DSP Robotics Support

Very cool tulamide. I have a few projects that might benefit from these classes. They always seemed a little mysterious to me, but now I think I finally understand what they are.

@tulamid: Wow - good work!
have to study it later...

@exo: could you add a PDF download link for these Tutorial parts because may print preview looks awfully - thanks!

OK, nix requested "filters for imbeciles", so here it is Digital filters: the basic logic behind…
The article is an introduction to filters and explains with very simple to understand example, how they work.

KG_is_back wrote:OK, nix requested "filters for imbeciles", so here it is Digital filters: the basic logic behind…
The article is an introduction to filters and explains with very simple to understand example, how they work.

Awesome, KG! Martin also already tried to make all those things more understandable. I hope you both keep doing this for dummies like me.
The real issue I have is not the math itself. For example, going to your article, without reading a single line of text I saw the image of the dsp module and understood immediatly what it does just from looking at the code. My real problem is all the fancy special terms, that nobody explains. On the web, even the simplest of filters are described in a way that I am expected to already be an audio engineer. As if the filter guys invented some secret language that nobody understands but themselves.
Let me give you an example of my problem. From my understanding I can't detect a frequency without the factor time. A sine wave that cycles every half second has a twice as high frequency than a sine cycling every second. As a non-filter-guy my logic tells me to feed an array with incoming values of the sine wave, and as soon as the first value I recorded appears for the third time, a new cycle begins. So I can now calculate the frequency from the number of samples stored in the array. Let's assume 100 Hz. I want my filter to cut everything above 60 Hz, so I just don't send the array out. Fine.
1) This worst of all filters would introduce a noticable delay (for very low subbasses at 36 Hz for example it's almost 3 hundreds of a second). But there is no delay in real filters. How do they detect a frequency from just one sample?
2) This worst of all filters would only work with sine waves. Additionly, it would only work if there's just one wave at a time. How do real filters detect a frequency even with 3 or 4 different waveforms overlaying themselves, again from just one sample?
3) This worst of all filters just cuts off everything at a certain point. But how would I introduce some kind of fade, where the frequency above the cutoff point will be dampened more and more the higher it is?
4) When looking for filters there always are terms involved like magnitude, real, phase, etc. But nobody explains those in an understandable way. Where do they come from, what is their task, why are they needed, etc.?

In short, I can do simple filters - as long as they don't need frequencies. Which means, I can obviously do no meaningful filters.

How about making a dictionary?! ... I had this idea for a quite some time. Some therms mean several different things depending on a context and usually have several synonyms, which are used depending on a context.

KG_is_back wrote:How about making a dictionary?! ... I had this idea for a quite some time. Some therms mean several different things depending on a context and usually have several synonyms, which are used depending on a context.

I don't know how others see it, but for me it means a lot. You can't imagine how thankful I would be for such a dictionary, especially if it's explaining the terms rather than describing them. You would be my hero (but not for longer than a month, I don't want you to get cocky )!

tulamide wrote: I don't know how others see it, but for me it means a lot. You can't imagine how thankful I would be for such a dictionary, especially if it's explaining the terms rather than describing them. You would be my hero (but not for longer than a month, I don't want you to get cocky )!

OK, Exo and I have set up aresourcepage at FS guru where links to different useful pages will be added. Guys, feel free to add suggestions...
Also a DSP dictionary is in the process of creation. I've quite literally started from "zero"...
http://flowstone.guru/dsp-dictionary/

... as an armature at this programming stuff, let me just say how fantastic it is to have you guys sharing your knowledge and experience.

I'm much to old to back to formal school ...although the 'hot babes' there could easily change that thought

I've no illusion of being some guru programmer. But I would like to create programs that can perform things that I just don't find available, or to my liking/needs.

Many of my 'ideas' are above my current skill level ... but more importantly ... being able to learn/understand this stuff better definitely changes how to approach constructing a program solution. Even being the slow learner that I am, I have looked at some of my early attempts ,,,, and shutter

So I must be learning something And YOU guys have been instrumental in that evolution.

THANKS !!

OTOH I would reapproach some explanatory stuff. In Flowstone, it's all about the NODES ...and more general concepts, from which you go into details when you need it (and only if you need it). Start from there - and you are the winner.

How many of you are teachers? The main problem is - who is the target reader, especially in environment like Flowstone (visual, dataflow, modular). If you wish to be understood - you may need to let go of some details in the main flow and re-emphasize other stuff, to give an entry point later. To be honest - try to explain such stuff (like filters) to my dad, who is good in analog/digital electronics, and tries to grasp the FS. You will hit the wall. So my suggestion is - test your tutorials on someone who knows what the items (filters, compressors, etc.) are - it will be someone with background in electronics or sound engineering or like that (or a kid - FS in education) - someone who isn't a programmer but is logical and organized person. If they can understand it - you wrote a good article. If they didn't - well - "cool tutorial, so... what is it about again please?" (or "I like when you are screaming" ).

Puff... I had to say that. I see too many tutorials that are cool only for those who already know.

tulamide wrote:(...)Let me give you an example of my problem. From my understanding I can't detect a frequency without the factor time. A sine wave that cycles every half second has a twice as high frequency than a sine cycling every second. As a non-filter-guy my logic tells me to feed an array with incoming values of the sine wave, and as soon as the first value I recorded appears for the third time, a new cycle begins. So I can now calculate the frequency from the number of samples stored in the array. Let's assume 100 Hz. I want my filter to cut everything above 60 Hz, so I just don't send the array out. Fine.
1) This worst of all filters would introduce a noticable delay (for very low subbasses at 36 Hz for example it's almost 3 hundreds of a second). But there is no delay in real filters. How do they detect a frequency from just one sample?

Well, actually all filters do introduce delay, and all filters use more than one sample. So your intuition is quite right on that part. However, filters do not exactly detect frequencies, they sort of attenuate signal components without ever knowing their exact frequencies. Think of a sieve that will hold back large grains without actually detecting each grain's size.

tulamide wrote: 2) This worst of all filters would only work with sine waves. Additionly, it would only work if there's just one wave at a time. How do real filters detect a frequency even with 3 or 4 different waveforms overlaying themselves, again from just one sample?

The math term for this feature is linearity, meaning basically that when you filter a bunch of tones (signal components with distinct frequencies) as a whole, you'd get the same result if you do them one by one and recombine them afterwords. Every signal can be decomposed into a set of (sometimes very many) sine waves, and that is a linear operation. A filter doesn't necessarily perform this decomposition, it acts on the signal right away. But the result is as if it did attenuate each individual component separately and then put it all together again.

tulamide wrote: 3) This worst of all filters just cuts off everything at a certain point. But how would I introduce some kind of fade, where the frequency above the cutoff point will be dampened more and more the higher it is?

That kind of filter is sometimes referred to as a brick-wall filter. It is not feasible in a strict sense in practice, however you can get close. In your scenario the filter daemon would not throw away frequencies above 60 Hz but only attenuate those components.

tulamide wrote: 4) When looking for filters there always are terms involved like magnitude, real, phase, etc. But nobody explains those in an understandable way. Where do they come from, what is their task, why are they needed, etc.?

I'm sure there is some easy reading stuff out there. Wikipedia is often a good start. Apart from KG's articles, that is.