Ebow Exposé Part IV - RESURRECTED!

[Paul Marossy]

I had this idea to compare the sensitivity/gain and frequency response of the real ebow and the rebuilt ebow from outside the box.



The set-up can only check the electrical phase of the coils.  However you can adjust the results based on the magnetic phase using a compass. [I forgot to mention this is OK for A/B comparisons at a given frequency but to get a flat response output you need to feed the output into an integrator.]

For your video you place the compass to the side of the ebow however what you should check is compass needle direction normal to the coil faces (as shown in the pic).   It should be evident which direction the magnets are strongest from the snap in the compass needle.

That's an interesting test. I can try doing that sometime in the near future. I just did the compass test the way you suggest and if I'm holding the Ebow as I would normally, north is up for both coils.

An idea to increase the sensitivity of the rebuilt unit was to place a small magnet *on the back* of the existing magnets.   The way they stick together naturally is the correct orientation.   However, if you have strong rare-earth magnets I first suggest adding a small spacer between the two magnet as those rare-earth magnets an completely changed the magnetization (and direction!) of other magnets when they touch them.   You want to make sure you know exact what direction the existing magnets are before you start putting those rare-earth magnets near them!

Noted. I didn't keep the rare earth magnets near the button magnets for very long.

It occurred to me later this set-up might be a pain in the butt for testing the Ebow.  I have used it to test pickups in the past and it works fine.  It's low inductance coil fed with a current and you correct the response with an integrator.   The thing about the Ebow is we would like to inject a constant voltage into the LM386, ie. we want a constant voltage out of the receive coil.   The low inductance coil won't do that unless you put the integrator before the test coil.

I'm sure I can figure out a way to set up a test jig. Would be interesting to do.

It seems our minds are all on the same page because I noticed you posted a pickup test idea in the lounge,
https://www.diystompboxes.com/smfforum/index.php?topic=129998.0

In your test set-up the coil is a high inductance coil fed with a voltage.   That would be much more convenient to test the Ebow than the low inductance coil.

Thanks to you helping me fix my old Heathkit o'scope a year or two ago, which opened up a whole new world to me. I didn't even know anything about X-Y mode capabilities before that!

The initial thinking was normal model has the coils in phase and produced the fundamental.  Harmonic mode drove the coils out of phase, and the harmonic depends on the spacing of the ebow coils.

I believe there is an explanation to why the second harmonics is produced in normal mode!  Look at the string motion of the harmonic and the second harmonic.


The pickups and ebow are on the right half of the string.    When you place the ebow in this area *both* coils see the same polarity of the string for fundamental and second harmonic.   In both case the string is up or down for both coils.  Normally you would expect the fundamental to win but since the amplifier response isn't flat it might help to promote the string oscillation to be the second harmonic!   You could argue that with more dramatic frequency shaping it might favor the third harmonic - hence the idea that a high-pass filter promotes harmonics mode.

If one of the coils has the phase flipped then the position of the ebow would need to be placed so that the drive coil was on he left of one of the string nodes and the receive coil was on the right of the same string node - sort of promoting rocking about that point.   The fact it depends on position makes the results more variable.

I think you're onto something there. I did some calculations for 25.5" scale guitar, and found that the area where the Ebow would normally be used is in the zone where the anti-node occurs for the 2nd and 3rd harmonic. The anti-nodes of 2nd and 3rd harmonics are 1.5" apart. The Ebow appears to be taking advantage of string wave characteristics. The change in waveforms at the output coil seems to promote one vs. the other. Those extra humps on the waveform when in harmonic mode I think would be all it needs to make it lock onto the 3rd harmonic. Also, when in harmonic mode there's kind of parallel paths to ground - one is battery return and the other one is coil going to ground thru the diode, as it is doing full time. The last big question in my mind is why the oscillator frequency of 2.4kHz & 2.6kHz?

I guess the other reason for similarities is they are both from the pickup and both from an ebow exciting the strings.   The two waveforms have very similar origins.  Even if we don't know what is happening we know they come from the same place.  When you go to LTspice you can't only trust the waveforms unless you *know* you have captured all the physical and electrical effects in the LTspice.  We also know that a simple spice simulation probably doesn't do that,
so we shouldn't expect them to be the same or be surprised if they aren't!

Yeah, I know that a program like LTSpice is only going to be an approximation of real world. I think it does a pretty good job in this case. It is hard to model something like this and even more so when you "know enough to be dangerous" 

[Paul Marossy]

You might fine-tune magnets with pieces of the plastic magnetic tape/strip used for fridge badges etc. It won't be strong enough to alter the existing magnet a great deal but depending on its polarity can add or subtract to the field.

If you can go back to the one-string test bed, clip the scope probe neg on the battery neg (diode cathode) and see what signal is present on the diode anode. Whatever the diode//cap is doing should show as some kind of signal, although it might only be small mV amplitude.

That would be interesting to see. I can try doing that sometime in the near future.

Funnily enough the coil I settled on for my experimental sustainer had a DCR of 90. Mine is body-mounted like a pickup so had to be driven from a higher supply voltage because the distance from the string is variable when played. Also, to get harmonics seem to need a stronger drive anyway. At the moment the driving LM386 gets 11v.
It produces the fundamental.
Harmonic mode is polarity reversal but the interesting thing is that the harmonics repeat on either side of the 12th fret. That is the 3rd fret is the same harmonic as the 15th fret!

That would be very trippy having the 3rd fret same as the 15th fret! 

Here's a question. Does LTspice sim a battery fully? You added a series resistor for the battery internal resistance, but is the capacitance represented? What even would the capacitance of a battery be? Is there capacitance built into the simulated DC supply?

LTSpice has a box where you can enter a number for the battery internal resistance, but I have no idea what the program actually does with that info.

[anotherjim]

Well, you may know that some pedal designs put a diode in series with the +9v feed for polarity protection. This is an alternative to the reverse polarity diode placed between +9v and ground and the advantage is that the series diode will not cause a short circuit with a wrong polarity supply and all that happens is - nothing! The diode will not pass current in reverse. The design must add a supply bypass capacitor, large enough to have a low impedance at the lowest frequency the circuit works with. Incidentally, it makes no difference here if the series diode is in the + or - supply feed.

The series diode prevents the bypass capacitance of either a battery or DC power adapter from fully bypassing the AC currents produced in a circuit. It may increase the impedance of the bypass path in one polarity which can affect the operation of the circuit. Adding a supply bypass cap on the circuit side of the series diode solves this. The Ebow has no such supply bypass capacitor, there is only the battery, and that is on the other side of a series diode in harmonic mode.

Now, a battery cell has a lot in common with a capacitor (two plates with something separating them) and is often relied upon as the only path for circuit AC currents. Capacitance comes with the territory, it simply has to exhibit it. It's more important than the battery's internal resistance which is trivial when the battery is good but increases as the battery becomes discharged. So you can only really put a likely value to the resistance and move on.

So, an experiment in the sim might be to add a capacitor across the battery, maybe a few different values from 1uF to 100uF and see if there is any change in waveforms in harmonic mode.
In case the penny hasn't dropped, I'm trying to see what the diode is for. If it isn't there to make things asymmetric by rectifying AC currents and so increasing 2nd harmonic, then I don't know what!

[Paul Marossy]

Well, you may know that some pedal designs put a diode in series with the +9v feed for polarity protection. This is an alternative to the reverse polarity diode placed between +9v and ground and the advantage is that the series diode will not cause a short circuit with a wrong polarity supply and all that happens is - nothing! The diode will not pass current in reverse. The design must add a supply bypass capacitor, large enough to have a low impedance at the lowest frequency the circuit works with. Incidentally, it makes no difference here if the series diode is in the + or - supply feed.

Yes, usually I see the diode across the power supply. Seems like generally speaking, people don't want the voltage drop of a series diode, so they put it across the power supply.

The series diode prevents the bypass capacitance of either a battery or DC power adapter from fully bypassing the AC currents produced in a circuit. It may increase the impedance of the bypass path in one polarity which can affect the operation of the circuit. Adding a supply bypass cap on the circuit side of the series diode solves this. The Ebow has no such supply bypass capacitor, there is only the battery, and that is on the other side of a series diode in harmonic mode.

Now, a battery cell has a lot in common with a capacitor (two plates with something separating them) and is often relied upon as the only path for circuit AC currents. Capacitance comes with the territory, it simply has to exhibit it. It's more important than the battery's internal resistance which is trivial when the battery is good but increases as the battery becomes discharged. So you can only really put a likely value to the resistance and move on.

So, an experiment in the sim might be to add a capacitor across the battery, maybe a few different values from 1uF to 100uF and see if there is any change in waveforms in harmonic mode.
In case the penny hasn't dropped, I'm trying to see what the diode is for. If it isn't there to make things asymmetric by rectifying AC currents and so increasing 2nd harmonic, then I don't know what!

I don't know to model it in LTSpice. Rob mentioned Warburg Impedance, and I have no idea how to model that. I get the concept but doing the math on that is over my head. I tried doing the capacitor across the battery but that doesn't really affect the waveforms at all. I do see a difference in the current at the output coil between modes but voltage-wise looks pretty much the same no matter what value I plug in for capacitance. I think part of the problem is that the model for the LM386 may still be missing some items... as Rob had mentioned in one of these posts. There may be a protection diode or two in that chip somewhere that doesn't show up on the manufacturer data sheet (where they show the schematic of the IC chip). Currently the bottom half of the of the output waveform has a little hump in it when it should just be flat like a clipped sine wave. That kind of messes up one half of the output waveforms on these models. 

[anotherjim]

I don't know to what extent the modelling deals with DC supplies, but as many circuits can be lashed up without bothering with any power caps the voltage sources probably act as much like capacitors as they need to keep the voltage stable.

Is it possible to place a "differential" voltage probe across the harmonics diode? This so that the circuit ground doesn't ignore the fact that the battery ground will be a diode drop negative from the circuit ground. I mean, if the diode is doing anything, it must have a signal across it, but you won't see that if the normal voltage probe has its negative connection on circuit ground.

[Rob Strand]

I just did the compass test the way you suggest and if I'm holding the Ebow as I would normally, north is up for both coils

Essentially opposite to the arrows on my drawing.   What that should mean is you can treat the electrical phase as the overall phase.   The round trip voltage gain should be a very good indication of the sensitivity.  And it will show just much lower the gain of the rebuilt unit is.   I was thinking if it's only a factor of 2 you could tweak the LM386 with some external components to double the gain.

Noted. I didn't keep the rare earth magnets near the button magnets for very long.

The change is a very fast process!   What you can also do is place the existing magnets between two large rare-earth magnets to re-magnetize the existing magnets to maximum.   The two rare earth magnets are orientated so they attract, effectively creating a field in between.  (This set-up isn't quite as good as it could be as ideally the backs of the two rare-earth magnets should be linked with a magnetic path.  I've see people using a vice as a permanent jig where the vice frame links the backs of the magnets.   I wouldn't use a vice for a once-off since the process will magnetize the vice which would be extremely annoying.  It's also difficult to demagnetize!)

I'm sure I can figure out a way to set up a test jig. Would be interesting to do.

I'm sure you will come up with something.

Thanks to you helping me fix my old Heathkit o'scope a year or two ago, which opened up a whole new world to me. I didn't even know anything about X-Y mode capabilities before that!

No problem good to see it up and running.  I used XY mode a lot on my analog scope.  There was point in time where digital scopes didn't support XY mode and it made me move to different methods.

The last big question in my mind is why the oscillator frequency of 2.4kHz & 2.6kHz?

In the case where the unit self oscillates the lower frequency it makes it easier to lock to the string.  (I explained the details in the other thread).

In the case where the unit doesn't self-oscillate it does something as well.   When the circuit is adjusted just below self-oscillation the frequency response has a large peak at 2.5kHz.   This provides a gain boost.  The gain will increase from low frequencies upto 2.5kHz.  That increases sensitivity and emphasizes harmonics.

I don't know to model it in LTSpice. Rob mentioned Warburg Impedance, and I have no idea how to model that. I get the concept but doing the math on that is over my head.

There's nothing special about LTspice's internal resistance.   It's just like adding a resistor in series with the voltage source.  It defaults to 1 milliohm.   (LTspice's Caps and Inductor also have a series resistance, which doesn't default to zero.  FYI, if you don't set it to zero it screws up high Q circuits like my string model.  However, my string model was rough so I didn't bother setting the series resistance.)

The problem with 9V batteries is the impedance is all over the map.

Heavy duty new:  25 ohm
Alkaline new:   1 to 5 ohm

When the battery runs down the resistance goes up quite a bit and the open circuit voltage actually drops a bit as well (which you can ignore an let the impedance do the voltage drop).    Just how high the resistance goes up depends on the circuit you are powering more than the battery itself.

For example at heavy loads (1hr discharge) a 9V alkaline might reach the end of useful life at 10 ohm to 20 ohm.    Notice that this is less than the resistance of a *new* heavy duty battery, meaning a heavy duty battery cannot supply such heavy currents.

When the alkaline battery in my DMM comes up with low battery and I leave the battery sitting around I've measured 10 to 20 ohms, even though the DMM is a low current device.  (Simple low battery indicators use voltage, not impedance.)

In a light load device which can operate down to low voltages a 9V battery might reach end of life at 20 ohm upto 200 ohm; for either alkaline to heavy duty.

So up front, the range of impedance is enormous!     I think 10 ohms would be a good starting point for the DC resistance.    However it wouldn't take much to bump that out to say 33 ohm if you kept using a flat battery.

The AC impedance can be crudely approximated as follows:



- Choose what DC resistance you want to emulate, say 10 ohm
- ignore Ld
- ignore the LTspice source resistance, or set it to zero
- Add series resistor R0, value about 0.1 to 0.2 times the DC resistance
- Set Rct to 0.9 to 0.8 times the DC resistance.  (you want R0 + Rct to add to the DC resistance.)
- Set the cap to  Cd = 1 / Rct.   This will be an enormous cap and sets a 1 second time constant
  I just pulled cap value from the "7th planet", might need refining.

Since there is an average DC load you should still see a voltage drop determined by the DC resistance.
So arguably the circuit will show the same droop without the cap however, with amplifiers a large series
resistance can cause the amp to oscillate so the impedance model will help stop that happening to some degree.

Something else, imagine the AC output swinging +/-3V peak and the output load is 8 ohm.  That's 3V/8 = 375mA which will cause a 10*0.375 = 3.75V drop across a 10 ohm battery impedance.   So quite a bit of drop on a 9V rail.   If you backed the load off to 16 ohm, 3V/16 = 188mA, drop across 10 ohm = 1.9V.   As the battery impedance goes up drop the amount of swing of the amp will drop and there's a point where you won't get +/- 3V peak.  If the load impedance is increased you will get less drop and more output swing.   So for a *given* battery DC resistance there will be a load which gives the maximum amount of field.   The thing is for a 10 ohm battery impedance the maximum field won't end-up with a 79 ohm coil, it will be a lower impedance.

[Paul Marossy]

[Essentially opposite to the arrows on my drawing.   What that should mean is you can treat the electrical phase as the overall phase.   The round trip voltage gain should be a very good indication of the sensitivity.  And it will show just much lower the gain of the rebuilt unit is.   I was thinking if it's only a factor of 2 you could tweak the LM386 with some external components to double the gain.

I see. That makes sense. With this new Ebow test that you propose, do I just get it to self-oscillate? Can't really get it close enough to a guitar string for it to work but a screwdriver ought to have enough mass to get it going, maybe. I would be using an unmolested one for this test.

In the case where the unit doesn't self-oscillate it does something as well.   When the circuit is adjusted just below self-oscillation the frequency response has a large peak at 2.5kHz.   This provides a gain boost.  The gain will increase from low frequencies upto 2.5kHz.  That increases sensitivity and emphasizes harmonics.

I went back and re-read those posts. So maybe they picked that frequency because of the gain boost and it also seems that this frequency range works best for exciting the string in the right frequency range (2nd & 3rd harmonics). Too high of an oscillator frequency and it just doesn't do anything. Not without tweaking the circuit I guess.

There's nothing special about LTspice's internal resistance.   It's just like adding a resistor in series with the voltage source.  It defaults to 1 milliohm.   (LTspice's Caps and Inductor also have a series resistance, which doesn't default to zero.  FYI, if you don't set it to zero it screws up high Q circuits like my string model.  However, my string model was rough so I didn't bother setting the series resistance.)

The problem with 9V batteries is the impedance is all over the map.

Heavy duty new:  25 ohm
Alkaline new:   1 to 5 ohm

When the battery runs down the resistance goes up quite a bit and the open circuit voltage actually drops a bit as well (which you can ignore an let the impedance do the voltage drop).    Just how high the resistance goes up depends on the circuit you are powering more than the battery itself.

For example at heavy loads (1hr discharge) a 9V alkaline might reach the end of useful life at 10 ohm to 20 ohm.    Notice that this is less than the resistance of a *new* heavy duty battery, meaning a heavy duty battery cannot supply such heavy currents.

When the alkaline battery in my DMM comes up with low battery and I leave the battery sitting around I've measured 10 to 20 ohms, even though the DMM is a low current device.  (Simple low battery indicators use voltage, not impedance.)

In a light load device which can operate down to low voltages a 9V battery might reach end of life at 20 ohm upto 200 ohm; for either alkaline to heavy duty.

So up front, the range of impedance is enormous!     I think 10 ohms would be a good starting point for the DC resistance.    However it wouldn't take much to bump that out to say 33 ohm if you kept using a flat battery.

The AC impedance can be crudely approximated as follows:



- Choose what DC resistance you want to emulate, say 10 ohm
- ignore Ld
- ignore the LTspice source resistance, or set it to zero
- Add series resistor R0, value about 0.1 to 0.2 times the DC resistance
- Set Rct to 0.9 to 0.8 times the DC resistance.  (you want R0 + Rct to add to the DC resistance.)
- Set the cap to  Cd = 1 / Rct.   This will be an enormous cap and sets a 1 second time constant
  I just pulled cap value from the "7th planet", might need refining.

Since there is an average DC load you should still see a voltage drop determined by the DC resistance.
So arguably the circuit will show the same droop without the cap however, with amplifiers a large series
resistance can cause the amp to oscillate so the impedance model will help stop that happening to some degree.

OK I used your simple battery impedance model (based around 10 ohms) with what are probably close to the actual values used in the Ebow, and the waveforms are getting closer still.

Something else, imagine the AC output swinging +/-3V peak and the output load is 8 ohm.  That's 3V/8 = 375mA which will cause a 10*0.375 = 3.75V drop across a 10 ohm battery impedance.   So quite a bit of drop on a 9V rail.   If you backed the load off to 16 ohm, 3V/16 = 188mA, drop across 10 ohm = 1.9V.   As the battery impedance goes up drop the amount of swing of the amp will drop and there's a point where you won't get +/- 3V peak.  If the load impedance is increased you will get less drop and more output swing.   So for a *given* battery DC resistance there will be a load which gives the maximum amount of field.   The thing is for a 10 ohm battery impedance the maximum field won't end-up with a 79 ohm coil, it will be a lower impedance.

I see. That makes sense. I remember early on I was seeing quite a drop in the supply voltage when I measured it. Maybe that would be my beacon for knowing when I am on the right track with the driver coil?

I'm wondering if there is a simple way to see what is actually happening at that diode. Sounds like I need a differential probe? Can I DIY one? If I follow what anotherjim was getting at, it seems like the diode might be rectifying AC current and the cap is there to smooth out the ripple? This is a new way of thinking for me... I am used to thinking that ground is always zero so to speak, but I guess that is not the case when it comes to current.

[anotherjim]

On the real circuit, you don't need a differential probe, you just clip the scope probe ground on the battery negative. If the rest of the circuit doesn't have the same ground as the scope, the probe tip on the diode cathode will give you a reading "across" the diode. A differential probe can give a better picture as it tends to cancel common-mode noise, just like a balance audio feed can. A 2-channel scope can be made to work differential with 2 probes by selecting "add & invert". Channel 2 is inverted and added with channel 1 and magically you have a differential probe. A Single channel scope will not have this function.

In the sim, you could maybe attach a different ground symbol to the battery negative and if the software allows it, select this battery ground for the probe ground. I don't know how LTspice works this, but I've a feeling it may be stuck using the circuit ground.
However, I don't see why you can't remove the extra ground points in the sim and just draw wires making all the ground connections going via the switch to a single ground symbol on the battery negative. Then it's more like the real circuit. Any signal source you feed in to replicate the string pickup must ground via a wire to the circuit ground, not a ground symbol. Now the sim probe will have to reference the battery negative as ground so a probe on the diode cathode will be across the diode.

[amptramp]

There is an article called "Vibrating Wire Audio Filter and Oscillator" in the May 1968 issue of Radio-Electronics:

https://worldradiohistory.com/Archive-Radio-Electronics/60s/1968/Radio-Electronics-1968-05.pdf

starting on page 52.  They didn't use any IC's, just 2N2953 and 2N2925 transistors but they had current running along the wire (guitar string) and a U-shaped magnet over the wire.  This may be easier than using a second pickup as a driver.  The vibration is in the direction of towards and away from the guitar body in the example they give, so you may need to rotate the magnet to get a better output.

[Rob Strand]

OK I used your simple battery impedance model (based around 10 ohms) with what are probably close to the actual values used in the Ebow, and the waveforms are getting closer still.

Looks convincing to me.

I see. That makes sense. I remember early on I was seeing quite a drop in the supply voltage when I measured it. Maybe that would be my beacon for knowing when I am on the right track with the driver coil?

I'm not quite sure but early on I think you were using a Heavy duty battery?    Later you switched to the Alkaline, and I think you had the Alkaline battery when you did the experiment of adding more turns - as per the video with the "version 2" output coil.

In your experiment add more turns of the *same* wire should give some more output since it reduces the battery loading.  However, looking at your results you are getting a stronger effect than I would expect.  So that makes me think there's more to the sag story than the basic battery impedance.   Here were are talking electrical effects.

Another aspect of your experiment is the magnetic effect.   When you add more turns is that helping the field drive the string.    A larger coil tends to "project" the magnetic field further out.  So that could show up as better string drive.  By the same token, the effect of the small diameter iron core coil overshadow the coil shape aspect.   However, both you 8 ohm coil and the later 96 ohm coil had more or less the same final coil OD so the projection theory doesn't hold up.

I don't know what's going on.    My gut feeling is something is sagging more than expected, maybe the LM386 output.

As far as playing with the turns on the coil and the battery DC resistance, the following set-up looks OK to me.   It does not factor in the possibility of the larger diameter coil having an effect.



The way it is set-up is the Bout output tries to model the strength of the field from the coil.     You can adjust the number of turns using the variable "N".   At the moment the simulation steps through three different number of turns.  The inductance and resistance are calculated from the turns.    There two cases for the resistance.  If you use "Rx" for the value of the output resistor the resistance is based on on filling-up the  bobbin with the given number of turns.  This would be considered an optimal coil.   The second  case is "Rx2" which is the case where you add turns of the same wire diameter - I've put that in there so so can see it produces a different result to "Rx".     You can also see that adding turns in your experiment should increase the output, but the opposite is true for Rx!

I've also got the battery impedance in there to model the sag.   It's set to quite a high value.

So here's the result when you change the turns but you also change the wire diameter keep the bobbin full.
[N increasing green -> blue -> red]


And here's your experiment of adding more turns of the same wire.



The traces overlap so here's a zoomed version.


Conclusions:
- if the bobbin is kept full it seems even with a high battery impedance the output increases with a lower impedance coil (less turns).
   (I'll check this again to see if we can increase the battery impedance further to see if there is a minimum load.)
- if the same wire is added to we get a small increase in output with more turns.  The increase is somewhat smaller than the results in your experiment - that's a mystery to me!

There's a few other options on the schematic, for example the output cap can be automatically chosen to resonate with the output coil inductance at a specified frequency.   ATM, the cap is fixed at 220uF.

'm wondering if there is a simple way to see what is actually happening at that diode. Sounds like I need a differential probe? Can I DIY one? If I follow what anotherjim was getting at, it seems like the diode might be rectifying AC current and the cap is there to smooth out the ripple? This is a new way of thinking for me... I am used to thinking that ground is always zero so to speak, but I guess that is not the case when it comes to current.

Because the unit isn't grounded you can just wire the Oscilloscope across the diode.    For LT spice you can plot the voltage across the diode directly.

As an aside, this old post would be something which would stress the output of the LM386 and show any output sag/drop under load.   By running the circuit through LTspice we can see if the LM386 model works correctly when the LM386 output is stressed.
https://www.diystompboxes.com/smfforum/index.php?topic=124739.0

Here's the set-up for the simulation.   The input source is a stiff power supply, not a battery.



Unfortunately, the original article put a spanner in the works because they talk about adding 1 ohm resistors in series with the IC outputs to measure current but they don't make it clear if those resistors were in place when they produced their results tables.

I can match their results if I place a 1.3 ohm resistor in series with the IC output.

The question is, did their results have the resistor or not?  ahhh!

If not, then I would need to add 1 ohm emitter resistor to the output stage of the LM386 model.   This is something I looked into earlier but I found no information if the LM386 has emitter resistors or not.   I did find the LM380 schematics which show 0.5 ohm emitter resistors.    Just to be clear, we can't just add the 1 ohm emitter resistors to my model the AREA parameter of the biasing diodes would probably need to be tweaked.  For now, as a hack, we can just add 1 ohm in series with the output of the IC.   Keep in mind at this point we don't know if we should add it or not.

[Paul Marossy]

It was a real pain in the rear but I managed to get some scope traces across that diode. Not sure they can be trusted 100% as it's not really a proper fully functional unit but it might give us some idea of what it is doing. 

Regular mode, scope set to AC:


Harmonic mode, scope set to AC:


Regular mode, scope set to DC:


Harmonic mode, scope set to DC:


Not exactly sure what it's doing here, but it appears that in regular mode it would be suppressing harmonics?


I also made that snooping coil and did what I could with it. Couldn't use it as originally intended but I'm sure one of you brainiacs can find something useful in it. The main challenge was creating the Ebow Levitation Device so I could use the coil. That was a fun exercise making that appendage to attach to the testing jig. I didn't look at the guitar pickup output as I already have an idea of the output voltages I would be seeing there. Should I also look at that too? I still have the jig set up, so I could also do that if need be.

Specs on the coil are as follows: 20 turns of 0.40mm wire, coil O.D. 23mm, 0.77mH, 100 ohm series resistor on + connection of BNC cable, total DC resistance 104 ohms, wound CCW with starting point being + on the BNC cable. Coil was at bottom of Ebow, so about 2mm or so from the bottom of the coils (actually a little more than that if you count about 0.75mm-1mm for the thickness of the plastic housing). String was about 5mm above the pickup, which puts the snooping coil about 2.5-3mm above the pickup. Input/sensing coil always measured 2mV.

[Rob Strand]

Not exactly sure what it's doing here, but it appears that in regular mode it would be suppressing harmonics?

It looks quite different to what I see on the simulation.

Regular mode has 20mV swing (due to the diode),



Harmonic mode I'm seeing about about 6Vp-p.



The above sims were with zero battery resistance but oddly enough adding a significant battery resistance didn't change things much.   At 400Hz your "Version 2" output coil has an impedance of about 100 ohms so it will take quite a large battery resistance to produce a significant drop.

The 6mV swing on your normal mode waveform looks like the voltage could be across the switch itself (?),  perhaps the CRO ground was on the battery -Ve terminal and not the circuit ground?    The general shape of the waveform is OK but that originates from the clipped output of the LM386.

I'm seeing a larger swing in Regular mode.   It's possible to see that if the actual LM386 output is dropping under load more than the simulation.

I haven't tried a zener in place of the diode.

I tried a zener and the waveform shape is wrong.

but ...

If the 4.7uF cap was 10uF it would fixed the voltage swing across the diode. ... and then I realized your harmonic mode is at 770Hz, which effectively does the same thing as making the cap larger.   However, it's doesn't quite work out like that, since the voltage is still high,

[anotherjim]

I'm puzzled. Should there be anything meaningful across the diode/cap in normal mode? I thought they were out of circuit then.

[Rob Strand]

I'm puzzled. Should there be anything meaningful across the diode/cap in normal mode? I thought they were out of circuit then.

Yes, it's a bit confusing.

Up to now we have kind of put the diode + cap to the side.

Here's my take on what the Ebow is:



If you go to reply #4 in this thread you can see the PCB.
The connector at the top left goes:  GND, (+) VCC, VBN

At least that's how I think it all goes.

Paul's rebuilt unit has different coil specs.
(Just realized my schem shouldn't have "V2" coils, they are "original" coils).

[anotherjim]

Thanks, that is clearer. So either the whole circuit (except the output coil) returns via the diode or just the output coil does. Paul's waveforms make more sense to me now.
This is really hard to understand because although the 4.7uF bypasses the diode for AC, it is just small enough to have some impedance. The diode//cap arrangement suggests it is meant to represent an asymmetric impedance for AC, but the way it's switched over between modes, frankly makes my brain hurt!

Could it be, that in normal mode, the diode//cap has only a trivial effect and it was just convenient to switch it like that?

[Rob Strand]

This is really hard to understand because although the 4.7uF bypasses the diode for AC, it is just small enough to have some impedance. The diode//cap arrangement suggests it is meant to represent an asymmetric impedance for AC, but the way it's switched over between modes, frankly makes my brain hurt!

Could it be, that in normal mode, the diode//cap has only a trivial effect and it was just convenient to switch it like that?

I don't get it 100% either.

Since a long way back, the way I've been looking at it as:
- Normal mode is just the trivial diode drop.    The thing that bugs me is why have a diode drop at all?  You could easily arrange the switch to bypass the diode!   Maybe a halfway reverse protection?
- Harmonic mode is some sort of high-pass filter with the aim to promote the string oscillation to occur at higher frequencies. (We know the high-Q string doesn't care so much about phase, so amplitude wins for oscillation.)
The way I saw the diode is a small cap alone looses too much signal so the half passing and half high-pass filter
is kind of "in between".

I'll admit upto this point I haven't put a lot of time trying to confirm those thoughts at all.

[Paul Marossy]

Is it possible that the 4.7uF cap is actually bad? Or is that how it's supposed to be with the very squashed sine wave at that diode in regular mode? I still suspect that 4.7uF cap, since it was damaged by the person that de-gooped the circuit board. It seems to test OK but I've had cases where electrolytic capacitors seemed to test OK but they really weren't.  I guess I could put a different 4.7uF cap in and see if I get different result.  Also am wondering if that diode is perhaps not a 1N4148. I can't imagine what else it could be if not.

When I was initially figuring out the schematic, I saw it as regular mode battery return path as being forced thru the cap/diode and in harmonic mode there was a direct battery return. The driver coil also has to go thru the diode/cap to ground, but that is full time. For the battery return it's the path of least resistance. In AC design, the air always takes the path of least resistance. I'm thinking of this in more of a mechanical engineering mindset as that is what I do every day - I design HVAC for enormous 15,000-25,000+ sq ft houses in fancy gated communities in the middle of a very hot arid desert, which are centered around big green lawns where people try to whack a little white ball into a tiny hole from many yards away. No one wants to see any of my HVAC equipment, including the air devices, but I will for sure hear about it if something doesn't work as expected!

[anotherjim]

That's a very clear view of the PCB.
If you do change the cap, how about checking what happens if it's out and there is only the diode?

[Rob Strand]

Is it possible that the 4.7uF cap is actually bad? Or is that how it's supposed to be with the very squashed sine wave at that diode in regular mode? I still suspect that 4.7uF cap, since it was damaged by the person that de-gooped the circuit board. It seems to test OK but I've had cases where electrolytic capacitors seemed to test OK but they really weren't.  I guess I could put a different 4.7uF cap in and see if I get different result.  Also am wondering if that diode is perhaps not a 1N4148. I can't imagine what else it could be if not.

If you do change the cap, how about checking what happens if it's out and there is only the diode?

The cap is definitely doing something.  There has to be some resonant effect, even if the diode makes it non-linear.  Also I'm sure the cap there is doing something because the measured swing across the diode depends on it.   I think the cap is working even though it doesn't match the sim.  It's possible the cap isn't exactly 4.7u, that's no surprise for an electrolytic.

No one wants to see any of my HVAC equipment, including the air devices, but I will for sure hear about it if something doesn't work as expected!

You can say the same for just about anything.   There's plenty of $100,000 loudspeakers in such places.  The owners have no idea of the enormous technical knowledge and 50 years of development history it took to produce it.   Who thinks about why the steering wheel on your car is what it is?  size, shape, profile, material ... someone has to make those design choices!  It's easy to change something but only a few people in the world know what it takes to *improve* the design from one model to the next.

[Paul Marossy]

That's a very clear view of the PCB.
If you do change the cap, how about checking what happens if it's out and there is only the diode?

That's the CAD drawing that I made of it. Been tweaking it over the last month or so. I typically do this for projects I work on/reverse engineer/clone, just for my own benefit, and sometimes for the benefit of others. 

I know that reversing that diode just killed the power in one of the modes, but I don't think I ever checked to see what happened if not in the circuit.

The cap is definitely doing something.  There has to be some resonant effect, even if the diode makes it non-linear.  Also I'm sure the cap there is doing something because the measured swing across the diode depends on it.   I think the cap is working even though it doesn't match the sim.  It's possible the cap isn't exactly 4.7u, that's no surprise for an electrolytic.

Yeah seems like it's doing something but not sure if it's working as intended. I have a feeling once that cap has to do some "hard work" it craps out. Could also be my driver coil is too big and loading down the LM386 output too much?

No one wants to see any of my HVAC equipment, including the air devices, but I will for sure hear about it if something doesn't work as expected!

You can say the same for just about anything.   There's plenty of $100,000 loudspeakers in such places.  The owners have no idea of the enormous technical knowledge and 50 years of development history it took to produce it.   Who thinks about why the steering wheel on your car is what it is?  size, shape, profile, material ... someone has to make those design choices!  It's easy to change something but only a few people in the world know what it takes to *improve* the design from one model to the next.

Yeah, good points. I have worked on some projects where I was told that the AV room equipment alone was $250,000.  I can't even imagine that. I live in a far different world than these people do.