Ebow Exposé - Part III

[Rob Strand]

Out of curiosity I found this website which shows a 3.6M resistor between the output and the non-inverting input.

Based on Paul's receive coil and latest info on the coil I ball-park the receive coil at say 50mH and 200 ohm series resistance.

From Paul's ebow schematic the input circuit has Rf = 13k and C = 33n.

The coil series resistance is important to ball-park things since that represent loss on the feedback loop.   Given the non-laminated steel parts the expected losses could be very much higher than just the coil resistance.   Nonetheless it's interesting to try the 200 ohms alone as puts an ball-park value on the required value of feedback resistor Rf to make the unit self oscillate.

So I found with that set-up the feedback resistor only needs to be about 1MEG for self oscillation.   Even if the core losses were 10 times more that means Rf = 100k.   Oscillation frequency 3.8kHz.

We expect a resonant/oscillation frequency of approximately 1/(2*pi*sqrt(LC)) = 4kHz.

The unit has RF = 13k, that allow the  LC oscillator to oscillate even with high losses. 

The final test would be to put an oscilloscope across a guitar pickup (or some coil) then slowly bring the ebow closer to the pickup/coil and see if there is a 4kHz or so signal.   Too much metal around the ebow could stop the oscillation - for example the covers on a Humbucker will further add to the losses.

[Rob Strand]

I did a quick knock-up in LTSpice.

The aim was to have a look at the effect of changing the amount of feedback via the feedback resistor (13k on current schematic) and the amount of coupling from the ebow pickups.  The amount of  coupling is controlled by the coil design and also how close you place the unit to the strings.

Here's the basic schematic (yes I know I have an extra input cap).

Rf feedback resistor, controls the self oscillation.
kf string coupling constant, controls the coupling to the string.

Please read the conclusions in the text.  I didn't create plots for every case I tested.

At this point it looks like Rf=13k is too low.  However, I'm not seeing good locking behaviour for any Rf values.
The signal doesn't become clean until I increase Rf to 1.3M and then it's pretty much like Rf has been removed.

String frequency is set to 400Hz.

The way the circuit is set-up is as follows:
- output coil current determines magnetic field
- output coil magnetic field determines force on the string
- voltage on string model represent the velocity of the string
- receive coil output voltage responds to sting velocity

I have not set-up the model to represent true physics variables.  I just have the basic physical construction then
I threw in kf to roll-up all the physical scaling factors.   I have no idea if the kf values do in fact
match the physics since I only cared about getting it to oscillate via the string.



1) Rf = 1M3, kf = 7.5e-3





2) Rf = 1M3, kf = 1.5e-3

Reducing the coupling to the string shows a cleaner signal, more like those seen on an oscilloscope.



3) Rf = 130k, kf = 7.5e-3

Even with Rf=130k the oscillations don't look like they transition well.  I've got strong string coupling which should
help the string kick-in but even that the doesn't do so well.

First half is 3.8kHz self oscillation then it transitions to 400Hz string.



FYI, the LM386 model follows the schematic in the datasheet.  It roughly matches some of the specs in the datasheet but I can't vouch for it working like the real device when pushed.    It does look reasonable.   I'm not sure if the crappy vout waveform for kf=7.5e-3 can be attribute to a problem with the LM386 model.

Update: the "inversions" on the negative output swing was traced to the bias current being too high on the gain stage (in the LM386 model).   The bias current needs to be set so it prevents the inversion behaviour but still has good positive output swing under load.   However looking closer into the cause it looks like when the positive input
is below -0.75V it starts stuffing up the works - starting to suspect there a few protection diodes missing on the internal schematic.

FYI:

Here's the new waveforms with the tweaked LM386 model. 

1b) Rf = 1M3, kf = 7.5e-3

You can see the reduction in the -ve swing inversion.



1c) Rf = 1M3, kf = 7.5e-3

I then added a 4k7 resistor in series with the LM386 non-inverting input to prevent the non-inverting input buffer robbing base current from the voltage-gain stage on large negative inputs.    You can see it cleans-up the waveform a bit more.

[Paul Marossy]

That's interesting stuff Rob. I really barely know how to use these sophisticated modeling programs, and all of this has been a great learning experience for me. I've used Circuit Maker in the past to look at tube circuits and whatnot, but this is a whole new world opened up to me (kinda like when I got a smart phone and learned what wifi is, and no that's not in the recent past  ).

Regarding these LTSpice LM386 models, I don't think anyone will ever get them right. I doubt that the schematic on the datasheet is really that simple... there must be other things in that chip besides what is shown on that little schematic.

With a working Ebow I attempted to do some measurements using your coil suggestion to try and determine frequency, but my test equipment is just not sensitive enough...  One of these days I should pick up a decent digital scope and get out of the dark ages? 

I did however find something interesting: If you stick a small screwdriver in that channel on the bottom of an Ebow, across the coils, it will squeal audibly! The switch changes the frequency. So maybe what that harmonic mode is actually doing is changing the frequency of the oscillator and what it says in the patent doesn't really apply today? Maybe they found another way to accomplish same end result? I don't know, I'm just looking at it from outside the box so to speak.

Anyway, I made a little test fixture for the one I'm attempting to resurrect so I can easily measure voltages, etc. which is where I am starting. As far as I can tell, it's sitting there oscillating at around 8K, and in the other mode it's quieter so I can't really tell if same or not. Operating under the assumption that my output cap is bad, they might all be a little off from what they should be (whatever that actually is), but these are the voltages I recorded using 9V non-alkaline battery that measures 9.3V

Reg mode:         

Pin 1 -- 1.0V
Pin 2 -- GND
Pin 3 -- 0.85V
Pin 4 -- GND
Pin 5 -- 2,82V
Pin 6 -- 6.18V
Pin 7 -- 3.03V
Pin 8 - 0.98V

Harm mode:

Pin 1 -- 1.16V
Pin 2 -- GND
Pin 3 -- 0.03V
Pin 4 -- GND
Pin 5 -- 3.15V
Pin 6 -- 5.26V
Pin 7 -- 3.46V
Pin 8 -- 1.11V

[Rob Strand]

That's interesting stuff Rob. I really barely know how to use these sophisticated modeling programs, and all of this has been a great learning experience for me. I've used Circuit Maker in the past to look at tube circuits and whatnot, but this is a whole new world opened up to me (kinda like when I got a smart phone and learned what wifi is, and no that's not in the recent past  ).

IMHO, it's not easy at all.   If you pullup a few models together and do a simulation and it works, all well and good.  However, there's a great universe of knowledge under the hood to get that to happen.  When it doesn't work it's an uphill battle to work out why.   I've used these things since 1990 or so which gives me an unfair advantage.

Regarding these LTSpice LM386 models, I don't think anyone will ever get them right. I doubt that the schematic on the datasheet is really that simple... there must be other things in that chip besides what is shown on that little schematic.

Agreed, you can only do your best. I've actually spent many hours researching transistor based models for the LM741.   I got models from papers, books and companies.  While the datasheet shows a schematic it doesn't elaborate on the details of the transistors.   They aren't all the same, and they aren't BC108's or 2N3904's.   Some of the papers go into fine details of the transistors and to be honest the results aren't great.   There's also many internal schematics for the LM741, not all the same!   A lot of opamp models do not use transistors inside, they use a "macro model".   It's very easy to get a macro model to match a datasheet but it's very difficult to get a transistor based circuit to match the datasheet - simply because you have to get so many details correct and you are given no information.

With a working Ebow I attempted to do some measurements using your coil suggestion to try and determine frequency, but my test equipment is just not sensitive enough...  One of these days I should pick up a decent digital scope and get out of the dark ages? 

Honestly, there's nothing wrong with an analog scope.   The measurements on digital scopes are convenient but not necessary - accuracy isn't great anyway.   IMHO the biggest advantage of a digital scope is capturing transient waveforms. 

I did however find something interesting: If you stick a small screwdriver in that channel on the bottom of an Ebow, across the coils, it will squeal audibly! The switch changes the frequency. So maybe what that harmonic mode is actually doing is changing the frequency of the oscillator and what it says in the patent doesn't really apply today? Maybe they found another way to accomplish same end result? I don't know, I'm just looking at it from outside the box so to speak.

That's *yet another* way to cause coupling.   I should imagine it does self-oscillate.  It's very much like the 13k feedback resistor.   Except the "connection" is done via the magnetic field from the output coil to the input coil.  If you keep the screwdriver as far away from *each of the coils* as possible it should act like similar to the feedback resistor.  When you place the screwdriver very close to either coil is starts to screw-up the inductance and then the frequencies you see are no longer representative of what happens in the real unit.

Anyway, I made a little test fixture for the one I'm attempting to resurrect so I can easily measure voltages, etc. which is where I am starting. As far as I can tell, it's sitting there oscillating at around 8K, and in the other mode it's quieter so I can't really tell if same or not. Operating under the assumption that my output cap is bad, they might all be a little off from what they should be (whatever that actually is), but these are the voltages I recorded using 9V non-alkaline battery that measures 9.3V

The frequency doesn't look too far off the rough estimate I got of 4k.   I trust that 8kHz just as much as my 4kHz, they a lot of scope for variation in the details.

Reg mode:         

Pin 1 -- 1.0V
Pin 2 -- GND
Pin 3 -- 0.85V
Pin 4 -- GND
Pin 5 -- 2,82V
Pin 6 -- 6.18V
Pin 7 -- 3.03V
Pin 8 - 0.98V

That thing must be pulling a fair bit of current.   Pin 6 is getting pulled down to 6.18V.   The non-alkaline batteries are never great for regulation but still 6.18V looks pretty low.

[Rob Strand]

FYI, here's my LM386 setup.   I just started from scratch and didn't try to dig-up any of my old stuff (actually the initial file I got from the web, tried to save some typing.)  There's a few things I still don't like about it.

Here's the schematic in terms of transistors.   I've scrawled notes over it in case I have to come back to it at some point.


Here's the model.   It should be the same as the schematic but I actually had the model file first then drew-up the transistor schematic afterward so I could debug it easier.   They should be the same but I might have missed something, like matching the part numbering.

The node numbers in the model should match this pic.   Keep in mind I've replaced the current source with a transistor circuit (as shown in previous schematic) - the part numbers are all 20's.



This is the contents of the model file, with some notes scrawled in there:

Code Select Expand


* LM386
*
* Does this behaviour show up on the real device?:
* For negative inputs on the NI-input we see
* an inversion on the negative swing.
* Tweaking the gain stage current source helps
* However the primary issue is the NI-input
* buffer will remove base drive from the voltage gain
* stage under large -ve input (say < -0.75V)
* An external workaround is to add a 4k7 resistor
* in series with the NI-input.
* - Rob S, 2022/12/14
*
* Is = 4.6mA @ Vs=9V
*
***********************************
* IC pins:   DIP8 package order
* Pins: 1 GAIN, 2 INM, 3 INP, 4 GND,
* 5 OUT, 6 VS, 7 BYP, 8 GAIN
.subckt lm386 1 2 3 4 5 6 7 8
***********************************

* input emitter-follower buffers
q1 4 2 11 pn
r1 2 4 50k
q2 4 3 15 pn
r2 3 4 50k

* differential input stage
q3 13 11 12 pn
q4 14 15 1 pn

* diff amp tail resistors
r3 6 7 15k
r4 7 12 15k

* feedback resistors
r5 12 8 150
r6 8 1 1.35k
r7 1 5 15k

* input stage current mirror
q5 13 13 4 np
q6 14 13 4 np

* voltage gain stage & rolloff cap
q7 18 14 4 np
*c1 18 14 15pf
* 35pF (match -3dB) to 50pF (match 10dB @ 1MHz)
c1 18 14 35p

* Current source for gain stage
* Approx 1.7mA @ Vs=9V
* *** Even with RE added, this source dependency on Vs
* is too high to match the Is vs Vs spec in the datasheet.
* (Also, could derive bias current from through 15k input
*  biasing like LM380.)
R21 21 4  50k
q21 21 21 6 pn
q22 16 21 22 pn 20
R22 6 22  10

* bias diodes
* Set to 2.55mA @ Vs=9V
q11 16 16 17 np 100
q12 17 17 18 np 100

* output stage
q8 6 16 5 np 100
q9 19 18 5 pn
q10 5 19 4 np 100

* Need to rejig with Nc = 1
* generic transistor models generated
* with MicroSim's PARTs utility, using
* default parameters except Bf:
.model np NPN(Is=10f Xti=3 Eg=1.11 Vaf=100
+ Bf=200 Ise=0 Ne=1.5 Ikf=0 Nk=.5 Xtb=1.5 Var=100
+ Br=1 Isc=0 Nc=2 Ikr=0 Rc=0 Cjc=2p Mjc=.3333
+ Vjc=.75 Fc=.5 Cje=5p Mje=.3333 Vje=.75 Tr=100n
+ Tf=1n Itf=1 Xtf=0 Vtf=10)

.model pn PNP(Is=10f Xti=3 Eg=1.11 Vaf=100
+ Bf=100 Ise=0 Ne=1.5 Ikf=0 Nk=.5 Xtb=1.5 Var=100
+ Br=1 Isc=0 Nc=2 Ikr=0 Rc=0 Cjc=2p Mjc=.3333
+ Vjc=.75 Fc=.5 Cje=5p Mje=.3333 Vje=.75 Tr=100n
+ Tf=1n Itf=1 Xtf=0 Vtf=10)

.ends



FYI,  the 100's after transistor models is the same as the AREA used in the transistors on the schematic.  What that means is these transistors are 100 times larger than the basic transistor.    A value of 100 might be excessive but the model is trimmed to work with it.

[deadastronaut]

probably not relevant at all but i found it interesting regardless.

when i was messing with sustainers a guy on here called artemis years ago

tried using inductors and small magnets as a coil driver. i tried it and it worked quite well....

magnet under the inductor. inductor wired as the coil. voila' nice simple ready made coils.

put on a tiny piece of vero done...

anyway as you were.....great thread. 

[Paul Marossy]

That thing must be pulling a fair bit of current.   Pin 6 is getting pulled down to 6.18V.   The non-alkaline batteries are never great for regulation but still 6.18V looks pretty low.

Yes, I also noticed that and is why I gave a bit of a disclaimer about that output cap maybe throwing off the numbers. Could be a factor here I think. The LM386 was not getting warm so I am not sure what exactly is pulling so much current.

I had a bit of insomnia last night and thought I'd see if I could fix a problem that my counterfeit DSO Shell was having and it seems that I did something to improve it. I still think it gets the frequency wrong sometimes but the thing I did find out is that it appears to self-oscillate at around 2.2-2.4kHz. Seems like one setting is lower in frequency but I can't tell... maybe that's the harmonics I am hearing? Sure seems to me like the frequency kinda doubles. Anyway, video for that demonstration is below.

Here is my test fixture that I created yesterday from things I scraped up in my laboratory

And here is Part I of my Ebow Resurrection Attempt.

[johngreene]

That thing must be pulling a fair bit of current.   Pin 6 is getting pulled down to 6.18V.   The non-alkaline batteries are never great for regulation but still 6.18V looks pretty low.

Yes, I also noticed that and is why I gave a bit of a disclaimer about that output cap maybe throwing off the numbers. Could be a factor here I think. The LM386 was not getting warm so I am not sure what exactly is pulling so much current.

I had a bit of insomnia last night and thought I'd see if I could fix a problem that my counterfeit DSO Shell was having and it seems that I did something to improve it. I still think it gets the frequency wrong sometimes but the thing I did find out is that it appears to self-oscillate at around 2.2-2.4kHz. Seems like one setting is lower in frequency but I can't tell... maybe that's the harmonics I am hearing? Sure seems to me like the frequency kinda doubles. Anyway, video for that demonstration is below.

Here is my test fixture that I created yesterday from things I scraped up in my laboratory

And here is Part I of my Ebow Resurrection Attempt.

The reason you see such high voltage on the output is due to inductive kick. The same principle that allows boost converters to work.

[Paul Marossy]

The reason you see such high voltage on the output is due to inductive kick. The same principle that allows boost converters to work.

That could be, but in LTSpice I just tried removing the resistor on the LM386 output AND the output coil, and it's still at about 9 volts on Pin 5 (out). Per the data sheet the output should automatically be 1/2 the supply voltage. Therefore, the models that I have been using all seem to be missing something because that just doesn't happen. I need to try Rob's model when I get a chance and see if that behaves any differently.

[johngreene]

The reason you see such high voltage on the output is due to inductive kick. The same principle that allows boost converters to work.

That could be, but in LTSpice I just tried removing the resistor on the LM386 output AND the output coil, and it's still at about 9 volts on Pin 6 (out). Per the data sheet the output should automatically be 1/2 the supply voltage. Therefore, the models that I have been using all seem to be missing something because that just doesn't happen. I need to try Rob's model when I get a chance and see if that behaves any differently.

Right, I was just commenting on the large AC voltage you were seeing. If the model is missing one of those diodes as you mention in your video, that will cause what you are seeing as they are part of the biasing network.
Huge inductive kicks and near perfect filter performance are always an illusion when using pspice if you are using ideal inductor models. I can't tell you how many times I've sent simulations back to other engineers that think they designed the best filter ever and it doesn't work in real life and I can simply add a resistor to match the Q (at only one frequency) and suddenly their simulation more closely matches their results.

[Rob Strand]

I had a bit of insomnia last night and thought I'd see if I could fix a problem that my counterfeit DSO Shell was having and it seems that I did something to improve it. I still think it gets the frequency wrong sometimes but the thing I did find out is that it appears to self-oscillate at around 2.2-2.4kHz. Seems like one setting is lower in frequency but I can't tell... maybe that's the harmonics I am hearing? Sure seems to me like the frequency kinda doubles. Anyway, video for that demonstration is below.

Thanks for the videos, really great stuff.   You did a great job on the rebuild too.

The first thing I noticed in your screwdriver test is when the screwdriver isn't place across the coils we don't see any self oscillation - is that right?    That really makes me think the 13k resistor isn't 13k.    As per my simulations in reply #41 we start to lose oscillations when that resistor is about 1.3M; the value might need to be revisited based on the new receive coil estimates, see below.

We can also get a good estimate for the inductance:

If you are getting say 2.3kHz with the screwdriver coupling the coils then we can back-engineer the receive coil inductance,
                   L   =  1 / ( (2*pi*f)^2 * C)
                   C = 33n from the schematic

which gives,

                  L = 145mH.

That's assuming the screwdriver  isn't messing with the inductance too much.   Since you aren't touching the coils with the screwdriver that's a pretty good assumption.

If we go back to the inductance estimate I made back in reply #10,

Code Select Expand


N 1187.5 1888.3 3002.5 number of turns
dw [mm] 0.080 0.063 0.050 bare wire diameter
AWG 40.0 42.0 44.0 wire AWG

R [ohm] 88.3 223.3 564.6 Coil resistance

La [uH] 5876.7 14858.2 37565.9 Air-cored inductance, no core
Lm [mH] 31.8 80.3 203 Inductance with core

Which inductance to use depends on whether the magnet is ferrite or Alnico.   I'm thinking ferrite since the chipped magnet in your video is black on the inside.    That's going to drop the inductance estimates a bit.

The inductance will probably end up around 3.5 to 5 times the air-core inductance.   For the 42AWG wire that's 53mH to 75mH, which doesn't quite make the expected 145mH.   However with the 44AWG wire, we get 130mH to 180mH which is very much in the ball-park. (Keep in mind this is a complicate set-up and I'm only making estimates, so we can't totally write-off 42AWG.)

So it looks like the original coil is 44AWG and that actually agrees with the wire diameter you estimated from the original coil. 

Given your build is probably 40AWG the inductance will end up around 20mH to 30mH  the self oscillation frequency of the rebuilt coil will be f = 1/(2*pi*sqrt(LC)) = 5.5kHz.    You could increase the 33n cap to drop that frequency but it's probably not worth it at this point.

Since the rebuild coil has a lot less turns than the original you will lose some sensitivity.   That will mean putting the coils closer to the strings to get the unit to oscillate at the string frequency.   Increasing the cap 33n cap might actually help improve the sensitivity because the resonance will peak the gain.     That's something you should keep in mind.

FYI, when you do simulations keep in mind there's two cases now:  original receive coil and rebuilt receive coil.  They have different R and L specs and will behave differently.

[Rob Strand]

probably not relevant at all but i found it interesting regardless.

when i was messing with sustainers a guy on here called artemis years ago

tried using inductors and small magnets as a coil driver. i tried it and it worked quite well....

magnet under the inductor. inductor wired as the coil. voila' nice simple ready made coils.

It's a good way to get something working without stuffing around winding coils.

Hacking the buzzers like in the video I posted in Reply #20 is a good idea as well.
The pole diameters are pretty small so the magnetic field won't extend out very far but it seems to work.
https://www.diystompboxes.com/smfforum/index.php?topic=129920.msg1258007#msg1258007

[Paul Marossy]

I had a bit of insomnia last night and thought I'd see if I could fix a problem that my counterfeit DSO Shell was having and it seems that I did something to improve it. I still think it gets the frequency wrong sometimes but the thing I did find out is that it appears to self-oscillate at around 2.2-2.4kHz. Seems like one setting is lower in frequency but I can't tell... maybe that's the harmonics I am hearing? Sure seems to me like the frequency kinda doubles. Anyway, video for that demonstration is below.

Thanks for the videos, really great stuff.   You did a great job on the rebuild too.

Thanks. This has morphed into a bit of an obsession.... but my curiosity won't let it be until I know as much as I can. 

The first thing I noticed in your screwdriver test is when the screwdriver isn't place across the coils we don't see any self oscillation - is that right?    That really makes me think the 13k resistor isn't 13k.    As per my simulations in reply #41 we start to lose oscillations when that resistor is about 1.3M; the value might need to be revisited based on the new receive coil estimates, see below.

We can also get a good estimate for the inductance:

If you are getting say 2.3kHz with the screwdriver coupling the coils then we can back-engineer the receive coil inductance,
                   L   =  1 / ( (2*pi*f)^2 * C)
                   C = 33n from the schematic

which gives,

                  L = 145mH.

That's assuming the screwdriver  isn't messing with the inductance too much.   Since you aren't touching the coils with the screwdriver that's a pretty good assumption.

If we go back to the inductance estimate I made back in reply #10,

Code Select Expand


N 1187.5 1888.3 3002.5 number of turns
dw [mm] 0.080 0.063 0.050 bare wire diameter
AWG 40.0 42.0 44.0 wire AWG

R [ohm] 88.3 223.3 564.6 Coil resistance

La [uH] 5876.7 14858.2 37565.9 Air-cored inductance, no core
Lm [mH] 31.8 80.3 203 Inductance with core

Which inductance to use depends on whether the magnet is ferrite or Alnico.   I'm thinking ferrite since the chipped magnet in your video is black on the inside.    That's going to drop the inductance estimates a bit.

The inductance will probably end up around 3.5 to 5 times the air-core inductance.   For the 42AWG wire that's 53mH to 75mH, which doesn't quite make the expected 145mH.   However with the 44AWG wire, we get 130mH to 180mH which is very much in the ball-park. (Keep in mind this is a complicate set-up and I'm only making estimates, so we can't totally write-off 42AWG.)

So it looks like the original coil is 44AWG and that actually agrees with the wire diameter you estimated from the original coil. 

Given your build is probably 40AWG the inductance will end up around 20mH to 30mH  the self oscillation frequency of the rebuilt coil will be f = 1/(2*pi*sqrt(LC)) = 5.5kHz.    You could increase the 33n cap to drop that frequency but it's probably not worth it at this point.

The screwdriver I think was right under the coils. I was more focused on trying to keep the screwdriver off the pickup, the magnet was strong with that pickup 
I debated those resistor color bands for a while. Definitely did not look like BRN-ORG-GRN (1.3M). That's OK, at least we now roughly know what its function is. If someone wants to donate an Ebow to science, let me know.  I'd love to pick the goop off just the botoom of the PCB and measure stuff.
As I mentioned before, that wire on the input coil is very very tiny. Almost like a human hair!

Since the rebuild coil has a lot less turns than the original you will lose some sensitivity.   That will mean putting the coils closer to the strings to get the unit to oscillate at the string frequency.   Increasing the cap 33n cap might actually help improve the sensitivity because the resonance will peak the gain.     That's something you should keep in mind.

Yep, I'm expecting that. My aim is just to get it working again. I think by the end of this process we will know a helluva lot about this intriguing circuit than we ever did before. 

FYI, when you do simulations keep in mind there's two cases now:  original receive coil and rebuilt receive coil.  They have different R and L specs and will behave differently.

Yeah I have been doing parallel simulations, just focused on the rebuild at the moment. With your SPICE model I think we are light years closer to reality than when I started out with those flawed LM386 models I've been working with.

[Rob Strand]

Thanks. This has morphed into a bit of an obsession.... but my curiosity won't let it be until I know as much as I can.

Well you roped me in, so there you go  !

The screwdriver I think was right under the coils. I was more focused on trying to keep the screwdriver off the pickup, the magnet was strong with that pickup 

Over the years I've a done a lot of tests where I put things in front of magnets and coils.  I always found it a headache dealing with the distances between things varying and being inconsistent - not to mention the annoyance of magnets smacking things together.   A trick I worked out was to place objects between the two objects under test.  For example sheets of cardboard, sheets of plastic, wood, even plastic bottle tops.   You just put it between the coil/magnet and the external object.   That way every test is has fairly consistent spacing (and behaviour).    It makes life a hell of a lot easier!   You can also measure the thickness of the plastic with the calipers.   When I did calculations, I could use those thickness measurement in the calculations.

I debated those resistor color bands for a while. Definitely did not look like BRN-ORG-GRN (1.3M).

I'll take your word for it.  I can't claim to know anything about it.  All I'm doing trying to juggle the parameters to match-up with what we see.

I think by the end of this process we will know a helluva lot about this intriguing circuit than we ever did before. 

No doubt about it.

[Rob Strand]

I found this video the other day.  It's a DIY ebow based on using those buzzers.

https://www.youtube.com/watch?v=lFSfWysFvuY

The first part of the video he repairs some broken PCB connections.

In the last part of the video he does a demo.

The circuit used is the simple LM386 ebow schematic on the web.

Given it is the basic schematic it has no feedback resistor and no resonating cap on the receive coil (it could have a cap 1nF, not sure, but the there's no feedback resistor so it doesn't matter).

What's interesting about the video is the DIY ebow has no problem driving the strings.
So you can get these things to work without self oscillation.

[anotherjim]

I think we would all expect a basic feedback-only system to be able to work. However, you might have to pluck the string to start it up, and that would ruin the illusion of bowing.
Somewhere I read a quite scientific investigation of the traditional horsehair/rosin bowing system and of how there is a continual stick/slip excitation of the string. I'd assume the self-oscillation of the E-bow is to approximate this. But, I'm surprised the oscillation frequency is what surely can be audible as a whine although that should stop when the string takes over.

[Paul Marossy]

I think we would all expect a basic feedback-only system to be able to work. However, you might have to pluck the string to start it up, and that would ruin the illusion of bowing.
Somewhere I read a quite scientific investigation of the traditional horsehair/rosin bowing system and of how there is a continual stick/slip excitation of the string. I'd assume the self-oscillation of the E-bow is to approximate this. But, I'm surprised the oscillation frequency is what surely can be audible as a whine although that should stop when the string takes over.

I'm in the camp of it continuously oscillating. Ebow patent mentions it will excite the string from a resting state to its resonant frequency via positive feedback, but that doesn't really tell us much other than it has feedback loop. Blocking noise from the oscillator might be one purpose of those coated steel rings that I originally assumed were ferrite. I'm thinking that maybe you don't want it oscillating too quickly, probably needs to be something not too far away the frequency range of the guitar strings... I mean if it was oscillating at 100kHz, could that even get the string going? That's so far above the highest frequency on a guitar that it seems to me that the guitar string would be unresponsive to an oscillator running that fast.

[johngreene]

I think we would all expect a basic feedback-only system to be able to work. However, you might have to pluck the string to start it up, and that would ruin the illusion of bowing.
Somewhere I read a quite scientific investigation of the traditional horsehair/rosin bowing system and of how there is a continual stick/slip excitation of the string. I'd assume the self-oscillation of the E-bow is to approximate this. But, I'm surprised the oscillation frequency is what surely can be audible as a whine although that should stop when the string takes over.

I'm in the camp of it continuously oscillating. Ebow patent mentions it will excite the string from a resting state to its resonant frequency via positive feedback, but that doesn't really tell us much other than it has feedback loop. Blocking noise from the oscillator might be one purpose of those coated steel rings that I originally assumed were ferrite. I'm thinking that maybe you don't want it oscillating too quickly, probably needs to be something not too far away the frequency range of the guitar strings... I mean if it was oscillating at 100kHz, could that even get the string going? That's so far above the highest frequency on a guitar that it seems to me that the guitar string would be unresponsive to an oscillator running that fast.

If it is basically an injection locked oscillator as I was suggesting earlier then 100kHz may be outside of its capture range. Injection locked oscillators do have an extremely wide capture range but I don't think it is over 10 octaves!
If it's not an oscillator it would need to be a design that is on the verge of oscillation needing only the steel string as the final contribution to the positive feedback for it to oscillate. This also means a high dependency on the frequency of the vibrating string.
I wonder if those coated metal rings are actually mu-metal?
https://www.kjmagnetics.com/blog.asp?p=mumetal

[Paul Marossy]

I think we would all expect a basic feedback-only system to be able to work. However, you might have to pluck the string to start it up, and that would ruin the illusion of bowing.
Somewhere I read a quite scientific investigation of the traditional horsehair/rosin bowing system and of how there is a continual stick/slip excitation of the string. I'd assume the self-oscillation of the E-bow is to approximate this. But, I'm surprised the oscillation frequency is what surely can be audible as a whine although that should stop when the string takes over.

I'm in the camp of it continuously oscillating. Ebow patent mentions it will excite the string from a resting state to its resonant frequency via positive feedback, but that doesn't really tell us much other than it has feedback loop. Blocking noise from the oscillator might be one purpose of those coated steel rings that I originally assumed were ferrite. I'm thinking that maybe you don't want it oscillating too quickly, probably needs to be something not too far away the frequency range of the guitar strings... I mean if it was oscillating at 100kHz, could that even get the string going? That's so far above the highest frequency on a guitar that it seems to me that the guitar string would be unresponsive to an oscillator running that fast.

If it is basically an injection locked oscillator as I was suggesting earlier then 100kHz may be outside of its capture range. Injection locked oscillators do have an extremely wide capture range but I don't think it is over 10 octaves!
If it's not an oscillator it would need to be a design that is on the verge of oscillation needing only the steel string as the final contribution to the positive feedback for it to oscillate. This also means a high dependency on the frequency of the vibrating string.
I wonder if those coated metal rings are actually mu-metal?
https://www.kjmagnetics.com/blog.asp?p=mumetal

Interesting idea and a very good question. I thought it was just a coating on a steel ring... but perhaps could be MuMetal underneath. I wouldn't know how to make that determination.

[Rob Strand]

I'm in the camp of it continuously oscillating.

I'm not convinced either way yet.

The only thing pushing me toward the Ebow self oscillating is the size of the 13k resistor - it's a shame that resistor got damaged!  My previous sim showed 13k to definitely cause self oscillations.  I haven't tried redoing the simulation with new resistance and inductance estimates for the (original) receive coil to see how low that resistor needs to be before self oscillation starts.
[EDIT: original receive coil 145mH 565 ohm
  1.3M does is just below oscillation but is is close to the critical value.
  1.2M oscillates.
  13k will definitely oscillate.]

What's clear is the DIY units people make work.  They don't oscillate (no feedback resistor).  They work for the reasons I posted in reply #35.    FWIW, these types of kicks to start oscillators are not present in an LTspice simulation (unless you add it) as they are outside influences.
https://www.diystompboxes.com/smfforum/index.php?topic=129920.msg1258065#msg1258065

The question is more like,  does the *real Ebow* oscillate?

If you have a real Ebow it's a very simple test:
- wire a pickup to the oscilloscope
- place the transmit coil *only* over the pickup but far away as possible.
  Keep the receive coil in the open space, not over the pickup.

  The caveats ensure the metal structures in the pickup don't cause
  oscillation due to false/deliberate magnetic coupling- otherwise
  they would act like your screwdriver test.   A sign this is happening
  is the oscillation will suddenly startup only when you move the Ebow
  close to the pickup.

- if you see a signal 2kHz to 10kHz (or whatever) signal on the pickup then it's oscillating.

Sorry, I edited the edit as I still had the 4k7 input resistor "mod" in place.