You don't need to be an 'investor' to invest in Singletrack: 6 days left: 95% of target - Find out more
I have started a new job with a company that uses fiber optics and does signal processing. I am developer and have been told I don't really need to know about signal processing, but I would like to understand what topics I should know if I am to learn about it in my own time.
I'll be honest and say my maths is really rusty and I struggled helping my daughter with her GCSE maths 🙁
Any pointers would be useful.
Thanks
Gosh, I used to know about that sort of thing in another life (actually it's not that long ago I was developing software for it, but mostly relying on algorithms designed by other people). TBH if you're starting from nothing then you could do a lot worse than just use good old google eg https://www.google.co.uk/search?q=signal+processing+introduction . Another obvious thing to search for is "Fourier transform", but then if you read any of the articles from the first search it will give you pointers for other things to research. Though signal processing is a huge field and if you want stuff which is more specific to what you're doing you're going to need to get more details of what that is.
Though as you've clearly recognised a lot of this stuff is very maths heavy - I suspect if it's something you struggle with it might be useful for you to spend a bit of time revising that - otherwise you're going to find very basic signal processing stuff quite hard. Whatever you might have been told, I suspect if you're developing software for it you'll have to be able to understand at least some of the maths.
Fourier transform is the thing to learn if you are worried about the maths then get a copy of is Stroud engineering mathematics, or her red book further engineering mathematics.
As a developer you with probably be using FFT, fast fourier transform s look that up too. Probably using http://www.fftw.org library
+2 for Stroud, thats where to start.
Why not get an open source DSP kit and try it out? Cheaper than the Intel dev kits?
Sorry auto correct killed me again K A Stroud is the author.
As others have said Time-domain models, Frequency-domain models, Fourier analysis, Laplace transforms, Z transforms.
Signals and Systems by M L Meade and C R Dillon is an excellent introduction to the subject.
Both K A Stroud books are excellent for mathematics in general with really good worked examples.
+3 for K A Stroud and the co-author Dexter Booth taught me maths at uni.
A relatable way to learn about signal processing is digital audio. In just learning the analogue to digital sampling process for audio you'll cover quite a lot of the fundementals of digital signal processing. Sample rate, aliasing, number of bits, quantisation, Nyquist frequency, basic filter design, etc.
EDIT: I also think you can learn quite a lot without any maths to begin with.
EDIT 2: See if you can get a copy of Matlab to have a play with.
These have stood me in good stead to blag DSP for the last few years. Click around a bit, as the links are not necessarily the complete info the sites have. One of these will suit if you're lucky...
https://betterexplained.com/articles/an-interactive-guide-to-the-fourier-transform/
http://www.dspguide.com/ch1.htm
if you get stuck, the STW of the DSP world (sort of)...
http://www.kvraudio.com/forum/
HOLY CRAP. This new forum just added "mega"* to my post! <- and having though about it, I'm not sure I like the intrusion.
* = the picture of the website is auto generated.
>Signals and Systems by M L Meade and C R Dillon is an excellent introduction to the subject.
That was the course book to the Signal Processing module in my degree back in 1991! Still got a copy somewhere....
I used to love Laplace transforms, Z transforms etc (maths geek at heart).
<span style="color: #444444; font-size: 16px; -webkit-text-size-adjust: auto; background-color: #eeeeee;">EDIT 2: See if you can get a copy of Matlab to have a play with.</span>
numpy and scipy can do it all for free. Unless your company already use Matlab I wouldn’t bother.
You could do worse than have a (virtual) thumb thorough this book. Lots of context to go with the maths,
+2 for Stroud, thats where to start.
Having flashbacks to 20+ years ago.
This is a good primer I found the other day whilst searching for something.
https://jackschaedler.github.io/circles-sines-signals/index.html
Stroud for the Maths, Castro for the signal processing
numpy and scipy can do it all for free. Unless your company already use Matlab I wouldn’t bother
Yes, that was my gist, Matlab isn't cheap. Octave and Scilab are alternatives I know of
Lots of code on musicdsp.org bit far from basic and nothing I ever managed to grasp.
The book that UrbanHiker linked to looks quite good really http://www.dspguide.com/ch1.htm
.
I would be tempted to buy it rather than read it online.
jambo's link looks excellent as well - https://jackschaedler.github.io/circles-sines-signals/aliasing.html
You might want to make sure that you can understand that before going too far down this road. The introduction looks pretty terrible compared to the rest which seems quite readable and contains a lot of words that you might come across. I don't know exactly what you are doing and you might not need to understand what the maths does but I would suspect that you might need to be able to read it to implement it
I'd forgotten how much fun that stuff was :). It's been 30yrs since I've had to do it in anger. Hope you have fun with it
Yes, that was my gist, Matlab isn’t cheap. Octave and Scilab are alternatives I know of
tried both looking for a matlab alternative. Settled on python using the Spyder IDE, it even has a matlab mode to look and feel like the matlab interface if that’s what you are used to. Best way to get is to download and install Anaconda.
My father was Professor of Electronics and Communications at Southampton University in the late 70's early 80's until his untimely death. One of the books he wrote was called Signal Processing Modulation and Noise and was used as a text book for the BSc students. Probably a little out of date now, but he was pioneering digital communications back then, mainly for the MoD.
Cheers all. Like has been mentioned on here. DSP is a massive subject area and it's knowing where to start. My maths is certainly lacking and whilst I have a copy of Stroud that I purchased last year to go through I never got around to it. The book is massive and it's knowing what branch of maths you need to know for DSP.
I'll peruse the links and recommendations this weekend.
I'd say that helping with gcse maths to Fourier transforms is quite a leap
But Fourier transforms is one of the most visual fields. I think you would gain a huge amount fin a purely visual explanation.
I still remember my lectures on this. So much cool stuff. Like NASA taking out the block lines in the moon landing footage by block one spatial frequency in Fourier space
I bet there are some great videos in you tube
Can't add much other than another vote for K A Stroud.
I found:
Mathematics For Engineers
An interactive approach
Anthony Croft and Robert Davison
Very useful.
Perhaps check out some of the big fpga websites for dsp app notes. xilinx for example. Microchip might also have some stuff.
This book is a fun intro to Fourier...
https://www.amazon.co.uk/Who-Fourier-Mathematical-Transnational-College/dp/0964350432
One thing i often find missing in interviewees is a proper understanding of digitisation and how an analogue signal is descretised. Often fundamental insight into fundamental parameters such as signal to noise ratio, non linearity, sampling rate vs bandwidth, even an understanding of what a decibel actual is, is missing! And if you can't get that bit correct, then any subsequent DSP is rather academic.... Increasingly the people dealing with, coding, and developing the actual DSP stuff have little insight and often no visibility of the (often tricky) process of getting the raw digital data in the first place!