curtisa

Uhh...Isn't this what I said anyway? Limiting = compression except for a much higher ratio. Whatever! Compression. Limiting. Doesn't matter - we're just attempting to average out the driver signal Well, the LM13700 isn't exactly reknowned for it's audio performance. The companies that produce commercial compressors and limiters use much more esoteric circuits based around higher cost dedicated VCA chips. I did actually try breadboarding up the circuit in the LM13700 app note last night, but it refused to work - all I was getting was really nasty distortion as soon as the input exceeded a certain threshold which seemed to have nothing to do with the "output amplitude" pot. I don't have a second LM13700 to swap out to see if the first one was faulty though. I want to try a variation on the compressor idea using a completely different IC, a SSM2018, which is a professional dedicated VCA chip. I'm expecting much more predictable results using it, but the unfortunate thing about this IC is that it's particularly hard to get and quite expensive, at least here in Australia. I think you will need some form of preamp gain because the AGC will expect to see higher input levels than what a guitar can typically deliver. Shouldn't need too much though, say 10-12dB extra. I'm finding the really frustrating thing about my setup is that I'm 99.99999% sure that adding compression to the driver signal should even out a lot of inconsistencies, but each time I go to test the theory I always get better results using the straight LM386 circuit. D'oh! Pete: I just noticed your request for audio samples in one of your earlier posts further up - I'll try to record some samples of my setup this weekend coming so that you can compare my driver performance to yours. Cheers, Curtis. current sounds right, voltage isn't what I expected, maybe I told you to measure the wrong thing... maybe curtis can help us here ? Col Are you changing the meter probe positions when you change from AC current to AC voltage? Current measurements need to be taken with the meter in series with the circuit you're measuring. Voltage measurements need to be done with the meter across the circuit. Cheers, Curtis.

Not quite. Limiting and compression both use the same "threshold" system. The only difference is that once the input exceeds the threshold in a compressor, the output continues increasing but at a slower rate. You can have a compressor set up with a 2:1 ratio, which means for every 2dB that the input exceeds the threshold, the output only increases 1 dB. Limiting is set such that any increase in input level over the threshold results in no increase in output, which in compression ratio terms is infinity:1. You can add gain to the input of the compressor/limiter so that the quieter audio is louder and the louder audio is compressed - all you're doing is raising the average level of the audio by reducing the "distance" between the loud and quiet bits. The impression I'm getting from the app note is that the threshold is defined by the 100K variable resistor (labelled "output amplitude"). It sets the absolute "ceiling" of the audio, above which nothing shall pass. I haven't built and tested this yet so I don't know exactly how it would behave in real life. It's possible I guess. Otherwise how would connecting up more than two stompboxes together work? They would each have completely unrelated earths to each other. My query stemmed from the fact that we're sharing the same common power supply Anyway, if it works, it works Edit: looking at your schematic, I wonder if you'd get better performance by changing the order of the buffer and preamp to: Input from guitar -> buffer (buffered output here) -> preamp -> AGC etc... ...instead of the buffer and preamp in parallel, the way it's currently drawn? And would the shared earth work properly if it was? No, what you've got there is fine - you only need to couple once between stages. Absolutely! From the National Semiconductor Audio Handbook 1977: "...Shown or not, bypass capacitors are always required. Ceramic disc capacitors (0.1uF) or solid tantalum (1uF) with short leads, and located close (within one inch) to the integrated circuit are usually necessary to prevent interstage coupling through the power supply internal impedance. Inadequate bypassing will manifest itself by a low frequency oscillation called "motorboating" or by high frequency instabilities. Occasionally multiple bypassing is required where a 10uF (or larger) capacitor is used to absorb low frequency variations and a smaller 0.1uF disc is paralleled across it to prevent any high frequency feedback through the power supply lines..." Whether it actually makes much of a difference in our application though..? Doubt it. IME I've never had to do anything like that with vero boards. If we were playing with RF circuitry it'd be a different matter, but in our application I reckon it's more work than it's worth. As long as our finished layouts are neat and tidy (be it vero, perf or full PCB) I reckon we're fairly safe. Cheers, Curtis.

The app note describes it: "...As VO reaches a high enough amplitude (3VBE) to turn on the Darlington transistors and the linearizing diodes, the increase in ID reduces the amplifier gain so as to hold VO at that level..." That to me says limiting - reducing gain to keep the output voltage at a constant level. Still, if it's working for you... Hmmmm...technically the LF358 and the LM13700 need the same earth. It should've worked. It's possible that all three stages earthing through the same voltage divider was too much for it and was pulling the earth "off centre". You could try making the voltage divider using smaller reisitors, say 2x 1k to allow for a slightly better current margin. The LM386 will be quite happy referencing back to the 9V negative terminal, just remember to capacitively couple the input to the previous stage. Gotta dash, more later. Curtis

Yep, I see now, thanks - stage 1 = brick wall limiter (thou shall not pass!), stage 2 = further gain reduction based on gain reduction of first stage. ...err...hang on a tic, aren't you just limiting the limited signal? That's not so much dynamic range inversion, just limiting two times over. Isn't it just as easy to run this with one stage of the LM13700, with extra gain at the input buffer (LF358), and get the same effect?. I think I need to sit down with this one and play with it. Have you tried connecting a CRO to the output to see exactly how your circuit behaves? Ahh, I see. But surely if you insert buffers and whatnot into the current feedback loop to change the response, all you need to do is convert from voltage back to current again by inserting a resistor at the output of the feedback loop buffer? I did notice that on your circuit. You actually have two different derived "earths" in your circuit - the 2x 33k voltage divider for the LF358 on the far left, and the 2x 3.9k voltage divider for the LM13700 in the middle. I would think that you'd get better performance by having one "ground" point for everything, not two different ones. IME multiple un-referenced grounds is certainly asking for trouble in almost any sort of audio application in terms of noise performance as it can create earth loops. My thinking is that you want the whole lot referenced back to the guitar's earth, which will then be earthed back at the guitar amp, but you want to keep the sustainer/circuitry separate as much as possible and earth it back at the furthest point in the chain, which in a guitar would be the output jack. With any audio gear that does digital and analog mixed signals (think any digital fx processor) the grounds of the digital and analog side of things are run completely separately, but are commoned at one place only - usually back at the power supply. This is to prevent/minimise any digital signal muck from being carried through the analog earths and into the audio path. I reckon the sustainer might have to be treated in this way as well - certainly sounds like the Sustainiac does by your comment above. The sustainer and it's circuitry are "dirty signals" in comparison with the natural sound of the guitar. Cheers, Curtis

Col, I'm just trying to get my head around your work-in-progress AGC circuit that's been posted a few times. I've found the entry in the LM13700 applications note that describes the AGC element you've used in your circuit. The way it describes the action of the element is to act purely as a limiter, ie output voltage is kept at a fixed level regardless of any increase in input voltage. A few queries/ideas: What is the function of your second stage of LM13700 in your circuit (on the far right of the diagram), and what is the purpose of feeding it the same control voltage that the first stage uses? The 10 ohm/47nF combo on the output of pin 8 forms a low-pass filter with a cut-off frequency of circa 30kHz. The one in the app note uses 50 ohms/20uF giving 160Hz. Is it possible that by applying different combinations of filtering (or even inserting an EQ in the control voltage path in the feedback loop between pin 8 and pin 2) might affect the operation of the AGC element so that it's more responsive to input signals based on frequency content? For example, by adding a high frequency boost to the control voltage the AGC will respond by clamping down the input voltage harder if it's a high frequency sound (or has high frequency content). That could have implications for evening-up the string-to-string response by changing how the AGC responds to pitch.

No, I don't think you could "use" this energy for anything much. The transformer action of the driver/pickup combo might be stepping up the voltage through the pickup windings, but it would have an extremely low current capability. At best you'd get (winding ratio) x (max LM386 output) volts, with a current delivery maximum of (volts) / (winding resistance of pickup). Probably somewhere in the vicinity of a few volts at a couple of milliamps, depending on how you've got it loaded up. Because you've shorted the secondary of the transformer. With the secondary shorted out there is zero resistance across the terminals. Ohms law will tell you that V/R=I, and in this case secondary volts / 0 ohms will equal lots of current. The transformer will try to put out this current, but won't be able to do it (due to winding resistance and other losses) and will just heat up. The short on the secondary is reflected back to the primary by the transforming action of the device and you have excess current in the primary, heating that up aswell. With the combo pickup/driver you're probably not going to damage anything, but I would think that you'll get pretty poor driver response and a heavily distorted output from the LM386 with a shorted "secondary". That could be one of the reasons why the Ebow uses such a simple circuit compared to the Sustainer - with the Ebow you have the option of moving the device up and down the length of each string to find the best place to get things going. With the Sustainer you're stuck with an immovable driver in a fixed location. Each string probably has different "sweet spots", and the resultant phase differences from string-to-string at a given driver location may dictate the use of complex phase correction networks to keep the whole thing running? I am not yet convinced...is possible as a guess. I am still not sure that the noise is a factor was ever a concern in these phase correction circuits...they only suggest string response concerns. Yes, I'm thinking that too - I can get excellent results by moving my driver up and down each string to find the "sweet spot", and I get passable results with the driver in a fixed position. By moving the driver up and down the string I am effectively tweaking the phase relationship of string-to-driver in subtle ways which makes a big difference to the overall effect. Which is where Col's circuit is useful aswell. Dynamic range limiting/reducing can help, both in string response and power consumption of the driver circuit, but I reckon there's more to it than just that. I'm sure the Fernandes Sustainer is complicated because it has to be to work reliably, which is a bugger for us DIY-ers! This is the most salient point...even the most basic and crude experiments that achieve the "effect" of driving the string will provide this thrill of seeing a driver in action. I will never forget the reaction of my young daughter when I ushered her in to see this happen just after I got such a thrill. She help this little 1cm disc above a string and the thing just sang. Everyday for a week she asked or sneaked in to witness it again... Hehe...yes, I agree. I could've gone out an bought a brand new Ebow for $250, but I've built one that works for the cost of one IC. Now THAT'S cool. Experiments shall continue! Curtis.

One more thing before I sign off for tonight - when I was doing my tests I was leaving the pickup leads for the combo driver/pickup disconnected. When I shorted the two pickup leads together, I could change the pitch of the EMI feedback. It occurs to me that with a combined driver/pickup sharing a common core, what we have here is a transformer, with a primary of around 100 turns (the driver) driven by the LM386, and a secondary (the pickup) with a few thousand turns generating voltage determined by the ratio of primary to secondary turns. Shorting the pickup leads together is obviously shorting out the "secondary" of the transformer and resulting in excess, and wasted energy in the primary and secondary windings. Leaving the "secondary" open gives us a pair of leads radiating all sorts of who-knows-what into our sensitive guitar electronics. Catch 22? Over and out Curtis.

I think what's more likely is there's just too much gain, and the resultant extreme emphasis of the upper harmonics because of the introduced distortion just makes the whole system unstable. I tried the same thing with the TL071 by making it a high-ish gain non-inverting amp and got similar results. Compressors tried were an 1176 clone and the bog standard one in my Digitech RP6. EQ used was again the one in the RP6. Noise wasn't the problem with the compressors, just no real improvement in the sustaining action. And the EQ's only added extra "peakiness" to the sustainer - depending on the EQ settings it just made certain notes sustain harmonically instead of fundamentally, kinda like a cross between harmonic and normal modes. Obviously boosting certain frequencies pre-driver "encourages" the strings to sustain at the harmonic that is accentuated by the EQ boost. I also tried incorporating distortion (again the RP6) but ended up with an extreme version of the sustainer in harmonic mode, no matter what note I played. I'm finding that once the note has started, provided the sustainer can excite the string it'll happilly sit there for ever and a day singing away. Adding extra processing in the signal chain seems to change the way the strings want to sustain, but not always in a good way. My thinking is that perhaps the positive feedback is some sort of AGC control - the smallest-level input to the LM386 will immediately force it to start swinging it's output as hard as it can. Any further increase in input level results in no appreciable increase in output level, as the output cannot swing any further. I can see building a foolproof circuit will in some ways be just as tricky as the driver itself, mainly because the biggest variable in the whole system is the most important one - the driver. We all currently experience different levels of success with our drivers, even with the same circuits. The one big difference is that we cannot control exactly how we're constructing our drivers, and we're all ending up with something slightly different that works better with some circuits than others. Tricky Cheers, Curtis.

Back from the weekend! Had a shot at trying to connect the driver to the LM386 circuit with a transformer with a 1:1 ratio. Results - needs more testing. It seemed to work, but I got carried away with other experiments and didn't get a chance to see if isolating the driver from the circuitry made any real difference to noise levels. Also had a play with different arrangements for the buffer feeding the LM386 circuit. The Fetzer/Ruby arrangement surprisingly made things worse - EMI feedback was far too extreme to be usable. There's obviously way too much gain in the Fetzer for me to use. I found the best performance when I used a single FET-input op amp, a TL071, as a non-inverting buffer. The buffer added no extra gain to the LM386 circuit. Finally got a chance to hear the sustainer amplified too, and also got the opportunity to hear the EMI fuzz everyone's mentioned. I found that I had quite a lot less fuzz when the buffer at the front end of the LM386 circuit had no extra gain. Also tried adding compressors and EQ's to the signal path for the driver - still had better performance with all of it left out and running direct: op amp buffer -> LM386 -> driver. I'm curious to see what happens if we add positive feedback to the LM386, as shown on the Ebow drawing much earlier in this thread. Has anyone tried this to see what effect this has on sustainer performance? Cheers, Curtis.

Errr...good point...dunno! Like I said, my idea could've been a red herring! In that case, the only real advantage is that the wiring from amp to driver would no longer be a source of EMI, which has caused havoc for some of us. But you're right, the costs of employing a shielded audio transformer and specially-wound driver will probably outweigh the benefits. Can't deny the potential to step-up current into the driver by using a transformer though. The weekend is approaching - I shall experiment and get back to you all. Cheers, Curtis

Here's a quick sketch of the balanced blade driver idea that I was talking about: You're looking at the driver side-on. The two dots at either end of the transformer indicate the polarity of the transformer secondaries. The two magnets are the cores, and the wire is wound around one clockwise, and the other is wound anticlockwise. The magnets/blades are opposite polarity. The induced magnetic field at the point where the blades are exposed to the strings is enough to move them, but as soon as the distance from the blades increases the two mag fields start mixing together and cancel each other out as they're equal and opposite. I've got absolutely no idea whatsoever if it'd work, it could just be another crazy idea bound for failure, but I'm hoping this'll help explain what's circling through my head at the moment. Cheers, Curtis.

The really good audio transformers come with intgrated shields to prevent, or at least minimise this sort of thing. Hehe, yeah like I said, it works in my head... I agree, and was the main driver (pardon the pun) for me to pick your idea up and run with it. I like the idea of sustainers and their ilk, dislike the idea that I have to sacrifice a pickup for it (particularly one like the neck pickup), and prefer not to be tied to a special handheld doohickey to generate these effects, particularly if it's only monophonic (E-bow). Cheers, Curtis.

Yeah, it's quite possible that my potting isn't great. It's more than likely the windings are rattling and carrying on and being generally antisocial. However, the fact that the thing works first time, and works quite well if positioned away from the pickups is encouraging! Cheers, Curtis

In an oven at 50 deg C won't hurt it. IME I have yet to melt the enamel insulation on copper wire with anything much approaching 100 deg C, and when you think about it some of our commercially available amps power transformers get pretty hot under normal operating conditions without showing signs of internal flashovers from shorted turns. It's the melting temp of the tape that will matter more, hence my low baking temparate and short time period. Haha! The trick is to do it while she isn't at home Seriously though, once it's wrapped in tape there's nothing oozing out of it anyway...unless you go seriously overboard with the potting! And I certainly wouldn't do it with any solvent-based potting stuff like epoxy. True enough, I'm pretty sure I didn't over-stress the wire as I was winding, just keeping the slack off the spool to make it easier to control. Besides, with such an elongated winding window it'll be nigh impossible to keep it too taut as there's so much slack in the longer part of the bobbin - that's what the potting's for. Also something worth considering too I guess, and something no doubt that has been considered in the Fernandes Sustainer circuit - I seem to recall there being phase compensating networks. Has anyone tried the DIY sustainer with an isolation transformer? If not, I have a bunch lying around at home that I can fiddle with to see what difference it makes. When I mentioned transformer driving I was thinking two-fold: 1. To isolate the driving signal completely from the guitar's earth, preventing potential sources of induced earth garbage. 2. Driving the coil in a balanced fashion to self-cancel the mag field around the driver except at the point of origin, ie the blade(s). The driver would be a little more complicated to manufacture, but perhaps would have all the EMI cancelling advantages that we're searching for. In my (guff-filled) head I imagine our little LM386 amp driving a step-down isolation transformer with a centre-tapped secondary, maybe 4:1 ratio (to step up the drive current and force the driver to work harder without taking extra battery grunt to do it), and the driver would be a dual-blade contraption with one coil wound clockwise, the other wound anti-clockwise and the magnets mounted upside down in one of the blades. The transformer is connected with the outer-most taps wired to the outermost ends blade winding and the centre tap wired to the point where the two windings meet in the middle (don't have the ability to post a mock-up sketch at the moment, sorry!). Whether this would work or not in practice is completely beyond my thinking without the assisstance of a caffeine injection though! Cheers, Curtis

Admittedly I have yet to try the setup with the guitar plugged in to an amp. I was just testing the guitar/sustainer acoustically, so the squealing I was hearing was the driver, not the output signal of the guitar through the amp. I've got no idea what other noise nasties are in the signal yet! OK, so potentially you could shift the centre frequency up and down by decreasing or increasing the cap value respectively. I reckon you could also insert some extra series resistance in the output of the LM386 circuit to "flatten and widen" the response of the resonant circuit (ie, lower the Q) and potentially regain the "lost" low notes/high notes. The only drawback to doing that is that you lose some power to the coil and you'll need extra amplification to restore the lost gain in the driver. Now that could be an interesting alternative/addition to the normal/harmonic mode switch - different sized output caps could be switched in and out of circuit to changes the way notes sustain at the pitch extremes. Have you heard of a guy called Michael Manring? He's a bassist who's released a couple of solo albums where he creates a lot of interesting effects using multiple E-bows on his fretless bass. I only have one album of his, "Thonk", but it's got a piece callad "Adhan" where he creates multiple drones and melodies on different strings with lots of morphing harmonics in much the same way you've described above. Cheers, Curtis.

Hi psw (and everyone else) Just finished burning through all 120 pages of this monster thread, and got inspired enough along the way to build my own thin coil driver! I figured I'd join the discussion and contribute a few ideas and thoughts. My driver is the same stacked pickup/driver combo that Pete came up with back in circa post 45. The construction of my single coil pickup was a little bit different to yours though, in that the individual pole pieces were sunk into separate holes, whereas yours were sunk into a slot that ran the full width of the pickup. I actually removed the poles, cut a slot in the bobbin (effectively joining all the polepiece holes together) and inserted a new 3mm thick steel blade into the bobbin window. The blade top is curved to match the strings curvature. The top of the existing bobbin forms the bottom of the window for winding the driver coil, and so all I had to do was cut a new bobbin top to keep the windings from sliding off the end of the steel blade. I used a narrow piece of masking tape to insulate the winding wire from the steel blade as I wound it on. I found an easy way to keep the windings nice and tight as I was winding them on was to pass the enamelled wire between a pair of books and draw the wire between them as I wound. The weight of the books applies a bit of pressure to the wire as it comes off the spool without gripping it too much, keeping the wire nice and taut. For potting I used a version of Selleys Aquadhere that is rated for exterior use and heat resistant to 110 degrees C, applying it to the driver with a matchstick once every 10 turns. The whole thing was wrapped in electrical tape, and the wire ends tacked in place with a small dot of superglue. I baked the assembled driver in the oven for 30 minutes with the oven set to the lowest temperature it'd do (50 degrees C in my case) to accelerate the setting time of the PVA - I noticed that some of us here had trouble with the innermost layers of PVA not setting properly and staying soft. "Baking" our drivers may be a good solution to preventing this from happening. My driver ended up measuring slightly higher than most - 14ohms with 100 turns - but this still falls in the "usable" range for the LM386 IC. I guess the wire I used as a little thinner than the recommended 0.2mm - I just used a big roll of stuff I had lying around that looked like it'd do the job! Anyway, the thing works! Just using the bridge pickup on my guitar feeding the "Champ" circuit that Pete posted, and holding the driver in my hand over the 20th fret I had sustain on most strings in most positions, even harmonic mode works! No microphonics either, so the potting must have worked too. High-E refuses to budge, but it looks like we're all having (or had) some trouble moving the high E string. Next experiments will involve building a FET input preamp, like the Fetzer, to prevent loading the pickup signal and hopefully getting better performance yet. Some thoughts/ideas: I noticed in your prior posts that you changed the output cap of the LM386 circuit and got improoved high-E string response at the expense of the lower strings. The output cap and the driver coil itself forms a series LC resonant circuit, the centre frequency of which is governed by the values of inductance (the driver coil) and capacitance (the output cap) of the circuit. At the centre frequency of the resonant circuit the impedance of the coil and cap are equal and at a minimum, and current flow will be at a maximum. Maximum current equals maximum magnetic field in the driver and best efficiency at a particular frequency of operation. So by reducing the value of the cap you're shifting the resonant frequency higher, which is why you're getting improved high-E sustaining action at the expense of low strings. Making C even smaller shifts the resonant frequency higher which will eventually stop being of use as it travels higher in frequency beyond what the string will actually do. The other thing about resonant LC circuits is that they all have a "Q", which is the amount of peaking in the frequency response of the circuit. I am thinking that one of the reasons why some of us are getting better results with our drivers than others is that some of our driver/LM386 circuits have high Q's (very sharp, narrow response at the resonant frequency) while others have a much lower Q (wider, flatter peak at the resonant frequency). The amount of DC resistance in an LC circuit governs the degree of Q - more resistance = lower Q, less resistance = higher Q. In order to get maximum efficiency and even response from our drivers, we may have to find the best LCR trade-off somehow. The C bit is easy, just substitute different values of C in the output cap position of the LM386 circuit, as some of us have already tried. R and L are a little more tricky - R is determined by the DC resistance of the coil once it is wound on the bobbin (size of the wire and length), and L is primarily determined by the number of turns and the material that forms the coils' centre, be it magnets, steel, air, ferrite etc. The popping of the circuit when turning switches on and off (like the normal/harmonic or power on/off switches) could be reduced by adding large value resistors (say greater than a couple of meg ohms) across the poles of the switch. Pro recording gear uses this technique to prevent pops when throwing switches. One more thing that's worth considering before I stop rambling, I noticed that when I moved my driver back towards the bridge pickup I got squeal as I approached the neck pickup even when I was using the bidge pickup as the signal source. This circuit is obviously quite susceptable to inducing all sorts of EMI garbage into both the signal and earth conductors. I wonder if there is such a way to run the driver in a balanced 3-wire configuration (ie, the coil driven by two equal and opposite signals swinging around a centre point) to reduce the extraneous EMI generated by the driver? All professional audio equipment uses this technology to reduce unwanted induced noise picked up in cables from being amplified at the input of the next connected device. The LM386 circuit feeding a transformer with a centre-tapped secondary would give you a balanced output, but the driver would need to be specially wound to accomodate this. I wonder if the transformer in the Fernandes sustainer is performing some similar kind of balancing operation here? Enough rambling for now! Cheers, Curtis.