TECHNICAL HIJACK!

"To me, it looks like offset heterodyning makes matters far worse at the far / bass end.. that the standard null tuning gives the best far field result."  - FredM

True if you are doing simple heterodyning, where any offset will kill the far field.  But if you are pursuing a digital route then you can do a second numerical offset (subtraction) which will restore the far field sensitivity by restoring the null.

Period measurement vs. frequency measurement is the same for where these two measurement modes make sense on a higher end frequency counter that can do period measurement.  When one has a high speed reference and one is trying to measure a high input frequency, one divides the reference by some large fixed number to get an interval like 0.1s or 1s and then counts the input frequency edges in the interval.  When one has a high speed reference and one is trying to measure low input frequencies, it makes more sense (i.e. one obtains more information) if one counts the reference clocks in the input period.  In the absence of averaging, frequency measurement resolution is limited by the input frequency period, whereas period measurement resolution is limited by the high speed reference period.

"But if you are pursuing a digital route then you can do a second numerical offset (subtraction) which will restore the far field sensitivity by restoring the null." - Dewster

Ok, I can see how one can get rid of the offset numerically by subtracting a value equal to the offset period.. I can see that adding offset is a requirement for digital theremins in order to get period values quickly in order to reduce latency, and I can see that subtracting the offset restores the same sensitivity as if one was going to null on an analogue theremin..

But I still cannot see how any improvement in linearity (either far, mid or near field) can be obtained through this scheme.. The best linearity one can get (without more complex numerical stuff acting on the 'captured' values) is whatever linearity the (analogue) VFO / antenna front-end gives AFAICS.

In order to get numerical values from the period of a post-heterodyned signal, one needs a complete difference cycle - with "to-null" (no offset) tuning, a 20Hz cycle will take 50ms to resolve, and even 100Hz will take 10ms.. Using offset reduces the maximum time to 1/(offset frequency) so in the case of the Open.Theremin (600Hz offset) the longest latency is in the order of 1.7ms and quite acceptable.

So I fully see and understand (and have used) the offset for purpose of reducing latency, but until I can see / understand the mechanism for the proposed linearity improvement, I remain in doubt - and wonder if ive been beating myself up needlessly over not exploiting the "improved linearity" of offset tuning -

"The only time one needs to have audio from the beat frequency is if you actually use this audio - if you use this audio, then you need to increase the difference frequency using PLL's or whatever so as to get difference at low latency for any digital stuff (or if one wants to generate CV that doesn't screw up below about 100Hz).. But if you dont use the difference for audio, the last thing you ever want is low difference frequencies or a null difference frequency... By "offset tuning" I am not referring to the clever realization that one can exploit this for improved linearity." - FredM 2nd last post,pg 15 theremini thread.

Except that I was sure I had seen the mechanism for this linearity improvement, but I cannot see it now, and cannot remember what I "saw" !  - Arrrgggghhhh!

Fred.

http://www.mediafire.com/download/bvvvm2yyl5bw7f9/Rod_2014-06-17.xls

That's it in action.  160MHz digital reference, ~2.55MHz NCO heterodyned with ~2.5MHz LC to 50kHz far field, period measurement, then subtraction to restore far field null.

Thanks Eric,

I started to see it again when I stopped thinking analogue! Got myself real muddled trying to apply the Open.Theremin parameters to an analogue scheme, and lost the plot completely, LOL..

I am working my way to full comprehension (I hope ;-) and think I am starting to actually see it, rather than just seeing the figures / charts - and starting to see (I think) why you have been after such a (compared to analogue / what I am used to) massive change in VFO frequency, as this looks to be essential for correct operation (?)..

The numbers are way out of my comfort zone! ~2.5MHz VFO, 50kHz offset, 160MHz counting clock - This is uncharted ground for me! - Actually, you should ignore any "advice" I give regarding antenna voltages or anything else related - I have absolutely no idea how any long-range capacitance sensing behaves above about 1MHz!

In truth, the digital theremin concepts (or at least the way you are doing things so as to truly exploit digital potential, rather than just tacking some digital hardware onto an essentially analogue theremin) are (I now realize) so far removed from my paradigms as to make them nearly mutually exclusive.

Certainly no way to exploit offset on an analogue design I think! ;-)

Fred.

"I am working my way to full comprehension (I hope ;-) and think I am starting to actually see it, rather than just seeing the figures / charts - and starting to see (I think) why you have been after such a (compared to analogue / what I am used to) massive change in VFO frequency, as this looks to be essential for correct operation (?).."  FredM

I don't think massive delta F with high res sampling are essential to making a linear digital Theremin, but they are likely necessary to enter pro response time territory.   If one pads the antenna down with too much capacitance the hand delta gets pretty small, which must then be pumped back up with heterodyning / long sampling intervals / averaging, and I believe this will magnify drift / intrinsic noise / instability / environmental interference in the oscillator.

"The numbers are way out of my comfort zone! ~2.5MHz VFO, 50kHz offset, 160MHz counting clock - This is uncharted ground for me!"

To make that spreadsheet more like the UNO, use the following values:

Tank L: 1mH

Stray C: 100pF

System Clock: 16MHz

Beat Minimum: 0.6kHz

The above gives ~500kHz VFO operation and very similar linearity.  Though raw response time is already down to ~600Hz, which will drop like a rock with the application of even moderate LPF.  I haven't experimented along these lines, but I feel uncomfortable with a sampling frequency in the audio range.

Thats a really great spreadsheet, Dewster ;-) and yes, I have plugged the UNO values in, but the results are (as I see it) no where as impressive as with your values - perhaps im reading them wrong..

On your implementation, I see an (adjusted offset) count going from 0 @ the far end (count 3200) which is obtained in a period of about 20us, to 2705 (count 495) at the near end, which is obtained in about 3us.. From these numbers I believe you could get 16 bit pitch resolution from accumulating the count of 20 cycles, and the worst case acquisition latency doing this would be 400us (Would range from 60us to 400us).

With numbers like that, IMO, you are well within "Pro" territory - even if you had to lose a ms or two with filtering or processing or whatever!

Fred.

(16 bit is extreme! you could probably go to 14 bit without anyone being able to notice, and this would drop the acquisition time to 100us, which could perhaps be important if you are dynamically updating the wavetable to give real analogue emulation - This is the area that IMO is probably the most important for a Pro instrument, one IMO needs to generate audio which is 'alive' the way a heterodyning analogue theremin is.. where the waveform changes as pitch moves from one interval to another.. IMO you seem to have solved all the headaches with the sensor side - perhaps you could regenerate a "reference" and "variable" HF output from your FPGA, and feed this to an analogue mixer ?  LOL ;- )

hi fred and dewster,

(or should i say hijack?) just uploaded a picture in the photo-albums, showing how i understood things. 

xtheremin8 this is one beautiful picture of yours and I have no doubt the guys need your help because they have been at it for a total of ten years between them and still are not getting anywhere. hum...

                                              Testing the Fredew Theremin

thanks touchless,

for the kind and funny reply. 10 years.. not sure i'm a big help, at least a scribble sheet so we could speak about the same things.  your posted "sympathetic coke-swing" video explained another certain movement. there are so many possibilities, building a simple device something, just for yourself, for experiments in the shed.  if one really needs something like that. i don't know why building a entire body and such, if a varicap could do the trick. but what do i know? i'm a tinkerer, i admit.  and sometimes a bit interfering. but my belief is: if one sees a problem and doesn't help, he's part of the problem.(no, not since the internet). and then: i can't keep my mouth shut. so it's a bit hard to interfere with the big-boys, and i bow deep in front of so much knowlege and the exceptional enthusiasm on a subject.  ;-) and mostly it's hard to explain yourself, when your not native english/american. even after reading krazy kat. http://en.wikipedia.org/wiki/Krazy_Kat 

miaou

xtheremin - you have creative youthful enthusiasm and I like that. We have a saying here in the States, that some people become so smart they get stupid. Now I hope that is not going on here because I want dewster to win. I just can't place my bet on him anymore.

Now Thierry is a man that commands my respect even though he is French! 0-:

PS: There are no big-boys here, we just get louder as we get older because no one will listen. (-'

T