Let's Design and Build a (mostly) Digital Theremin!

Posted: 3/9/2014 1:01:03 AM
FredM

From: Eastleigh, Hampshire, U.K. ................................... Fred Mundell. ................................... Electronics Engineer. (Primarily Analogue) .. CV Synths 1974-1980 .. Theremin developer 2007 to present .. soon to be Developing / Trading as WaveCrafter.com . ...................................

Joined: 12/7/2007

Hi Dewster,

That is a lovely oscillator!

I have just run it on my simulator using my 74HCU04 "analogue" model (which I constructed from the schematic and used the best transistor models I could find) and with 5 in series, it runs superbly.. (a minimum of 3 are required to get enough gain - but with 5 the gain is more accurately defined by the input and feedback resistors)

I also ran it at supply voltages from 3 to 5, and with non-ideal inductors (series resistances up to 3 ohms (simulation crashed at 4 ohms) and its great - low current (peaks < +/- 150uA at 5V, < +/- 80uA at 3.3V) great antenna voltages (~50V P-P @ 3V3,  ~70V P-P @ 5V)

I think you are right when you say "this one looks like a keeper!"

Congratulations!

Fred.

(this oscillator has a superficial resemblence to the Lev oscillator, but I dont think there is much real similarity - I might be wrong though..

I cant get my head 'round it right now, but I wonder what the effect would be of coupling L1 and L2.. On the Lev oscillator, this coupling hugely increases the total series inductance of the tank, and yet allows the antenna resonator to act on only one section of this inductance - but I will certainly play with coupling just to see what happens ;-)

ADDED ->

Been probing the waveforms and understanding the operation better.. the more I see, the more I like it - its so clean! .. I think your use of capacitive potential divider for the input tapping is one of the things that turns the quite crude "open" oscillator into a something special - well, actually, theres a lot of things that you have done that transform this design... With the 3k3 resistors, a simple discharge tube on the antenna should provide adequate ESD ptotection - Yeah! Its a cracker!

Its actually different enough to anything I have seen that I think you have full right to "ownership" of it.. Certainly the best simple CMOS oscillator I have ever seen! - Oh, I have absolutely no doubt that it will work as well or better than the simulations.

Using 5 CMOS inverters, 1 ohm per inductor, and 5V supply, all other components unchanged, I got 1.378MHz with 10pF antenna C, and 1.4284MHz with 9pF, about 50kHz/pF

More ->

Ok, playing with values etc, I think this oscillator is better for digital applications than for analogue - I think (not sure) that when two inductors are used, there is a slight increase in near-antenna sensitivity - the antenna LC I think is working to "amplify" the effect of the antenna C which is in || with the L1 -- divider C.. Only marginal, but the wrong bias for heterodyning -> audio. Using only one inductor one gets back to a simpler relationship, but lose some of the antenna amplitude. To optimise this oscillator for analogue audio the Lev topology would probably be better..

But this was designed for digital, and for that its great - Without the 2nd L, its still a good oscillator for analogue - better and simpler I think than some of the standard oscillators like those in the SC theremins - but I need to build it to be sure.

Posted: 3/9/2014 2:07:58 PM
dewster

From: Northern NJ, USA

Joined: 2/17/2012

"I cant get my head 'round it right now, but I wonder what the effect would be of coupling L1 and L2."  - FredM

I've wondered this too.  This double coil feature has been in my previous designs.  I haven't looked into it extensively, but from my general understanding of inductors (admittedly weak) my feeling is pretty much nothing (other than a reduction in the windings necessary).

"With the 3k3 resistors, a simple discharge tube on the antenna should provide adequate ESD ptotection..."

And perhaps a littelfuse SP721 directly on the CMOS I/O.

"...when two inductors are used, there is a slight increase in near-antenna sensitivity - the antenna LC I think is working to "amplify" the effect of the antenna C which is in || with the L1 -- divider C."

What do you mean by "amplify"?  ;-)

From my spreadsheet simulations I don't think I've ever seen anything outside of a general increase in sensitivity, though it could perhaps alter things a bit depending on the bulk C of the antenna.  The extra inductor also makes it run at a lower frequency, which can enable one to use coils with higher self-capacitance / lower self-resonance.  The ratio of L1 and L2 is fairly non-critical.  If L2 quite a bit larger than L1 the sensitivity will approach the theoretical, but then other issues crop up which aren't worth the trouble.

(I've often wanted to re-wire my EWS so that the tank had the same L and roughly the same C as the EQ & antenna - could use a fixed L in the tank and eliminate all the finicky tuning nonsense, though the near field might be a bit less linear.) <= Don't do this, it's boneheaded.  A second inductor only works for series tanks!

=============

Breadboarded it this morning.  It works and seems quite stable!

- Only tried one L value so far, a 50mH Bourns I had laying on the bench, and without the same value series EQ inductor (L2=0).  It oscillates at 158kHz with a 250mm long 10mm diameter antenna.  Changing the L value(s) shouldn't require any component changes (try to keep L2~=L1 if you use L2). 

- Increased C3 to 220pF to reduce V swing into the CMOS input (safety margin - you want this amplitude large for noise reasons, but to not exceed the supply range - if C3=C1 I believe it is safe under all conditions, even if the antenna is removed, but I need to check this).

- Tried a single inverter 74HC04, 4069UB, both work at 3.3V with the Bourns coil. 

- I can get it to stall if I grab the antenna tightly with my hand, but it starts right back up. Excess gain seems to help here which makes sense.

- See some phase noise with super delayed trigger.  I believe this is mostly environmental but it obviously needs characterization in isolation (NP0 capacitor instead of antenna).

- R2 is necessary to initiate oscillation, the value seems fairly non-critical.

- Note that C4 in the schematic represents the antenna static capacitance, and is not an actual component.

I'm anxious to try air-core coils with it.  One thing I want to investigate is counter-wound coils on the same former to eliminate RF emission / reception.  Could put 4 windings on a former and make the tank and EQ all at once.

Posted: 3/9/2014 2:46:21 PM
livio

Joined: 2/2/2014

I promised that I would analyze any new proposal but will not write much for do not disturb. If you like, you can open this file. Otherwise forget my post. Some new video here.

Posted: 3/9/2014 3:20:50 PM
dewster

From: Northern NJ, USA

Joined: 2/17/2012

"I promised that I would analyze any new proposal but will not write much for do not disturb. If you like, you can open this file."  - livio

Thanks for the analysis livio!  And I don't mind your commenting here.

Sensitivity is % change in frequency / pF change in capacitance for a given bulk capacitance.  You have kHz / pF which doesn't mean much.  I'm absolutely sure your oscillator is less sensitive than mine (~1/2) in that respect due to the topology and component values.  This assumes both have a 10pF series capacitance going to the antenna, which is something I won't be doing.  I will likely use 100pF or more which would make my oscillator >3x more sensitive than yours.

And, while I understand that noise can very much be an issue when trying to quickly resolve tiny changes in frequency, I believe once the intrinsic noise of the oscillator is around that of the environmental noise, it becomes rather moot (diminishing returns, and of course it depends on the spectral characteristics of the noise).  Have you characterized the environmental noise?  I haven't yet and so can't comment beyond that.  But I will say that there is a danger in any project of overly focusing on the optimization of one aspect of the design.

Stalling is a relative thing.  Does your oscillator start back up after a stall?  Mine does but I have no experience with yours.  This is where the extra gain of the CMOS inverter really helps.  You are throwing away ~1/2 of your sensitivity with the 10pF in series with the antenna in order to keep it from stalling.

For all I know I will end up using your oscillator, but at the moment it seems wrong to throw out signal in order to reduce intrinsic noise.  It's very nice to have other options though, and your oscillator is one of the best I've seen - it certainly Spices well and the component values seem optimal.

=========

I really do want to use the best oscillator I can find.  If I can design it, that makes it a lot funner and more exciting, but that's not the point.  The last thing I want to do is hold onto a sub-optimal circuit for bad reasons.  If you can persuade me that I'm doing something fundamentally wrong I would genuinely appreciate it, but so far your technical arguments seem rather scattered and unfocused.  And please don't take that personally, I certainly don't mean it that way - we are having a technical discussion about technical things and critical thinking is necessary.

Posted: 3/9/2014 4:04:05 PM
dewster

From: Northern NJ, USA

Joined: 2/17/2012

Also livio, you should include antenna voltage swing in your analysis.  Your oscillator does ~14V p-p.  My oscillator can easily generate ~50V p-p or more.  I don't know, but I suspect that in some instances being able to swamp environmental noise with a big antenna voltage swing may be more important than reducing the intrinsic noise of the oscillator.

You might also want to breadboard my oscillator and try it with your setup.  I'd be quite interested in any practical experience anyone has with it.

Posted: 3/9/2014 4:48:22 PM
livio

Joined: 2/2/2014

"I'm absolutely sure your oscillator is less sensitive than mine (~1/2) in that respect due to the topology and component values. This assumes both have a 10pF series capacitance going to the antenna, which is something I won't be doing.  I will likely use 100pF or more which would make my oscillator >3x more sensitive than yours"
Please download the five simulations I have provided to help you, add 1 pF and see yourself. 
Compare the two oscillators both without the antenna isolation capacitor, and your solution is less sensitive. 
Compare the two oscillators both with the antenna isolation capacitor, and your solution is less sensitive. 

"Have you characterized the environmental noise?"
Apart the 50Hz (easely removed by the 10pF antenna isolation capacitor) and the noise generated by the oscillator components. There are not other noises desturbing the oscillator. Maybe you refer to noises produced by mechanical vibrations, on the large inductors and amplified 100 times and more, by the heterodining process?

"You are throwing away ~1/2 of your sensitivity with the 10pF in series with the antenna in order to keep it from stalling."
The capacitor is not for "stalling" it is for the 50 Hz noises. Without the capacitor your oscillator stalls too. Test the simulations and you will see this. You can remove the 10pF capacitor also from my oscillator if you prefer.

STALL
Yes it starts back after any stall and it never stalls, also with many square meters antennas. Try it and you will see.

AMBIENTAL NOISES
You can see my video tests near to monitor and keyboard, we never noticed environmental noises. And no musician has reported environmental or temperature problems. If needed, to increase the output voltage is not difficult, but we prefer to use a moderate voltage to reduce emissions.

----------------

I had planned not to bore you more, but it is difficult to accept assertions such as those of previous posts. If you want me quiet you should not write things like "noise is not important", "extra gain of the CMOS inverter," "stalling" and "I'm absolutely sure your oscillator..."

Noise in the band 1Hz to 10Hz is important, the real limit is the noise and no filtering can reduce it later.

CMOS inverters are incredibly noisy, they are designed for fast digital commutations, not for low noise.

 

 

Posted: 3/9/2014 5:37:53 PM
dewster

From: Northern NJ, USA

Joined: 2/17/2012

"Please download the five simulations I have provided to help you, add 1 pF and see yourself."  - livio

Please read what I wrote.  What you are calling "sensitivity" is wrong.  You need to compare %F change / capacitance change for a 10pF antenna.  kHz / pF is not a meaningful definition of sensitivity because it is dependent on the operating frequency.

By this measure your oscillator has a sensitivity of ~1 %F/pF.  Mine has a sensitivity of ~3.4 %F/pF.  This is comparing both circuits as they are built and intended to be used, and not adding or subtracting components to make the numbers look better or worse.  You could put a series inductor on yours and improve the sensitivity, but is that a fair way to compare our circuits if you don't in practice?

OK, if you remove L2 from my circuit and put a 10pF in series with the antenna the sensitivities are very close, but mine is still a bit more sensitive than yours in this scenario.  But it's only about a 20% difference so who cares?

"Maybe you refer to noises produced by mechanical vibrations, on the large inductors and amplified 100 times and more, by the heterodining process?"

No, I'm talking about electrical noise in the environment, not the oscillator.

"CMOS inverters are incredibly noisy, they are designed for fast digital commutations, not for low noise."

They are often employed in crystal oscillators.  Here is an interesting paper on it:

http://www.icbase.com/pdf/add/ti/AN-101-00039en.pdf

Posted: 3/9/2014 6:32:44 PM
dewster

From: Northern NJ, USA

Joined: 2/17/2012

"EQ" L on non-EWS and similar?

Fred, on top of a general sensitivity increase, another reason to use L2 in just about any any series tank Theremin oscillator is to have another low pass component between the oscillator and the environment.  I believe I see a noise and stability improvement on my bench when the second 50mH Bourns is in the L2 position on my oscillator (as opposed to L2=0 or a wire).  Scope trigger is out as far as it can go (50ms) and I'm zoomed up 250ns/div.  Noise is maybe 50ns with a compact fluorescent 0.6m away.  Open / closed hand @ ~0.75m clearly shows ~0.5ms delta.

Posted: 3/9/2014 6:38:15 PM
livio

Joined: 2/2/2014

Sensitivity (without input capacitor)

CapSensor    55.79  KHz/pF   (at 2.1 MHz)
Dewster V2   49.89  KHz/pF   (at 1.46 MHz)

Also considering the different base frequencies, the two oscillator have very similar sensitivity.
With the frequency correction your oscillator has a 28% more sensitivity, not so different.
Try to raise your oscillator to more than 2 Mhz and they will be also more similar.
You can not work at 1.46 MHz, at the and of the medium waves, near to mega watt broadcasting stations.

"electrical noise in the environment"

If documentation says this, maybe they were receiving broadcasting stations in the Long and medium waves bands. (100 KHz to 1.7MHz). Also in the 2 to 3 MHz band there are some nautical stations but with a power incoparably lower. Maybe some 100 or 1000 Watt not 500 MegaWatt.

"They are often employed in crystal oscillators.  Here is an interesting paper on it:"

Chrystal oscillators are monstrously stabilized by the chrystal itself. A limited amount of noise is not influent in a chrystal oscillator. We are making a VFO not a chrystal oscillator. We are making a VFO that must measure incredibly low capacitance variations.

- - - - - -

An extreme example to understand the problem (I do not propose to work at 5 meters)

Why we can not see hand movements at 5 meters? 

Because of the digital resolution? No, also analogical Theremins can not do this.

Because of the ambiental noises? No, a Faraday cage could demonstrate this.

After a certain distance the signal is so little that the noise of the oscillator components becomes predominant.
You can hear the effects of the noise in this video: http://www.youtube.com/watch?v=N6IW2Q0ZRKA

- - - - - -

Have you tried to work at 1.5 meters with your oscillator, you do not hear noise effects?

Please, answer to this, it is very important for me... if you can go to 1.5 meters and more without hearing the noise effects, then your oscillator is incredibly good and you have to explain to me how to make it.

 

 

Posted: 3/9/2014 7:18:36 PM
dewster

From: Northern NJ, USA

Joined: 2/17/2012

"Have you tried to work at 1.5 meters with your oscillator, you do not hear noise effects?"

With a 250mm length / 10mm diameter rod antenna, or with a plate antenna? 

From what I've seen today it seems stable enough to work with a 250mm length / 10mm diameter rod antenna out to maybe 0.75m or more.   This is ~450kHz with a single air-core.  I haven't tested it at higher frequencies yet.

[EDIT] I just tested it with a double coupled air coil and it looks stable out to a meter or so with a 250mm length / 10mm diameter rod antenna.  Even with my "rat's nest" hookup.  This is ~460kHz.  Having trouble getting it to oscillate with a 0.5mH inductor but it's probably my setup.

"You can hear the effects of the noise in this video..."

Did you keep the tank capacitors about the same?  By this I mean: is the sensitivity and antenna voltage swing very low like it is on the Open.Theremin oscillator?

You must be logged in to post a reply. Please log in or register for a new account.