ฝัง
- เผยแพร่เมื่อ 13 มิ.ย. 2024
- In this video I do some DSP Phase shift processing with the Teensy 4.0. Some other DSP videos I have:
STM32 DSP: • DSP with Nucleo 144 F7...
ESP32 DSP: • ESP32 IoT Radio - Part...
Big shout out to Charlie Morris' channel for all the inspiration. Lots of great stuff on his channel
/ @charliemorriszl2ctm
The Teensy and the Audio Shield
www.pjrc.com/store/teensy40.html
www.pjrc.com/store/teensy3_au...
Correction: I meant to say "I2S" not I2C
Here's the Iowa Hills software site
www.iowahills.com/
I am using the Hilbert Filter from this page
www.iowahills.com/8DownloadPag...
Note I missed out a critical parameter in the Hilbert filter configuration and that is you have to put in the desired phase offset in the "Phase Add" field (i.e. +45 and -45)
Here's some DSP theory pages. The convolution algorithm
shulgadim.blogspot.com/2012/0...
A good reference source
www.dsprelated.com/freebooks/...
Finally the code and coefficients
github.com/thaaraak/TeensyDSP - วิทยาศาสตร์และเทคโนโลยี
![จังหวะนี้ - เม้ก อภิสิทธิ์ [ Official MV ] จอนนี่มิวสิค](http://i.ytimg.com/vi/4gQRZCkRyiU/mqdefault.jpg)
ref: 'Notes and Volts' Teensy Synth tutorial series. 👍
That's where I learned.
CORRECTION: Although the SGTL5000 has I2C controlling the CODEC, the audio communications between the T4.x and SGTL5000 are strictly I2S. Many folks don't understand the difference between I2C device control and I2S audio protocol.
I've used 2 SGTL5000 audio boards, with one modified with a unique I2C address for quad in/out, as well as the CS42448 6in/8out CODEC for a laser synth. But, with the CS42448 @ EOL, I'm now developing a new PCB, using the 6in8out PCM3168A CODEC.
JS
Just subscribed as well .... this is gold for me! HAM radio experiments to next level. Even for audio phaser application super interesting!
Thank you Egbert!
You got yourself a new subscriber with this video ! really interesting, looking forward to explore your channel :) thank you for sharing your work !
Thank you very much - glad you enjoyed the video!
Excellent Video. So interesting. Yes Charlie Morris is my inspiration as well.
There is a *pile* of code running behind the Teensy arduino environment - particularly in the audio processing side of things. I have even seen a few videos of hooking up the NXP programming/debugging environment to the Teensy through the debug lines. I'd love to do that because I am not a huge fan of the Arduino programming environment
@@na5y Yes I too am not a big fan of that environment, but it is a great starter. I wish they stop calling library's they are classes. Library is where all the collection is. Oh well , I use Atmel Studio. Intellisence is very nice. Keep your video's coming I enjoy them. Thank you. 73
@@rfdave3980 I have been using STM32CubeIDE. Would like to use Keil, because it has broad support, but it is expensive for the hobbyist and the "freeware" version limits you to around 32kB binary files.
This is really impressive!
Thank you very much - glad you enjoyed!
beautiful tutorial
Thank you for you kind words - I am glad you enjoyed the video
Re-watching the video in 2023. Thanks for sharing! 👍 Couldn't find much information about Hilbert Filters either. The book "Understanding Digital Signal Processing, 3rd Edition" by Richard Lyons has a good description of Hilbert transform in chapter 9, and there is a math for calculating the coefficients in section 9.3. The author does 0/90 degree phase shifting instead of +45/-45 one which obviously doesn't make much difference for us. The chapter does describe how to preserve the *relative* phase shift for a pair of signals, see Fig 9-12. I found the chapter somewhat difficult to follow because the author seems to have other uses cases of the Hilbert transform in mind. But from what I can tell it should serve our purposes.
Thanks for the lead Alex I will check it out. I ended up figuring out how to produce my own 0/90 phase shift using python libraries to generate the coefficients: th-cam.com/video/BUGQaTJ6aWI/w-d-xo.html
Its not perfect and you have to make sure you run the signal through a band pass filter as the delay filter doesn't do that. I still haven't been able to find a way of generating coefficients for an arbitrary phase shift though (say 0 and 89). Let me know if you find one!
Nice video. Thank you
Thank you Phillip
i would love to see SDR based on the teensy , did you make one yet ? i only just discovered the device and ordered myself a Teensy 3.2 and the TEENSY3_AUDIO board
Hi Steven - thank you for the comment. I never did a video of my Teensy SDR - I did post the code up at github here github.com/thaaraak/IQ-Radio/blob/master/Tayloe/Tayloe.ino - that radio was receive only. I know its not quite what you are after but I did do an STM32 based SDR here th-cam.com/video/I8ZwlgjXAZM/w-d-xo.html
Charlie Morris has a really good series on his SDR which uses the Teensy starting here th-cam.com/video/J7FEJeCYBpY/w-d-xo.html
Thanks again for the comment - please let me know if you have any questions
Thanks for this video, been working on phaseshift, but hung up on converting the coefficients for teensy. Some people convert it in a spreadsheet, which I was trying to figure out. I see you do it in the program, that doesn't effect the CPU usage compared to pre converted?
You mean the multiplication of the coefficients by 32768 in the array? That 's only going to be done once on initialization. It will slightly delay startup but it won't have any runtime effect beyond that. Thank you for the comment and question!
@@na5y oh. Ok, that makes sense then, thanks for quick reply!
@@DavyOneness You are welcome - good luck with the Teensy!
The low cut-off for Hilbert filter with 160 taps and 44.1kHz sampling rate would be 44100÷160=276Hz. It is typically usable up to 1.3-1.5x of that value, which is ~400Hz.
I am surprised that only 200 coefficients are allowed by Teensy. Is this a hard value or a library limitation?
Thank you for your comment. The library used by the teensy for FIR is the CMSIS DSP library. In particular this is the call from the teensy arduino code (filter_fir.cpp)
arm_fir_fast_q15(&fir_inst, (q15_t *)block->data,
(q15_t *)b_new->data, AUDIO_BLOCK_SAMPLES);
There's no limitation in that library the limitation is just purely the amount of processing power you have on the uController
@@na5y Thank you for reply and video. Subscribed.
@@na5y To improve SSB carrier suppression I would try to scale left or right FIR coefficients by small amount
@@odissey2 I will have to give that a try - thank you for the tip!
Very nice video, pls I am new inro this, looking for low budget audio dsp 2 in 2.1 or 4 out, low , mid high pass, and sigma studio, (Less programing)
Thank you for the comment. I am not familiar at all with Sigma Studio. I did do a separate video on DSP using STM32 F767 and the pMod board. Check it out if you are interested - programming is relatively straightforward
th-cam.com/video/jvmejxsQzn4/w-d-xo.html
Don't you lose a bunch of precision converting your double to a short int?
You do but the FIR filter algorithm used by the teensy (arm_fir_fast_q15) uses 16bit signed integers for the coefficients, so it doesn't matter
Would it be possible to programme a FM stereo encoder signal using this setup?
I must admit I haven't done a lot of FM - just some VCO stuff (have the audio signal waveform as input to control an oscillator frequency). I haven't done any FM demodulators.
I am mainly using DSP here for the 90 degree audio shift required for unwanted sideband suppression. Upshot is I would say probably not useful for FM modulation - but its an intriguing question. There's a paper here about using DSP for FM demodulation might be worth a read
www-elec.inaoep.mx/~rogerio/Digradio.pdf
@@na5y Thanks for your reply. I will look at quite some of your vlogs in the near future. Just discovered it.
Main question is actually can you write code that makes a physical balanced mixer circuit superfluous?
@@egbertgroot2737 The tayloe mixer series I do elsewhere - here for example:
th-cam.com/video/bOyliyEWw8U/w-d-xo.html
Does away with a traditional double balanced mixer and replaces that with a quadrature controlled FST3253 switch. Could it *all* be done in code without specialized hardware (frequency synthesis as well) - all I am sure of is that *I* couldn't do it ;)
Thanks for the comment!
Bro What is the Mixer ADE ......
I'm sorry which mixer were you referring to?
Very interesting video! I have bought a Teensy4.0 and the corresponding audio board. Your displayed hardware setup is very useful for me! I want to make an automatic gain control for my TV audio output so I can gain-equalize the various sequences (commercials, films, or my various mp3 songs which have non-normalized gains). I have a question. Where did you get the command language shown in your sketch shown below? I know signal processing and the Arduino but I'm a new guy for the Teensy. There must be a manuel describing the audio library commands somewhere? THANX!!!
from your Hilbert Arduino sketch:
#include
Si5351 si5351;
AudioControlSGTL5000 ctl;
AudioAmplifier amp;
AudioInputI2S i2s1; //xy=233,194
AudioFilterFIR fir2; //xy=494,294
AudioFilterFIR fir1; //xy=498,93
AudioOutputI2S i2s2; //xy=743,192
AudioConnection patchCord1(i2s1, 0, fir1, 0);
AudioConnection patchCord2(i2s1, 1, fir2, 0);
AudioConnection patchCord3(fir2, 0, i2s2, 1);
AudioConnection patchCord4(fir1, 0, i2s2, 0);
..
..
bool i2c_found = si5351.init(SI5351_CRYSTAL_LOAD_8PF, 0, 0);
si5351.drive_strength(SI5351_CLK0, SI5351_DRIVE_2MA);
si5351.drive_strength(SI5351_CLK2, SI5351_DRIVE_2MA);
..
...
ctl.enable();
ctl.volume(0.8);
ctl.inputSelect(AUDIO_INPUT_LINEIN);
amp.gain( 10 );
fir1.begin(plus45, NUM_COEFFS);
fir2.begin(minus45, NUM_COEFFS);
There is an online manual Bill - go here
www.pjrc.com/teensy/gui/index.html?info=AudioInputI2S2
As you click on the class (e.g. I2S2) help text appears over to the right. I think the rest of the message was cut off though. If you are interested in Finite Impulse Response filters work there area a variety of resources out there e.g.
eclipseaudio.com/fir-filter-guide/
The software I use to create the coefficients is here
iowahills.com/
@@na5y Thanx. So I imagine that you used this GUI tool to create those commands? It seems to be very useful for 100% forward processing operations such as a Hilbert filter. However, in my case, I will have to provide feedback information for the volume gain, according to the output of the averaging filter. I noticed that there are not any closed-loop algorithms (LMS filter for example) in this GUI tool. Anyway, I'll start my project when I get my long header pins that I forgot to order. I'll see how that goes. Thanks again!!!
I usually use the GUI just to get started and then refer to the class information from then on. There's a project page on that site I sent you with a whole pile of other audio stuff people have done with their teensy's. Its a fun little board. Let me know how you get on and thanks for the comment.
The teensy uses CMSIS-DSP under the covers. That has IIR filters in it so if you are feeling brave you could mod the teensy audio code. I think you could use that for least mean squares couldn't you? Not a DSP expert here by any means. Also there is an RMS component in the library which you could potentially use
@@na5y I'm learning more & more about Teensy thanks to your useful answers! No, I won't need an LMS (it's an adaptive FIR) algo for the AGC. I was just citing examples of closed-loop functions. Yeah, the Teensy looks like a fun processor. Reminds me of my DSP days when I was working. I'll let you know how my project goes. Hardest part is starting it.