Ebow Exposé - Part III

[Rob Strand]

Yeah that will be my next headache... rewinding the input coil. I rewound the output coil by hand yesterday using wire from a trashed inductor. I initially was just going to look at it with the curve tracer "octopus" I made for such things. When I was done with my third or fourth attempt I finally got it wound without breaking the wire. I measured it and it was exactly 8 ohms before I put it back on that mangled PCB! Pure luck!  Still couldn't get it to show any signs of hysteresis or saturation in the audio frequency realm.

Well done!   If you got 8 ohm then there's a good chance you have both the turns and wire diameter close to the original.

If that's a ferrite magnet it is very weak. Does not appear to be attracted to anything ferrous whatsoever.

That's weird.   Some of those "plastic" ferrite magnets are quite weak.

FYI, the ring can also be a magnet.

Knowing all the materials is important to work out exact what magnetic field comes out the front, what the coil inductance is, and how the receive coils senses the string (also how drive coil drives the string).   The "working out" part isn't easy after that anyway.

Now when I simulate it using 110 ohm/4mH input coil & 8 ohm/3mH output coil I see waveforms that really start to make much more sense.

Cool.

The way you have it set-up ATM has a sine-wave input.   This is very easy to get to work and to play around with.    The real device is much more complicated:   the output coil *current* causes a force, the force moves the string (act on the mass and tension), the velocity of the string creates the pickup voltage.   The string is like a comb filter.  In order to raise the bar a bit on the simulation you have to add a lot of complicated stuff.   The string response would need time delay and a mixer.     The difference in this case is the circuit forms an oscillator and there's no sine source.

I also tried to simulate couple of the circuits shown in the LM386 application notes, and some won't even do anything - like the Weinbridge Oscillator, it does nothing

Simulating oscillators is tricky.  An oscillator needs a kick to get it started.   In a real circuit this comes from the amplifier and power supply.   In a simulation you can add a current pulse to kick-start the circuit.  Another way is to set the voltage on one of the oscillator caps using Spices "initial conditions", then you must set use initial conditions on the .tran simulation.  (On a real ebow the mechanical vibrations and any hammering on the frets kicks off the oscillation.)

I was thinking the same thing this morning. Maybe it's to "focus" the energy towards the string, preventing "cross talk" with other nearby things. And may also have a secondary effect of providing shielding from EMI/EMF.

The focusing thing is important to how it works for sure.   That's why it makes a big difference knowing what all the materials are, what are magnets, what materials are magnetic.

These things are what they are.   There's also behaviours which people are used to, like getting stronger signals when you place the ebow close to pickups.   No doubt in this case the drive coil field is feeding straight into the guitar pickup.  If you mess with the construction or geometry it won't behave the same.   For example a radial field magnet with all components magnets probably won't have such a strong field going into the guitar pickup.

You can see a similar looking coil design in these old buzzer-speakers,
https://www.youtube.com/watch?v=UkQhRFS6sRM

While these look similar in construction to the ebow they may or may not actually be similar (From your description of the parts they aren't these same).  The buzzer-speakers have ring magnet on the outside.   The back side of the ring magnet goes to a metal plate then the metal plate goes to a the central metal pole face - the whole idea of all that is to "conduct" the magnetic field out of the front of the pole piece.   Most of the field is between the pole piece and front edge of the ring magnet.

How much the field "projects" out the front is closely related to the diameter of the pole face.   These buzzer-speakers have both very small pole face diameters and also ring magnets, so the field won't project too far out.   On the other hand the ebow uses a larger diameter pole face, and if it uses an axially polarized magnetic then the field will deliberately extend further out.

By the same token, no matter how you build it you will get some basic ebow behaviour  - for that you only need some form of drive coil and some form of receiver.

If you want to be devious you don't even need a receive coil.   Use the guitar pickup itself!    Of course now you will need wires from the guitar output back to the ebow amplifier.    Also an external ebow won't get a receive signal if the guitar volume is down.   An on-board ebow unit would get around that.  It doesn't get around the wire to the drive coil - that needs to be in your hands.   It does save winding a receive coil if you want to play.    (I wouldn't be in favour of using the guitar pickups as motors.)

[johngreene]

Is the ferrite there to aid immunity to RF?

I started looking at this thing with the aim to try to back-engineer some of the unknowns through devious means.  I wrote a whole heap of analysis and thoughts but I haven't got back to it.  It keeps growing.  Initially I just wanted to give Paul some estimates to cross-check the coil resistances but it's blown out to bigger issues!

To summarize a few thoughts:

That shielding would only make sense for the receive coil.   Maybe hum prevention.  However since we aren't listening to the receive coil I suspect you could handle quite a bit of hum.

The shorter magnetic path around the coil (due to the outer ring) sets the inductance.   This may be the driving factor for the ring, especially say resonating with the 33nF cap to improve receive gain.   When I looked at this I noticed the Q is very high.  The 13k resistor doesn't look bad but it could be lower.    Not having a reliable resistance or inductance for the receive coil put a spanner in the works - sort of concluded you would need to build something and tweak it.   Then make sure it worked in normal and harmonic mode.

I haven't finished looking at the effect of using an simple axially polarized magnet *and* having the drive/receive coils with outer rings.   What I was doing is looking at the magnet as creating a magnetic field through the string then that moving "magnet" is inside the gap field between the pole-piece and the outer ring.   The "magnet" in the string points point in one direction to left of the coil center and in the other direction to the right of the coil center.   I don't even know it that will add anything as a motor/sensor - if not the outer ring would *only* be used to set the inductance.

Having a ferrite sleeve will increase the inductance but more importantly it contains the magnetic field a lot more than if there wasn't any. So my guess is that similar to the reasons you use coils with ferrite sleeves in old RF designs with tuned circuits, the sleeves are to prevent magnetic coupling between the input and output coils. And not much more.

[Paul Marossy]

OK HERE'S ANOTHER UPDATE:

I managed to rewind the input coil using a clever DIY winding jig that I made out of stuff lying around. It worked remarkably well! It's not the tiny wire gage of the original but I did manage get to 108 ohms DC resistance using wire from some little transformer I've never been able to find a use for, so that's probably good enough to for testing. Not trying to make a perfectly functioning Ebow, just attempting to resurrect a completely hacked apart one and see it function again. Thanks Shango66 on YouTube  for giving me this idea... working on TVs and radios is a helluva a lot easier as far working on them goes!

Working on this thing requires the patience of a saint and the dexterity of a neurosurgeon.  I tried to get it to work with the original IC chip but I think it may have been dead (I seem to remember that the person that originally dissected it said it was dead before they hacked it apart... but that was back in 2007 sometime and don't remember for sure). So I cut the old one off and soldered in a new LM386 but that had problems... got very hot a couple of times so I think it got fried somehow. So I replaced it with another one and verified all of my connections once again and got rid of the crap sockets from Amazon which I think was causing some of my problems. This time it did not get hot but it was STILL dead as a rock, and the voltages were I measured obviously very wrong. After that I came to the conclusion that perhaps the electrolytic caps were all suspect, in spite of looking OK on a curve tracer, and after all they did have substantial physical deformities in them. I don't have a 220uF axial cap on hand but I did replace the other two. BTW that cap in parallel with the diode is acrually a 4.7uF, not a 10uF like I originally thought. I replaced it with a 10uF, don't think that is too critical. When I was done replacing those two caps I still had a circuit that did nothing to excite the strings but this time I could hear a faint whrring sound! It was emanating from the circuit board. Sounds like it's about 10kHz or so.

I came the conclusion that I was on the right track as far as the LM386 being an oscillator. Because of the fact that you can put an Ebow over a non-vibrating string and it will excite the string into vibrating at its resonant frequency, I was thinking that it was probably oscillating all the time and when you put it over a string it then assumes the frequency of the vibrating string. I modeled my circuit as it is now in LTSpice and found something interesting: With 0mV and 0Hz on the input, it flatlines until somewhere around 6.5mS and then goes into an oscillation - very high frequency and it's square waves! All of the LM386 LTSpice models I can find all seem to have the same flaw with completely unrealistic output voltages on Pin 5 but if I ignore those voltages and do some hacks to make output voltage realistic the waveforms look kinda realistic to me. The LM386 data sheet states that the output automatically assumes 1/2 the supply voltage but all of these LTSpice models are whacked when it comes to that.

Also discovered that the switch messes with supply voltage that the LM386 sees. IIRC it knocks it down about 2 volts. I'll remeasure all that stuff again when that time comes.

So at this point I am one capacitor away from possibly having a functioning Ebow circuit. Once I round up a 220uF cap I'll see if that fixes the problem. I have a hunch that is it because I have literally replaced every other component on that circuit board.

Last, I discovered that what I thought were ferrite rings are actually steel rings with a coating on them. Maybe it's some kind of thin ferrite coating? In any case, I was thinking about how they could act as shielding if they are not connected to ground (AFAIK) but since there is an audible whirring noise from the LM386, maybe it's to prevent that from being induced into the coils? I'm sure those rings are probably serving a few purposes simultaneously. I surmise that in an Ebow as it comes out of the factory you can't hear that because it's potted in that goop that smells a bit like gasoline when you pick at it.

I'll make another video that documents my process up to this point in the near future.

[johngreene]

.....

Last, I discovered that what I thought were ferrite rings are actually steel rings with a coating on them. Maybe it's some kind of thin ferrite coating? In any case, I was thinking about how they could act as shielding if they are not connected to ground (AFAIK) but since there is an audible whirring noise from the LM386, maybe it's to prevent that from being induced into the coils? I'm sure those rings are probably serving a few purposes simultaneously.

Magnetic shielding doesn't need to be grounded to help shield the magnetic field. Steel, aids the magnetic flux, reducing the inductance, but gives it a path of less resistance for it to follow.

[anotherjim]

Why not breadboard it? Build the circuit on the BB and temporarily mount the coils on something flat with wires long enough that you can sit the whole thing on top of a guitar and hold the coils over a string.

[deadastronaut]

/\ this...

i did the same when playing with sustainers ages ago...

nice work by the way paul. very interesting. 

[anotherjim]

Quick math says a good 9v battery into 8ohm is going to reach 2w and higher still if it's doing square waves. Current is getting on for 400mA. PP3 not going to live long. LM386 is going to get hot and trapped in goop, not going to get any cooler.

[Paul Marossy]

Quick math says a good 9v battery into 8ohm is going to reach 2w and higher still if it's doing square waves. Current is getting on for 400mA. PP3 not going to live long. LM386 is going to get hot and trapped in goop, not going to get any cooler.

Yeah that's true. I'm not sure that LTSpice is simulating it correctly anyway. One reason I want working circuit is so I can see the waveforms on the scope. And I will for sure be very closely monitoring the LM386 for any signs of meltdown before it happens. If all is well, then I can move onto putting it over some strings.

Why not breadboard it? Build the circuit on the BB and temporarily mount the coils on something flat with wires long enough that you can sit the whole thing on top of a guitar and hold the coils over a string.

Because I want to resurrect it and have it be usable in the plastic housing too. I seem to like difficult challenges. 

[anotherjim]

You still can resurrect it but it could be a surer path if the basic circuit components are proven first.

[Paul Marossy]

You still can resurrect it but it could be a surer path if the basic circuit components are proven first.

In my mind, assuming that I had an accurate measurement on the output coil, the only real question remaining is what does the input coil need to be?
if an 8 ohm speaker works why wouldn't an 8 ohm coil work? I guess it's the difference in current? Speaker is small compared to the coil?

[anotherjim]

8ohm obviously works as a pickup, in principle, but a small turns count means low output voltage. And isn't it going to be proportional to flux strength, angles, rate of change, distances etc? Given 50k load impedance for the amp (give or take any other bits on the input), you can have a more sensitive pickup with more winding for a higher output without being loaded down significantly.

[Rob Strand]

Thanks for the update.  I haven't had time to put my mind into it over the last couple of days.

the sleeves are to prevent magnetic coupling between the input and output coils. And not much more.

It's certainly possible.  This is where the engineering of such a device comes in.  What effects are significant and what are not.

I managed to rewind the input coil using a clever DIY winding jig that I made out of stuff lying around.

Way to go Paul!

It's not the tiny wire gage of the original but I did manage get to 108 ohms DC resistance

Looking at the tables I posted earlier the resistance is close to the 40AWG.  Which agrees with your comments about the wire.   It also implies the original coils must be in the 200 ohm to 600 ohm region.   The inductance will be high also.

I don't have a 220uF axial cap on hand but I did replace the other two. BTW that cap in parallel with the diode is acrually a 4.7uF, not a 10uF like I originally thought. I replaced it with a 10uF, don't think that is too critical. When I was done replacing those two caps I still had a circuit that did nothing to excite the strings but this time I could hear a faint whrring sound! It was emanating from the circuit board. Sounds like it's about 10kHz or so.
...
Also discovered that the switch messes with supply voltage that the LM386 sees. IIRC it knocks it down about 2 volts. I'll remeasure all that stuff again when that time comes.
...
I discovered that what I thought were ferrite rings are actually steel rings with a coating on them. Maybe it's some kind of thin ferrite coating?

Interesting info.  Thanks for posting it.

When placing a magnet on the back of something the steel rings make more sense to me.   Ferrite magnets will usually saturate ferrite.   In the current scheme, other than the magnet everything else is steel - a common set-up for a motor system.

The coating is probably a blackening oxide coating.  It's very common on magnetic parts.  The main purpose is to prevent corrosion.

I came the conclusion that I was on the right track as far as the LM386 being an oscillator.

No doubt about it.   The string is the resonant element, that why it follows the note.   No need for the circuit to oscillate by itself.

The LM386 data sheet states that the output automatically assumes 1/2 the supply voltage but all of these LTSpice models are whacked when it comes to that.

I don't trust spice models of IC s much at all.  There are good and bad ones out there and no way to work out what is what without a lot of testing.   I have built-up LM386 models from the internal schematic.  I found the gains and biasing worked out OK.

In my mind, assuming that I had an accurate measurement on the output coil, the only real question remaining is what does the input coil need to be?
if an 8 ohm speaker works why wouldn't an 8 ohm coil work? I guess it's the difference in current? Speaker is small compared to the coil?

Not sure what is going on with your coil.

If you have an coil with a DC resistance of 8 ohm then the resistance limits the current.   If you have a coil with a significant inductance then the impedance can only go higher than that.  FWIW, if the magnet is ferrite my inductance estimates could be high.  If we take the lower inductance estimate of around 2mH then at 300Hz the impedance due to the inductance is 3.8 ohm, so the total impedance at 300 Hz is sqrt(8^2+3.8^2) = 8.9 ohm.   Not a great deal higher than the DC resistance.   If the core saturated then inductance would drop but even then the impedance can't drop below the DC resistance of 8 ohm.

With 9V the LM386 should handle 8 ohm.     There's a couple of things I'd be thinking at this point:
- There is DC getting across the coil.  That could be caused by the 220uF output cap in the wrong way
   or the cap has a short.  It could also be a short on the PCB bypassing the 220uF output cap.
- The LM386 is oscillating.     The circuit doesn't have Zobel Network so there is a risk.   Maybe your batch
  of LM386's need one.   You can see HF oscillation on the oscilloscope.

[Paul Marossy]

I came the conclusion that I was on the right track as far as the LM386 being an oscillator.

No doubt about it.   The string is the resonant element, that why it follows the note.   No need for the circuit to oscillate by itself.

Right. But if the circuit is not oscillating with no input, then how does it excite a string that is doing nothing into resonating at its resonant frequency? Is it by some other mechanism? Is it like being too close to a loud amp and it feeds back? Ebow calls what their product does "direct string synthesis" and the Fernandes Sustainer system calls it "feedback sustain". Maybe they don't have the Zobel network because they want it to oscillate? We can only speculate about it. The fact that mine is faintly audibly oscillating at least tells me that the LM386 is functioning. 

The only way to really find out is to partially dissect another one, not by doing the harsh overkill that other people have done, but by only getting the goop off the foil side of the PCB where you can get some scope probes in there to look at what it's doing. I recently saw that Joyo is making a handheld sustainer now for half the price of the Ebow. I wonder if they basically did just that, or if they just figured it out on their own.

[johngreene]

I came the conclusion that I was on the right track as far as the LM386 being an oscillator.

No doubt about it.   The string is the resonant element, that why it follows the note.   No need for the circuit to oscillate by itself.

Right. But if the circuit is not oscillating with no input, then how does it excite a string that is doing nothing into resonating at its resonant frequency? Is it by some other mechanism? Is it like being too close to a loud amp and it feeds back? Ebow calls what their product does "direct string synthesis" and the Fernandes Sustainer system calls it "feedback sustain". Maybe they don't have the Zobel network because they want it to oscillate? We can only speculate about it. The fact that mine is faintly audibly oscillating at least tells me that the LM386 is functioning. 

The only way to really find out is to partially dissect another one, not by doing the harsh overkill that other people have done, but by only getting the goop off the foil side of the PCB where you can get some scope probes in there to look at what it's doing. I recently saw that Joyo is making a handheld sustainer now for half the price of the Ebow. I wonder if they basically did just that, or if they just figured it out on their own.

If the eBow can get a string vibrating on its own without the player having to pluck the string to get it started in anyway then yes, the circuit itself would have to be oscillating all the time with the feedback coupling being between the input and output coils. When brought close to the guitar strings and it starts vibrating this 'takes over' the oscillation. This is also known as an 'injection-locked' oscillator.

[Paul Marossy]

I came the conclusion that I was on the right track as far as the LM386 being an oscillator.

No doubt about it.   The string is the resonant element, that why it follows the note.   No need for the circuit to oscillate by itself.

Right. But if the circuit is not oscillating with no input, then how does it excite a string that is doing nothing into resonating at its resonant frequency? Is it by some other mechanism? Is it like being too close to a loud amp and it feeds back? Ebow calls what their product does "direct string synthesis" and the Fernandes Sustainer system calls it "feedback sustain". Maybe they don't have the Zobel network because they want it to oscillate? We can only speculate about it. The fact that mine is faintly audibly oscillating at least tells me that the LM386 is functioning. 

The only way to really find out is to partially dissect another one, not by doing the harsh overkill that other people have done, but by only getting the goop off the foil side of the PCB where you can get some scope probes in there to look at what it's doing. I recently saw that Joyo is making a handheld sustainer now for half the price of the Ebow. I wonder if they basically did just that, or if they just figured it out on their own.

If the eBow can get a string vibrating on its own without the player having to pluck the string to get it started in anyway then yes, the circuit itself would have to be oscillating all the time with the feedback coupling being between the input and output coils. When brought close to the guitar strings and it starts vibrating this 'takes over' the oscillation. This is also known as an 'injection-locked' oscillator.

This is exactly what I think it is doing.

[Rob Strand]

Right. But if the circuit is not oscillating with no input, then how does it excite a string that is doing nothing into resonating at its resonant frequency? Is it by some other mechanism? Is it like being too close to a loud amp and it feeds back? Ebow calls what their product does "direct string synthesis" and the Fernandes Sustainer system calls it "feedback sustain". Maybe they don't have the Zobel network because they want it to oscillate? We can only speculate about it. The fact that mine is faintly audibly oscillating at least tells me that the LM386 is functioning.

Oscillators always need a kick to start them off.

When you simulate an oscillator in spice often you find it doesn't oscillate.  That's because it's a perfect world in the simulation.   The only noise is perhaps numerical noise from the calculations.  There's nothing to kick it off.

In the real world there's usually something that kicks an oscillator into working.   Think of a guitar against a cranked amplifier.   If you damp the strings with your finger, the act of removing your hand moves the strings a bit and kick it off.  People walking around puts vibrations through the floor.   Even sound in the room can kick it off.

Once you start an oscillator the amplitude grows *exponentially*.

For the ebow:
- The biggest thing that kicks it off is when you fret the note.
- Next would be tiny motions of your finger on the fret.
- Moving the receive coil towards the string causes the magnetic field to change near the string.
   That will create a small pulse in the receive coil which gets amplified and that signal gets to output coil.
- When you move the ebow close to the strings the magnets push the strings a bit.
- Your body moving the guitar transfers bumps to the strings.

There's always something to kick it off.

If the eBow can get a string vibrating on its own without the player having to pluck the string to get it started in anyway then yes, the circuit itself would have to be oscillating all the time with the feedback coupling being between the input and output coils.

Early on I was thinking the ebow creates and LC oscillator with the 33nF cap  and the input coil inductance.  Later I had doubts that was the case.  You would see some high frequency oscillation there all the time and you would hear the stray field come out of the pickups (provided < 10kHz) and it would be very annoying ; eg. place ebow near pickups and damp the strings.   I don't recall hearing anything like that.   The constant oscillation would pull current all the time unless you can push the frequency up and the drive coil had a high inductance (so high impedance and lows current).

[johngreene]

Right. But if the circuit is not oscillating with no input, then how does it excite a string that is doing nothing into resonating at its resonant frequency? Is it by some other mechanism? Is it like being too close to a loud amp and it feeds back? Ebow calls what their product does "direct string synthesis" and the Fernandes Sustainer system calls it "feedback sustain". Maybe they don't have the Zobel network because they want it to oscillate? We can only speculate about it. The fact that mine is faintly audibly oscillating at least tells me that the LM386 is functioning.

Oscillators always need a kick to start them off.

When you simulate an oscillator in spice often you find they don't oscillate.  That's because it's a perfect world in the simulation.   The only noise is perhaps numerical noise from the calculations.

In the real world there's usually something that kicks on oscillator into working.   Think of a guitar against a cranked amplifier.   If you damp the strings with you finger the act of removing you hand moves the strings a bit.  People walking around puts vibrations through the floor.   Even sound in the room can kick it off.

Once you start an oscillator the amplitude grows *exponentially*.

For the ebow:
- The biggest thing that kicks it off is when you fret the note.
- Next would be tiny motions of your finger on the fret.
- Moving the receive coil towards the string cause the magnetic field to change near the string.
   That will create a small pulse in the receive coil which gets amplified and signal gets to output coil.
- When you move the ebow close the magnets push the strings a bit.
- Your body moving the guitar transfers bumps to the strings.

There's always something to kick it off.

If the eBow can get a string vibrating on its own without the player having to pluck the string to get it started in anyway then yes, the circuit itself would have to be oscillating all the time with the feedback coupling being between the input and output coils.

Early on I was thinking the ebow creates and LC oscillator with the 33nF cap  and the input coil inductance.  Later I had doubts that was the case.  You would see some high frequency oscillation there all the time and you would hear the stray field come out of the pickups (provided < 10kHz) and it would be very annoying ; eg. place ebow near pickups and amp the strings.   I don't recall hearing anything like that.   The constant oscillation would pull current all the time unless you can push the frequency up and the drive coil had a high inductance (so high impedance and lows current).

In Pspice it is common practice to step the power supply from zero to the supply voltage after the simulation starts to accurately simulate startup conditions. This is usually enough to kick off any kind of oscillator design.

Having a working eBow would definitely be helpful in determining how it 'should' work. There is the possibility that it does not have enough output -> input coupling to oscillate on its own but will start oscillating when it experiences a load from a steel string coming close to the coils. As you bring the ebow close to the string it will induce a small change in input voltage which would kick the output enough to get things started. Basically, just the movement of your hand may be enough to kick things off and start the string vibrating if there is a string close to the coils. 

[johngreene]

Out of curiosity I found this website which shows a 3.6M resistor between the output and the non-inverting input.
https://academia.jansensan.net/imca-470/no-signal-generated-so-far/
If I simulate this circuit, and step the supply voltage, it takes off oscillating if the input inductance is at least 800mH. This is due only to the 3.6M resistor providing the positive feedback. So I would imagine that will less inductance, it wouldn't self-oscillate, but would be highly sensitive to any 'movement' on the input amplifying that and hitting the string with more pull in the same direction.

This guy was able to get it to work using the commonly shared schematic and PC speakers for pickup and drive coils.
https://youtu.be/UkQhRFS6sRM

[Paul Marossy]

In Pspice it is common practice to step the power supply from zero to the supply voltage after the simulation starts to accurately simulate startup conditions. This is usually enough to kick off any kind of oscillator design.

That's a good trick to know but it didn't work for me on the one circuit I can't get LTSpice to start. 

Having a working eBow would definitely be helpful in determining how it 'should' work. There is the possibility that it does not have enough output -> input coupling to oscillate on its own but will start oscillating when it experiences a load from a steel string coming close to the coils. As you bring the ebow close to the string it will induce a small change in input voltage which would kick the output enough to get things started. Basically, just the movement of your hand may be enough to kick things off and start the string vibrating if there is a string close to the coils.

That could also be the case. LTSpice won't be able to simulate that aspect of it. It may be that mine is just sitting there oscillating because the output cap is bad. I will find out the answer to that question in the next few days. Still is a good sign though because I know that LM386 is doing something.

Out of curiosity I found this website which shows a 3.6M resistor between the output and the non-inverting input.
https://academia.jansensan.net/imca-470/no-signal-generated-so-far/
If I simulate this circuit, and step the supply voltage, it takes off oscillating if the input inductance is at least 800mH. This is due only to the 3.6M resistor providing the positive feedback. So I would imagine that will less inductance, it wouldn't self-oscillate, but would be highly sensitive to any 'movement' on the input amplifying that and hitting the string with more pull in the same direction.

That's interesting, haven't seen that one before. It seems that this Ebow circuit is a carefully orchestrated dance of several factors that all have to come together exactly a certain way for the magic to happen.

I wonder if any of these schematics on the web are pre-Plus Ebow, so they may be correct for ones manufactured in a certain time period. Most people seem to be looking at the Plus Ebow which was released in 1998. I do have a working Plus Ebow and a 3rd generation black one with white lettering but I don't want to donate either of them to science. 

[johngreene]

In Pspice it is common practice to step the power supply from zero to the supply voltage after the simulation starts to accurately simulate startup conditions. This is usually enough to kick off any kind of oscillator design.

That's a good trick to know but it didn't work for me on the one circuit I can't get LTSpice to start. 

hmm, because if I use a "Step Voltage" for the 9V supply and set it to delay 1mSec after the start of the simulation, it works perfectly to start it up oscillating. But I am not using LTSpice.

It seems to me that this is setup such that it is right on the brink of oscillating and as soon as there is some kind of 'reactive' connection between the input and output coils (i.e. the guitar string) it will start oscillating.