"Build the EM Theremin" - altered component placement - comments?

Yeah - I have no idea how you managed to get it to run even in simulation! ;-) .... How far in time did you run the sim, and how did you set the start conditions? ... Wouldn't surprise me if what you were seeing was a transient anomaly..

The only "magic setting" needed was "start power supply at 0V". I let it run for 2ms, and at the beginning the oscillation was rather small and it increased to max soon (within 1/8 of the time maybe). On the left transistor's collector there was an almost clean sine (well my "let's try what happens if we do this" way of changing the circuit may not be beautiful ;) )

you could probably get the board to produce a few hundred cycles before it petered out if you whacked the supply just right ;-) .. did you check the amplitude of the oscillations? sure they weren't 50uV rather than "real" ?

Well the board actually runs "forever" with 5V after I changed all resistors to smaller ones - other than the resistance of the coil which I guess is the problem in my naive way of changing the circuit? But here it's the wrong coil (or at least it would be a coincidence otherwise) anyway and I might be lucky.
But it's enough voltage apparently to feed another transistor to buffer this so I can hook the scope to it.
If I knew a good place to upload such things to I'd post the ltspice file.

Actually, in spice, I also have a file where I added the reference oscillator and the detector and a lowpass filter, and looked at the result, while playing with the variable osc's capacitance, to see the heterodyne freq jump around in audible range. But it (the antenna capacitance having influence) worked only when I set the antenna coil from 40µH to 0.0000040µ or so i.e. basically nothing. And the needed changes in capacitance to really see something seemed rather large (tens of pF or more), on which an EE I know commented "spice and reality are not best friends, don't be fooled by fair weather harmony"

But dont go trying to make Bob's oscillator run at 5V!

Oh that's only for the experimenting board, since I didn't feel like hauling the bench PSU over here to the computer & scope place (yes very suboptimal "lab" setup, I don't have nearly enough space...) - I didn't touch the circuit for the PCB I was about to make, I even kept the ugly "let's use one half wave per rail!" dual 12V PSU :-D

Oh, you cannot add an EQ coil to this circuit,

Why is it called that, "EQ coil", btw?

A small solderless plug-in breadboard is extremely useful to have for quick tests.. Often I work with both at the same time, particularly if I dont know what im doing ;-) ... Build a small section on the plug board, F about with component values, measurements etc, move to breadboard when its working, build next doubtful section on plug-board, interface with wires to previously transferred section if building a complex circuit, and so on.

That sounds like a time saver for some situations, I might think about getting a small board then. I just think I might not trust to be able to control the quality of connections with more than 5 parts involved, which could result in eliminating the otherwise saved time vs. (de-)soldering everything, and numerous palm impressions on the forehead.

If I knew a good place to upload such things to I'd post the ltspice file."  - tinkeringdude

I use MediaFire and am pretty happy with it.  It's got ads but it's free:  http://www.mediafire.com

"Actually, in spice, I also have a file where I added the reference oscillator and the detector and a lowpass filter, and looked at the result, while playing with the variable osc's capacitance, to see the heterodyne freq jump around in audible range. But it (the antenna capacitance having influence) worked only when I set the antenna coil from 40µH to 0.0000040µ or so i.e. basically nothing. And the needed changes in capacitance to really see something seemed rather large (tens of pF or more), on which an EE I know commented "spice and reality are not best friends, don't be fooled by fair weather harmony""

Without access to your sim I can't be 100% sure, but it's been my experience that the EW oscillator is almost impossible to understand from a simulation perspective.  Looking at the circuit I understand how it the drivers get and keep things oscillating, but the action of the large series ("EQ" - so called because it has some effect on the pitch field response) inductor on the tank LC is mysterious.  Thierry has a nice mathematical treatment of it that seems to account for some linearization, and Fred understands it from that angle, but I've never fully grasped it without resorting to technical flights of fancy.  I believe what happens is the tank LC stimulates the oscillation of a second LC, that of the EQ inductor and the antenna & hand capacitance.  The resonance of this second LC swamps the resonance of the tank and gets it to do its bidding, but if you don't get the relative tuning just right it doesn't behave well and can draw a lot of current and cook the transistors.

The "magic oscillator" angle is the toughest thing to move past and see for what it is (mostly a canard IMO) when doing Theremin research.  Just because an oscillator appears complex, or has alien drive compared to other oscillators, or has unexpected stuff hanging off of it doesn't mean it will behave any differently or have better linearity than other more conventional oscillators.  Having said that, the EW oscillator is likely an exception to this rule, but the benefit of a small amount of potential linearization (if tuned just right) is more than negated IMO by the setup and long-term maintenance difficulties it will likely give you.  The resonance of the EW coil and the antenna & hand capacitance can be quite variable depending on environment, so I imagine many (most?) EWs out there are running around at least somewhat mistuned.

The EW oscillator is a semi-successful attempt at linearizing pitch response in the near field (the zone ~6" from the antenna - pitch normally "cramps up" here for mathematical reasons).  The EW Pro has even more EQ circuitry and reportedly is more successful in this regard, and perhaps more cantankerous as others have reported that the oscillator can stall (which to me is a no-no).  The TVOX addresses linearity not in the near field but in the far field via oscillator coupling, but I believe Thierry said it behaved best with this mechanism defeated?  I'm not a fan of the sound nor the action of oscillator coupling, and recommend you buffer the EW oscillators before the mixer to minimize it.

I may have missed something, but so far all of my experience so far has followed the Barkhausen criterion, which put simply is: you get oscillation when the delay around the feedback loop = 360 and the gain > 1. 

Provided your skills are up to it, your initial take is IMO the sanest, simplest, and most likely to give good results: construct one simple high voltage LC oscillator per antenna, and generate the frequencies to beat against them in digital logic.  If you go this route you should pick inductors that are relatively stable with temperature (this is why I go with air-core).

I am not going to put my head on any block over this - Found Thierrys explanation.

In its most simplistic terms (the way I think about things, LOL ;-) if one has a series LC circuit (resonator) it has two 'zones' - one below resonance, one above resonance.. (see here) - Below resonance, the series inductor with antenna (capacitance)  is "seen" as primarily a capacitance in parallel with the oscillators tank, Above resonance its "seen" as primarily an inductance in parallel with the tank.

Oscillators (well) designed for direct antenna connection (no series inductance between tank and antenna) they have a big inductance || small capacitance (680uH || 180pF for example) so change in the || connected antenna capacitance gives a much bigger change in oscillator frequency (higher sensitivity)

If you look at tanks (well) designed for use with series inductance to antenna, you will find a large capacitor || a small inductor - If you connect such a tank to the antenna without a series inductor, you get extremely low sensitivity, because the few pF antenna capacitance is || to a massive tank capacitance (10pF || 2200pF || 100uH for example) ..

So what does the series L (+ antenna C) do ?

Well, operating above its resonant frequency (which changes as the antenna capacitance changes, and goes lower the closer one gets to the antenna) one ends up with effectively a changing INDUCTANCE across the tank - this is why a small inductance (and larger capacitance) is needed for suitable oscillators.. The series L effectively converts the changing antenna capacitance to a changing tank inductance, and has a large 'amplification' factor and other benefits.. (including increased antenna voltage)

In order for the above to work, the series resonator must always be operating at a frequency above its resonance as in, the oscillator must always be driving at a frequency above the antenna circuit resonant frequency - the fact that the oscillator frequency is (strongly) influenced by the antenna resonator is probably what makes tuning the two such a pain in the arse! - its SO much easier when one can tune the antenna resonator and leave the oscillator frequency alone, as one could do with early theremins or can do with an adjustable antenna.. I have got rid of the tuning on my EW board and just tune by adjusting antenna length (well, actually, am messing about with using the tuning of the reference as a linearity control - but thats a different topic ;-)

But Thierrys article goes into the maths etc of how linearization works - I see it in visual terms - superimposing the linear slope of the the "effective inductance" on the non-linear antenna capacitance-distance relationship... But I am not able to verify the maths or conclusions thereof, nor dispute it - all I can go on is what I believe ive seen (and with theremins I no longer believe what I see ;-) and "accept" explanations that I am too dumb to fully understand - or reject them if I find physical reasons to do so.

Fred.

 || = Parallel with

Oh - I have not mentioned the least employed topology, which Dewster has promoted (and perhaps "invented" ) and has IMO many benefits.. That's having a series LC Oscillator.. As in, the series antenna resonator us the oscillator tank. This topology doesn't have the characteristics of || tank with series resonator, and doesn't on its own give linearity improvement, but its certainly (IMO) better than || tank without linearization, and is the topology I am now using. (thanks Dewster - it solves many issues!)

Oh - Cooked transistors isn't something I have seen from ||- configurations, but that's an aside.. I cannot see how this can happen, but not saying it doesn't.. I am more inclined to think ESD takes out transistors.

"I am not going to put my head on any block over this..."  - FredM

Fred, I hope you don't think I was trying to put you on the spot!  You (and Thierry) just seem to have more of an intuitive grasp of the EW EQ.  I'd look into it more and try to figure it out if I thought I might use that approach in my own stuff.

"Oh - Cooked transistors isn't something I have seen from ||- configurations, but that's an aside.. I cannot see how this can happen, but not saying it doesn't.. I am more inclined to think ESD takes out transistors."

Interesting.  I've seen high currents in my spreadsheet sim, but ESD could certainly account for failure as well.

"Fred, I hope you don't think I was trying to put you on the spot! " - Dewster

Oh no! - I never thought that for a moment!  .. Just preempting others who seem to try to "put me on the spot" at any chance they get ;-) .. "Paranoia it strikes deep - into your life its gonna creep - and it starts when your always afraid.. step out of line and some troll is gonna shoot you down .." - Niel Young (I think ;-)

I truly dont know how effective / ineffective the the series LC || tank scheme is - All I know is that for some designs its certainly more effective when correctly tuned! ;-) .. The object is to increase the rate-of-change on the VFO frequency in the far-field, and reduce it in the near-field, and visually I can percieve this function from the series LC...

However, keeping the the idea of "increase the rate-of-change on the VFO frequency in the far-field, and reduce it in the near-field" in mind, this can easily be achieved by deriving a VFO frequency related control signal, shaping this signal, and using this to control a VCO (or feed back to a CV node in the VFO) to achieve the required correction - And this I know works, can be seen to work just by looking at the control signal, and is the way I am now going (in combination with a variant of your series LC oscillator).

Added ->

"Interesting.  I've seen high currents in my spreadsheet sim, but ESD could certainly account for failure as well."

When you say "high currents" are you talking about currents (and related dissipations) capable of exceeding the 2N3904 continuous current rating (200mA) or dissipation (625mW) ? .. I haven't seen anything close to this.. I may be missing something (particularly in terms of the relationship between the series effective inductance across the tank) - I can see that at resonance this could nearly short out the tank.. But even so (with the EW circuit) - there's a 2k2 emitter resistor, so even under a worst-case scenario I cannot really imagine (a stall due to tank inductance reducing too greatly, or a rapid sequence of stalls and re-starts) the current (C-E) through the transistors (or particularly Q1) cannot, AFAICS, exceed the current available from the voltage on Q2:C to -12V via R2 (2k2)..

But R2 has a constant voltage across it (-0.6 on emitters, so about 11,4V across R2) so current through it is regulated at about 5mA... And therefore I cannot see the Ice of either transistor exceeding 5mA by much..

I looked into this because I was thinking about putting beefier transistors into my EW board, due to tales of cooked transistors - but could find no mechanism to cause these transistors to operate outside their specifications (or even close to this)..

The current through the series (antenna) LC is sourced (and sunk) AFAICS from the supply via the tank, and the current through Q1 is merely enough to excite the tank - it is not the source of the 'antenna' current as far as I can see from every simulation I have run and every measurement I have made.

So (until shown a mechanism that contradicts this) I believe that cooked transistors are most likely caused from electrostatic discharge to the antenna, and that putting a discharge tube on the antenna will solve that problem.

But please - If I am wrong and anyone can see why I am wrong and explain this - Let me know! I could easily be wrong!

Fred.

>>

I have been focusing perhaps too much on collector current - the problem (if there is one) might have more to do with dissipation.

If we allowed peak current of 10mA (Ic) and peak voltage of +24V, we get 240mW .. The part is rated 625mW at ambient of 25C.. So the transistor should be able to pass ~25mA continuously (DC) with 24V across it without exceeding its specification.. And if operating correctly, it will only be dissipating only a small portion of the time, and not have 24V DC across it, but be conducting for a portion of the 24V sine on the collector - average dissipation probably in the order of 100mW.

However, the ability to dissipate the heat is a function of ambient temperature (the hotter the surroundings, the less the dissipation efficiency) and if the voltage on the collector was by some mechanism (presumably related to resonance, but I cant see it) elevated, then I suppose the transistor could get warm...

I really dont see it - no matter how I run the numbers, I dont see a circumstance where Ic or Pd can exceed the specifications of the part, or even get close to doing so.. But...

More ;-) --->

Collector-base breakdown voltage is 60V ... So any -Ve transient on Q1:C >(+/-)60V could destroy it. But what could cause such a condition other than ESD or some other abnormal outrage? - I really dont know!

So I've read several threads here about the Etherwave and that it's recommended to use the newest component values, i.e. from the "HotRod" PDF.
I was just updating my schematic in DipTrace with those new values (to generate a BOM later), to be shocked by a change in circuit, while having my freshly arrived PCB using the "EMtheremin" circuit laying next to me ;-)

Maybe there's even more, but what I've seen so far:
In the old doc, C11 is a polarized cap, with + going to GND.
In the new doc, C11 goes to +12V instead, and is no longer polarized.

What is the change for? How important is it?
I'm thinking about soldering only the other end of the cap and connecting the one in question to +12V with a bodge wire...

By the way - these linearity improving addon modules, is the schematic available somewhere or are they commercial-only?
(some more bodge wires won't hurt then, as the old schematic also does not have the extension connector :D )

There are lots of changes -

IMO, dont get confused or worry about them.. C11 works fine connected to +V or GND, its like this - for RF, both +V and -V and GND are all GND (0V) because they are all capacitively coupled to each other... I prefer to go to GND rather than a supply rail, but both work.. The important thing is that whatever capacitor you use, it should have low ESR / ESL -

On the linearity improvement 'modules'..

They consist of a couple of transistors that buffer the oscillators prior to the mixer - This eliminates (or reduces) the coupling between the oscillators so changing far-field (bass) linearity.

One is commercial and has never been published, there are others that have been published .. The commercial one is fitted to many instruments used by Pro's - This MAY and also MIGHT NOT mean that its any better than any other - but at least you know its been tested by Pro's and not found wanting... But I am absolutely certain that alleged 'dangers' of Dewsters circuit are utterly bogus, and this alone would make me opt for his circuit even if I was a millionaire!

This thread is the most tested non-commercial one - Read the thread from page one (where the schematic is) and make up your own mind about any risks, or whether there's something else going on.

I think I may have been the first to suggest buffering the EW oscillators years ago on TW.. I wouldnt dream of designing a theremin without buffers, and would strongly advise you to fit them - but read all the posts because there are some important issues.

Fred.

Ah, thanks!

Yeah, even I as a "semi advanced of n00b" :D wondered why they weren't buffering them, IIRC the article was published in the 90s, transistors weren't expensive anymore for a long time, and I couldn't think of another reason, well other than "this has worked in the past, let's keep it that way".

Well, but first I'll try to get the basic version running on my non-Bob-approved PCBs and with the variable coils I found. Some east european old radio things adjustable 45..100µH shot off ebay... considerably less pricey than ordering some used toko, which I only find in the US -> expensive shipping, and higher per pcs price.

Well. couldn't find any 47µ for the newer schematic that uses fixed + variable in series. Also couldn't find 68µ. But if this doesn't work with the lowest adjustment of those 45..100µ for the volume osc, I may just open the thing, try to melt the wax or whatever is on there, and unwind some... Fortunately, the seller provided a logarithmic sheet for number of windings -> inductance for the cores. I just hope I can open the thing without injure ;)
Btw, could a second, much smaller coil over the 100µ one (10 or so windings vs ~ 100) have any disturbing effect if left unconnected? (those things have 2 coils in them)
And should the metal cans be grounded? It looked like that on an EW PCB foto, but I've never held one in my hands.

Oh! Speaking of grounding - is the metal front plate a must, or can it be wood / plastic?

And, I suppose with just a 10x probe I won't be able to meaningfully look at the unbuffered oscs with a scope, eh?

One thing at a time ;-)

"And, I suppose with just a 10x probe I won't be able to meaningfully look at the unbuffered oscs with a scope, eh?"

No - and this is another damn good reason to buffer the oscillators! ;-)

"Oh! Speaking of grounding - is the metal front plate a must, or can it be wood / plastic?"

Adhesive aluminium tape works wonders at turning wood / plastic into metal ;-)... I have one roll of copper tape with conductive adhesive (expensive) - solder wire to a small piece of this which I stick to the (cheap) aluminium tape..

As for inductors / coils - I use 42IF106 IFT transformers for almost everything - you can get loads of variable inductance values from these .. search TW ! ;-)

"Adhesive aluminium tape works wonders at turning wood / plastic into metal ;-)."

Hehe, yeah for another project I planned to stick regular alu foil to the wooden case with glue... but if there's already adhesive one, even better. And then a screw with a washer pressing on it for contact.

If the damn mouser (germany) site wasn't on strike right now the remaining parts would be "on their way" already, tsss...

One part hard to come by so far is that mic stand mounting flange... almost any US online music store seems to have them, but not here... no equivalents either, I've asked e.g. thomann directly...

Is the placement of screws in the case critical? I guess I'd opt for the smallest screws that could possibly hold it together, with some wood glue also.