Some sustaining questions

[Mark Hammer]

I was impressed as heck with forum member cortezthekiller's DIY sustainer, when I was over at his place the other day.  Impressed enough that I set aside other projects and decided to make one.  I had some 32-gauge magnet wire (the type recommended), some 386 amp chips, and all the other suggested parts.  I wanted a "stealth" type that could fit between the neck pickup and the end of the fingerboard.  Using some slugs from old Japanese pickups, I made a nice slim profile holder for the wire, attached some small bar magnets to the bottom, attached some wooden spacers to confine the windings to the top end of the structure, wound a coil that measures 8.2ohms, potted it, wired it all up as per the Ruby specs, and fired it up.  Didn't get all that much of use.  A bit of sustain on the unwound G, but that was it...with the exception of some howling oscillation (which I learned is likely why one uses it with the bridge pickup).

So, a couple of questions for those who have had more success:
a) I used a soft iron one-piece slug.  Do the slugs need to be individual, or is a single-piece still workable?  Is the composition of the slug important?
b) I'm using a pair of strongish rubber magnets, glued to the bottom of the slug, with a decent "tug" at the top, where it would be close to the strings.  Is the strength of the magnetic source at the bottom critical?
c) Does the coil need to be only 3mm tall for some reason or is that stipulation only to assure that the coil is far away from the magnet at the other end of the slug?

[EBK]

I know that you are in the process of posting some pics of that thing you described.  I'll just wait patiently without pestering you so you can focus on what you are doing.   

Is this something like an EBow?

[deadastronaut]

page 1 of many coming up....... good to see you doing this mark...in my experiences with diy sustainers

there are many variables that make a sustainer not a ''one size fits all''

string gauge, pickups, driver coil, amp, action, interference from pickup next to coil....

and sausages of course...

i made one of my coil drivers from 6  3mmx2mm round neo magnets, glued to a flat thin plastic base
then wrapped the wire around the the 2mm height till it reached the top..around 8-10 ohms - ish..IIRC.....

i removed my neck pup, as it interfered with the driver too...

lots of sustainers will work, no problem....but getting the top B / E is the most desirable to me but also the most problematic...

looking forward to seeing what your results are....

[Mark Hammer]

Fine, here's a pic.  The bottom shows the slug and the magnets used.  I think the slug is from old Univox humbucker pickups.  The top one has the spacers and coil installed and teflon tape around it for protection.  It is quite possible that the magnets do not have the orientation I believe they do.  I'll have to try a single long ceramic magnet.

And, as I learned yesterday, getting the wound strings to sustain is easier than getting the thinner unwound to do so.  It seems counterintuitive until one considers that the larger diameter of the wound ones presents a bigger and easier target for the magnetic field.

[cortezthekiller]

Hey Mark, great to hear you are giving this one a go.

I found out something that helped when experimenting last night:
Using the ruby amp design, the input stage should be changed to a booster instead of buffer, ala "fetzer ruby" amp.

Also I had used a pickup bobbin with alnico magnets and stuck on a fairly size-able rare-earth magnet on the bottom for even more strength.

This now seems to allow all of the strings to sustain properly up and down the neck.
If the gain of the amp is too high using the trimmers then there will be unwanted distortion and oscillation,
so it is a balancing act as well.

[Mark Hammer]

I had originally built it using the basic Ruby, with the JFET input buffer; mostly because I took a quick look at this site - http://diy-fever.com/misc/diy-sustainer/ - and remembered "Ruby amp" rather than the particulars.  When I took a second look for more clues, I realized there needed to be a booster up front, so I fixed that.  Still no dice.

I'm going to play with the magnet/s a bit.  I'm not confident the polarity of what I'm using is oriented the way I believe it to be.

I should also note that I'm using an LM-386-1.  Have others had success with that particular 386 or do I need to step up to a higher output one?  I did not implement the adjustable gain, opting instead for a simple 10uf cap bridging pins 1 and 8.  I figured that fixed gain could be easily offset with the 10k level trimmer.

[PRR]

> A bit of sustain

You need gain. The '386 gives some. The *magnet strength* is also a form of gain, though a car-lifter magnet might damp more than gain-up. I'm not shocked you may need a booster in front of the '386.

And yeah: big signal on a winding near pickup winding will just howl-round before the string gets excited.

Keep playing around.

[anotherjim]

I've no experience, but if it comes to getting more raw power in the exciter coil, then simply increasing voltage gain may not be enough. It won't do more magnetism if the amp is clipping. You could try two of those 386's in bridge mode which is getting on for x4 power.

[pinkjimiphoton]

i look forward to seeing ya work this out too!

that said, have ya tried a digitech freqout sustainer/feedback generator yet?

unfathomably, they finally got something RIGHT. like a polyphonic ebow with multiple harmonic nodes and just the right amount of drive. kinda pricey, but sometimes its worth it.
imho, this one is worth checking out.

now back to the op, sorry for the hijack

[deadastronaut]

^ yeah that is really nice....cheers jimi.

gut shots of it..
https://www.youtube.com/watch?v=EQaioPsH20U

[Mark Hammer]

i look forward to seeing ya work this out too!

that said, have ya tried a digitech freqout sustainer/feedback generator yet?

unfathomably, they finally got something RIGHT. like a polyphonic ebow with multiple harmonic nodes and just the right amount of drive. kinda pricey, but sometimes its worth it.
imho, this one is worth checking out.

No luck yet.  I have tried out the Freqout, though, and it's a terrific little pedal.

[R.G.]

Some thoughts on sustainers -

Strings are not permanent magnets, so there is no way to make magnets repel them (well, other than running a substantial current through them near a permanent magnet!) so there has to be a permanent magnet somewhere in the structure to pull the string a little off center toward the magnet. Coils can then oppose the PM field and release the PM's field-hold on the string a little to get things started, and then aid the PM when the string oscillates back toward center.

I've wondered if it's possible with enough field drive from a sustainer to not use a permanent magnet at all. There would be no output other than thermal noise with the string at rest, so nothing much would happen until the string was excited in some way, and then with enough magnetic drive from the sustainer, it ought to be self-sustaining.

The string is a lossy mechanical resonance. Pulse it - like with a pick - and it then resonates, losing the energy imparted by the original impulse to moving air molecules, moving the bridge, fret, etc., and in an electric, moving the M-field of the pickups to create the voltage and currents in the pickup windings. All this together means that you could model a guitar string as a spring-mass resonance, with energy leaking out by a mostly-fixed damping resistance. The damping resistance varies with the construction of the guitar that holds the string, but for a given string, guitar, and fretting style, it's likely to be closely bounded. Just thought - there is some energy fed back in if you amplify the string motion, feed it to a speaker, and have the speaker drive the string either directly (in an electric) or through the guitar body, bridge, and fingerboard.

A sustainer works by using an amplifier to feed back in at least as much energy as is lost to the damping. The feedback for this path is string->pickup->amplifier->sustainer driver amp->sustainer coil ->magnetic field->string.

All that's required for continuous oscillation is that the energy lost from the string are balanced out by an equal amount fed in from the sustainer driver. With a "loop gain" of exactly one, the string will oscillate as long as nothing changes. This is an exact analogy to an L-C oscillator. If the gain is less than one, the oscillation will die away. If the gain is more than one, the oscillation will grow in size until something interrupts the loop - like banging against a power supply or fretboard, depending on whether you're using electrical or mechanical analogies. How fast the oscillation changes, either dying away or increasing, depends on the ratio of energy gained from feedback to the energy lost from the resonance.

So the gain of the loop from pickup to sustainer amplifier to sustainer coil and back to the string, and its phase determines how hard an oscillation is to start, and how quickly it will "bloom" out to full volume or decay. I think you can make a string decay faster, but it's trickier because a guitar string will happily resonate at frequencies other than the main one - harmonics. The phase of the sustainer feedback loop, its exact position along the string length, and the tension and materials of the string will select for other harmonics, so simply inverting phase doesn't necessarily cause a string damping.

In any case, higher gain in the drive loop will make things happen faster.

The drive coil interacts with the permanent magnet field to alternately aid and oppose it. The M-field is a function of the area of the driver coil, the number of turns, and the current in the coils. With the field being a function of the current, you can probably do better driving as few turns as possible with as large a current as practical. That kind of implies using the driver amp in constant current mode, or faking constant current mode with a frequency-slanted voltage drive. The LM386 ought to be great as a coil driver.

There are questions in my mind about whether a single coil across all the strings, a single coil per string independently, or a coil per string in series or parallel would be the best format for the coil drivers.

[anotherjim]

I've given it a little thought too. For DIY, the coil is the tricky part, but consider a relay. Coil is energised and pulls the contact armature towards a steel slug pole piece. It isn't too hard to get the assembly out of a miniature pcb relay. The coil DC resistance is usually given in the specs and of course we get an operating voltage although don't assume that voltage is what it'll take to wiggle a string. Generally, higher operating voltage means higher coil DC resistance. The coils are small enough to be used one per string.
Mostly the relays are intended for DC coil operation. If you supply AC, it will have a doubled frequency effect on string pull since there is no permanent magnet to bias it. That is, both positive and negative going polarities will produce an attraction. There is no repulsion. The best excitation waveform may be a narrow pulse train to allow the string time to naturally relax away from the attraction.
While I was typing that, a  stepper motor drive flashed in my mind...hmmmm... It maybe doesn't need to be driven by a linear amplifier?

Don't forget if DC coupling inductive loads, the coil flyback voltage can fry your parts.

With even numbered stringed instruments, it may help to alternate the polarity of the coils to cancel the overall radiated field.

As a side note, if you stick a permanent magnet on the relay coil pole, you have a magnetic pickup (don't even need to take it out of the case to try that).

[PRR]

> there has to be a permanent magnet somewhere

No. There has to be a polarizing source. It does not have to be Alnico.

Just run DC in the coil. Idle at say 100mA. Swing tickle-signal down to 10mA and up to 190mA. This could actually mean "simply" leaving the output cap off a LM386 type power amp.

That's quite "wasteful" in this day of good permanent magnets, an ongoing power drain which can be avoided. Battery suckage. But DC bias was used in several classes of speakers. Mostly on a second winding (the huge field-coil speakers) but occasionally in the signal winding (some early telephone receivers).

Perhaps not a good field for seeding, unless an incredibly synergistic combination comes in dreams.

Most of the rest of your theoretical teachings are spot-on standards.

The number of turns has no theoretical effect. Purely what is convenient. You find the Power required (a very tough question). You see what voltage is convenient. That tells you an Impedance. Some dinking in a wire table says a wide range of resistances are equally practical. You work out a trial winding on a core with a string and see what inductance and capacitance do against you. A 100K impedance winding will drown in stray capacitance within the audio band. A 1 Ohm winding is a poor fit for even short cables.

[R.G.]

Dang. I wrote up another long, finger-blunting post and it didn't "stick" here.

Yeah, it's possible to use the drive coil as its own "bias" but that's wasteful. As I mentioned, I actually suspect that a drive coil without a static magnetic field, PM or DC, could work, but would not self start. That might actually be an advantage in some situations. You'd have to pick (or touch) the string to start it. After that, the drive could grab it as it comes closer. Probably a different harmonic response.

Yes, theoretically the number of turns is immaterial. As a practical matter, what you're driving is a magnetic field, and the field will be proportional to the number of turns and current in them, as ampere-turns is one measure of field intensity. Single turns need X current for the necessary field intensity. Ten turns need X/10 amperes. 100 turns gets the same field intensity with X/100. So you need to be able to change the field intensity up to max at least as fast as the string vibrates.

The high E string vibrates at 329.62...Hz in standard tuning, so the 12th fret is 659.255...Hz and a theoretical 24th fret is 1318.51... You need to be able to ramp the field up to max in 1/2 of the period of the wave, arguably about 750uS.

For an inductor, V = L* di/dt and L = k * n² . So the magnitude of the drive voltage can be written for some given field strentgh of Idr (the drive current)_and frequency. Field strength increases linearly with N and Idr, but the voltage needed to drive the coil increases as the square of the number of turns N.

There is a balancing act going on here, where you have a relatively fixed drive voltage, arguably 9V or 18V for bridged drive, where you can ramp the number of turns up until the inductance gets big enough so you can't get to max drive within the cycle time. So for each power supply voltage, there's going to be a limit to the coil inductance you can drive, and that translates into a limit on the coil turns and coil area.

The lower limit is about one (although I suspect a half turn would work, just a conductor parallel to the string) but the voltage across the coil at such low inductances and resistances is nearly zero. That means you're burning up power in the amplifier to control the current in the coil for a linear amplifier.  The more turns, the lower the current and the higher the voltage across the coil at high frequencies at least. On top of this, there is the effect of adding wire resistance. That goes up as the square of the number of turns too, for the same winding area, but it has the side effect of burning up some of the drive energy and moving the M-field phase with relation to the driving voltage to the coil driver.

If I loved writing equations more than I do, I might sit down and try to solve this mess for turns and voltage.

However: we have both class D amplifiers driven from 9V now, and we have the string as a very highly selective tuned circuit. The class D amplifier use is obvious. But we might get away with simply driving the coils with a PWM signal, as A.J. mentioned. In fact, we >might< get by with driving the string with one pulse per half cycle and letting the string filter out the resulting harmonics. After all,  a pick is an impulse applied to the string.

[PRR]

> work, but would not self start

Without bias, both N and S magnetic field pull the string. The force is at twice the intended frequency.

With typical spacings, at small amplitude, the two peaks cancel out. As you say.

When amplitude is significant, say the N swing is closer and dominates the S swing. Now amplitude could build. But the force-factor rises with amplitude. It will try to vibrate to infinity (and of course something clips first).

All not a shock: 2nd Harmonic Distortion is nil at small signal and rises with signal.

Don't wear-out your abacus on the impedance thing. We are simply subdividing say 1 pound copper. Everything multiples-out right. "Too many" turns implies both a high voltage (inconvenient here) and also paying more to the copper-beater. A single turn is hard to bend, may need fat bars to a bizarre amplifier, and fully load-up on a 1V battery which is awkward. We really want a concept of "power" before thinking about impedance. Ah, yes, we need an impedance for the experiments to determine power needs. Bell Labs' work-around in 1927 was to build a 100 Watt amp, fiddle the speaker, then go back with a 1W amp.

We can NOT get good coupling from driver to string; this would damp the string's natural decay. Several thought experiments suggest coupling must be 1000:1 at most. The standard pickup structure should be as good as any, and is readily available. The winding must accept 1,000,000 times pickup power. Say pickup delivers 0.1uW, this is 0.1W into the driver. At 3V signal (9V batt) we want around 30 Ohms. A few hundred turns filling the bobbin. 

[amptramp]

Check out the title article "Vibrating-Wire Audio Filter and Oscillator" in the May 1968 issue of Radio-Electronics:

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

It begins on page 52.  They use a horseshoe magnet surrounding a wire.  The oscillator is interesting.  The output is taken from the ends of the string so either a nut that insulates each output from the other is necessary (and is normal) and the frets would have to be non-conducting (not so normal) and the bridge would be conductive and at ground potential or the frets would be all grounded together and the bridge would have to be insulated.  There would be six individual inputs and there should be six drive coils so you can sustain a complete chord.

It seems you need a bias magnet but a drive coil that was offset from the centreline of the string or halfwave drive would overcome the tendency to drive the string at twice the frequency.  But if you notice, the oscillator uses no drive coil - the string and magnet do it all.  It would require some modifications to a guitar but would require no coils.

[PRR]

> The oscillator is interesting.

Indeed. It (like most oscillators) raises the Q of the tuned circuit over infinity. So it "sustains" even before you pluck. This is probably the simplest example of what we want to do. Question is can the Q be adjusted just-short of infinity, so it does not sing before its cue, yet sings a long time after pluck.

[amptramp]

> The oscillator is interesting.

Indeed. It (like most oscillators) raises the Q of the tuned circuit over infinity. So it "sustains" even before you pluck. This is probably the simplest example of what we want to do. Question is can the Q be adjusted just-short of infinity, so it does not sing before its cue, yet sings a long time after pluck.

Most oscillators are not set to a given amplitude by adjusting the gain to provide no growth or decrease of signal strength, they usually use some non-linear element to reduce the gain above some amplitude so they are set for growth until they flatline.  A sustain would do something similar - it would allow the signal to decline to some value then hold the level constant.  Maybe the non-linear lamp adjuster of the original Wein Bridge oscillator from the first Hewlett Packard audio oscillator should be added to this oscillator.

[PRR]

The classic amplitude control, when you do not listen to the raw output, is to just let it clip. Set gain to 1.1, when it clips 10% the average gain falls to 0.99 and amplitude falls-off, it finds its own gain-1.0000 point.

In this case, if we wanted a true oscillator, we'd just let the '386 clip. Because the audio signal is taken from the resonator, not the exciter, we won't hear all that clipping. (And in guitar, we might not object if we did.)

But we don't want the string to sing before we pluck. So it is not the classic oscillator (where sure-start is demanded).

We could try to trim gain to 0.99999, which is obviously difficult.

I'm thinking the amplitude control must also suppress low-level gain. Essentially "crossover distortion". If amplitude is low, gain is 0.9+. Above small levels gain rises to 1.1. For too-large level clipping limits the level.

Still not musically perfect. Do we want "all" sustain to tend to the same level? (In some musical styles all notes are max-loud; other styles have dynamics.) How do we tell the dumb ringer we want a small ring here and a big ring there?