Ebow Exposé - Part III

[Rob Strand]

I did a small case study on Paul's rebuilt receive coil: ability to lock to the string

1) As it is now

Here I find what largest feedback resistor to cause self oscillation:
23mH, 108R, 33n
max 1MEG, 5.7kHz

Then I work out what string coupling is required to maintain string oscillation
without any feedback resistor.
Rf=100MEG
k=1e-3, osc in 0.75s

Now I make the self feedback 10 times stronger, by reducing Rf, and
see how much I have to crank-up the string coupling for the oscillation
to transition from self oscillation to the string frequency (400Hz).
Rf=100k
kf=9e-3, osc in 0.73 sec

Same idea but now the feedback is 100 times stronger.
Rf=10k
kf=110e-3, osc in 0.7sec

Case 2)  Larger cap - cap sets self oscillation frequency to 2.1kHz

23mH, 108 ohm, 220nF
Rf max 180k, fosc =2.1kHz

Rf=100MEG
kf =1e-3, osc in 0.75s

Rf=18k
kf=6e-3 for transition in 0.7s

Rf=1.8k
kf=22e-3 for transition in 0.7s

Conclusions
- with no feedback present a string coupling of kf=1e-3 works in both cases

- when we increase the feedback 10 times over minimum
   the smaller cap needs kf to increase 9 times but the larger cap
   only needs kf to increase by 6 times.
   Improvement in locking is a factor of 1.5

- when we increase the feedback by 100 times
  the smaller cap needs kf to increase 110 times but the larger
  cap only needs kf to increase by 22 times.
  Improvement in locking is a factor of 5.

So clearly dropping the self oscillation frequency improves the ability
to lock to the string frequency.    Keeping the feedback resistor
high as possible also helps since the string feedback doesn't have to
"fight" with such a strong local feedback.

[Rob Strand]

Refinement on results of reply #80.

The problem with reply #80:

When Rf is large, stepping down Rf by a factor of 10 increases
the feedback by a factor of 10.  But when Rf is small compared
to the inductor resistance that's not quite true.

The amount of feedback is the divider ratio I introduced
few posts back.   I'm going to call that
  ko = Rp / (Rp + Rf)

I then compare ratio ko/kf where kf is the amount of coupling
from the string.  If ko/kf is large that means the the oscillator
is locks to the string easier in relative terms.

Here's all the equations.
f= 1/(2*pi*sqrt(LC))      ;self oscillation freq.
Rp = XL^2/Rs = (2*pi*f*L)^2            ;equivalent parallel resistance of inductor
Rp' = Rp // 50k                                  ; effective parallel resistance with LM386 loading
ko  = Rp' / (Rf + Rp')                         ; output feedback level

1) Rebuilt unit, as it is now
Rs = 108 ohm
L = 23mH
C=33n
f = 5.777kHz
Rp = 6.45k
Rp' = 5.71k

Rf   ko   ko/kf
1MEG   5.68e-3   -   
100k   54.0e-3   6.0
10k   363e-3   3.3

2) Rebuilt unit with larger cap to drop self oscillation frequency.
Rs = 108 ohm
L = 23mH
C=220n
f =2.237kHz
Rp = 968 ohm
Rp' = 950 ohm

Rf   ko   ko/kf
180k   5.25e-3   -
18k   50.1e-3   8.4
1.8k   345e-3   15.7

Conclusions:
So luckily,
- ko steps up by the about the same amount
for case 1 and case 2.    So the prevous conclusions
still apply.
- the ko/kf is clearly better for the large cap case.

If anyone finds a bug please let me know.

[Paul Marossy]

I see y'all have come up with more interesting details while I was absent. 

TIME FOR ANOTHER UPDATE!

I got a little closer to it actually working this weekend but still don't have a working Ebow circuit yet... but it is almost working.

I decided to rewind the input coil, which was a royal P.I.T.A. but I got it done. New specs on the coil is 446 ohms DC resistance, capacitance 0.1uF and the apparent inductance is 46mH. I used the wire from a Mouser 42TM018 10K:10K center tapped audio transformer, which has a DC resistance of 500 ohms so I knew I had enough wire to wind as much as I could on the input coil. Was difficult to work with, think it's 44 AWG and it breaks if you look at it wrong. 

So with the new input coil it's now oscillating at around 3.2-3.3 kHz. FAR more responsive to the screwdriver test. I suppose if I coulda got it to 600 ohms it might actually work? I could try reversing the wires on one of the coils and see if that works. It didn't work when I tried that with the previous 108 ohm input coil but I suppose we have to eliminate that as a possibility with this new one. It seems that the feedback resistor actually sets the frequency of the oscillation in addition to how "strong" it oscillates. I thought about maybe changing that input cap but I have a hunch that the input coil still isn't producing enough voltage to get a feedback loop going.

The most interesting stuff in the video below starts at about 10:45 in.

[johngreene]

........

So with the new input coil it's now oscillating at around 3.2-3.3 kHz. FAR more responsive to the screwdriver test. I suppose if I coulda got it to 600 ohms it might actually work? I could try reversing the wires on one of the coils and see if that works. It didn't work when I tried that with the previous 108 ohm input coil but I suppose we have to eliminate that as a possibility with this new one. It seems that the feedback resistor actually sets the frequency of the oscillation in addition to how "strong" it oscillates. I thought about maybe changing that input cap but I have a hunch that the input coil still isn't producing enough voltage to get a feedback loop going.

I think that since you are clipping the output with what you have coming in from the sensing coil, you couldn't get any more output. But that input needs to be 'stronger' than the self-oscillating feedback in order to take over the frequency of oscillation. Which means your feedback could be too strong. Also, if it is unable to excite the string then the problem may be with the output driver coil not being a strong enough electro-magnet. I don't think I've seen where you measured the actual sensitivity to a guitar string the circuit has....
If it was me, I would dial back the 'on-board' feedback until it no longer self oscillates, then put it on a guitar string and manually start the string vibrating and measure the output of the 386 to see how well the circuit is acting as an amplifier. One thing to consider is that when a circuit is being driven into saturation, it loses gain. So the 'on-board' feedback should only be enough to get the circuit oscillating without saturating the output too hard. IMO.

[anotherjim]

I think you need to see if the output coil is strong enough first. It probably is, since the coil value seems reasonable, but I suggest checking.
Can you disconnect the input coil and inject a variable frequency signal in its place? Then hold the output coil over a guitar string and see if, with a generated frequency to match the string frequency, it starts the string up. There is a risk of direct coupling to a guitar pickup, but muting the string should prove if that's happening.

If you are happy the output coil can work, I'd mess with polarities.
As far as the magnets go, can you tell the polarity of a working E-bow with a simple compass or were the originals marked in some way?

I'll mention that for experimental purposes, you can steal coils on bobbins from miniature PCB relays of similar size. Higher coil operating voltages have higher DCR and the smallest 6v ones start around 100R give or take, 24v are going to get close to what's needed for the input?

[Rob Strand]

I decided to rewind the input coil, which was a royal P.I.T.A. but I got it done. New specs on the coil is 446 ohms DC resistance, capacitance 0.1uF and the apparent inductance is 46mH. I used the wire from a Mouser 42TM018 10K:10K center tapped audio transformer, which has a DC resistance of 500 ohms so I knew I had enough wire to wind as much as I could on the input coil. Was difficult to work with, think it's 44 AWG and it breaks if you look at it wrong. 

Epic!

So with the new input coil it's now oscillating at around 3.2-3.3 kHz. FAR more responsive to the screwdriver test.

The inductance doesn't match the oscillation frequency.

I think the screwdriver test indicates an improvement.

When I watched the video I made notes.   There's a few issues in there.

Tests on Old Receive Coil

- 8:10  oscillating vs non-oscillating any issues with the magnetics or coils
  will affect both cases.  A weak coil will not let the string oscillations beat
  the self oscillation.

- 12:40 Feedback resistor
   Maximum feedback resistor 180k.
   We expect the actual circuit to require a lower "maximum"
   feedback resistor to the simulation.

   The steel pole piece causes more losses and you need a
   lower feedback resistor to compensate.

   For that reason the spice sim will always work with a high
   feedback resistor.   The normal/"correct" way to handle this
   is to add resistors to the spice to account for the
   loss in the steel pole piece. 

- 16:00 Harmonic mode,

    The 4.7uF cap will roll-off the low
   frequencies and allow on the high frequencies to
   pass.   The high-pass frequency would be,

   fhp = 1/(2*pi*8*4.7uF) = 4.2kHz

   The parallel diode screws that up a bit.

  When you pass a square-wave through a high-pass
  filter it will make the top of the square-wave tilt.
  Usually downward.

  Are we sure the diode is 1N4148/1N914 and
  is the direction of the diode on the schematic
  correct?

- 16:40  String over the coil.

   This test *won't* create a string oscillation.   The string
   needs have tension and be tuned to a note.   The ebow
   then excites the string and it produces *that* note.

   By placing the the string in the field all you are doing
   some magnetic material in the magnetic field.
   That decreases the inductance of the receive coil
   and consequently it drops the *self oscillation* frequency
   a bit.   If you place a small steel nut on the receive coil
   it would do a similar thing.

   The screw driver test does that as well but what it's also doing
   is making the *self oscillation* stronger.   The screwdriver
   test still doesn't approximate what happens with the string.
   The screw drive doesn't have a frequency/note.

-23:00 Input waveform

The input waveform is always going to be sinusoidal as the
inductor and capacitor act like a filter to remove the harmonics.

-27:16 Spice Sim not comparing apples to apples.

The spice sim has a strong 247Hz input signal.
The real circuit is operating via a self oscillation at 4kHz.

The spice sim is doing a completely different thing to the real circuit.
You need to do a spice sim which is also self oscillating.

I can't quite see the schematic but you have what looks like a 5k
resistor between the coil and the 33nF cap.
The resistor value should be the DC resistance of the coil.
  A 5k in that position could stop self-oscillation.

  If you want to put in the 4.7k "work-around" for the model I
  mentioned some time back then the 4k7 needs to be between
  the 33nF cap and the LM386 input.

-29:00 From the pic it makes me think the last band is orange.
Perhaps the feedback resistor is 130k? instead of 13k

Going forward:

Test the receive coil by wiring it to an amplifier.

Check how much output when you place it over the string
and pick a note.  If the receive pickup output is
is too low it will not have enough signal to sustain string
oscillation.

You could even wire the receive coil to the LM386.  Look
at the output of the LM386.

If plucking the string cannot clip the LM386 I think
you have a problem.  It's unlikely the unit will
sustain string oscillations.

That's along the lines of what Jim noted as well.

[Paul Marossy]

I think that since you are clipping the output with what you have coming in from the sensing coil, you couldn't get any more output. But that input needs to be 'stronger' than the self-oscillating feedback in order to take over the frequency of oscillation. Which means your feedback could be too strong. Also, if it is unable to excite the string then the problem may be with the output driver coil not being a strong enough electro-magnet. I don't think I've seen where you measured the actual sensitivity to a guitar string the circuit has....
If it was me, I would dial back the 'on-board' feedback until it no longer self oscillates, then put it on a guitar string and manually start the string vibrating and measure the output of the 386 to see how well the circuit is acting as an amplifier. One thing to consider is that when a circuit is being driven into saturation, it loses gain. So the 'on-board' feedback should only be enough to get the circuit oscillating without saturating the output too hard. IMO.

Yes I had to build a one string jig with a guitar PUP yesterday so I can do more testing. It's not really possible to do any testing using a guitar, too problematic.

I think you need to see if the output coil is strong enough first. It probably is, since the coil value seems reasonable, but I suggest checking.
Can you disconnect the input coil and inject a variable frequency signal in its place? Then hold the output coil over a guitar string and see if, with a generated frequency to match the string frequency, it starts the string up. There is a risk of direct coupling to a guitar pickup, but muting the string should prove if that's happening.

If you are happy the output coil can work, I'd mess with polarities.

After I fabricated the one string jig (in the quote right above this comment) I was able to get it to vibrate a B string, but it's kinda weak. Also doesn't vibrate at the fundamental, so yeah it's not able to overcome the oscillator. I was thinking that perhaps the output coil isn't quite strong enough but haven't yet made a determination on that. This morning I added a couple more wire appendages attached to the PCB so I can add cap in parallel to the 0.033uF via capacitor substitution box and mess with the oscillator frequency. I currently have a device that can inject the oscillation into a guitar PUP, so obviously that's not right. 

The inductance doesn't match the oscillation frequency.

That's what I measured. Not sure if other things mess getting an accurate measurement on that.
EDIT: I measured again in circuit and it was 11.1mH. Not sure what happened there.... I measured right after I wound it. Attack of the one-eyed steel ring monsters?

- 8:10  oscillating vs non-oscillating any issues with the magnetics or coils
  will affect both cases.  A weak coil will not let the string oscillations beat
  the self oscillation.

Yes came to that conclusion last night while messing around with my one string jig that I created yesterday. (mentioned above)

- 12:40 Feedback resistor
   Maximum feedback resistor 180k.
   We expect the actual circuit to require a lower "maximum"
   feedback resistor to the simulation.

   The steel pole piece causes more losses and you need a
   lower feedback resistor to compensate.

   For that reason the spice sim will always work with a high
   feedback resistor.   The normal/"correct" way to handle this
   is to add resistors to the spice to account for the
   loss in the steel pole piece. 

OK, good point. What kind of resistance would that be? Would it be in series with coil, parallel?

- 16:00 Harmonic mode,

    The 4.7uF cap will roll-off the low
   frequencies and allow on the high frequencies to
   pass.   The high-pass frequency would be,

   fhp = 1/(2*pi*8*4.7uF) = 4.2kHz

   The parallel diode screws that up a bit.

  When you pass a square-wave through a high-pass
  filter it will make the top of the square-wave tilt.
  Usually downward.

Noted. I put a 10uF in there because I didn't have 4.7uF cap that was physically that small. How does that affect things?

  Are we sure the diode is 1N4148/1N914 and
  is the direction of the diode on the schematic
  correct?

Not 100% sure about type of diode, but reversing the diode does not work. I can demonstrate in another video I will make soon. I added a reversing switch to see what happens.
On my nearly destroyed example the diode was too busted up to tell anything about it, and on the only other known dissected PCB picture I can find, it appears that the band points to upper right hand corner of the PCB as I have it drawn. So we know correct polarity (for a 1N4148 type) but we really have no idea what it actually is. 

- 16:40  String over the coil.

   This test *won't* create a string oscillation.   The string
   needs have tension and be tuned to a note.   The ebow
   then excites the string and it produces *that* note.

   By placing the the string in the field all you are doing
   some magnetic material in the magnetic field.
   That decreases the inductance of the receive coil
   and consequently it drops the *self oscillation* frequency
   a bit.   If you place a small steel nut on the receive coil
   it would do a similar thing.

   The screw driver test does that as well but what it's also doing
   is making the *self oscillation* stronger.   The screwdriver
   test still doesn't approximate what happens with the string.
   The screw drive doesn't have a frequency/note.

Yes, I am aware of that. Was just testing to see if the sensitivity improved from the nearly completely unresponsive state it was in before I rewound that coil. I still have PTSD from that. 

-23:00 Input waveform

The input waveform is always going to be sinusoidal as the
inductor and capacitor act like a filter to remove the harmonics.

Noted.

-27:16 Spice Sim not comparing apples to apples.

The spice sim has a strong 247Hz input signal.
The real circuit is operating via a self oscillation at 4kHz.

Yeah I know. I was trying to see if I could get something in LTSpice that looked anything like real world observation and then compare how far apart the two are.

The spice sim is doing a completely different thing to the real circuit.
You need to do a spice sim which is also self oscillating.

Tried that but I can't get it to look like what I observed on the scope.

I can't quite see the schematic but you have what looks like a 5k
resistor between the coil and the 33nF cap.
The resistor value should be the DC resistance of the coil.
  A 5k in that position could stop self-oscillation.

  If you want to put in the 4.7k "work-around" for the model I
  mentioned some time back then the 4k7 needs to be between
  the 33nF cap and the LM386 input.

You can ignore that, was just messing around to see what happens in LTSpice when I do certain things.

-29:00 From the pic it makes me think the last band is orange.
Perhaps the feedback resistor is 130k? instead of 13k

I had same thought, but it was just not possible to determine 100% what it is. Sustained too much damage and the goop obscured the last tiny bit of that band which was left to see.

Going forward:

Test the receive coil by wiring it to an amplifier.

Check how much output when you place it over the string
and pick a note.  If the receive pickup output is
is too low it will not have enough signal to sustain string
oscillation.

You could even wire the receive coil to the LM386.  Look
at the output of the LM386.

If plucking the string cannot clip the LM386 I think
you have a problem.  It's unlikely the unit will
sustain string oscillations.

That's along the lines of what Jim noted as well.

Can't really do that... the wires are just too delicate. 

Here is a demonstration of how I don't have to do anything to the string to get an Ebow to do its thing, and something I noticed about the harmonic mode.

[Rob Strand]

After I fabricated the one string jig (in the quote right above this comment) I was able to get it to vibrate a B string, but it's kinda weak. Also doesn't vibrate at the fundamental, so yeah it's not able to overcome the oscillator. I was thinking that perhaps the output coil isn't quite strong enough but haven't yet made a determination on that. This morning I added a couple more wire appendages attached to the PCB so I can add cap in parallel to the 0.033uF via capacitor substitution box and mess with the oscillator frequency. I currently have a device that can inject the oscillation into a guitar PUP, so obviously that's not right

The thing which is a bit puzzling is your new coil should be similar to the original now.   The main reason to play with the cap would be to tune the self-oscillating frequency to be similar to the original Ebow, about 2.2kHz.   I think your best bet would be to use the largest feedback resistor you can which still self-oscillates.   That gives the string the best chance to over come it.

That's what I measured. Not sure if other things mess getting an accurate measurement on that.
EDIT: I measured again in circuit and it was 11.1mH. Not sure what happened there.... I measured right after I wound it. Attack of the one-eyed steel ring monsters?

The inductance still doesn't make sense.   The old inductor had less turns and it was 17mH to 25mH.  I would have thought the new coil would be 40mH to 100mH, something like that.

If you measure the self-oscillation frequency with a known cap (33n or 2x33n), whatever you have in there then calculate L = 1/( (2*pi*f)^2 * C) it should be close to the measured inductance. 

What caps do you have in there? and what is the self oscillating frequency?

What kind of resistance would that be? Would it be in series with coil, parallel?

It's very difficult to work out.   It's best done by measurement at the desired frequency. 
A simple way:  You find the largest feedback resistor that causes self oscillation in the real circuit.   Now you enter all the correct inductance, DC resistance, and cap values into spice along with that feedback resistor value.  Then you add a parallel resistor across the inductor and reduce the value until self oscillation stops, then bump it back up.   

Trying to calculate the resistor value by maths and physics is very difficult.

Noted. I put a 10uF in there because I didn't have 4.7uF cap that was physically that small. How does that affect things?

I'm not sure!

Something still not so clear is if the cap is acting like a filter then that would make the gain lower and make it more difficult to lock in harmonic mode.  Maybe that's why they added the diode, to bump up the output level.

Not 100% sure about type of diode, but reversing the diode does not work. I can demonstrate in another video I will make soon. I added a reversing switch to see what happens.
On my nearly destroyed example the diode was too busted up to tell anything about it, and on the only other known dissected PCB picture I can find, it appears that the band points to upper right hand corner of the PCB as I have it drawn. So we know correct polarity (for a 1N4148 type) but we really have no idea what it actually is. 

Yes, I later I remembered I had checked the diode polarity in the pic many posts back.


Forgot to mentioned I think your output coil is fine.  It matches all the specs of the original coil.

[Rob Strand]

I've updated the LM386 spice model to prevent the output inverting when there are strong negative
swings on the non-inverting input.

I added two resistors to the base of Q3 and Q4.

In the real device, these low valued resistors can be part of base of the transistor itself.  I was
tempted to set the RBB parameter of the transistor model but I didn't want to screw-up all the
other transistors so I just added real resistors (R33, R34) to the circuit.

Here's the idea, (not neatly drawn)

Code Select Expand


* LM386
*
* Rob S, V1, 2022/12/14
* 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, V2, 2022/12/23
* A fix is to add 220 ohm series base resistors
* to Q3 and Q4.
*
*
* 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

* Base resistors to prevent output inversion
* with strong -ve INP inputs
R33 11 33 220
R34 15 34 220

* differential input stage
q3 13 33 12 pn
q4 14 34 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


[Rob Strand]

Something else,  when you clip the LM386 heavily the output signal *before* the output cap has a rising top like the waveforms in your last video.

[Paul Marossy]

Thanks for the new SPICE model Rob, very useful. With my circuit as it is currently, I am finally getting something in LTPspice that isn't so far off from real world.

CHECK THIS OUT! IT WORKS! Well, for the most part. 

Problem appears to have been the output coil. I rewound it yesterday. I used wire from some little transformer I've had for many years that I couldn't really use for anything but didn't want toss it... so it finally had found a purpose after all these years. I believe it has 22 AWG wire in it.

Anyway, I started out intending to wind 8 ohms but that barely had any wire around the pole piece, so I thought that's not right and thought that perhaps they used some other value like 16 or 32 ohms. I figured that maybe 32 ohms was what I should try. That worked but it was not vibrating the string very intensely, so I figured maybe they did a 10:1 ratio thing and I added some more wire to make it 45 ohms. That was a big improvement but still not enough. So I added yet more wire to it and I figured maybe 75 ohms would do it after modeling it in LTSpice. Ended up with 79 ohms and that works really well! The new output coil specs are 79 ohms DC resistance, 1.14uF capacitance & 9.4mH apparent inductance.

The next problem I had to tackle was the oscillator injecting noise into the pickup. Also couldn't get the string to vibrate unless I have the oscillator set at the same frequency as the string, but it would maintain it once I got it started. That seemed like some real progress at least. After quite a lot of experimentation I found that it worked best with the oscillator running at 2.12kHz. To get it to run at that frequency I am using .056uF cap in parellel with the .033uF cap. Feedback resistor ends up being 177K for it to work optimally (as possible). With these settings the oscillator I guess is not running(?) and when I place the circuit close enough to the strings it comes to life. So I guess that is how they get it to be quiet? I surmise that the oscillator starts up and the string quickly takes over? Turns out that the full time oscillator doesn't really work unless you're focusing on a single frequency and it's far too noisy to be usable. Was an interesting learning experience. It would appear from my experimentation that the Ebow probably uses a 130K feedback resistor.

My jig doesn't allow me to put the circuit directly over the string as it would be when I am using a Ebow but I saw voltages of up to 100+ mV so I think the output is probably comparable to my unmolested Ebow (it will put out 200mV if it's right over the pickup).

So really the only remaining problem is that the harmonic mode doesn't really work. I think it's trying to do something but can't quite get there. When I switch between the two modes I have to fiddle with the feedback resistor whenever I switch between modes (one works and the other one is some distance off). I suspect that the 4.7uF cap needs to be some other value because my coils are different that whatever is actually used. From one of my experiments, it appears that when in harmonic mode it's tripling the fundamental or at least that is what is sounds like. I can get it to double, but not triple.

All in all, I am pretty satisfied that I have been able to get this far. The coil winding jig that I made has been invaluable. I re-wound both coils a few times and it sure made it less of a P.I.T.A.

Maybe I could stop here but I am curious why the harmonic mode doesn't want to work. Also am wondering if there are any other possible diodes besides a 1N4148 that it could be. I saw a "8" on the broken diode I removed from the PCB, so I assumed it's probably just a regular old 1N4148. BTW if it's reversed, it just kills the power to the circuit.

Will put out another video in the near future of the whole testing process for those who might find it interesting.

[anotherjim]

I always admire the determination required to DIY a coil. The last time I did it was in engineering school 50 years ago and that was with a winding machine and a tutor - and I only had to do it once.

In harmonic mode, do the harmonics have a predictable relationship with string pitch throughout the fretboard? Could the distance between sense and drive coils have a bearing on the harmonic pitch?
I'm equally fascinated/perplexed by the E-bow harmonic method. Is it somehow like the pinch-pick technique that Billy F Gibbons uses?

[deadastronaut]

on sustainers you just flip the driver coil for harmonic mode.

god know how they do that on an ebow with just an spdt....hmmmmm....


ive been inspired by this thread, i was going to knock up an ebow, but then found my sustainer test board so am messing with that again....cheers for the inspiration guys.   

have a great xmas all. 

[anotherjim]

Yeh, I just used polarity. But I did it with SPDT just to select between + or - input of the 386 with a centre-off switch.
Mine didn't need a biasing magnet. Harmonics seem to need a stronger drive. Being fixed to the instrument, it can self-start but that's probably from the switch-on thump of the 386.

[deadastronaut]

SPDT just to select between + or - input of the 386 with a centre-off switch.

ooohh yeah makes sense, i,ll try that...nice one jim.. 

[Paul Marossy]

I always admire the determination required to DIY a coil. The last time I did it was in engineering school 50 years ago and that was with a winding machine and a tutor - and I only had to do it once.

Yeah you need some real determination alright. It's time consuming and tedious, requires steady hands and a light touch. Harder to do when you're taking wire from a winding in an inductor or small transformer. You also have to devise a way to do it. I learned how to repair broken wires, it happened a few times. 

In harmonic mode, do the harmonics have a predictable relationship with string pitch throughout the fretboard? Could the distance between sense and drive coils have a bearing on the harmonic pitch?
I'm equally fascinated/perplexed by the E-bow harmonic method. Is it somehow like the pinch-pick technique that Billy F Gibbons uses?

I haven't tested it extensively but I am pretty sure that it triples the fundamental.

on sustainers you just flip the driver coil for harmonic mode.

god know how they do that on an ebow with just an spdt....hmmmmm....

Years ago that is what I thought they did in the Ebow... a centertapped driver coil or something along those lines. I guess they accomplish that with the diode and 1N4148 diode. Still seems very mysterious to me how that even works.

[anotherjim]

SPDT just to select between + or - input of the 386 with a centre-off switch.

ooohh yeah makes sense, i,ll try that...nice one jim.. 

I think it was Duckarse who planted the idea in that Opto-guitar thread.
Not applicable in the E-bow as it's also the power switch.
I do wonder if they didn't consider a  momentary push button for power. A "deadmans handle" like on some power tools. Let go and the power is off.

[Rob Strand]

Thanks for the new SPICE model Rob, very useful. With my circuit as it is currently, I am finally getting something in LTPspice that isn't so far off from real world.

It took lot of mind bending to work out how to fix it without making unrealistic mods to the internal circuit.   I did some poking around on the web and I found an example were a guy got the signal inversion of the output.    He was pumping 5V into the input of the LM386.   The change I made was very subtle, it's so subtle I'm thinking some real devices invert when the input is overdriven and some don't.  (I'm trying to resist unpacking my archived parts.)

CHECK THIS OUT! IT WORKS! Well, for the most part. 

Really good progress.   Especially for Christmas when things a suppose to be quiet  "Not a creature was stirring, not even a Mouse."

Problem appears to have been the output coil. I rewound it yesterday. I used wire from some little transformer I've had for many years that I couldn't really use for anything but didn't want toss it... so it finally had found a purpose after all these years. I believe it has 22 AWG wire in it.

Anyway, I started out intending to wind 8 ohms but that barely had any wire around the pole piece, so I thought that's not right and thought that perhaps they used some other value like 16 or 32 ohms. I figured that maybe 32 ohms was what I should try. That worked but it was not vibrating the string very intensely, so I figured maybe they did a 10:1 ratio thing and I added some more wire to make it 45 ohms. That was a big improvement but still not enough. So I added yet more wire to it and I figured maybe 75 ohms would do it after modeling it in LTSpice. Ended up with 79 ohms and that works really well! The new output coil specs are 79 ohms DC resistance, 1.14uF capacitance & 9.4mH apparent inductance.

Awesome effort!

Normally when you want to make the magnetic field the strongest you try to make the have the lowest resistance.   Low in the sense that for a given supply voltage you set the resistance so it pulls the most current the amplifier can handle.   If you look at relays offered in different voltages the high voltage coils take less current by the power (I*V) is about the same similar relays.   If you want a more pull from the relay for a given voltage then you use a lower coil resistance.  (It's possible to show all this with maths.)

The fact you got more pull from a higher resistance coil means something wacky is going on.   What comes to mind is with 8 ohm the current is higher and it's causing something to sag like the output voltage or the power rail.

The next problem I had to tackle was the oscillator injecting noise into the pickup. Also couldn't get the string to vibrate unless I have the oscillator set at the same frequency as the string, but it would maintain it once I got it started. That seemed like some real progress at least. After quite a lot of experimentation I found that it worked best with the oscillator running at 2.12kHz. To get it to run at that frequency I am using .056uF cap in parellel with the .033uF cap. Feedback resistor ends up being 177K for it to work optimally (as possible). With these settings the oscillator I guess is not running(?) and when I place the circuit close enough to the strings it comes to life. So I guess that is how they get it to be quiet? I surmise that the oscillator starts up and the string quickly takes over? Turns out that the full time oscillator doesn't really work unless you're focusing on a single frequency and it's far too noisy to be usable. Was an interesting learning experience. It would appear from my experimentation that the Ebow probably uses a 130K feedback resistor.

When I played around with this in spice a few posts back I concluded making the feedback resistor as large as possible should help locking so your results are showing that now.

For your new output coil I get  1130 turns of 40AWG and the resistance is 79 ohm and inductance roughly 15mH.

If I take your 2.12kHz and 56nF+33nF = 89nF  I calculate the inductance of your receive coils as 1/[(2*pi*f)^2*C] = 63mH.

I have a feeling the iron core inductors are screwing up the measurements on your DMM.

My jig doesn't allow me to put the circuit directly over the string as it would be when I am using a Ebow but I saw voltages of up to 100+ mV so I think the output is probably comparable to my unmolested Ebow (it will put out 200mV if it's right over the pickup).

It's going to be tricky without a string.  I was think about a tight elastic band with a small off-cut of guitar string on it to make it magnetic. However sticking things onto strings often kills the "note" completely.  Try sticking a tiny ball of blue-tack on a guitar string, it goes dead.

So really the only remaining problem is that the harmonic mode doesn't really work. I think it's trying to do something but can't quite get there. When I switch between the two modes I have to fiddle with the feedback resistor whenever I switch between modes (one works and the other one is some distance off). I suspect that the 4.7uF cap needs to be some other value because my coils are different that whatever is actually used. From one of my experiments, it appears that when in harmonic mode it's tripling the fundamental or at least that is what is sounds like. I can get it to double, but not triple.

Yes, no doubt the whole thing is fine tuned.   Especially since the Ebow doesn't flip the polarity of the coils in Harmonic mode.  I can only think relies on high pass filtering of the 4.7uF cap and the corresponding phase-shift that results to promote oscillations at higher frequencies.  As mentioned before, a high-pass filter in the output is going to drop the drive level so if the whole string motor and receive pickup combination aren't at 100% max sensitivity it might not work.

Since your new output coil is 79 ohms compare to the original 8 ohm I suspect you might need to reduce the 4.7uF cap so that it provides an equivalent amount of high-pass filtering and phase-shift.

Maybe I could stop here but I am curious why the harmonic mode doesn't want to work. Also am wondering if there are any other possible diodes besides a 1N4148 that it could be. I saw a "8" on the broken diode I removed from the PCB, so I assumed it's probably just a regular old 1N4148. BTW if it's reversed, it just kills the power to the circuit.\

Yes, it make sense.  Reversing would only make sense if the diode was a zener.

Will put out another video in the near future of the whole testing process for those who might find it interesting.

That would be cool.

In harmonic mode, do the harmonics have a predictable relationship with string pitch throughout the fretboard? Could the distance between sense and drive coils have a bearing on the harmonic pitch?
I'm equally fascinated/perplexed by the E-bow harmonic method. Is it somehow like the pinch-pick technique that Billy F Gibbons uses?

I posted a pic back in reply #13.   When you flip the one of the coils you drive at a peak and receive at a trough.   When whatever wavelength that is matches the spacing between the two coils it will oscillate at that frequency - which is much higher than the fundamental frequency when you the two coils are in-phase.

[Paul Marossy]

It took lot of mind bending to work out how to fix it without making unrealistic mods to the internal circuit.   I did some poking around on the web and I found an example were a guy got the signal inversion of the output.    He was pumping 5V into the input of the LM386.   The change I made was very subtle, it's so subtle I'm thinking some real devices invert when the input is overdriven and some don't.  (I'm trying to resist unpacking my archived parts.)

You did a great job. Once I discovered a dumb mistake I made on the LM386 schematic it was looking much more like real world.

Really good progress.   Especially for Christmas when things a suppose to be quiet  "Not a creature was stirring, not even a Mouse."

I have 11 days off with not much else to do this year. My mouse was quite active! 

Awesome effort!

Normally when you want to make the magnetic field the strongest you try to make the have the lowest resistance.   Low in the sense that for a given supply voltage you set the resistance so it pulls the most current the amplifier can handle.   If you look at relays offered in different voltages the high voltage coils take less current by the power (I*V) is about the same similar relays.   If you want a more pull from the relay for a given voltage then you use a lower coil resistance.  (It's possible to show all this with maths.)

The fact you got more pull from a higher resistance coil means something wacky is going on.   What comes to mind is with 8 ohm the current is higher and it's causing something to sag like the output voltage or the power rail.

Rewinding/adding more windings to the driver coil is easy. The wire is so much bigger! You'll likely think there's some more things that don't make sense with respect to the driver coil.

When I played around with this in spice a few posts back I concluded making the feedback resistor as large as possible should help locking so your results are showing that now.

For your new output coil I get  1130 turns of 40AWG and the resistance is 79 ohm and inductance roughly 15mH.

If I take your 2.12kHz and 56nF+33nF = 89nF  I calculate the inductance of your receive coils as 1/[(2*pi*f)^2*C] = 63mH.

I have a feeling the iron core inductors are screwing up the measurements on your DMM.

Yep that seems to be correct as far as the FB resistor is concerned. With what I believe is 22 AWG wire, my 79 ohm winding was about 8.5 mm x 6 mm high. Don't know how many windings that would be but it didn't take that many to get to that DC resistance.

It's going to be tricky without a string.  I was think about a tight elastic band with a small off-cut of guitar string on it to make it magnetic. However sticking things onto strings often kills the "note" completely.  Try sticking a tiny ball of blue-tack on a guitar string, it goes dead.

I came up with a clever one string device that worked pretty well. Even found a simple way to flip it upside down. You'll see it in next video.

Yes, no doubt the whole thing is fine tuned.   Especially since the Ebow doesn't flip the polarity of the coils in Harmonic mode.  I can only think relies on high pass filtering of the 4.7uF cap and the corresponding phase-shift that results to promote oscillations at higher frequencies.  As mentioned before, a high-pass filter in the output is going to drop the drive level so if the whole string motor and receive pickup combination aren't at 100% max sensitivity it might not work.

Since your new output coil is 79 ohms compare to the original 8 ohm I suspect you might need to reduce the 4.7uF cap so that it provides an equivalent amount of high-pass filtering and phase-shift.

May be the case, but it still works. Just is harder to get it going. I've been messing around with LTSpice for an embarassing number of hours trying to come up with something that looks like real world. I think I'm pretty close now but I still don't understand how the harmonic mode works. At least the program isn't doing anything that makes any sense to me in that regard. Real device appears to be making use of the 2nd & 3rd harmonic of the fundamental.

Yes, it make sense.  Reversing would only make sense if the diode was a zener.

Agreed.

I posted a pic back in reply #13.   When you flip the one of the coils you drive at a peak and receive at a trough.   When whatever wavelength that is matches the spacing between the two coils it will oscillate at that frequency - which is much higher than the fundamental frequency when you the two coils are in-phase.

I believe mine are in phase. Curiously, the circuit does not really behave any differently if I reverse the wires on one of the coils. I could understand it if it literally reversed the windings on the coil but doing it with just a diode is mysterious to me. I wonder if somehow it takes advantage of the lowered supply voltage along with the battery resistance (which I have assumed is 2 ohms after Googling it).

I'm going to start new thread with the next video so we can start with a more or less working circuit and maybe figure out the rest of it out from where I am at with it now. It seems currently my input coil still isn't quite strong enough and I don't know that I could rewind it by hand again with even tinier wire. I feel pretty good that I have been able to get as far as I did with it. It's a long video but a few people should find it very interesting.

[Rob Strand]

May be the case, but it still works. Just is harder to get it going. I've been messing around with LTSpice for an embarassing number of hours trying to come up with something that looks like real world. I think I'm pretty close now but I still don't understand how the harmonic mode works. At least the program isn't doing anything that makes any sense to me in that regard. Real device appears to be making use of the 2nd & 3rd harmonic of the fundamental.

It's doubtful anything with show up in the simulation. You have to add more stuff to account for the spacing between the coils and behaviour of the string.

In the simulations I posted before I have a circuit which represents the string so I can see both the self oscillation and the string oscillations.  Even that isn't enough for harmonic mode.   You would need to add a circuit string which had more than one resonant frequency.   A real string has many resonant modes, see reply #13.

Getting spice simulations to match reality is tricky.  It's much more difficult when you have you want to incorporate the behaviour of physical objects.   Physical objects aren't electronics.  You have to come-up with whole circuits which act like the physical objects.   In my simulations I used an parallel LCR circuit to represent the string.   The spacing between the coils could be represented by a time delay between the input an output coils. The disturbance of the string from the output coil takes time to propagate up the string to the receive coil.