Compressing variable values

Projects Programming
1

hello
I have created a class that allows me to create wave (sine, square, saw,etc). My intention is to modulate a carrier wave with other waves (modulators), my program works fine and I have been able to achive this. I am also using potentiometers to change the frequency, offset and amplitude of the modulating waves.

I am planning to use the modulated wave to control the parameters of an LED object.

Now in order do this I need to ensure that the modulated carrier values are:
1 data type int
2 within 0 to 255
3 positive values.
clearly this can be easily achive with restrain function, absolute, etc, but the result would not be what I want. ie any value above 255 would be 255 . What I would really like is to be able to compress the values to limit them to the range required and enure that they are positive. then cast the result to int.

the difficulty is that my modulated wave would vary when I use pot to affect the modulating waves values so maximum and minimum would be variable. I am thinking that I need a way to track max and min and then go from there (using map()). But I am sort of a bit stuck... I am pasting a picture to show the modulated wave (but this is fixed, not moving potentiometers, so the real modulated wave couuld end up being much wilder).

wave
wave2840×539 221 KB

1 Like
2

@ 1 and 2: use uint8_t
@ compressing: use some kind of compressor function, e.g. arctan()

3

yes that's exactly it

if you have your calculated value and you know it's always between min and max and you want to see

  • min ➜ 0
  • max ➜ 255
  • waveValue in between ➜ linear interpolation between 0 and 255
    then you can do

uint8_t displayValue = map( waveValue, min, max, 0, 255);

1 Like
4

@zwieblum Thanks!
@J-M-L great thanks for the explanation! I am gonna give it a try! :slight_smile:

5

as it seems to be working fine,
as for you suggestion:

  auto carrier =  carrierWave.sineValue(); 
  auto modulator1 = modulatorWave1.sineValue();
  auto modulator2 = modulatorWave2.triangleValue();

 resultingWave = (carrier + modulator1 + modulator2);        
if (resultingWave < minValue) {
  minValue = resultingWave;
}
else if (resultingWave > maxValue){
  maxValue = resultingWave;
}
uint8_t displayValue = (uint8_t) map(resultingWave, minValue, maxValue, 0, 255);

I have used unit8_t and casted the result, so far it's working ok but I am right in thinking that this method could cause errors due to overflow?

this is the result so far

pic
pic851×404 33 KB

1 Like
6

no... I am talking rubbish... it cannot overflow as it is bound to be within 0 and 255 after the mapping :stuck_out_tongue_winking_eye:

7

map() does not constrain (there is the constrain function for that) but here because the input value is guaranteed to be in between minValue and maxValue, the result is guaranteed to be between 0 and 255 so it will fit a uin8_t

8

Yes that's what I meant, but you made it clearer... Thanks again!

9

But, are you adjusting minValue and maxValue for each point of the wave? They should be the same for all the wave, the absulute max and min. Otherwise you are applying different scaling factors all the time, until you reach the absolute max and min.

10

You have point... I have not had the chance to see it working to see the side effects of this method, but I see a possible issue there. yes there would be a different scaling applied as the max value increase and the minimum decreases. And that should not a problem for me. In fact the problem will be the opposite. You made me think that once I have increased the max and min, if i then decrease the modulations again, the scaling applied will be the same as when the max and min were at extreme values....it may not be a huge problem for this project, but it would be interesting to find a way to avoid this.....:roll_eyes:

11

You can keep absolute max/min and also the max/min of the last N samples, by storing also the timestamp of when they occurred.

But it means that when the signal changes the amplitude overtime, the amplitude will adjust itself up and down. You could smooth the changes in max/min to avoid jumps.
I suppose that it depends on what you want to achieve.

Another option, if you know the carrier and modulators amplitudes, or can estimate them, would be to calculate the total max and min at any time when you change the signals.

12

That's probably the best way to approach this. I will need to see how this works in practice and whether it is a noticeable problem. If it is I will try this way.

Thanks for your help :+1:

13

aside from min/max and linear ramp you can use a compressor function that adjusts [0...inf] to [0..255] - like the atan() that I mentioned.

1 Like
14

Ok thanks! I'll look into it as I don't know anything about this function or how to use it to compress. Cheers.

Update... And looking around I found this which at first sight seems pretty handy:

15

That would significantly skew the wave shape near min / max. Might not be desirable/ similar to just trimming.

But indeed without an absolute min/max the signal would change dynamically

16

Actually that's what the different audio compresser plugins in e.g. OBS do. As your perception is log functiopn it's quite OK.

17

I did not understand the modulated wave were for audio but you might be right indeed.

18

Actually it is not for audio, it is for LEDs. But I had in mind audio compression when approaching the problem anyway, and I think it would work fine in this scenario as well.

19

OK - visual perception is not linear either

20

Yep, it's log, too