DSP Robotics Support

MyCo and Martin Hi!

Unfortunately I'm not really able to fully understand your comments; my lack of education and not your fault!
In principle I don't understand how any algorithm could determine what was correct and accurate, other than maybe a neural net based AI perhaps (I do have one in my skull). Then of course there's time-criticallity to worry about in any processor. I imagine than any software would involve recursion and polling but how would it know when it had got the correct answer?
Many years ago (I'm 61) I made a hardware device that you sang or whistled into and it would display the note on a keyboard shaped array of LEDs. The zero-crossings were timed and the result fed into a ROM I programmed to convert the period value into the LED position code. The tricky part was, of course, to get the zero crossings accurate to the fundamental. My solution then was to use a low-pass VCF. It's output amplitude was compared to the incoming raw signal and the result was fed into the VCF's control voltage input. I found that by setting the ratio of about 20:1 I could get the filter to pass just the fundamental, i'e when the VCF's output was about 5% of the audio input amplitude.
I bet some of you uber-clever people could program such a system but I worry about the feedback path that would be involved and,as I found experimentally, the timing and response of the VCF's control signal was all-important and very critical.

Cheers

Spogg

Spogg, I had this schematic lying around which does pretty much what you describe. It uses a tunable 4th order Butterworth lowpass filter, a dual envelope follower and a comparator in a control loop setup. The lowpass cutoff adapts to the spectral content of the input, such that the amplitude is reduced to 70%. That's a higher value than your 5%, however it provides good results with my (steeper?) filter. There are other adjustable parameters in the loop which one can play with, and as you point out the feedback timing is crucial. The lowpass output is then analysed for zero crossings, using linear interpolation for subsample resolution. All in all it works reasonably well, not great.

Wow Martin!
That's a great piece of work and so much more than I'm capable of yet. Worked straight out of the box.
Many thanks for sharing this.
I shall study it and hopefully learn something
Cheers
Spogg

Hi Martin,
YES, and wow, very cool. Keep up the great work! I think we can all learn from Martins great work and of course a few others that have been with Synthmaker and now Flowstone for some time.
Take care, BobF

Many thanks rocknrollkat,

for this and all yours others. Keep them comming, some good stuff here. They all go right into my tool box! May not need them now, but when I do they will be very much appreciated

Till next time. BobF.....