Recent content by tomas

It's hard to say without seeing the flicker on the video, but sometimes display may be having problems with keeping up with reliable transfer at high SPI clocks or the wires that you are using to connect to display are too long to transmit high frequency SPI clock. You can try lowering SPI clock...

Typically flickering is because you first erase content and then you draw. Between erase and new draw there is nothing on screen and that causes flicker. To avoid that you should "never paint the same pixel twice". Don't erase the content of screen then draw, but draw the new content over...

Yes using FIR is good idea, but if you want to go IIR way a simple biquad structure implements all kinds of filters (LP, HP, BP, BR, halving, etc), you can use calculator here to find out coefficients needed https://www.earlevel.com/main/2021/09/02/biquad-calculator-v3/ And C++ source code...

Every non-linearity introduces new harmonics. You should do Taylor series expansion to see what is the amplitude of components generated. For example single sine gets expanded to sin(x) = x − x^3/3! + x^5/5! − x^7/7! + · · · Each power component of Nth degree creates Nth harmonic. So x^3 term...

MQS is essentially noise-shaped PWM/PDM output so you should not expect much. NXP docs specifically say that over 10kHz you are going to get bad THD+N.

It converts linear scale that your potentiometer has to logarithmic frequency scale (how human hearing works), so equal knob movement represents one octave movements say from 100 to 200Hz and 1kHz to 2kHz (for example)

Everything that involves convolution is implementable on TPUs. Other algorithms too, check out some science on the subject: https://arxiv.org/abs/2107.05473

Just for what it is worth NXP i.MX RT1176 -based products start to emerge: https://coral.ai/products/dev-board-micro/

I was just about to post that info, but it looks like TeensyWolf was faster :- Product page for imxrt1176-based Dev Board Micro is now available: https://coral.ai/products/dev-board-micro/ I don't know how easy would be to program the Coral Edge TPU on this board, but they advertise this...

Just wanted to say Thank You, Paul Stoffregen, for doing all the work you do. I can imagine how stressful keeping up with all things is, as I am running the some other tech forum. Thank you for sharing the info about manufacturing side. T4 is fantastic piece of design and manufacturing and your...

CMSIS provides ready-to-use implementation. For interpolation you need call arm_fir_interpolate_init_f32 (once) and arm_fir_interpolate_f32 ( in the loop) For decimation you need call arm_fir_decimate_init_f32 (once) and arm_fir_decimate_f32 (in the loop) Look inside CMSIS source files they...

If you did oversampling "by the book" you would need polyphase FIR interpolator at the input and polyphase FIR decimator filter at the output. These are costly. Not doing things by the book results in aliasing. FWIW D'Angelo filter (Arturia) is running at 176kHz rate...

With max resonance you can clearly hear that analog filter has some "gritty" quality to it (2:52 of your video), while digital is sterile and lacks this effect completely.

I noticed that you are using copy of the same input sample when oversampling and averaging output. You may consider using linear interpolation on input and dropping on output.

Sorry, I did not want to sound harsh, I realised that I was too quick to post feedback, so I deleted it as it requires more time. I will run newest code and do more tests. UPDATE: Now I am running your newest code and I can enjoy increased frequency range :-)