high gain would generally be directional, like a dish, so the gain tells you how much more power you can get when you point it correctly towards the source.

[wary arrow]

Ahh thanks!

I wish old school Radio Shack stores were still around

[granite spear]

I actually saw one recently. Not very big, but it did still exist and have a small section of random components, too. 😁

[grizzled vapor]

Depends on frequency, but if the buck switching frequency is not near your target receiving frequency, it's fine. Maybe add a filter inductor/bead and some caps just in case

Aluminum foil shielding won't hurt. I think it should just be wrapped around the buck to reduce radiated interference

And grounded

[untold needle]

I mean receiver is 27.5MHz with a 1MH, bandwidth

[grizzled vapor]

Well

[untold needle]

How do you ground foil

[grizzled vapor]

That's a lot of difference

Shouldn't be much

Interference

Actually rtlsdr won't even pick up 1mhz

[untold needle]

Yeah I didnt think so either

[grizzled vapor]

Testing is the best way

[untold needle]

rtlsdr can receive 1MHz

[untold needle]

I can test on Monday when the buck arrives I guess

[grizzled vapor]

Connect to negative

[untold needle]

Of the buck?

[grizzled vapor]

Kind of a tinfoil hat for your components lolol

[untold needle]

Hmm

Hopefully it doesn't need anything

[grizzled vapor]

Yes

[untold needle]

27MHz is pretty rare

[untold needle]

By the way @primal warren @young cove @granite spear we are launching it tomorrow, the only problem would potentially be if the transmitter would reach the ionosphere or not and then reflect, so range is the only issue. All the testing is done and we've given the pi in to be integrated into the balloon. Just wanted to update you guys.

[granite spear]

Good luck! 🤞

[untold needle]

the transmitter is only 15mW, is that strong enough?

to reach the ionosphere and back

we dont know the gain

its a yagi antenna though

homemade with some long booms

[primal warren]

That depends on your antenna gains, bounce loss, path loss, and (especially) receiver sensitivity. https://resources.system-analysis.cadence.com/blog/rf-link-budget-calculation-guide

Since you're not pushing information over the link, I suspect you can use averaging and filtering techniques to improve your link budget.

[untold needle]

how much power do we need to hit the ionosphere and back

[remote pond]

what part of the ionosphere? generally, you would do a link budget calculation, and you can look up various rough guidelines. is the balloon going to reach the ionosphere, or are you counting on somehow distinguishing an ionospheric reflection from the direct path from the transmitter?

[untold needle]

using modulation for the last part yes

@tropic fractal came up with that idea

[granite spear]

I'd ordinarily expect "watts" rather than milliwatts for that kind of distance at the least, though. Like ham radio sorts of gear.

[grizzled vapor]

Although things like WSPR beacons only have miliwatts of transmit power, but the signal can be picked up relatively far away if the ionosphere is good

Also because it uses FSK and a very low bitrate

[young cove]

i use WSPR with 1W and get heard 8000km away with a non-fancy antenna, and that is with ionosphere skip. Since this is line of sight, at least to begin with, it may work out

[primal warren]

Like I said, it depends on antenna gain, receiver sensitivity, bandwidth, modulation, sample time, etc.

[wary arrow]

I read somewhere that the ideal antenna length is 1/4th the wavelength. If I wanted to have a really tiny antenna (like a pinhead) for a 450MHz signal, would I need a lot of power to make it transmit?

[granite spear]

You would probably want a "chip antenna" or a coil in that case... the same effective length but wrapped up into a smaller space. Though honestly a literal pinhead size becomes pretty infeasible.

Like, the wire trace to connect to the pinhead becomes a better antenna than the pinhead itself.

[wary arrow]

Oh the antenna doesn’t need to be straight, I didn’t realize 😬

[wary arrow]

Maybe I want a spiral antenna

[primal warren]

Not pinhead size, but this is the general idea https://www.adafruit.com/product/4394

[untold needle]

Ok we launched today

From Belgium we are recovering the balloon from somewhere in the Netherlands or Germany

It's all intact

Anyways I will give you guys results after we analyse

I may have forgotten to let the antenna dangle loose, it might be taped to the side, I may have forgotten before launch. @granite spear @primal warren @young cove what will this do to it

[granite spear]

Is it coiled up?

[primal warren]

It will tend to reduce the gain and distort the pattern, depending on the shape it's in and what it's near

[untold needle]

Sort of

[untold needle]

How bad are we talking

[primal warren]

Standard engineering answer: it depends.

[untold needle]

It's coiled with a piece of tape sticking it to the exterior

[primal warren]

Coiled how tightly? What radius? What spacing between turns? Flat coil or conical? Is it near anything? Is that anything conductive? Is that anything radio circuitry?

[untold needle]

Well all circuitry is on inside and dipole on outside

It's coiled like just normally

It's a flat coil

Not very big radius

[primal warren]

I'm going to guess a 24-31dB loss

[untold needle]

Are you joking

[primal warren]

Yes. Yes I am.

[untold needle]

Oh ok

That's good

Actually how bad it is

[primal warren]

I simply don't know enough about the situation to make even an educated guess, other than "it will tend to reduce the gain"

[untold needle]

My team is furious with me

[untold needle]

How much do you actually estimate

It's a flat coil, 1.75m of wire, 7cm radius but together at the centre, its a flat coil and taped at the centre

@primal warren @granite spear please help

My team will kill me

Tightly coiled

[primal warren]

I'm not sure how we can help at this point. As Julius Caesar put it "Alea iacta est." It will be what it will be. Collect the data and hope for the best.

[untold needle]

How bad do you think it is

Is it very bad

Or medium bad

Or minor issue

[primal warren]

There are a few possible viewpoints here, and they're all guesswork. One is that you're already flying a bunch of untried technology, and it may have been marginal to start with. If so, it's still going to be marginal, so in that depressing scenario, it isn't much worse. The other is the "more signal, better results" mindset, and in this case, you're going to miss out on "more signal", so any results you do get will tend to be impacted.

[untold needle]

Well we don't actually know if the transmitter was ever going to work

But my team don't think if it like that

They essentially think it's all my fault now

Unless we get data

[granite spear]

If there's an easy way to boost the transmit power, that might be something to consider.

[young cove]

did you test the transmitter and receiver on the ground at all?

[untold needle]

Yep

[primal warren]

Unfortunately, that's true, and it's a long-standing problem. Like the person who misses one basket gets blamed for losing the whole game.

[untold needle]

Too late

Payload been recovered

[untold needle]

We tested it in the same room

And it all worked obviously

But transmitting in the same room ≠ doing a skip off the ionosphere

Even my professor is angry

RIP

[granite spear]

That's good! The first time I tried getting sky-brightness photos from a balloon, it landed in a swamp. Recovery isn't easy sometimes, heh heh.

[young cove]

i think it's a bit crazy you only get one chance at a launch. If this were other than a student project there would be tons of testing, and multiple runs.

so did the recovery get the balloon too?

[granite spear]

The balloon probably just popped.

[untold needle]

No everything was recovered

[young cove]

so was the tape still on the balloon

[untold needle]

The people who launch it have been doing it for years, they are very advanced

[untold needle]

Yes

[young cove]

i wouldn't take this as a personal failure, it is a system failure; the launch procedure needs a checklist and cross-checks

[untold needle]

Recovery went well, from Belgium it went to the Netherlands/Germany and they recovered it.

But now our results.m...

[untold needle]

My team don't see it that way

How bad is the damage though

[young cove]

you just have to wait and see the data.

no chance to try it again?

[granite spear]

Your team is who should have been doing the cross-checks...

[young cove]

when commercial pilots prepare for a flight, they have checklists. It's not just seat of the pants

[untold needle]

In summer maybe

[untold needle]

Well it's too late now and it was my fault

[untold needle]

We should have had a checklist

[young cove]

this is a learning experience. my impression is that you did about 80% of the work. There was a lot of success: the transmitter was built, the receiver was built, the software was developed.

[untold needle]

And the thing is, it's more than likely the transmitter didn't work but we will never know because of my mistake

[young cove]

and you still don't know about the data, don't beat yourself up.

I thought you said you tested it in the same room?

[untold needle]

That's true, that will come later

[untold needle]

We don't know if it works for long range

Just 2 pieces of wire can transmit short range

We never tested a skip off the ionosphere

[young cove]

so there's a missing thing, which is that you should have done a rrange check outside across a considerable distance. so whose fault is that.

[untold needle]

Probably mine as a sort of team leader

That's so poor from me

[young cove]

how do you know it skips? it could be line of sight. I'm confused why you thought it had to be ionosphere bounce

The essence of learning is making mistakes.

[untold needle]

To be fair the dipole was actually not properly in the terminal block today, one of the wires came out and I saw that right before launch and quickly fixed that. Well it could be line of sight but we would like it to be a skip which is why we modulated.

[untold needle]

Professor supposedly said we wasted time and money as well as resources

I left as I have something to do back in the UK

[young cove]

i think you need to expect more from your team as well. Part of this learning experience is about the management structures and team cooperation that make for success

[untold needle]

So I won't be there for data collection

I gave them all the code though

Literally all they have to do is give my code a directory from the pi and it will create the graphs for them and display them one by one

If we don't get data I'm going to be blamed, if we get some data but its bad then I'm going to be blamed. If we get good data then everything is all right

[young cove]

there are many things that can go wrong here. For the professor to blame you alone is short-sighted. The professor should not expect success. The professor should also be guiding you toward success. It sounds like the professor does not necessarily have as much systems engineering experience as needed.

[young cove]

I think the transmitter people should have proposed that

[wary arrow]

“Sink or swim” is not a good way to teach, except in special situations

[grizzled vapor]

@untold needle What organization is the high altitude balloon project for?

[untold needle]

Not sure

@young cove @granite spear @primal warren only a few of the files have data, but a lot of them are showing concurrent data at 27.1-27.15MHz but its extremely strong. We must have caught a radio station for sure as it looks like it has proper shape and everything.

[young cove]

that frequency band is part of "CB radio", which is used extensively in many countries

so you could easily have been picking up CB'ers

[untold needle]

while everything else is just like noise, it is so strong

look at that

its unreal strength

[young cove]

some CB'ers use illegal amounts of power

what freq was your transmitter using?

[untold needle]

its like -70dB, so strong

[untold needle]

cycling through 27.2-27.8MHz

I dont think we got much from our transmitter

[untold needle]

likely we got a CBer

[granite spear]

Congrats! A whole lot had to go right in the system design for you to get real data like that! Pat yourself on the back... 😁