help-with-radio

[untold needle]

Also surely SciPy fftpack just does it for me @granite spear, or reduces the work significantly

[granite spear]

Fairly low-level, so a little assembly-language-ish, direct register access to chip peripherals, stuff like that.

The FFT is relatively easy, though, as there are libraries for that.

[untold needle]

Didn't you say advanced coding above

[untold needle]

That's what I thought, such as FFT pack from SciPy

Assembly language for what

[granite spear]

Yes, but I mean to get the RP2040 talking to the ADC. That's the advanced part.

[untold needle]

Oh

Right

Makes sense

[untold needle]

The RP2040 has to grab data from ADC

Doesn't seem impossible

Probably youtube tutorials for it

Or books

[granite spear]

You'll find tutorials about how to use ADCs in general, but not necessarily how to interface to your specific chip or how to reads at millions of samples per second.

Again, not impossible, just tricky.

[untold needle]

Yeah, seems pretty tough

As the ADC has to comply with the RP2040

[granite spear]

It's more like "your software has to MAKE the RP2040 comply with the ADC".

[untold needle]

Yes

I found tutorials and resources for 500kHz sampling

But I need 2MHz at least

[untold needle]

Doesn't the RP2040 have its own ADCs. ADC overview
The RP2040 has an internal SAR (Successive Approximation Register) ADC. The ADC uses an independent 48MHz clock, and a sample takes 96 cycles of this clock to complete (2us).

[granite spear]

Yeah, as you noted it's not fast enough for Nyquist.

[untold needle]

Yep

I need a far higher speed ADC

Something like 20MSPS

@granite spear so the logic is that the ADC converts voltages to what exactly

And then the RP2040 has to take those and store them

[granite spear]

It converts to numbers/data. An 8-bit ADC will give you a value between 0 and 255 for how big the voltage is versus some reference level.

[untold needle]

Yes got it

[untold needle]

That seems....trivial

Why can't RP2040 just grab one number and add to file

So trivial right

[half plover]

Because it takes a few cycles to write to the file.

[granite spear]

Any time you use the word "just" in an engineering design, it should set off red flags in your head. 😜

It's like asking, "Why can't you just throw a 1000 mph fastball? You just pick up the ball and throw it hard, right?"

Things that are otherwise easy become tricky when you have to do them millions of times per second.

[half plover]

“Just”, “that’s simple”, “that’s easy” all kind of fall into the territory lol

[untold needle]

I guess

[untold needle]

I don't think I fully understand the coding required at the moment

[young cove]

i didn't realize this discussion has been going on for weeks, but given the short time frame here, I am wondering about basically building a radio and sampling platform from scratch. If I were trying to do this in a hurry, I would start with a simple SDR, like an RTL-SDR dongle, get the baseband data it outputs, and just record it. If I'm worried about how much data that will be, I would sweep a narrower band across the frequencies of interest. I would probably do this with something like a Pi Zero that could act as a USB host to the RTL-SDR dongle, and record the data on the Pi Zero SD card.
I am probably missing something like the power budget here. My understanding is that just recording is good enough, because you will be recovering the balloon with its data.

[granite spear]

Yeah, that's a perfectly reasonable approach to consider. The project started just wanting a couple of signal-strength measurements but has grown in its requirements.

[wary arrow]

Probably naive question: why do DSPs have general-purpose processors, when DDSs don’t?

[young cove]

@untold needle ^ up a few posts

[half plover]

Classic engineering problem, scope creep 🙂

[young cove]

if the data can be recorded, then what to do with the data can be decided later

maybe also some filtering can be done to drop uninteresting intervals of data

[young cove]

I would say that a DDS chip is part of a larger system: what you want it to do is controlled by some smarter software running somewhere else. It might be easier for the control processor if a DDS could be programmed to do simple stuff like moving around in frequency autonomously, but the requirements vary widely. Also the silicon process needed for the DDS may not be such a good match for dropping in a processor core.

[untold needle]

do you see any problems in the current plan though, I'll outline it as follow:

Antenna --> Oscillator --> Mixer --> Filters --> Log detector --> High-speed ADC --> RP2040--> Fourier analysis on ground level

[untold needle]

Too true, this is likely to be the final project as we only have a month left essentially

[young cove]

the problem I see is the limited time you have to implement this, and that any changes later require knowledge of this particular design

that is why I suggested a building block approach with an existing radio and an existing compute engine

[untold needle]

we have a tight weight limit and current draw to

too*

[young cove]

what is the current draw, over what period of time? Are you counting the battery in the weight limit or is that outside the 300g I saw earlier

[untold needle]

200g

and current draw is 200mA

250mA maybe actually

[young cove]

yup, that is small, too small for something like an RPi zero

[untold needle]

not too small for an RP2040

plus I'm not afraid of working 5-6 hour days

and we have like 6 people

[young cove]

is the radio already built?

[untold needle]

only me and one other guy are particularly useful

[young cove]

the radio part

[untold needle]

transmitter?

[young cove]

is this a class project or a research project ... no, the receiver you outlined above, before the data acquisition

[untold needle]

Well we have build a yagi antenna and are connecting it to a function generator for the transmitter

[untold needle]

its an international competition basically

the receiver above

we have all components but not build yet no

almost all components

log detector is yet to arrive but should in 2-4 days

[young cove]

ok, so long term maintainability is not a requirement 🙂

[untold needle]

yep

long term maintainability is our last priority

I am not afraid to put in the work

its more about what work I need to d

do

[young cove]

there are just a lot of parts to this, so I was trying to minimize that, but i would say good luck!! the more off-the-shelf with existing designs you can make it, the more time you save

[untold needle]

yeah no, thanks a lot for your efforts, if you have any more tips/ideas I would be grateful

[wary arrow]

Thanks!

[half plover]

It’s more of a joke in engineering that no matter how hard you try, the scope of a project shifts. It’s not meant to be a slight of any kind, just a reality of engineering projects. (Take it from an engineer 😛 )

[short nimbus]

Most of the questions in this discord happens because the engineering scope of the asker changed

like they bought a basic mcu but as their knowledge grow they want to do more things which involved buying other parts or another mcu

[primal warren]

I've been musing on a somewhat related project, and it occurred to me that I could implement the digital part of DDS in an FPGA and just stream the data to an outboard DAC via SPI or parallel. And there are CPU cores for FPGAs. However, I have the distinct luxury of being free of any particular deadline.

[wary arrow]

Is a raspberry pi a good thing to use to connect to / manage things like ADCs and DSPs?

[primal warren]

Since the Pi is running a multitasking operating system, it's generally not a good choice for when you want accurate timing (which is often the case with ADCs). However, DSPs are CPUs in their own right, so you can just send them instructions and they'll run them (I built a board a while ago that basically did that, the board plugged into a computer, the DSP ran the show, and the real work was done by specialized ARM based CPUs).

[wary arrow]

Thanks!

[untold needle]

@primal warren @granite spear do you reackon SciPy FFTpack is fast enough to siphon through 20-30 million samples on 8 computers in a decent amount of time < 1 hour or will I have to use something faster like PyFFTW

[untold needle]

I think ill go with pyfftw

[granite spear]

Yeah, you shouldn't have any problem there. O(n log n) is fairly fast on a desktop computer when it's running at GHz speeds. I'd expect it to run "while u wait" on one machine, probably less than a minute.

[untold needle]

So it doesn't matter SciPy or PyFFTW

[granite spear]

Note that you can try this part of the project out ahead of time if you want, by creating fake data and trying to analyze it, before you even have a balloon or a radio built.

[granite spear]

Either should be fine, yeah.

[untold needle]

Also from what I'm reading, in SciPy and pyfftw only require like 30-40 lines of code

It requires more lines of code to get the data than to fft it

[granite spear]

Yep, that's what I'd expect. The FFT itself will just be a function call.

[untold needle]

Yep

Thank God there are modules for it

[untold needle]

That's to run the FFT, I guess to plot it like 40 lines total combined

Could just use matplotlib?

[granite spear]

Yep, that'd be a good choice. You can do the whole thing in like a Jupyter notebook environment if you're familiar with that.

[untold needle]

I'm more familiar with pycharm plus I have pycharm Pro due to me being a student

I have github copilot too

👍

And chat gpt as backhup

The Holy trinity of programming

Chat gpt isn't that good to be fair

[granite spear]

Heh heh, that matches well with Larry Wall's quote that the three cardinal virtues of a programmer are laziness, impatience, and hubris. 😁

[untold needle]

Fourier analysis and FFT is probably the single greatest algorithm ever

[short nimbus]

use the api version (playground) that has a choice of models and parameters it's much better and less judgemental but still affected by the I am right right about this and will make up urls and facts to prove it syndrome

chatgpt is general purpose for entertainment. Playground has 18$ of credit costs me about 0.10$ in credits per day for all my newbie questions...

To illustrate what I say I kept looking for the made-up references it provided me and kept telling it it was wrong and making stuff up up at one point it asked me to respect its beliefs 🤦

[restive fjord]

perl ! The Camel Book was legend.

[short nimbus]

Noah would be confused because of oreilly books today

"Do I also have to save 2 books of the same type from oreilly too?"

[untold needle]

@granite spear I've finished the fourier in python and am now basically need to just add a few noise filters and that is complete

I also found this for high-sped ADC with RP2040

https://forums.raspberrypi.com/viewtopic.php?t=316942

also I'm concerned that micropython may not be fast enough to sample that fast, I think it can but I am unsure, just wondering, as that means I will have to learn C in that case.

[untold needle]

@granite spear this looks so cool

created my own fake noisy signal, extracted the real one using FFT then inverse FFT, then found out frequenices of real one. Was exactly 51 lines of code.

I used pyfftw so it is fast as well

[untold needle]

now added axis labels which I thought was a smart idea

[granite spear]

Note that this link is discussing the larger Raspberry Pi Linux boards, not the RP2040 "Pico" chip. And yes, I expect that you'll need to use C for this to get the sampling speeds you want.

[untold needle]

right yes, oh no, I hate C, I dont even know it for a start.

[granite spear]

Sorry. I did warn you that this would involve some relatively advanced coding. 😅

[untold needle]

As you can see above, the fourier was quite trivial.

[untold needle]

are there any microcontrollers for which python would work

[granite spear]

Not really, unless someone has specifically implemented a C driver for the ADC you're using and provided a Python library for it.

[untold needle]

From what I'm looking at, writing to a file at above 1MHz is very hard

[granite spear]

You don't necessarily need to write the file that fast, since you're only taking short runs of data once a minute... if it takes you 10 seconds to write all the data out before the next sampling run, no problem.

Actually there miiiiight be a way to do this in Python, as there is access to the RP2040's PIO features through it: https://learn.adafruit.com/intro-to-rp2040-pio-with-circuitpython/overview

[untold needle]

You are the GOAT

[untold needle]

less

we will take 5-10ms of data one every 2-3 minutes

[untold needle]

@granite spear I've invited a few of my project team members here

such as @weak niche and @twilit drum

[weak niche]

Yo

[tropic fractal]

Hi

[twilit drum]

Hello

[fast isle]

cool stuff here

[crystal citrus]

Hello

[untold needle]

@granite spear we are planning to get this ADC https://www.digikey.co.uk/en/products/detail/analog-devices-inc-maxim-integrated/MAX1426EAI/2137405. Its 10MSPS and 10 bit but I'm not sure how we could connect this to one of the pico pins. Thanks a lot.

[primal warren]

That particular chip is a parallel output chip, so it supplies its results on a pin per bit: if you want to use all 10 bits, you'll have to connect (at least) 10 pins.

[untold needle]

Oh no, that doesn't work, what type of chip are we looking for?

[untold needle]

The RPI pico has 26 GPIO pins, does this also mean that if we only want to use 8 bits, we only have to connect 8 pins? So maybe it does work

[granite spear]

Yes, this was the approach skerr was suggesting. The RP2040 has a nice "PIO" feature where it can read multiple pins at high speed, so it can deal with parallel-output ADCs like that. This is why it's in "possible, but a little tricky" territory. You'll probably need a few extra pins beyond that to control the ADC too, but you should have enough pins for that plus the SD Card, etc.

[primal warren]

The disadvantage of a parallel interface is it uses a lot of pins, the advantage is it's fast. And I was thinking you might not need all 8 bit, which is why I added that qualifier. Then again, if you're buying chips for a project, and you only plan to use 8 bits, you can often save money (and other resources) by using an 8-bit converter.

[untold needle]

does this mean it draws a lot of current thoguh

though

[primal warren]

I'm not sure what you're getting at. While having a lot of pins does account for a little bit of current draw (to power the pin drivers), in general fast ADCs are going to be power hungry.

[granite spear]

The number of pins doesn't really affect the power draw one way or another in most cases. With fewer pins, they'd have to go faster, which also costs energy, so it evens out, but generally the I/O isn't where the power goes anyway.

[untold needle]

well the one above that is 10MSPS is only 0.3mA current draw

[untold needle]

thats fine though, thanks.

Need to research RPI pico PIO pins

since thats what we have

so do I have to connect 8 pins of the ADC to 8 GPIO pins

[primal warren]

That's what I was thinking

[granite spear]

Yes, and probably a couple more too, depending on the ADC. For example, the MAX chip above needs a clock signal to tell it when to take a sample.

[untold needle]

right right

so the Pico will basically just control my whole circuit

[untold needle]

I realised this now, I was a bit slow

[untold needle]

but then how do I tell the ADC through the pins to give its data to my pico

[granite spear]

I'd have to read the datasheet to be sure, but I'd expect that after each clock, the ADC would present the value on the data pins for the Pico to read. So the clock signal says "give me a sample", and then you read the sample a short time later (on the next clock, probably).

[untold needle]

oh

this is significantly easier than the other solutions

[primal warren]

I think it takes several clocks to do a conversion, but it looks like you don't have to do anything to start/request a conversion, just read data from the pins

[granite spear]

Yeah, it looks like it pipelines 5 or 6 samples, so you're reading a little bit back in time, but it's still "one clock = one sample" overall.

Different ADCs will behave slightly differently, if you end up choosing another chip.

[untold needle]

pipelines? Well this doesn't seem too terrible, does this mean I'm taking 1s and 0s from 8 different pins and then basically storing that in RAM then saving it to some sort of file on the pico?

[untold needle]

right

[granite spear]

Yep, that's the basic picture.

[untold needle]

is it in any particular order, I know thats probably hard to say without looking at the datasheet

mmm, somehow this doesn't seem ridiculously difficult but still lots of scope for error

[granite spear]

Yes, it's a binary number. So if it wants to send you the number 5 for the voltage it samples, it would be 00000101.

[untold needle]

surely there is some sort of reference voltage or something

I mean I guess the pico can calculate that

yes but how do I know which order its in

like it could return me two 1s and 6 0s in any order

and I wouldnt know what it is

it could be 10100000 or 01010000

[granite spear]

Yes, that's something you will often need to provide to the chip, to tell it how to measure. "I want to measure voltages between 0.0 and 2.0V, so here's a 0.0V and a 2.0V signal for you. Now give me a number 0-255 for the voltage you see between those limits."

[untold needle]

ah, so a bit trickier

[granite spear]

The order comes from the pin labels. The bits would go in order like D0, D1, D2, etc.

[untold needle]

oh silly me

thanks

so this software shouldn't be an ungodly level of difficulty

It has to take a 0 or 1 from 8 different GPIO pins in a certain order and save them to some file

[granite spear]

I mean, I do have to warn you that if you're asking questions about how binary numbers work, writing the code for this is not going to be easy.

[worn bridge]

This is highly subjective and I don’t think I should give this a response. If you’re comfortable with assembly, this is certainly manageable.

[untold needle]

Oh I know how binary works

we did that in computing class ages ago

I know how binary works well

unsigned and signed integers

8-bit, 10-bit

[untold needle]

assembly, can't I use micropython or circuitpython?

[worn bridge]

The outer shell is certainly python-able, but the PIO state machines are built in assembly-esque syntax.

[untold needle]

Now I'm confused, aren't I just taking the data from the GPIO pins

oh I see

[worn bridge]

You are, but python gpio reads may limit your rate of sample-recording. I don’t know what speeds you need or how many samples the adc can save, but you mentioned PIO so I assumed that you need it to be faster than the rate Python would spit them out.

[untold needle]

approx 10MHz

not even approx

it will be 10MHz

[untold needle]

yeah you are right, it will required PIO I think

10MHz is quite fast

[worn bridge]

Yeah, sounds like a job for PIO or C.

[untold needle]

I'm not even sure C is fast enough, though potentially

[worn bridge]

It’s pretty borderline. I would assume PIO would be the more reliable method here.

[untold needle]

yes I think so, I will try both though.

[untold needle]

Ah I found an issue, how do I test that the ADC is talking to the RPI pico without using a PCB, do I just have to breadboard it or something

[worn bridge]

Breadboard is usually the way to go, if you’re not ready to solder wires yet.

[granite spear]

Yep. Your ADC might not be available in a breadboard-friendly package, though. The one you linked to above is surface-mount.

[untold needle]

mm thats a problem

not sure how to go about that

[untold needle]

what if its surface mount though

there are quite a few adapters aren't there?

[remote pond]

for some SMD packages, you can buy breakout boards that you can plug into a breadboard, if you're up to the soldering

[untold needle]

soldering isn't a problem really

plus the pico we have is with headers already

[remote pond]

well, fine-pitch soldering if you're using a generic breadboard adapter to solder the SMD package to

[untold needle]

https://www.digikey.co.uk/en/products/detail/analog-devices-inc-maxim-integrated/MAX1426EAI/2137405 this is the ADC

28 pin breakout board ouch

[remote pond]

at least it's not QFN or BGA

[untold needle]

High-Speed ADCs are pretty complicated components

compared to mixers and oscillators at least

[half plover]

Thankfully a lot of high speed ADCs are in TSSOP or similar packages

[untold needle]

https://uk.farnell.com/texas-instruments/tlc7528cn/ic-8bit-dac-multiplying-dual-7528/dp/3005721 this looks quite nice too

[granite spear]

Note that this is a DAC rather than an ADC... it outputs an analog signal you write to it instead of sampling something you want to read.

[untold needle]

how have I missed that

whoops

https://www.mouser.co.uk/ProductDetail/Texas-Instruments/ADC1173CIMTCX-NOPB?qs=7X5t%252BdzoRHD7J6ePnaKJZA%3D%3D There we go

[untold needle]

yes, the one we are getting is in TSSOP, though I'm not sure how that helps

I guess TSSOP breakout boards are more universal

@granite spear can you check this part of the circuit

This is Antenna --> LO --> Mixer --> 2 Filters --> Log Detector

now that I come to think of it, do I even need a log detector with Fourier, the pico and the ADC???

[untold needle]

And then how do I tell which band pass filter it has passed through, because if I can't do that, it might as well just go straight to the ADC which does simplify the circuit to be fair

[untold needle]

oh, I have another question, what comes to the raspberry pi GPIO pin, is it a high or low, or an integer like a 0 or 1 for part of the binary

is it just 8 separate 0s or 1s, or is it 8 highs and lows

Yes, its a high or low for each pin I believe

[half plover]

Each pin will send either, that’s correct

[untold needle]

so I use PIO to check if each of the 8 pins is high or low

[half plover]

The corresponding value together makes up an 8 bit value that you then can easy convert into an integer.

[untold needle]

yes I know this

I need to figure out how to check if a pin is high or low using PIO

[half plover]

Yup, you use PIO to capture the state of each pin simultaneously

[untold needle]

no clue how to do that

😂

though it seems like something that I can google

https://raspberrypi.github.io/pico-sdk-doxygen/group__hardware__gpio.html is this the same as PIO?

[half plover]

Yup

[untold needle]

wait, so I can use this then?

[half plover]

These are technically for writing it in C

If you’re using C, yes

[untold needle]

ah, but is C fast enough

it will be going at something like 10-15MHz sampling

[half plover]

C is the fastest in the hardware world

[untold needle]

isnt C++ faster

[half plover]

C++ is a higher abstraction of C, so no

[untold needle]

o

[half plover]

Hence the ++ bits

C++ takes care of many of the nuances that can be gotchas in C

[untold needle]

right

[half plover]

But those conveniences come at a cost

[untold needle]

wait doesn't this just work then?

#include <stdio.h>
#include "pico/stdlib.h"
#include "hardware/gpio.h"

#define PIN 2

int main() {
// Initialize the Pico's standard library
stdio_init_all();

// Set up the GPIO pin as an input
gpio_init(PIN);
gpio_set_dir(PIN, GPIO_IN);

// Loop forever
while (1) {
    // Check the state of the GPIO pin
    if (gpio_get(PIN)) {
        printf("Pin is high\n");
    } else {
        printf("Pin is low\n");
    }
}

}

[half plover]

What do you mean?

[untold needle]

this checks if a pin is high or low

and prints it to the console

this cod

code

or at least I think it does

[half plover]

It probably would, I’m not near my computer or a Pico to try at the moment

[untold needle]

I thought this was going to be more difficult

[half plover]

I will note that it’s generally a better practice to store the pin value in a local variable.

[untold needle]

ah yes

To be fair, I am quite bad at C

[untold needle]

Thanks for your help, you've helped a lot so far.

[half plover]

C/C++ is my day job and I’m not great at is

[untold needle]

compared to me, you are a wizard

I am terrible

chat gpt is probably better than me

[half plover]

I’m proof that you can be not great at something and still make the big bucks because the world needs software engineers

Or engineers in general really

[untold needle]

how big are we talking........I'm joking, although you are being far too modest

[untold needle]

I'm talking to chat gpt and its adjusting the code

to check 8 pins at the same time

[half plover]

https://github.com/raspberrypi/pico-examples/tree/develop/pio/logic_analyser

[untold needle]

oh.

this seems useful

[half plover]

I’d work off something like this example

[untold needle]

right

[half plover]

Raspberry Pi invests a lot of money in making the SDK good by including examples so they’re great jumping off points.

[untold needle]

this looks very complex and long.

What does it do that my/chat gpt's code doesn't

[half plover]

Yeah, it probably will be.

[untold needle]

ouch

[half plover]

Sets up a PIO program, configures all pin states, adds the PIO program to the pio peripheral, manages the state machine.

[untold needle]

isn't that the same as the program I sent

this one looks so hard

[half plover]

PIO takes a lot to get going properly

[untold needle]

this looks crazy

[half plover]

That’s what makes it fast 😉

[untold needle]

so I need to adjust this to save it to a file

oh no

I don't even understand it

gulp

[half plover]

Anywho, need to finish my coffee and get logged in to work. Remember that we only grow by doing things wrong a few times. And yes, this will be hard but there are people here willing to help.

As long as you put in the effort and also ask the questions, you’re going to do great

[untold needle]

I've been asking too many questions almost

[half plover]

Nah

[untold needle]

571 questions

The only thing I managed to do fully by myself(pretty much) was the fourier analysis in a language I knew(python)

this PIO state machine stuff is hard

[half plover]

There’s not really a question limit

Doing new things can be hard, but that’s what makes them worth doing.

[untold needle]

true

I need to understand this PIO code

[half plover]

Anyway, build that example and then slowly remove lines from it and see how it works/doesn’t work. Deconstructive analysis is a great way to learn

It’s heavily commented too, so lean into those to derive understanding

[untold needle]

I guess so

[untold needle]

yes, that is what I am looking at

hopefully EdKeyes pops along

[half plover]

And reference that read the docs page too

[untold needle]

yes

[half plover]

https://learn.adafruit.com/intro-to-rp2040-pio-with-circuitpython

You might get some practice writing PIO programs using circuitpython as well

[untold needle]

python yes

I love python

I wish it was faster

[untold needle]

I understand some of it thanks to the commenting

[untold needle]

@half plover Chat GPT is giving me suggestions on how to modify the code to do only 8 pins and write a decimal value to a file. It seems quite reliable so far

[half plover]

Id take charGPT with a grain of salt

Better to follow the docs

Or known good examples. ChatGPT is only a tool for aggregating what it thinks is relevant.

[untold needle]

To modify the code to record the states of 8 pins and save it to a file called binary.txt in decimal form, converting them from binary to decimal, you will need to make a few modifications to the code. Here are the steps you can follow:

Modify the constants CAPTURE_PIN_BASE and CAPTURE_PIN_COUNT to reflect the base pin number and number of pins you want to capture. For example, if you want to capture pins 0-7, set CAPTURE_PIN_BASE to 0 and CAPTURE_PIN_COUNT to 8.

Modify the function print_capture_buf to convert each set of captured pins to a decimal value, and write that value to a file. You can use the following code as a starting point:

scss
Copy code
void write_capture_buf(const uint32_t *buf, uint pin_base, uint pin_count, uint32_t n_samples) {
FILE *fp = fopen("binary.txt", "w");
if (fp == NULL) {
printf("Failed to open file\n");
return;
}
for (int sample = 0; sample < n_samples; ++sample) {
uint32_t value = 0;
for (int pin = 0; pin < pin_count; ++pin) {
uint bit_index = pin + sample * pin_count;
uint word_index = bit_index / bits_packed_per_word(pin_count);
uint word_mask = 1u << (bit_index % bits_packed_per_word(pin_count) + 32 - bits_packed_per_word(pin_count));
if (buf[word_index] & word_mask) {
value |= (1 << pin);
}
}
fprintf(fp, "%d\n", value);
}
fclose(fp);
}
This function creates a file called binary.txt and writes the decimal value of each captured pin set to the file, one value per line.

Modify the main function to call write_capture_buf after the capture is complete.

[half plover]

It’s fun, but I wouldn’t rely on it to be your source of truth when writing this.

[untold needle]

chat gpt seems reliable though

for this at least

it gives good suggestions

[west perch]

Or THINKS it gives good suggestions

[untold needle]

while its code might be wrong

[half plover]

Eh, I would still lean far more heavily on resources like stack exchange or documentation for your code.

[untold needle]

the suggestions seem decent-ish

[untold needle]

I guess yes

[west perch]

It doesn’t actually know anything — just a lot of clever guessing

[untold needle]

yes, it is a predictive algorithm isn't it?

[half plover]

In my experience, ChatGPT is not a tool a programmer should use for their work.

[untold needle]

You are likely correct

[west perch]

Yeah — it tries to predict what’s right, but it has no idea if it’s right

[untold needle]

I guess I can just change the code based on what the comments say

like it prints all the binary values

I need to save them to a list

based on the original github coder

[west perch]

I agree with skerr, better to check previous human work

[untold needle]

yes

https://github.com/raspberrypi/pico-examples/blob/develop/pio/logic_analyser/logic_analyser.c]

this is the original

I need to change certain functions

like instead of printing, I need to save it as decimal form to a file

I need @granite spear, the man himself

[worn bridge]

Does it need to be saved as decimal? I imagine you could save a lot of cpu cycles and power if you converted it in post processing.

Storage, too.

[untold needle]

doesn't need to save as decimal necessarily no

so I guess let me change my statement

I need to save it as an 8-bit binary integer in a file rather than printing it and I need to do it for 8 pins rather than however many they have done it for @worn bridge

I know which function I need to change

to make it save as a file

[worn bridge]

Looks like CAPTURE_PIN_COUNT will adjust the number of pins to sample

[untold needle]

thats what ChatGPT said as well, change that to 8 and capture_pin

capture_pin_base to 0

[worn bridge]

The rest is just modifying/replacing the print_capture_buf() function to whatever format you want and have the pico write whatever that output is to a file

[untold needle]

yes that is what I thought as well, I just needed human confirmation as opposed to ChatGPT

[untold needle]

Thanks a lot!

[untold needle]

I greatly appreciate your help, I mean that with no sarcasm at all, this server has been the backbone of this project

[young cove]

what I wrote in #welcome re ChatGPT:

ChatGPT can give plausible-looking answers to questions. But you can't trust ChatGPT as a source of knowledge. It doesn't "know" anything. It could be right, or it could be completely wrong (and appear confident), and you can't tell. It's just cobbling together phrases from a large language model, and it doesn't do attribution. It's like asking a parrot with a very large memory a question. We recommend you use a search engine instead.

[untold needle]

Yes, I agree with this, it is a helpful tool but only for direction as opposed to actual substance like code itself.

[young cove]

if it would at least do attribution, you could go to the original sources. The Bing Chat thing recently demonstrated does do attribution.

[untold needle]

mm right. I mean ChatGPT has its uses but its still quite far off being a reliable source of information.

[west perch]

It’s more a toy right now

[half plover]

I got bored with it pretty quickly

[west perch]

I poke it now and then

[untold needle]

same

[untold needle]

Its funny to play chess against it, it plays many illegal moves.

[west perch]

Probably more fun to play against a monkey…

[untold needle]

this is true, I only play chess for fun though, I used to play competitively, but that was many years ago

[untold needle]

mmm, I found another problem

@half plover the ADC can't handle negative voltages(<0.3V) so how do I offset that issue

because since the frequency is a sine wave

some points will be below the X axis

[half plover]

Raise your reference voltage

[untold needle]

what do you mean by this?

[half plover]

Is this the RP2040 ADC?

Or that TSSOP one

[untold needle]

TSSOP

[half plover]

Okay

[untold needle]

I am connecting that to the RP2040 through PIO

but I need to avoid negative voltages below -0.3V or preferably anything below 0V

[untold needle]

yes, and I'm not sure how to stop the negative voltages...maybe an Op Amp but I'm not certain

[half plover]

You can pass your ADC input through a voltage divider

[untold needle]

but its negative

[half plover]

Okay, let me explain through an image

[untold needle]

ok

[half plover]

What this does is raises the “0V” to half VCC

[untold needle]

this just splits VCC in half yes

I know how voltage dividers work from my electronics classes

[half plover]

👍🏻

[untold needle]

ok but what happens to the signal

so VCC is 5 volts in this case

so it splits it to 2.5V

wait so 2.5V just gets added to every voltage?

[half plover]

In a way yes

[untold needle]

so -2 would become 0.5

this basically just shifts the graph upwards, which is perfect

[half plover]

Yup

[untold needle]

so this is like

y = sin(x) + 2

[half plover]

If you’re power conscious, consider larger resistors

[untold needle]

we might apply for a current draw waiver

at the moment, current draw limit is 200 or 250mA

i cant remember

we are accounting for the pico taking up 100mA, the ADC taking 50mA and the rest is 50mA

this is off a 5V DC supply

[half plover]

Yeah, makes sense

[untold needle]

slightly confused how that works

V = IR

voltage is constant

Oh

increasing resistance decreases current

I see.

[half plover]

Right

[untold needle]

I = V/R

[untold needle]

increasing resistance decreases current oh yes

silly me

[half plover]

You could probably use 2x 1k resistors for a nice small milliamp power draw

2.5mA

12.5mW

[untold needle]

mhmm, that idea of yours was quite genius

are you sure this works @half plover since I've never seen this idea before, most people opt for an op amp, I wonder why when your solution is so much more elegant and efficient.

I tried this with some sample voltages and it works perfectly after adding 2.5 to every voltage

it gives me back the same frequencies as expected

highest = 4.449083245V
lowest = 0.550916755V

[half plover]

I’ve used it on some optical ADC projects that saw voltage swing wildly

[untold needle]

That solution is so elegant electronically

wow

its actually beautiful

just put it through a voltage divider and move the sine wave upwards on a graph

[half plover]

Yup lol.

In your final design, make sure to use as good of resistors as possible

Like 0.5% or 0.1% tolerance

It’ll make sure your readings are more accurate, less variance in resistance and whatnot

For more reading: http://developer.wildernesslabs.co/Hardware/Tutorials/Electronics/Part5/Voltage_Divider_Practicals/

[untold needle]

ouch, my electronics professor has access to mainly E24 and E12

though I'm sure he has some 1% ones too

idk if he has 0.5 and 0.1%, he probably does but idk

[half plover]

Use the best you can get, just make sure they’re identical in value and tolerance

[untold needle]

Yes, so that it goes to as close as 2.5 as possible

wait can't I just use

Vout = Vin*R2/R1+R2

product over sum

[half plover]

If they’re symmetrical you know center is 1/2 VCC

[untold needle]

yes but with 1% tolerance, there is a range of possible centre values

its still minute though

so it shouldn't make the biggest different

difference

[half plover]

Right

[untold needle]

Also, I've been wondering, how are you so knowledgeable in the field of radio

Like, without your help and several others, this project would be nowhere

some of this stuff is like post-uni

you only start fourier analysis in uni

and nyquist limit is only talked about pretty late

[half plover]

Lots of projects, experimentation, working for a Visible Light Communications startup

[untold needle]

this project is probably my biggest yet

[half plover]

Also I have an undergrad degree in computer engineering

[untold needle]

also why is the word 'd a m n' censored

[half plover]

It’s generally not consider child friendly as we have younger makers on this discord

[untold needle]

That's pretty cool, and I take it your favourite language is C/C++

[untold needle]

I'm probably towards the younger/median age without getting into specifics

[half plover]

Hardly lol, I use python as often as I can

[untold needle]

really?

I mean C is pretty useful

its fast

thats basically its main positive

I only know python

[half plover]

It is, but I usually don’t need it for my projects. I use it at work mostly (C/C++)

[untold needle]

i guess some HTML

I did a C course once but I've forgotten most of it

apart from clang

and compiling code

I hate that you have to do that

[half plover]

It’s all useful, love it or hate it, it’s good to know.

[untold needle]

finding out how to add stuff to a file in C is going to be painful

void print_capture_buf(const uint32_t *buf, uint pin_base, uint pin_count, uint32_t n_samples) {
// Display the capture buffer in text form, like this:
// 00: ------------
// 01: --------___----
printf("Capture:\n");
// Each FIFO record may be only partially filled with bits, depending on
// whether pin_count is a factor of 32.
uint record_size_bits = bits_packed_per_word(pin_count);
for (int pin = 0; pin < pin_count; ++pin) {
printf("%02d: ", pin + pin_base);
for (int sample = 0; sample < n_samples; ++sample) {
uint bit_index = pin + sample * pin_count;
uint word_index = bit_index / record_size_bits;
// Data is left-justified in each FIFO entry, hence the (32 - record_size_bits) offset
uint word_mask = 1u << (bit_index % record_size_bits + 32 - record_size_bits);
printf(buf[word_index] & word_mask ? "-" : "");
}
printf("\n");
}
}

I mean this is what I have to change

I don't know why they are saving it as text to be honest

@half plover ok, I tried modifying it

Does this work?

oh wiat

wait

This

[half plover]

Best way to know is to build it and try 🙂

[untold needle]

right

what is your feeling though?

[half plover]

Hard to say at this moment, I don’t see anything glaringly obvious.

[untold needle]

what do you think of my new code

this one

I tried if an insufficient amount of bits are provided, just fill in the rest with 0 spam

well it isn't my code

I got a guy in my project team who knows a bit of C to rewrite it

[untold needle]

I changed it

@granite spear I'm not sure if you know C though I wouldn't doubt it, do you think this code works to save an 8-bit integer to a file called binary.txt based on 8 pins being high or low.

https://github.com/raspberrypi/pico-examples/blob/develop/pio/logic_analyser/logic_analyser.c

This is the base code which I have modified

[granite spear]

I think it's okay, yeah, though pin_base isn't used anywhere that I can see. The bit packing is a little unusual, but it seems to match what the logic analyzer is doing. pin_count would need to be 8 or less to fit the output into the byte-sized output.

[untold needle]

let me chec

[untold needle]

you are right, I don't actually see pin_base anywhere, I don't know why it is there

[untold needle]

@granite spear i feel like coding this on the pico would be hard

this is the clock pin of the ADC

also falling edge implies I have to create some sort of astable circuit

[granite spear]

It'll require a modification to the PIO code, since you'll want to output the clock signal in sync with sampling the input pins.

[untold needle]

oh lord

ouch

[untold needle]

wait so what would that entail, is there any sort of example or anything as well as other options for it. Also surely I can use micropython for this, as I only have to have the ADC turn on for 5ms every 5 minutes

[granite spear]

I thought you were using C above?

[untold needle]

yes

because speed is required

Oh right, is there no way to do both?

[granite spear]

Not easily, yeah. Your application would typically either be all Micropython or all C.

[untold needle]

Ah right, ok then, so I will have to use C

that's going to be hard

what do I have to adjust in that PIO code

[granite spear]

I don't know the details, but the general idea would be to have the PIO execute a loop like:1. Set clock pin high 2. Capture the 8 input pins 3. Set clock pin low 4. RepeatSo you get one capture per clock cycle.

Precisely how to do that with the PIO is something I haven't played with before, but it should be within its capabilities.

[untold needle]

but I want it it to be on for 5ms capturing samples at 15MHz(max ADC speed) then off for 5 minutes

also, what is this business of the 'falling edge' of the clock input

[granite spear]

That would be controlled at a higher level, where you tell it how many samples to capture. The loop above is just to get one 15MHz sample.

[untold needle]

oh ok

[untold needle]

but how does it know to capture only for 5ms

[granite spear]

The clock pin has to flip on and off at 15MHz to tell the ADC to take samples that fast.

To capture for 5ms, you'd tell it "please record 75,000 samples then stop".

[untold needle]

ok

[granite spear]

I assume the logic analyzer code has a way to specify that.

[untold needle]

yes it does

[untold needle]

mhmm yes

it does

at the beginning

but then how to make it turn of for 5ms

as it doesnt know that

either

[granite spear]

That would be part of your higher level application loop... record for 5 ms, save all the data, sleep for 5 minutes, repeat.

[untold needle]

oh i see

just shove it in a while True loop

or a for i in range loop

and add a time.sleep(300)

well that would be in python

but do that in C

[granite spear]

Yep, the same sort of loop structures and delay calls are available in C too.

[untold needle]

could I use asynchrous routines

like this

https://hackaday.com/2019/09/24/asynchronous-routines-for-c/

[granite spear]

You could, but I suspect that would just needlessly complicate your code.

[untold needle]

well it still needs to continue saving the data to a file, or does the code i created earlier do that as soon as the sample is collected

and how else would I do it

[granite spear]

I'm not sure how your overall code is structured, but you would typically have a main loop that would call functions to perform those actions, likec while (1) { take_adc_samples(data, 75000); // 5 msec at 15MHz save_to_file(data, 75000); delay_ms(5 * 60 * 1000); // 5 minutes }

But there are all sorts of details, like how to name the files you're saving and stuff like that.

[untold needle]

how do I make it go at 15MHz though @granite spear , apart from that I get it

in fact, what even dictates how fast it is

[granite spear]

I'm not sure how the logic-analyzer code selects the sampling frequency. I presume it has some sort of loop structure which determines how fast the PIO code is executed.

[untold needle]

I dont see that

[granite spear]

This might be the part that does it:sm_config_set_clkdiv(&c, div);It's setting some sort of a divider factor versus the Pico's system clock.

[untold needle]

hmm

[untold needle]

how do you do that code formatting

[granite spear]

Single backticks give you inline code like this while triple backticks give you a code blocklike this

[untold needle]

printf("Starting PWM example\n");
    // PWM example: -----------------------------------------------------------
    gpio_set_function(CAPTURE_PIN_BASE, GPIO_FUNC_PWM);
    gpio_set_function(CAPTURE_PIN_BASE + 1, GPIO_FUNC_PWM);
    // Topmost value of 3: count from 0 to 3 and then wrap, so period is 4 cycles
    pwm_hw->slice[0].top = 3;
    // Divide frequency by two to slow things down a little
    pwm_hw->slice[0].div = 4 << PWM_CH0_DIV_INT_LSB;
    // Set channel A to be high for 1 cycle each period (duty cycle 1/4) and
    // channel B for 3 cycles (duty cycle 3/4)

@granite spear this looks like something to do with it

I dont know why the logic analyzer has PWN

PW

PWM

[granite spear]

It might be doing that as test output to have a signal to capture?

[untold needle]

right

[untold needle]

so should I include that

// Grant high bus priority to the DMA, so it can shove the processors out
// of the way. This should only be needed if you are pushing things up to
// >16bits/clk here, i.e. if you need to saturate the bus completely.

16 bits per clock

thats fast

[granite spear]

You could maybe use it to test your code in the same way, saving samples from that signal instead of from the ADC, before you hook up the real chip.

[untold needle]

mmm, but how does that make the pins high or lwo

low

[granite spear]

I'm not sure of the details, but basically you write a 0 to make the pin go low, or a 1 to make it go high.

[untold needle]

mmm thats weird

this is pretty complex code

idk what the PWM does

I sort of understand the rest

[half plover]

Pulse width modulation just means you’re changing the duty cycle on a digital pin’s output

Usually variably to do a variety of things like move servos, play tones on buzzers, etc.

[untold needle]

what is it used here for and do I need it

[half plover]

In this context? It’s likely just telling the peripheral that the signal will be either PWM or PWM like.

[untold needle]

oh ok

[half plover]

I’m not sure why the logic analyzer would have this, it’s possible they are generating a PWM signal as an example to be processed by the logic pins you setup for reading

[untold needle]

I think so

Do I need the example

this code is so complex

[half plover]

I’d venture to say you don’t need it

[untold needle]

hmmm alrighty

[untold needle]

My teammate wants to write his own logic analyzer as he thinks this is far too complex and unnecessary, although it is likely his wont work so I am working on my own(this one) in the meantime.

He knows more C than me though

I dont really know what else I need to do to this logic analyzer

I modified it

it seems fine

@half plover if you dont mind having a look, I would be greatful

and @granite spear too if you dont mind. Thanks a lot.

[half plover]

The point isn’t to make a logic analyzer though so making their own isn’t going to solve the issue. What he’s going to find is that regardless of what he thinks, it’s the same process to read 8 inputs of a signal bus for logic analyzing as it is reading 8 inputs from a parallel bus for loading into a file for signal processing.

The key here is simultaneous acquisition

[untold needle]

Does the code above do that?

[half plover]

You only get that with PIO and FPGA in terms of signal processing

I don’t know, I am working my day job (still 3 in the afternoon for me)

[untold needle]

Wow sorry to bother you

forgot not everyone lives in the UK

🥲

[half plover]

Not an issue, I can respond in between compiling code

[primal warren]

When replicating Ben Eater's Arduino-based sort-of "bus analyzer" for breadboarding a CPU, I took the simple approach of using an 8-bit-wide port for basically simultaneous acquisition.

[untold needle]

I dont follo

follow

[half plover]

At slow enough speeds many microcontroller can do it. For instance you can do logic analyzers for I2C on a simple microcontroller

But 15MHz for 8 data lines simultaneously is going to be very hairy

[untold needle]

extremely hairy

I mean it has to be anywhere above 10MHz basically

[half plover]

That’s why PIO is great because it’s independent from the cpu and has instant access to pin states.

[untold needle]

we will probably run it at 10MHz

[untold needle]

mm yeah

Try and have a look at my code when you have time, I understand if its tough though. Can't believe you are working while helping me. Thanks a lot

[half plover]

👍🏻

[primal warren]

A lot of us are multitasking

[untold needle]

yeah, I struggle with that myself

[half plover]

Btw madbodger there is very knowledgeable as well, very creative solutions

[untold needle]

Yh I've seen

[half plover]

A lot of corner case things you wouldn’t think of.

[untold needle]

Well then, would you mind having a look at my code and suggesting anything

[primal warren]

Alas, I'm not up to it at the moment, I'm brain fried

[half plover]

https://tenor.com/view/bacon-gif-4287744

[untold needle]

oop

that is unfortunate

I'll repost it if you change your mind.

[short nimbus]

so one be able to build a duga style radar in their backyard with electronic scale power ? Or is it completely outside the realm of electronics ?

I always though that radio was a sensitivity thing not necessarily limited by power

[grizzled vapor]

It's also limited by the FCC/other government agency. Since a radar would be constantly transmitting (I think so), it might anger them. But it would be epic if you could pull it off.

[short nimbus]

yeah but I was thinking like with the evolution of technology you'd need much less power to detect a 1950 fighter aircraft with electronics ?

So you'd actually need less power and with more sensitivity for the same size due to the technology you'd still detect your own radar pings ?

[grizzled vapor]

I don't know RF. I'm trying to design a thing that harvests RF energy from FM radio broadcast signals and I have no clue how to design a matching network for the rectenna.

Can anyone help with the above question about designing an impedance matching network for a rectenna and energy harvester?

[half plover]

So it’s best to start with a 50ohm antenna. But there are calculators if you want to use smaller traces

You probably want a pi network

Pick a band like.. 88.3MHz

(8.83x10^7)Hz, a 16 mil micro strip trace on 1oz copper will have a source impedance of like 124ohm, plug into a network calculator like this from UC San Diego and pick probably a high pass low pass or pi network? https://home.sandiego.edu/~ekim/e194rfs01/jwmatcher/matcher2.html

University of Alberta has a nice paper too http://www.ece.ualberta.ca/~kambiz/papers/J22.pdf

[grizzled vapor]

Uhhhhh, what is source impedance?

I just now understand impedance

[half plover]

Sorry I meant load impedance since the matching network will be on the load side of the circuit

Source impedance is the impedance of the antenna

[grizzled vapor]

How do I integrate an antenna for FM broadcast bands onto a PCB? A dipole would need a lot of space. I'm not sure how to calculate the source impedance of that either

[half plover]

Usually a BNC or SMA connector

Or coax is common

[grizzled vapor]

That could work, but putting the antenna on the PCB would be even cooler

[half plover]

Yeah, I’m not confident they make chip FM antenna

[grizzled vapor]

lol

I didn't think so either

But there must be some way it is done. Old phones would use the headphone lines as FM antennas

Although a rubber duck isn't that bulky

[half plover]

Yeah, FM quarter wave is too long

There’s some portable radio antenna with SMA connectors so you can have an SMA board connector

[grizzled vapor]

I'm doing this for the Hackaday low power competition (goal is maybe an FM-powered clock) and just to see how low power I can get something.

https://www.mouser.com/ProductDetail/e-peas/AEM30940-QFN?qs=W%2FMpXkg%252BdQ6tc%2FiKrK2HeQ%3D%3D

I'm probably using this IC

I'm guessing that if the antenna source impedance is 50 ohms, the matching network converts that to a source impedance of something else, which should be the same as the rectifier impedance + the harvester IC impedance

[short nimbus]

Shame that you have a total power limit

because in the 1940s they used a passive FM radio to listen in on US ambassadors

[grizzled vapor]

The harvester IC has "ZMPPT". I guess that means it can adjust its own impedance so the power delivered is at a maximum

[short nimbus]

with a theramin-like device and it sent out AM signals when it was iluminated by FM signals

[grizzled vapor]

That's some pretty cool engineering

sounds like backscatter

[short nimbus]

it was called "the thing" in a wood seal of the soviet union

https://www.cryptomuseum.com/covert/bugs/thing/

[grizzled vapor]

Or should the matching network source impedance + the rectifier impedance equal the IC impedance?

[short nimbus]

the article says being illuminated by radio waves in the FM range powered it

might be worth looking into but I suspect it's the advanced engineering of the cavity/membrane in it that made it possible not the quality of the radio components in it...

[grizzled vapor]

That's cool that it was a completely passive modulator using no components like individual inductors or capacitors. Instead the components were the device.

[untold needle]

@primal warren @granite spear does the ADE-1 RF Mixer that we are using need a power supply, as it doesn't have a VCC or Vin pin

is it a pssive component

passive*

[untold needle]

also i built the circuit

[primal warren]

Yes, it's a passive component. It doesn't need a power supply.

[untold needle]

oh alright

[primal warren]

What are R3 and R4 for? It looks like they would just waste power.

[untold needle]

add 2.5V to every voltage

so I don't get negative voltages

[primal warren]

That won't work.

[untold needle]

it was suggested by @half plover

[primal warren]

You'd need a coupling capacitor to block the mixer from grounding out the signal.

[untold needle]

he suggested this

picture

[primal warren]

That will only work with a floating signal in. That mixer does not provide a floating signal.

[untold needle]

mm I see, how do I fix that

[primal warren]

Insert a capacitor between the mixer and the resistors.

[untold needle]

between the mixer and the voltage divider?

[primal warren]

Right

[untold needle]

what value?

0.1uF or 10uF or something

[primal warren]

The capacitance, along with the load impedance, determines the cutoff frequency

[untold needle]

right, so depending on the output frequency of the RF mixer, that determines the capacitance

is there any way to calculate the capacitance required then?

[grizzled vapor]

High pass filter calculator

[untold needle]

no, but for the capacitor in between

im not adding a band pas filter

[grizzled vapor]

Generally, a higher capacitance will mean more ac at lower frequencies gets through

[grizzled vapor]

Yes, for the dc blocking cap

[untold needle]

yes

i guess ill add a 10uF

[grizzled vapor]

What is this for anyways?

[untold needle]

I need to add 2.5V to every voltage coming out of the RF mixer

[grizzled vapor]

But why are you trying to read RF directly? I think it would be too fast, but maybe you could catch some of the signal. That is a parallel ADC, so maybe it's fast enough.

[untold needle]

it is 15MSPS

its fast enough i think

[primal warren]

1/2πRC

[grizzled vapor]

That gives you at most 7.5 MHz maximum

[untold needle]

yes due to nyquist

[grizzled vapor]

But a good rule of thumb is to divide by 5

[untold needle]

it is reading 0-3MHz

[grizzled vapor]

So 3MHz

[untold needle]

perfect

it is mixed down to 0-2/3MHz

[untold needle]

oh right yes

[untold needle]

is R in this scenario 4K

4k

from the voltage divider

[primal warren]

Nope, it's 1k, since they're effectively in parallel

[untold needle]

but its 3 separate 2k resistors isnt it?

[primal warren]

2 separate 2k resistors

[untold needle]

yes

2 i meant

mistype

therefore C = 4/1000pi

[primal warren]

Hmm, nominal values of Vrb and Vrt are 0.36V and 1.56V, so 2.5V isn't the right center value anyway.

Also you'll need an RF amplifier.

[untold needle]

yes they are

oh thats a very god point

I will recalculate

so I need to hit in between 0.36 and 1.56

so all voltages need to be between 0.36 and 1.56 I guess

[primal warren]

You left F out of your equation. The cutoff frequency (F) is 1/2πRC

[untold needle]

yes

C = 4*10^6/Pi * 10^3

= 4*10^3 / pi

[primal warren]

No, you left F out of the equation again

[untold needle]

F = 2 mega

[primal warren]

That's way too low, maybe use 6 megahertz

[untold needle]

but out of the mixer it will be coming from 0-2MHz

LO is 27MHz

[primal warren]

You said 2-3 MHz

LO doesn't matter

[untold needle]

we are broadcasting 25.2 --> 28.8MHz

so the IF will maximum be 1.8MHz

2MHz to be safe

[primal warren]

And the cutoff frequency is the point where the response is 3dB down. You don't want to throw away that 3dB, so use a cutoff frequency comfortably higher than the signal you want.

[untold needle]

ok so 5MHz

[primal warren]

OK

[untold needle]

ok I'm rearranging

thats weird

I get a capacitance of 10^4/Pi

[grizzled vapor]

3db lower means 1/2 of input signal amplitude

For reference

[untold needle]

thats impossible

I cant get a 3kF capacitor

I have to increase R then

[grizzled vapor]

Definitely

[untold needle]

yes massively

[grizzled vapor]

LC filter is basically a voltage divider, but with a cap instead of a resistor. Since the capacitor's impedance decreases as frequency rises...

[untold needle]

ok so I need like a 3M ohm resistor

jesus

[grizzled vapor]

What is your cutoff? And high pass or low pass filter?

[untold needle]

well 5MHz cutoff I guess

I just did the calculations @primal warren

I need a resistor of 3M3 and a capacitor of 1F to give me a cutoff of 5.2MHz

so what do I change in the circuit

Do I put in a 1F capacitor and what values do I change R3 and R4 to

[grizzled vapor]

You can't easily get 1F cap

[untold needle]

We have 1 Farad capacitors

[grizzled vapor]

Maybe LC filter

[untold needle]

wait does it have to be electrolytic?

[grizzled vapor]

1F is a supercapacitor, and those usually have high esr. Might also have high parasitic inductance

Meaning not good for high frequency

Maybe LC filter would work

[untold needle]

yes, 1 farad is massive

ok we make the resistor higher

we spam resistor

[grizzled vapor]

Too much resistor means that you might get a lot of noise I think

[untold needle]

this is not good

We need a solution

[grizzled vapor]

Lc filter

[untold needle]

whats the difference

I dont even know if this is a filter

[grizzled vapor]

It is

But it doesn't really provide the voltage bias

You could maybe add an op amp

That's probably a good idea since RF is very low level

[untold needle]

@grizzled vapor the aim is just to add 2.5V to every voltage coming out of the mixer

[grizzled vapor]

IF amp ICs exist

[untold needle]

I dont understand why we have to filter things

[grizzled vapor]

It's because you inadvertently add a filter when doing the voltage bias

[untold needle]

oh right

whats the solution

[grizzled vapor]

You could reference mixer/RF ground to a virtual ground at 2.5v?

Just a resistor divider

Or use lmc7660 negative voltage supply

[untold needle]

hmm that seems interesting

[untold needle]

this

[grizzled vapor]

Another option is to use a transformer

[untold needle]

not an option for us

[grizzled vapor]

One that's good for HF

[untold needle]

I like this idea of grounding the mixer at 2.5

does that just automatically add 2.5 to every voltage @grizzled vapor

[grizzled vapor]

Also make sure your LO shares the 2.5v virtual ground

[grizzled vapor]

Maybe? Probably?

[untold needle]

brilliant

[grizzled vapor]

Although...

But virtual ground might be more stable

Or better in another way

[untold needle]

I need 5MHz not 1 hertz

But yes virtual ground seems like a good idea

@primal warren will this work, save me

[grizzled vapor]

But since it is a high pass filter that passes all freqs higher than x hertz

In this case 1Hz

[untold needle]

oh

but surely thats fine then

[untold needle]

you mean low pass?

[grizzled vapor]

Maybe run a few simulations first or just experiment with your circuit

[untold needle]

oh right

so I only need one resistor

rather than a voltalge divider

[grizzled vapor]

Use two resistor divider

[untold needle]

ok

[grizzled vapor]

And since they are effectively in parallel

[untold needle]

so capacitor --> 2 resistor divider

[grizzled vapor]

You just use R value of half of single resistance

[untold needle]

can you draw me a diagram please, I'm not fully understanding

[grizzled vapor]

Use Google

[untold needle]

I created virtual ground for mixer and oscillator

@grizzled vapor so what do I need to connect the IF pin to?

[grizzled vapor]

That's output

Also maybe add cap to vgnd

[untold needle]

for the mixer?

[untold needle]

yes but I need to add 2.5V to all of the voltages outputting from it, or does the virtual ground already do that

I adjusted the circuit to meet the requirements that you suggested @primal warren @grizzled vapor

[primal warren]

You did your math wrong, that's not even close

[primal warren]

The capacitor goes to the junction of the two resistors (that's your signal input)

[granite spear]

The power-supply portion of the circuit looks weird too. You don't want to use a voltage divider to create 3.3V like that, since the voltage will sag as the chip draws any significant power. You need a voltage regulator for that, or possibly you can tap off the 3.3V from the Pico. Also the capacitors are hooked up wrong, I think since, you've got both power and ground shorted together on one side of them, and the VDD connection to the chip isolated on the other side.

[grizzled vapor]

You need to correct some of the capacitors and power supply thins as edkeyes said. There are some shorts and reversed capacitors

[untold needle]

you mean in the middle of the potential divider

[untold needle]

mm ok, well the ADC requires a 'clean, quiet' source of +3V

[half plover]

Usually a LDO

[untold needle]

@granite spear im going off this for the capacitors

low dropout regulator?

[half plover]

Yup

But the Pico 3.3V DCDC should work just fine

[untold needle]

this circuit seems like whenever I want to add one component, I end up having to add another 10

[untold needle]

but ADC needs 3, unless 3.3 is fine

[half plover]

3.3V is pretty standard

[primal warren]

What you're trying to build is a fairly advanced project, and I'm guessing you're going to need several more components at a minimum

[untold needle]

what other components will be required?

[primal warren]

You'll need an RF amplifier, and probably some impedance matching networks

[granite spear]

That diagram looks good, but it doesn't match the way you have the capacitors arranged in your circuit, in terms of power and ground connections on each side of them.

[untold needle]

we have an RF amplifier(MAV11)

[primal warren]

I didn't see that on your board, just an "antenna in", so I didn't know if that went straight to an antenna.

[untold needle]

oh right

yes

I've done that separately

[untold needle]

doesn't it, the pins go to the bottom capacitor with a ground in the middle and the power supply connection to the positive side of the electrolytic capacitor?

[untold needle]

oh right

[untold needle]

mmm so to the middle then?

so this then

ok

[primal warren]

It looks like you're trying to power the ADC with a voltage divider instead of a regulator: that isn't likely to work

[untold needle]

yes that was the plan

ok so should I get the pico to power the ADC?

[granite spear]

Double-check which side of the capacitors the VDD to the chip should be connected to. In your circuit, it's all by itself, whereas in the other diagram, it's also connected to the power input, with ground only on the opposite side.

[worn bridge]

Welcome to hardware design for dummies.
(And no I’m not calling you a dummy)

[untold needle]

oh but I am a dummy

[primal warren]

Possibly, depending on the voltage/current it needs. And the way it's drawn, your power supply is short-circuited anyway

[untold needle]

how so?

[primal warren]

You have the + (left) side of the left hand capacitor group connected to both power and ground.

[untold needle]

that is a brilliant point

@granite spear @primal warren RF circuit version 165

I unshorted it

and I added the capacitor to the middle

and the voltage divider adds the middle of the nominal values, it adds 0.96V now which is in the middle of 1.56 and 0.36V

@half plover please have a look at it as well.

[primal warren]

Maybe consider something like this for the digital and analog power supplies for your ADC

[grizzled vapor]

Why are AVDD and DVDD connected to ground?

They should be connected to VCC

[untold needle]

it isnt connected to ground?

is it

[untold needle]

I haven't connected it to ground on my circuit diagram

[grizzled vapor]

The voltage divider should ideally also have resistors of equal values

[primal warren]

No, the voltage divider is designed to center the input in the range the ADC accepts

[untold needle]

yes, Vout in it is 0.96 which is the sensor

sensor's center

[half plover]

If the range of the ADC is 0-VCC, center would be half VCC

[untold needle]

but the nominal values of the ADC are 0.36V and 1.56V

[half plover]

I don’t think that’s right.

Plus the divider should be pulled up to VCC of the ADC, not 5V

[untold needle]

why is that?

then it would 'add' 1.5V to every voltage

[half plover]

Because you’re changing the reference

The divider being between VCC and gnd ensures there will be no negative values

[untold needle]

Didn't madbodger say to use it to centre the voltage

[untold needle]

yes exactly

[grizzled vapor]

They are. Capacitors are connected correctly, but AVDD and DVDD are connected to the other side of the caps.

[untold needle]

are they?

yep

they are

thanks

[half plover]

Let me provide a reference to what I’m saying

[grizzled vapor]

The hardest to diagnose errors are always stupid mistakes

[untold needle]

yep

[grizzled vapor]

Yes

[untold needle]

ok but what is the VCC of the ADC

is it the pico?

Circuit diagram V1800 after fixing all short circuits, main thing now is what do I change those voltage divider resistors to.

[granite spear]

The 3.3V pin on the Pico should be able to power the ADC, using its onboard regulator.

[half plover]

Yeah, there’s no need to tap the 5V rail either

[untold needle]

yep ok

[untold needle]

well the raspberry pi needs power

[untold needle]

what about the circuit

[half plover]

I’m talking about the voltage divider

[untold needle]

oh right

so I tap that to the Pi pico power supply to

[half plover]

Yeah

[untold needle]

okie

[untold needle]

is everything else alright?

[half plover]

Your voltage divider will change

[untold needle]

yes

but what voltage needs to come out of the divider

[half plover]

You will want it centered on the ADC supply since that’s its reference

So 1.65..V

[untold needle]

1.5 or 1.65

oh whoops

yes

3.3/2 not 3/2

yes

[untold needle]

so I just need 2 equal resistors

[half plover]

Right

[untold needle]

Version 2500 with no errors I believe

@half plover @granite spear @primal warren give it the godly signal of approval

🥲

[granite spear]

Heh heh. "Approval" is a strong word, but I don't see anything obviously wrong, at least. 👍

[untold needle]

So now I 'just'(I hate this word now, I use it sarcastically) need to add impedance matching networks and add the antenna circuit with RF amplifiers

any suggestions of where to add these things

[granite spear]

I'll have to defer to other folks about that, as I'm more of a digital guy rather than RF.

[untold needle]

oh alright

[half plover]

Same, my knowledge stops there as well

But you’ll want to use something with a wide GPBW

[untold needle]

GPBW?

I have 19 days left on the project now

[half plover]

Gain Product Bandwidth

[untold needle]

testing, building PCB(yes we switched again), building transmitter(other teammates doing that, its going well) and then programming the pico and getting that to work

19 days for all of that ^

[half plover]

Definitely doable

[untold needle]

Interesting

[grizzled vapor]

Rf is hard. It's described as black magic for a reason. A good place to find radio stuff might be some ham radio schematics. Like QRP labs' QDX

But what is the project?

You might want to use a pre-built IC or module that handles all the RF magic for you.

[untold needle]

such as?

[untold needle]

Create transmitter, bounce radio waves off the ionosphere to receive them on a gondola/balloon in the air. Collate/collect the data and save it to a file to be read at ground level. Then run fourier analysis on it using a program and find out the effect of the ionosphere on radio waves and see the received frequency.

[untold needle]

idk if that exists

plus we're into it

[grizzled vapor]

Make sure it's all good with the fcc. If you have a ham license and you transmit your call sign (like as Morse code or data by turning the transmitter on and off), it should be good. But I can't be sure.

[untold needle]

Yep, we have the go ahead

I'm in the uk

but its an international competition

it takes place in belgium @grizzled vapor so we have permission from them

[grizzled vapor]

Also, that's really complicated for a 19 day project. Really cool though. Hopefully, the transmitter can pump enough power into a directional antenna. Idk how to design an rf power amp. And recievers are hard too

[untold needle]

well skerr thinks its doable

we have 19 days left

transmitter is almost functional

very close

we built a yagi antenna

and are amplifying voltages from a computer

into the yagi

[grizzled vapor]

You could use similar circuitry to a 20W QRP transmitter

[untold needle]

we've got the transmitter

[untold needle]

interesting

transmitter going well, Fourier analysis I built a pretty solid program if I do say so myself