FIR Filter Design and Software Implementation - Phil's Lab #17

[@PhilsLab]

New hardware and PCB design course on mixed-signal embedded systems just released! ⏵Course content: www.phils-lab.net/courses ⏵Course sign-up: phils-lab-shop.fedevel.education

[@garthberry]

Do you have any plans to look at implementing IIR and FIR filters using the STM32G series FMAC?

[@dixon1e]

Thank you for your work! Also interested in the IIR Filter. FWIW I googled the FMAC and found this informational video: st-onlinetraining.s3.amazonaws.com/STM32G4-Peripheral-Filter_Math_Accelerator_%28FMAC%29/index.html

[@Sorenp-n5o]

Another great video Phil, thanks for clearing up how to actually obtain the filter coefficients, so far I have just trusted the calculators .. It makes complete sense that it would be the inverse FFT of the filters frequency response. It would have been nice to see the effects of scalloping prior to using the window function, just to really clarify what it does to a signal :) -Keep 'em coming, love me some DSP.

[@anotherwingover]

Your vid was totally intelligibly explained. IMO the depth of the theory was on point 👍🏻. Also, i am looking forward to hear how IIR differs from FIR

[@lucapanjer]

I would also love some more videos on control theory! Starting at PID, which you have, and continuing to LQR and Kalman.

[@georgetroulis]

Took my undergrad DSP final 4 days ago, this video would have been incredible to have for my earlier days in that class. To all students taking DSP, they are lucky to have your incredible videos, Phil

[@PhilsLab]

Thank you so much for your very kind comment, George! Hope all went well with your DSP final :)

[@hoodoooperator.5197]

I'm in the last 6 weeks of MSc Digital Systems Engineering at the University of York, they have provided _LITERALLY_ zero DSP content... So thanks to Phil & his lab for actually providing this video.

[@jackiesofir4660]

This class is increadible, so clear. Wow just wow.

[@vanucanel6335]

I just discovered this video now. Smooth as butter. It is like a good action movie, well constructed, fluent, interesting and also very informative. Thanks for the great effort.

[@PhilsLab]

Thank you very much for your kind comment!

[@robertnickels1527]

You have done in 30 minutes what my brain has struggled with for hours - trying to bridge the gap between the practical FIR filtering code that I've examined with the theoretical/maths in Lyons book. This is one of the best tutorials I've seen on any subject - perfectly paced, clear and just the right level. Thanks!

[@PhilsLab]

Thank you very much, Robert!

[@robertnickels1527]

@@PhilsLab One question after playing with your FIR Filter designer: what is the purpose of "Shift Samples"? Once you set the filter length (# of taps) isn't it implied that you will always shift by 1? Or is this a way to do decimation and filtering at the same time?

[@ThomasZander]

These videos are much clearer and more concise than many university courses and textbooks. Well done Phil!

[@trollgarten]

I just thought about exactly the same!!!!!! Clear problem description, a bit of theory and an easy to understand implementation which makes the solution crystal clear, very very well done

[@whereswayne]

Thank you Paul, you've made this complex topic so much more approachable.

[@rjgonzalez8108]

This is exactly the DSP video I needed to bridge the gap in my knowledge between theory and implementation. Keep these videos coming! I also like the fact that you share resources you found helpful.

[@ajiteshjs3389]

This is extremely useful as I'm currently working on an application based on imu data and require signals with very little noise. Thanks a ton !!

[@PhilsLab]

That's great to hear, thank you. Hope this can help out with your project!

[@MedicineForFuel]

Phil, your video libray of tutorials is absolutely amazing! Thank you for putting together these relatively short, practical, tutorials on extremely valuable information! 🙏

[@PhilsLab]

Thank you very much! I'm vey glad to hear your comments on these videos.

[@martinkjellstrand4071]

Wow, this will come in so handy on the automotive application I'm working on, which will be polling something like 10 signals from various analogue sensors. Thanks for yet another excellent video!

[@Isabellaa-ms5dk]

this is the best video I have watched to explain FIR filters so far. thank you

[@CraigHollabaugh]

Thanks for this quick FIR refresher and for mentioning t-filter. I found it to be extremely easy to understand and utilize for my STM32 projects.

[@WaveIO]

pure gold! I just learned at Special maths about convolution and Dirac impulse, Heavyside and so on but never though how to make the connection between theory and practice! As I said, pure gold! THANK YOU!!!!

[@PhilsLab]

That's great to hear, thank you very much, Lucian!

[@andreiefimov9666]

Thanks! Kalman filter in discrete time is pretty interesting. May be you can describe in details Madgwick filter?

[@MrZomhad]

Also very interested in that! Cheers

[@geovani20091]

Also interested. Cheers

[@Soupie62]

In a previous attempt to understand FIR, I used Excel to model a Low Pass Filter. Rather than impulse, I used a step (square wave) and altered the response until it matched a sine wave. Having the proper math would probably be better, but it worked as a one-off.

[@jimboloco539]

Cheers Phil! You make these topics very digestible and encouraging!

[@PhilsLab]

Thank you very much, very glad to hear that! :)

[@qwer.ty.]

Thank you so much for those very clear explanation ! Your voice is perfect for non English speaker !

[@PhilsLab]

Thank you very much!

[@zhitailiu3876]

Really really precious tutorial! The git repository is also a treasure as well! Great job, Phi!

[@44mod]

You are Awesome!! I have been looking at all of your videos, Just finished the ADXL355 build and took 20 pages of notes. I am trying so hard to understand all of this but it is hard for me because I am a self taught and just want to learn. You do a perfect job at teaching what you are doing.

[@PhilsLab]

Thank you! Really glad to hear that - many more videos to come :)

[@flywittzbeats4008]

This is amazing. I am still very new and naive to dsp, but it’s amazing to see your content. Thank you so much for your willingness to share. It will not be forgotten ❤

[@PhilsLab]

Thank you very much :)

[@andreasmiller5448]

Very good video and teaching. Very clearly explained. Thank you! Just FYI... a good way to increment a circular buffer index is like this: circular_index = (circular_index + 1) % BUFFERLENGTH;

[@anonymity66]

Another amazing video. Great job! Your stm32 coverage is better than any I've seen on youtube, and you have such a wide breath of topics from initial board layout to fab, testing and implementing drivers and filters!

[@amazagx]

Thank you very much for these two superb an clear videos. Please keep teaching us about digital filters with this “hands-on” aproach. Cheers!!

[@PhilsLab]

Thank you so much, Alberto! Many more videos to come :)

[@andrewobuchow]

Amazing video and as an Eclipse contributor it’s nice to see other people using Eclipse for C/C++ as well :)

[@selimsagr446]

Thank you, useful content again. Additionally, I suggest SerialPlot Real Time Plotting Software to plot your data.

[@suvapillay9358]

Just, Thank you for all your time and hard work to educate people. So refreshing to see complex concepts, explained thoroughly and implemented in a place where anyone can see the results. Merry Christmas mate.

[@MellexLabs]

Amazing video Phil! The only thing that was getting to me was that your voice was competing with your clicky keyboard, which made it difficult to follow the software as you spoke. But that is by no means an indication of the quality content you share!!... So please keep on sharing your knowledge with us.

[@xDR1TeK]

Really great video, watched the one after it about IIR as well. This is what KZbin has been missing. Put more math. If people run away then they are your wrong audience. I loved it, I tried to show some how this was done and had difficulty with tools to bring a proper demonstration. This video will be my goto from now on.

[@PhilsLab]

Thank you so much! Yeah, I'm always unsure of how much maths to put in with it still being accessible to most people. I'll try and sprinkle in more and more and see when people have had enough :)

[@LL-ue3ek]

A superb job in demonstrating from concept and theory to product.

[@vincei4252]

Where you format the accelerometer output before calling CDC_Transmit_FS(...) sprintf() returns the number of characters written to logBuf[] so you can save yourself the call to strlen() which is yet another pass over logBuf. I'll keep my comments about using sprintf() in ISR's for another day :-) ... Ok, well, this is relatively low rate stuff so formatting and transmitting from another ringbuffer in the main loop fed by the ISR would be another option to keep the ISR's lean and mean. Great stuff.

[@patrickhochleitner7754]

Thank you once again for the amazing job you put in your videos. First digital control theory, then firmware configuration + hardware design and now this. It's like you are reading my mind or smth. Really amazing stuff. I'm all for a Q&A video!

[@PhilsLab]

Awesome, I'm glad you like them, Patrick! Let's see if I can read your mind with the next video :D

[@nadie-qm8rq]

Thank you so much! this was so helpful, specially the firt part of the video, so I could better understand the concepts. This video was uploaded only 6 months ago, I'm glad I'm taking this course now

[@PhilsLab]

Thank you very much for watching!

[@PerchEagle]

I wish there was a bit of electronics demonstrations. I don't know actually where all you are talking about is in electronics world. 1. What hardware setup you are using? is it the microcontroller dev board or it's another piece of hardware? 2. What level of electronics this happens? audio, wifi, bluetooth signals...etc. ? What I liked: 1. The code explanation is very good especially following up other drivers libraries you've done in other videos 2. Theoretical information about is good even I didn't get a clear idea of where FIR should be implemented?

[@seankayll9017]

Excellent. Looking forward to your new course.

[@RohitSardessai]

Thank you so much for the content you create. These videos are immensely helpful and I appreciate the time and effort you put into this.

[@darkstar2111]

Great video man! FIR is really powerful, even my total bonehead implementation of 32 values moving average manages to increase DHT11 termometer resolution by more than order of magnitude.

[@heitormasson4818]

I'm stunned! Thanks for the class! Absolutly flawless! Keep up the great work!

[@PhilsLab]

Thank you very much!

[@VicManzo]

Programming tip: If you keep buffer length a power of 2, i.e. 2^4=16, you can reset the index faster using: fir->bufIndex = ++fir->bufIndex&(FIR_FILTER_LENGTH-1)

[@odissey2]

One comment: if filter length is L=2^N, then the group delay of the output comes in-between the samples: (2^N - 1) / 2. For example, for N=3, the group delay is 3.5. This can be important, if filter data need to be combined with the input data.

[@tempusFugit1337]

Even though the main topic of these videos seems to be µC programming, they still give the most understandable explanations about digital filters i ever watched. Could you perhaps a video on Fast Fourier Transform FFT and its Inverse (IFFT) implementation in a similar way of your FIR and IIR videos ? Greetings

[@deplinenoise]

Nicely presented. The C code can be made a lot more performant by storing two copies of the impulse response and dropping the branch in the inner loop.

[@nickhubbard3671]

Thanks for this - you've put a lot of effort into your video! Appreciated.

[@tiranoid]

And I'm gonna rewrite them using fixed point arithmetics to make life happier for low end MCUs ! go on mate !

[@paryska991]

This is awesome! Really hope for some more DSP videos!

[@PhilsLab]

Thank you, next video is on Z-transforms!

[@Jules-ki9nx]

Wow I wish I saw this video a year ago! I hope you continue this series. It would be neat if you did a video using DSP to do hand motion detection on a microcontroller.

[@PhilsLab]

Thank you! I'll be making a couple more DSP videos. Next ones on IIR filtering and Extended Kalman Filters.

[@skaalduggery]

12 hours of lectures in DSP < this video. Thank you for doing this, it has been incredibly helpful!

[@PhilsLab]

Very glad to hear that, thank you for watching!

[@ing.tomasferenc1276]

I just wana thank you Phill to sharing your knowledge with us! Appreciated that.

[@PhilsLab]

Thank you very much, Tomáš!

[@Gengh13]

Great video, perhaps would be nice to see a fixed point Implementation, or maybe compare the two on uC with or without a FPU.

[@PattysLab]

really clear much more clear than my teacher used to explain this!

[@poliproto2023]

Thanks for another awesome video! 😁 I'm love watching your channel while working on my projects 😇

[@PhilsLab]

Awesome, thank you very much - very glad to hear that! :)

[@mdhvdubey]

Excellent explaination and excellent keboard sound.

[@weirdsciencetv4999]

Amazing tutorial and very clean code. When i handle interrupts, what I do is set a flag when the interrupt is identified, then immediately clear the interrupt. Then that then triggers execution of the appropriate callback in the super loop. I do this to keep the interrupt vector free and clear. What do you think about this approach? Of course if I was using FreeRTOS, I would have some kind of thread safe semaphore or even a queue. But damn man your explanation of FIR filter was spot on and probably the best I have ever seen. Edit: also, subscribed! Keep up the amazing work!

[@bannay]

Great work as usual Phil! Keep up it mate 👍

[@PhilsLab]

Thank you very much!

[@cheesecake667]

Your channel is a treasure. Thank you

[@PhilsLab]

Thank you very much!

[@coltennabers634]

Another approach is using "ping-pong" buffers, where you have one buffer being filled with new data and the other being processed. You then configure DMA to transfer data between the 2.

[@theonlyari]

Your keyboard is very passionate.

[@titovs2100]

Thank you. Very instructive. I would've perform the convolution using the Discrete FFT.

[@donbrown7694]

I’ve been looking for such good explanations for a long time!! Thanks!

[@HelmutTschemernjak]

Very cool training course, so well done, I love it. Don’t forget to setup a PayPal URL to support your activities for me and others. Here is some minor programming feedback. - for the sprintf use always the snprintf function e.g.: snprintf(logBuf, sizeof(logBuf), …) This is a good programming practise to avoid buffer overflows. - For circular buffers the wrapping can be made easy via modulo operations. e.g.: fir->buf[fir->firbufIndex++ % FIR_FILTER_LENGTH] = inp; This way you never need to check if your are at the end of a buffer or at the beginning) Again really great stuff you are doing here. PS: I would write this as a C++ class where the constructor gets the length of the filter as a parameter and allocates the float array, however for the audience here keeping it simpler with basic C is better because everyone must understand the individual steps. Regards from Hannover, Helmut

[@natasha.]

WOW PAUL! I am so happy with this one! You are such a good boy!

[@Sovvyy]

Welp, I better pick myself up some stationery then! Superb video :)

[@Chillingworth]

Excellent work man. Thank you for sharing your knowledge in such a good way.

[@PhilsLab]

Thank you very much!

[@johnolson5622]

First, I found topic and the information very useful. Thank you! However, please consider a nice headset microphone. (A gaming headset blocks external noises nicely.) At times, the keyboard clicking exceeded the volume of the narration. It really broke my chain of thought and was quite painful. Not sure why, but around 19 min into the video the keyboard sounds seem to get extremely loud.

[@yaghiyahbrenner8902]

another good educational video, good work phil really enjoyed this.

[@PhilsLab]

Thank you, very glad to hear that!

[@Johannes-222]

very good video. Summarizes a lot of hours of explanation by my "signal processing" professor :D Thanks!

[@PhilsLab]

Thanks, Joey!

[@jonludwig8233]

Thanks, really great content. This is a legit intro to dsp. One thing I would suggest to help improve the clarity of your content is looking into typesetting math with LaTeX (or whatever people are using nowadays). I'm sure you will continue to have to present math and that really helps make the symbols and arrangement clear.

[@PhilsLab]

Thank you, Jon! I love Latex and use it for many more 'academic' things. However, I prefer the simplicity of handwritten notes for quick videos like this.

[@vikrampalsingh8239]

It was amazing to see this video. Keep up the good work. Please make more of such videos. :)

[@dansam1395]

Fantastic! Cant wait for the Kalman one!

[@IW2MXP]

It would be interesting to understand whether replacing the x[n] * h[n] convolution with the FFT product X(f)•H(f) and then returning back to y[n] with IFFT could be advantageous or not.

[@akhilzachariahmathew9072]

Hi Phil great video. Very informative

[@sanelleric8875]

I repeat what others said: WOW. GREAT STUFF!!

[@PhilsLab]

Thank you very much, Sanel!

[@Zuwwar]

Hi. Thanks for the great video. Two questions though; 1. You mentioned; anything below 10 Hz should have a DC Gain of about 1. What do you mean by DC Gain here in case of a low pass filter? 2. Before windowing, you mentioned that the filter response is not nice because at about 10 Hz, the reduction is only -6 dB or -5 dB. How much reduction should we expect for a great response after windowing? Thanks in advance.

This is really gorgeous. Thank you so much. God bless you.

[@sc0or]

Two last days I've being reading articles about PCM DSD, and now you have published this video. And this is not a random vid but rather one of my subscriptions. I wonder what does it mean ))) And how do you have time for all these cool things for us?

[@PhilsLab]

Haha maybe I can read your mind :D It's getting a bit tough with time but I'm trying my best to get two vids out a month :)

[@danielmartins3446]

Nice content Phil, it helped me out a lot

[@abdum1493]

Great video! Question: What options are available to minimize the time delay between the raw and filtered signal?

[@jojo150393]

Really usefull as there is'nt so much practicall explanation of this subject. Often, this is only how to use a function with python or matlab but not how to implement it.

[@yomboprime]

Very well explained, thank you. Waiting the next videos...

[@PhilsLab]

Thank you very much!

[@sumedhburbure4173]

This channel is a goldmine!

[@davidbrooks8621]

Thank you. The real question is why the sensor generates noise in the first place? When it is static there is no noise, but when you swing it we see a lot of added noise that shouldn't be there I think. If the noise is part of the true data it should be left.

[@elodiecemoi3057]

This is the best and more practical video that I have watch on this subject : congratulations ! I am a newbye regarding the FIR filters, so I just begin to learn... I have two small questions: 1-.@ 13mn27sec (and17mn07sec), what is the software or I.D.E. that you use for writing the C code in ? To plau with your program, we need to compile your C code... 2-@ 6mn40sec, I understand the mathematical part showing that the Output signal = Input signal convolved with Impulse response. But I am a little lost regarding how we define the Impulse response: the math part is not a problem for me, but we need to know the impulse response before doing the calculations, right ? (my goal is to implement a FIR filter prior an audio DAC conversion..) Thank you !

[@Avriox]

I've learned so much from your videos. You're awesome! Thank you!

[@nasone32]

another tool is included in matlab: just type filterDesigner. yours is pretty cool too! because it's easy to use.

[@kissingfrogs]

matlab @ $200 Phil's is way cool

[@NicozStrat]

Thanks a lot for the super interesting videos dude! Would you Say that trasforming a continuos time filter from s to Z domain, night be an equivalent workflow?

[@velvetsound]

Would love to see (and would be happy to pay for as part of a DSP course ;-) ) examples like this on how to do Butterworth, All Pass, Chebychev etc filters.

[@PhilsLab]

Thanks Jason! DSP course may be in the works :)

[@fp2634]

Makes's me think i should think about a way to implement a FIR on my stock investments, with a large pass band :). That Said, thanks for the vidéo, great job and nicely structured.

[@LiyangHU]

Been done already.

[@mychevysparkevdidntcatchfi1489]

Blackman-Harris window. Fred is so often forgotten...

[@warpigs330]

Really great video that is applicable to some things I am doing. Would love to see you design a circuit for recording and sampling audio. I have been trying to figure out how to make a circuit that can sample audio cleanly as well as dc input signals.

[@PhilsLab]

Thank you! That's actually exactly what the upcoming PCB design course is going to be on. So hopefully that could be of interest :)

[@warpigs330]

@@PhilsLab I'm very interested. It seems like audio input is fairly straightforward with a codec but having the same circuit able to accurately sample DC input is giving me some trouble.

[@supremeleader5516]

amazing video phil

[@FranciscoCrespoOM]

Hi Phil. This is great content. Thank you so much. I have a question, what's exactly the 'shift samples' concept? I got everything else but this thing.

[@rodrigomesquita8741]

Thank you very much, Phil!

[@PhilsLab]

Thank you for watching, Rodrigo!

[@coolestofcats3011]

Hi Phil, great video. I have a question though. Can you explain why you implement the filter using convolution instead of the FFT and multiplication in the frequency domain and then the inverse FFT? I am relatively new to DSP, so I might be asking something obvious here, but my understanding was that the FFT was quicker then directly computing the convolution. Thanks for the great video! Bas

[@PhilsLab]

Thank you, Bas! FFT-based convolution (via multiplication in the frequency domain) is only really faster when the filter length is >= 64 (approximately, depends of course), so one should stick to normal convolution for shorter filter impulse responses. Additionally, implementing an FFT-based convolution is far more complicated to code and would make for a much longer video.

[@coolestofcats3011]

@@PhilsLab Thanks, that's clear now. On a side note, I would watch the longer video on the FFT implementation. Keep doing wat you're doing! Bas

[@fb8673]

Clearer than sun. Thanks you.

[@PhilsLab]

Awesome, thanks!

[@wilsoncervantessierra6174]

Excelente video. Muchas gracias Phil's

[@PhilsLab]

Thank you for watching, Wilson!

[@xy3986]

Hey Phil - once more - congrats & 1000 thanks for Your videos... But: I'm wondering about wether in your "filtering-videos" You use always FLOAT... ...but I don't see, that You use the CMSIS-DSP-functions which are integrated in a Cortex M4 & M7? I think you know (of course) the examples delivered by ST (for M746-Disco-Board) - also with an example of a FIR-filter... So I have 2 questions: 1. Wouldn't it be faster an also accurate to work with 32bit integer to process a 16bit ADC-Result in the same way? 2. If it is not: Why you don't use the CMSIS-DSP-Functions? Thanks for your answer... ...und mach weiter so - Leute wie Dich braucht das Netz !! ;-)

[@DantalionNl]

Are you using a keyboard with Cherry MX Blue switches by any change?

[@ediabdullah428]

Thanks for sharing this awesome thing.

[@amw224]

Thanks a lot, I will try implement your solution, I have problem in ntc sensor!

[@woulg]

Really really great video, thank you for this. Great job

[@PhilsLab]

Thank you very much!