EM-Theremin, diode D4 and waveform

Hello everybody!

I've successfully build my first Theremin from the DIY Article, read the hodrodding manual and instructions from this forum. Now its working quite fine!!! Thank you guys!

After all, I've started trying to understand the stuff I've build which leads me to a handful of question, first is coming in this thread.

On low frequencies the sound is not very soft. I've managed to get an oscilloscope and poked around a bit. After I've build a testciruit with two oscillators running on different frequencies, mixed them through C2 and C6 I've seen a very nice mixed signal - as it should be. Well rounded waveform. Then I've attached the detector and my waveform was gone (top of the halfwave not in the middel but a lot more on the right side). After trying a bit I've changed D4 (1N4148) to the voltage regulator diode (something 1N400x) - and the signal was fine again. I've not tried over all frequencies. Just my first result - I think now I've to recalculate the divider R23/R24 because of the different signal level.
I've read that I'm not the only one that is unhappy with low frequencies...

Any suggestions why this happens and why the other diode seems to make a better Job? I can attach pictures from the scope if you want (and I 've found out how) than you can see the very(!) different waveforms.

edit: here are the pictures (http://private.net-fb.de/theremin/index.html)

This leads me to the second question: The mixed signal is rectified with an AM-Detector as far as I understand. But the signal is a signal, that can be seen here Wikipedia: Beat frequency (http://en.wikipedia.org/wiki/Beat_%28acoustics%29), which, if cut off in half, is not really a sin. And especially within the theremin circuit this leads to "peak, near nothing, peak,..." which sounds not very good. With the "right" detector, which also uses the lower part of the mixed wave, the frequency would be halfed. Which, if I tune the theremin to the needed range and have understood the posts right, can overload the transistors... Any suggestions? I'm thinking about to replace the whole (*lol* not very much parts) detector...

Or is this wanted for a "typical" Theremin sound?

edit: some typos

[i]"I'm thinking about to replace the whole (*lol* not very much parts) detector..." -fbarth [/i]

I will try to get back to the other questions later - after looking at your waveforms.. but to 'answer' this one..

Replacing the whole detector circuit does make a dramatic difference to the sound - but there are implications:

With reference to the EW (not EM) schematic, you will see that C2 and C6 effectively couple the VFO and RefO together, causing them to pull each other towards 'lock' as their frequencies become similar (at the bass end). The effects of this 'pulling' is A: to cause needed 'compression' of the distance between intervals at the bass end .. as in, linearity is affected, and B: To change the wave shapes dynamically as a function of frequency - particularly at the bass end.

I have played with changing the mixer - and did the following:

1.) connect a trimming capacitor between Q1:C and Q3:C.. A couple of pieces of insulated wire twisted together for a few pF does the job.. adjust the length / twists - this will adjust the coupling to taste.. You can also add a third wire to a 15k resistor to ground, to simulate the loading of R23+R24, but its not worth the bother!
2.) buffer the signals from each oscillator - Q1:C and Q3:C with a high-z FET .. The Fet buffers on the EPE-2008 theremin (http://www.element-14.com/community/docs/DOC-16872/l/epeanot2pdf) are a good choice, particularly if one wants to use the following mixer circuit (which works well).
3.) feed the buffered osc signals (apropriately scaled and biased) to whatever mixer you wish to use.. You could any 4Q multiplier, or any other transistor / diode mixer.. bufferring the signals gives freedom from the constraints of loading the oscillators and therebye changing the front-end performance.

IMO Bufferring the oscillators makes playing with theremin circuitry much easier - you get rid of all the interactions and problems.. you can probe buffered signals without causing loading - try putting a 'scope probe on Q1:C or Q3:C and the loading of the probe will utterly change the frequency and wave shape you are looking at - but you can probe the output from the buffer without problem.

Fred.

[b] Added (1) -> [/b] [i]" and have understood the posts right, can overload the transistors... "[/i]

I have not expierienced this problem, even with oscillators 'stalled' for several minutes.

looking at current paths through Q1 Q2, and imagining Q1:C connected directly to +12V - Maximum IC is determined by R2, so is 24/2k2 = 11mA. Worst case dissipation would be about 65mW when ZC was about 6k..

So , as far as I can see, no mechanism for destruction, both IC and Pd are well within the transistor specification. Likewise, base current is limited to safe levels by R1+R2 and R4+R2.

The only mechanism I can see would be a sudden halting of oscillation when the inductor(s) are most charged - one could get a transient event where the voltage on Q1:C rises above the Vce breakdown voltage (forward or reverse) (if you stop current when it is flowing through an inductor, the inductor will attempt to force this current flow to continue - this causes voltage across the inductor to increase - somewhat simplistic explanation, but it will do) -

But, as I say, I have never had this happen on EW type oscillators.. (and cannot actually see a transient condition which would cause this - except perhaps the disconnection of all capacitance in the LC circuit so there is no path for charge current to flow..).. I do, however, strap a small neon (NE1) discharge tube across my 'tank' circuits which discharge voltages > about 80V, and this may be enough to inhibit destructive events (I put these in circuit to protect against electrostatic discharge from the antenna).

If you want to play safe, look for a transistor with higher Vc rating.. I use ZTX450's for a lot of my oscillators, and these work happ

The mixer signal on the EM/Etherwave is not intended to be a pure sine wave, the clamp/clipping diode is there in order to pre-distort the signal in a (roughly) similar manner as it was on the old tube theremins which also took only parts of a half sine into account by either applying a huge negative bias to the grid of the mixer tube or by overdriving the mixer with an input amplitude of up to 70Vpp in order to obtain a more or less soft clipping.

Decoupling/Buffering the oscillator's outputs is a good idea in general but the EPE circuit (parts) will not work since it can't handle the 24Vpp + 12V DC offset output amplitude of the oscillators. Out of that I have done tests on my Etherwave Standard which is basically the same as the EM, and I found that removing the (wanted and inteded) interaction between the oscillators by inserting buffer stages will worsen the already bad linearity in the lowest register. And the gain of a somewhat less distorted half octave in the bass range does not justify the modifications and the loss of playability.

If you add some wanted coupling between the oscillators as suggested by Fred, you'll loose all what you intend to obtain since this coupling will anihilate the decoupling.

So Fred's theory is not bad but will not be applicable in this special circuit design.

Hi Thierry -
No argument that just lifting the EPE buffers without changing component values, will cause problems.. However, that is not what one would / should do... I should have elaborated though - I thought it was obvious - but if you did not understand, then I had better explain some more ! You are one of the most technically astute readers on this forum!

You say: [i]"Decoupling/Buffering the oscillator's outputs is a good idea in general but the EPE circuit (parts) will not work since it can't handle the 24Vpp + 12V DC offset output amplitude of the oscillators. Out of that I have done tests on my Etherwave Standard which is basically the same as the EM, and I found that removing the (wanted and inteded) interaction between the oscillators by inserting buffer stages will worsen the already bad linearity in the lowest register"[/i]

The issue of signal levels is resolved as follows: Place a resistor in series with the coupling capacitor one connects to the oscillator output (say Q1:C on the EW circuit) - So it would be: Q1:C -> 330pF -> 330k -> Gate of FET.. Fet Gate pulled to 0V via 100k. The exact values need to be determined - but the essential feature is the attenuation provided by the series resistance between gate and capacitor, and the shunting/biasing resistance between gate and 0V. One does not need to even think about the DC offset - the oscillator outputs must obviously be ac coupled to the buffers - one only needs to reduce the signal levels from ~24V P-P to a lower AC level for optimum FET operation.. One is also free to bias the fet as one chooses - this is a useful mechanism for adding distortion.[b] Added ->[/b] Not tried this, but one could put a capacitor across the 'shunting' resistor - this would attenuate the signal level, and smooth the waveform if one wanted a purer sine wave.

The reasons for having buffers is not to prevent oscillator coupling - it is to provide low impedence signals from the oscillators - signals one can look at, and work with, without the problems (influencing frequency and waveshape of the oscillators) that one has if the signals are not buffered.

The issue of linearity is addressed by having seperate coupling of the oscillators by a capacitance connected between Q1:C and Q3:C - using twisted wire, as described in my last.. You say: [i]"If you add some wanted coupling between the oscillators as suggested by Fred, you'll loose all what you intend to obtain since this coupling will anihilate the decoupling."[/i] But I do not understand your meaning! - You have two seperate buffers, each independently AC coupled to the signals from the oscillators, and presenting extremely small load (high Z) to these signals..

Then you have a seperate capacitive coupling between the oscillators, so that they affect each other. If we compare this to an unmodified EW, we have 2 capacitors connecting to the diode + 15k load.. These two capacitors share this load, and HF components couple back (greatly attenuated) to the oscillators, causing them to influence each other.

By coupling the oscillators together via a much smaller capacitance, one achieves the same objective - with the following advantage.. There is ONLY this deliberate [b](adjustable)[/b] coupling to think about - one does not need to account for the effect that changing the coupling capacitance will have on the performance of the mixer.. In the EW circuit, if one wanted to change the coupling, one would need to change the value of the mixer components in order to balance the levels etc.. Adding buffers and seperate adjustable coupling capacitance gives one the freedom to select the best (to ones taste) compromise between linearity and sound.

I do agree that there are better designs for buffers applicable to the EW than the EPE one - But the basic principle (FET Follower) is a standard circuit, and quite well suited to this application.. and a few tweeks to the EPE design does the job. O

I have just dug up my old notes relating to the experiments on bufferring the EW signals.. what I have posted in my last was all from memory, and was a bit sloppy/lazy! - I thought that just pointing in the general direction should be enough, and using existing examples (FET/EPE buffers) would save me time.

In fact, in my final build, I used NPN (ZTX450) voltage followers and not FET buffers, as there was no requirement for extremely high Z into the buffers..

If people want, I will post the schematic to Element-14.. Or you could just email me and I will put it into a .pdf and send it to you.

Fred.

I'm interested in your schematic! I'm not very experienced in electrical engineering so I'm happy about everything that helps me understand - thanks a lot so far! The idea with a buffer circuit is great. Because of my low experiences I've had no idea how to handle those problems. Now I've had a look at FET and Op-Buffers and have learned a lot more. Maybe I'll get those parts today than I can play around a bit more.

I've build the Theremin only to have one and make some noise with it (maybe someday even music) - and of cause it was a nice Christmas present for my girlfriend. Welle:Erdball, one of our favorite german bands is using it for efects, from there I've had the idea and my girlfriend loves the sound - and especially the way it gets played.

But now I'm fascinated from the technique used and its long over time for me to understand some electrical stuff. The circuit is very good for learning understand those stuff! I'm happy that I've ordered all parts twice. The second Theremin (mine) will be much more modular so I can play around with it and can easily exchange or add parts.

For everyone who is interested, I've managed to get the pictures from the scope uploaded (URL is my first post) - maybe someone can have a look. the last two photo lines show the difference between using one or the other diode on two frequencies.
The pictures were made with two FPO's without pitch tuning circuit, tuned two different frequencies.

I suspect that the different waveforms from 1N4148 and 1N400x might be due to the different capacitance presented by these diodes..

When a diode is reverse biased, it becomes, effectively, a voltage-dependent capacitance.. The greater the reverse voltage, the wider the depletion zone, and the lower the capacitance - When subjected to an AC signal, this produces distortion of that waveform.

The 1n4148 has a smaller 'chip' area so lower capacitance than the larger 1n400x power diode with its larger chip area.. Its capacitance is greater, and (possibly more important) I think the degree of change in capacitance will be greater.

I think this might (note - I say might.. I am not sure!) be the reason for the difference.

I am working on an oscillator / buffer at the moment - so will take some snapshots of the waveforms + post these and the schematic within the next week.

Fred.

Schematic for buffer is now posted Here (http://www.thereminworld.com/forum.asp?F=3&T=4805&cmd=p)

Fred.

Have there been any updates to adding a FET(s) or transistor(s) buffer to the etherwave or EM schematic just after the pitch oscillators and before the detector?

 

Thanks dave

(please excuse the interruption)

Hey, Fred. Great to see some postings from you again. I have missed trying to wrap my head around the electronics, and the Fred/Thierry dynamic. :-)

Hope your health has perked up a little.

 

</interruption>