curtisa

Just a couple of quick replies, not much time today Been a while since I've done any PIC work. From memory, the bare minimum to get a PIC going is a crystal, two 100pF caps (all for the self generating CPU clock) and a 5V supply. The A/D converters aren't fantastic, and they need a fair bit of support code to get them going, especially if you're also aiming to do D/A as most of the PIC's I remember don't have integrated D/A converters. But some of them do have PWM which can be cobbled together to form a crude D/A section. Power drain is pretty good from what I remember - a couple of mA when running from a crystal clock. The biggest prob will be spending time coming up with the coding and then burning the code onto the chip. I have the software and chip burner at home, but it's been a while since I've done anything with it. You can probably find PIC burning companies that will mass-burn chips for our purposes (assuming no-one can burn the chip themselves). Oh, I spoke too soon about the DPDT sustainer on/off switch - the absolute minimum switch I can do it with is a 3PDT switch. At the moment I have noiseless sustainer ON, but when switching OFF the sustainer makes a farting noise, for want of a better term Current drain in my sustainer is 7mA when switched on and no signal passing through. Wasn't able to take a measurement when playing. I used the guitar in the weekend at rehearsal. I've made a recording of it, and will post a clip of it soon to illustrate how it works in practice. I too get fizz in clean mode, but I'm actually not that bothered by it - I kinda like the sustainer more when it's working as an effect with a high-gain amp Cheers, Curtis.

None, it's just a preamp You'll still need a power amp IC to drive the sustainer. It's a bit of overkill too - probably not much use for us as a stereo IC with the tape bias selection stuff. Also, the ALC control is feedback, not feed-forward. Have a look on page 8 of the datasheet, you can see how the control voltage for the ALC is derived from the output from the circuit, not the input. Not a bad idea though, it'd be near perfect if we could find a single-channel, self-contained preamp/feed-forward ALC chip with a wide supply voltage tolerance. Keep searching! Ah...I am not sure if this is going to be effective...I am particularly concerned that you will get a pulse signal from the clock induced in the guitars pickups. If you want a pulse signal that follows the guitar strings frequency, the easiest way is effectively a fuzz circuit...a clipping amp, which is very easy to do! Also, this is likely to have problems with understanding any polyphonic sounds (ie chords) but is worth a shot... No, I think Pete is right. I don't think this'll work. Apart from the digital pulses being induced all through the guitar circuitry getting into the audio output (particularly when you're running it through a high gain amp), to get optimum performance you'd need to "ping" the sustainer at the same frequency as the string that's currently being played, requiring an input source from one of the pickups and extra circuitry to extract the fundamental frequency from the string and use it to clock the pulse generator. And as Pete rightly points out, it'd never work with with chords and evolving notes, at least not without a lot of complex circuitry and probably a polyphonic pickup. I've actually devised a switching arrangement for my sustainer that does all the necessary pickup bypassing and sustainer on/off control using just 1 DPDT switch I just need to double/triple/quadruple check it to make sure it all works properly and I'll post a diagram of how I did it. It could solve the problem of the popping driver when turning it on and off too In a nutshell, the key is to switch the battery -ve and earth, not the battery +ve - I'll leave you to stew on it for a day or two IMHO, In the interests of keeping the sustainer within reach of most of us with some practical construction abilities, I think we should keep the sustainer in the simple-as-possible basket. Doesn't mean we can't pursue the digital route, but at the moment the analog (LM386, FET buffers, opamps etc) system is certainly a lot easier to build and run. Although, if you've got a new idea going, by all means try it out and let us know how you're going with it I guess we also need to decide if we're devising a system that anyone could build from scratch by themselves, or if we're developing a fully-tested product that could be sold as a DIY kit, so the end user only has to put it all together. If it's the former, simpler probably equals better. If it's the latter, anything goes (digital, analog, mixed signal, DSP, embedded CPU etc), but it does mean that we have to do a lot more R&D before we can reach a failsafe solution for anyone to purchase and use. Cheers, Curtis.

There are now. Ladies and gentlemen, I present to you "The Bastard": . The pickups are nothing special, just the ones that came with the body. Stock Yamaha's I guess. The single coil has the driver wound on top, like Pete's stacked coil driver. Hardware is all Gotoh. Sorry, I don't have a closer picture of the sustainer - I'll try to take on or two over the weekend. It's a 3mm-wide steel blade with an arched top to match the curvature of the strings. The blade just replaced the original 6 pole pieces. Got the beastie working last night, amid a tangle of wires and aligator clips. With just a simple JFET buffer feeding the LM386 it actually works better than it ever did with a similar circuit with the driver held in my hand over the strings of my test guitar. Go figure One thing I have to do though in order to keep the whole thing stable, with the sustainer engaged, I have to short the single coil pickup wires together, otherwise the whole thing erupts into mad feedback. Still, with the sustainer working I have fundamental and harmonic mode almost perfect as-is. Based on the differences I've just experienced in holding a driver in my hand over the guitar strings, versus installing a driver permanently in a guitar, I'd say it's mandatory. All the circuit experimentation that I've been working on almost seems wasted as I can get almost perfect operation with a stupidly simple circuit, without any super preamps and compressors. That's not to say that there's no room for improvement though Don't build it just yet. While I have built it, tested it ,and got it working, others have not been so successful. And the improvement I've experienced just in permanently installing the driver into the guitar makes that circuit seem like overkill. Battling on... Curtis

OK, so you can fit 6 batteries inside your guitar to run the higher-power TDA2822 I think most of us would struggle to fit that many batteries and the circuitry inside our guitars though. In other news I've just about got my dedicated DIY sustainer guitar put together (code name: "The Bastard" ). It's a junk store special - the neck was a Strat copy neck with a busted headstock that I bought for $30 from a secondhand music store, and the body was a black Yamaha RGZ minus the Floyd Rose that I got for $12.50 at the local rubbish dump recycling shop. I blocked up the trem cavity and installed a Les Paul-style bridge, reshaped the headstock and body slightly to get rid of the "super pointy" look, and resprayed it navy blue. The end result isn't exactly the most pretty looking thing in the world, but it sounds and plays really nicely (surprisingly, for my first build!). Anyway, the driver has been slotted into the neck single coil pickup position, and I'm just in the process of installing the circuitry. The guitar actually came with a fairly large control cavity already done, and I thought I'd be pretty well off for space, but I'm still finding it hard to squeeze in the regular guitar circuitry, a 9V battery and the sustainer circuit inside. Consequently, at this stage it'll go together with a bare-essentials sustainer circuit of a simple FET buffer and the LM386 chip until I can figure out a way to fit in all these other compressors and preamps that we've developed. Cheers, Curtis.

That's a bugger, Col Dunno what to suggest? Like I said, it works for me in real life and on the sim. Regarding the output level from the guitar, surely you won't get 2V out when the sustainer is running? A couple hundred mV maybe. I'm sure I could get 2V out of my Ibanez RG7 if I hammered the low-B with a super hard pick, swinging my arm Pete Townsend style, but I reckon at that point I wouldn't be too worried about the sustainer! The dynamic range inversion I found only worked up to a point, afterwhich the compressor couldn't attenuate the signal any further and the output continued rising at a much slower, but linear rate. But this was well above the point at which the compressor stopped being usable. Don't forget that we're just looking at treating the input voltages up to a point - the point at which the sustainer can continue exciting the string, which will probably only be at a few hundred mV output. Anything above this is probably not worth worrying about. Anything up to and over 1V is probably us going mental onstage and bashing away with reckless abandon on an "A" power chord I would expect that this circuit would need to be tweaked if it were installed in a guitar anyway, as there's no way that our sustainers, guitars, and setups are going to be identical from person to person anyway. My parts drawers had 10 of the BF256B's in it. I tried three in the circuit to check component variances, and they all worked fine. I suspect in this case that's all we need - just close enough. We're not looking for super high fidelity, just a more even response from our sustainers. That said, I might take you up on that offer for the VTL5C2. I wouldn't mind trying it out if only to see what it's capable of in our application. Cheers, Curtis.

OK, results time This works, both in simulation and in real life: The whole thing is based around quad low-voltage opamp LM324. IC1a on the far left if the virtual earth generator, providing the midpoint for the 9V battery. IC1b is a high impedance input buffer, where the guitar pickup signal is applied. IC1d and the surrounding components is the precision half-wave rectifier with filtering provided by the 10uF capacitor. Finally, IC1c and the BF256B FET form the gain reduction cell that is used to vary the amount of drive. BF256 is not the only FET you could use here either, it's just the one I had handy in my parts drawer. The "output" goes to the LM386 chip (or similar). The FET/Opamp AGC actually operate as a gain "increaser" - with no control signal applied you get maximum gain, and as the control signal gradually creeps up the gain starts backing off. It never truly reaches zero gain, but it's not hard to get a gain swing of 30dB, which is a fair bit and probably more than enough for us I would think. The AGC transfer function is not linear either, more a gradual curve - as more gain reduction is called for, the steepness of the input/output ratio increases until it's actually starting to curve back around on to itself! Col's dynamic range inversion characteristic rears it's ugly head again! It would appear that the "curved" gain reduction curve negates the need for multiple thresholds from what I've seen with this circuit. In practice, if you really wanted to incorporate a threshold control/setting, you could just vary the 220k feedback resistor around IC1d. I found that varying the level being sent to the LM386 via a pot was enough to control the "speed" of the sustainer in getting going, and also how much harmonic bloom I could get (probably as I was starting to overdrive the LM386). In practice it works quite well with my handheld pickup/driver - the response over all strings and positions is quite even, and I can really back off the drive to the LM386 chip and maintain the sustaining action. Granted it's not as good as the SSM2018 limiter I was using before, but for a low parts-count, cheap, 9V battery operated sustainer compressor, it is a vast improvement over the raw LM386 amp I was initially using. Maybe... Bah! Wash yer mouth out with soap and water! Let's keep it analog at this stage before we go adding a further degree of complication Cheers, Curtis.

Ok, so perhaps the Orange Squeezer (and others like it) are geared towards more audibly "pleasing" compression, whereas what we're after is squeaky clean "accurate" compression. Ah. Could be two things happenning here - one is, like you say, the FET being operated in the non-linear area of it's characteristic curve, but also how we're generating the control voltage. If you use THATcorp's precision rectifier circuit you're actually only using a half-wave rectifier. So for every half cycle you're generating the control voltage, and for the alternate half cycle you're generating nothing. Also it's unfiltered, so if you apply it straight on to the control input of whatever gain cell we're using, you'll only compress every half cycle of the input signal, which WILL give us bags of asymmetry and distortion. Haha! Well I built it up last night and it works quite well! Very clean operation, and compresses quite nicely. I used a N-channel FET instead (BF256B) so I had to generate a negative control voltage instead of a positive one. No big deal, just reversed the diodes in the rectifier. The control signal had to be filtered though, otherwise I got loads of distortion and asymmetry (as above). The main problem I had was trying to get a sufficiently broad enough control voltage for the FET I was using, but this could easily be solved by using a different FET with a smaller Vgs voltage range (BF256B is something like 0V - 8V, which is too wide for 9V battery operation). Probably. Maybe. Dunno. I haven't looked into the LA-light compressor yet - I've been having too much success with the FET/Opamp version to try it out! Oooops - too late, already have! And clean FET attenuation utilising that circuit in the pdf document, in practice, works well. Just need to tweak it to get the operation correct. Full-wave rectification will probably help in that we'll get more accurate compression operation (we're generating the positive AND negative half cycles of the input waveform. I suspect that careful selection of the attack and release times will help aswell (filtering the control voltage). Sounds like we're on some sort of race - you take the LED/LDR route and I'll stick with the FET/Opamp system. Onelastgoodbye will take care of the drivers and we'll meet somewhere in the middle! BTW, nice work with that driver construction there, Onelastgoodbye Cheers, Curtis

I think it will matter if we want any hope of driving the high strings in high positions, or in harmonic mode. If the circuit is incapable of reproducing anything over a couple of kHz you'll never be able to excite the higher notes on the guitar into sustain, no matter how good the driver. What we want is an opto coupler. There's only one style that I know of. It's made by Vactrol, and it's not easy to get. VTL5C1 and VTL5C3 are examples of them - they're just an LED and a selected LDR encased in a sealed epoxy cylinder. A lot of amp manufacturers use them for channel switching and tremolo circuits. There's only one place I've found here in Oz that stocks them and they're 15 bucks a pop...if they have them at all... Well, I'm actually getting better, more-consistent performance with clean, compressed signal. For me the distorted signal worked, but introduced a lot of unpredictability into the sustainer - excess microphonics, harmonic sustain when in fundamental mode, excess EMI... That schematic of the Tangerine Peeler is pretty much it, just set the "FF/FB" switch into the "FF" position and you're away. Mind you, it does say on the diagram that it's an "untested trial design". I'm going to have a closer look at that FET/Opamp article that Col found over the next couple of days - long weekend for us down here, going to make the most of it! Interesting side note - I tried the sustainer with my acoustic last night. It's got a piezo bridge pickup and preamp all built in. With the feed-forward limiter and clean drive I could get plenty of sustain on the 1st and 2nd strings, but almost nothing on the 3rd-6th strings. Probably due to the bronze-wound strings having less magnetic material to play with. And I could move the driver all the way to the bridge where the piezo pickup is located and got absolutely zero feedback (driver didn't work so well here though as it couldn't "move" the string as efficiently). Cheers, Curtis.

In guitar amps they usually place a high voltage ceramic capacitor across the poles of the power switch (and/or standby switch) to suppress the pops. Don't know if this can be adapted to our purpose with the same degree of success? The monitor amp in my studio is completely silent when turning on and off, and that's without any fancy power-switch-capacitor malarkey. And it's a DIY job too (another Silicon Chip special). That's a shame - just last night I was breadboarding up the "Tangerine Peeler", but ran out of components before I could try it out. Maybe I should abandon it before I get too carried away? Yes, that's something I noticed when I sim'ed the Tangerine Peeler gain cell - there was an optimum range of control voltages that seemed to work best with the FET's available, and it tended to be quite non-linear. I suspect that the two FET's in the gain cell probably needed to be quite closely matched, which is probably more trouble than it's worth for us. Now that looks much more appealing, and reduces the component requirements down too. One FET, one (stage of an) opamp and a handful of resistors and caps. Might be time to give this one a go. Difficult to say. The light/resistance characteristic of LDR's can vary quite a bit from piece to piece, which means that our DIY'ed compressors will probably vary in performance from one to another. Cheers, Curtis.

No, I don't think that will work. Placing a variable resistance in series with the battery supply (or even wired as a voltage divider) to reduce the pop may work for a bit, but you're likely to burn out the pot if you leave it positioned just shy of full rotation. You may also find there's some instability in the sustainer circuit as you start winding the pot back, as you're introducing unnecessary and excess impedance in the power supply. I've just had another look at the LM386 datasheet and it seems that you can force the chip to mute by either connecting pin 7 to the supply line, or shorting pin 1 to ground. I reckon there's a possibility there for making popless on/off switching by somehow sequencing the power supplies around the chip - say when turning off the sustainer, pin 1 is shorted to ground and a moment later the power supply shuts off (or drains slowly, over a few hundred milli seconds or so). Yet more things to try out... Cheers, Curtis.

Farnell and RS Components will stock it, though it may be a bit more expensive than over-the-counter sales. Cheers, Curtis.

Nah, not the same, though I do know of the one you're talking about. Mine is the 5-knob jobbie produced by Jaycar for all of about 6 months in 2000 or so. I don't have the design notes and original magazine article anymore though, so I'm going to have to reverse engineer it to find out how it works. It's more the topology used, not so much the chips used. Pretty much just proving Col's idea of using feed-forward as a control method, and my insistence of using the precision rectifier (ala THAT corp) as the control voltage generator. The SSM2018 is pricey but not unattainable, but the main drawback is that it's minimum supply voltage is +/- 5V, which pretty much eliminates it's use for us in battery operated systems, unless we want to consider phantom powering or perhaps dual 9V batteries. I'm sure there are other options out there that will work. We've just got to find them. That's something I am working on at the moment, but it will take some time before I can get a hands-free sustainer going. The pickup/driver is currently removed from a guitar that I'm building that will eventually have the sustainer permanently installed. I don't have any other guitars that I can install a single coil pickup in to try it out in situ, so until I've got the guitar built I have to continue experimenting by holding the driver in my hands Ebow-style. I don't think the compressor will help the driver in terms of proximity to the guitars' pickups. If anything I'm expecting to be worse simply because the system has to incorporate makeup gain for the quieter strings/notes - there's more gain present (and more risk of feedback) when there's no signal passing through it. Perhaps this is a good excuse to incorporate downward expansion or noise gating into the driver system? Getting too complicated...I'd vote for minimising the potential for feedback at the driver-end of things - more development of the self-cancelling-humbucker-bilateral-dual-rail-super-duper driver. Well, there probably is a reason why they're using these systems, and I'm sure we're just discovering the same pitfalls that they went through during the R&D of their products. Might be time to order one in and try it out Haha! How's this: Thankyou, Col I haven't checked this yet, so I have no idea what, if any improvements or reduction in battery performance has been achieved by incorporating the limiter into the design. Besides, the limiter is currently running off +/- 12V on a separate supply to the 9V LM386. I don't really have any way of getting the whole shebang running on the same battery at this stage. This is probably more in relation to changing the resonant frequency of the driver, and hence the best band of frequencies at which the driver works. That's not to say that phase differences aren't a factor, but I suspect it's not the largest variable, particularly at the lower strings Yeah, that's what I'm using too, but my version must be a few revisions older than yours as I don't have a LM13700 model. Cheers, Curtis.

Ok, so I've just been fiddling with my (still handheld at this stage) pickup/driver sustainer, and I was running it through a limiter that was built from a kit published in an Australian electronics magazine (Silicon Chip if anyone is interested). The kit is sadly out of production now, but it featured the SSM2018 VCA chip at it's heart and also a very similar precision rectifier circuit as the THATcorp appnote. And it's a feed-forward design. AND it makes a helluva difference! I can get plenty of sustain on every string in just about every position. What's more is I can get the same performance at a greater distance from the strings than just using the plain old LM386 circuit, which means we don't necessarily need to have super-low action anymore. For some reason I'm getting better low string performance too, not sure why. Beforehand I could only go down to the note "E" on the 5th string/7th fret before it started sustaining harmonically instead of fundamentally. Now I can get fundamental sustain almost all the way down to low "G", 6th string, 3rd fret. Could be that I'm now no longer pushing the LM386 into clipping and accentuating the upper harmonics of the note? The limiter I'm using isn't really adaptable into a built-into-a-guitar arrangement though - power supply requires +/-12VDC, and the PCB board it's made up on is about 3" square, but as proof of concept it works extremely well. My next trick will be trying to tailor the response of the sidechain to emphasise the high-note drive. Good progress, we're on to something here! Dunno if I can do it using the trees/chainsaw story With feedback compression (eg, Col's original schematic, the Aussie comp etc) the output of the VCA is the signal that tells the compressor to start turning down the volume, so effectively the compressor doesn't care what's happening at its input, it just "blindly" goes about making it's output quieter everytime the threshold is exceeded. With feed-forward compression the input signal is what is used to tell the VCA to turn it's output down. Might seem like the end result is the same - signal exceeds a threshold, output turns down - but the difference in our situation is that there's another feedback loop outside the compressor: the driver, the string and the pickup. The compressor needs to know if the signal it's pumping to the driver is sufficient to maintain the sustaining action, and the way it does that is to "listen" to it's input. Feedback compression can't do that. Ok, thought of an analogy (possibly quite crap too!): Imagine a guy filling a bucket of water from a tap. He needs to turn the tap off when the bucket is completely full to the top, so he puts his hands around the sides of the bucket and waits until he can feel water trickling over the sides of the bucket to signal that he needs to turn the tap off. He's being reactive - that's feedback compression. Imagine the same guy filling up the bucket of water, but this time he's watching the water rise slowly as it fills. When he sees the water lapping at the rim of the bucket he knows it's time to turn the tap off. He can anticipate what the water will do as it nears the top. as he can see what it's doing. He's being proactive - that's feed-forward compression. ...And they lived happily ever after Cheers, Curtis.

Call it a moment of madness... The reason why we want to use the precision rectifier is because the diodes have a forward voltage which must be overcome before they start conducting. With a 1N4148 the forward voltage drop is in the vicinity of 0.6 - 1V depending on the current flowing at the time. The problem with that is that the normal signal voltage generated by the pickup is anything from a few mV to a volt or so - some of those signals are obviously too small to overcome the fwd voltage drop of the diode, so for small input signals you generate no control voltage, causing unpredictable compression action. Even if you add gain to the circuit before the diodes you still are left with at least some signals that will have trouble overcoming the fwd V drop, and you're also starting to run out of headroom in the rectifier circuit, limiting the upper range of input signals you can rectify. With the precision rectifier, and the diodes sitting inside the feedback loop of the opamp, the fwd V drop of the diode is negated by the action of negative feedback, and you can accurately rectify input signals much lower than the normal fwd V drop of the diode - practically to 0V input. So, am I right in assuming that the LM13700 VCA reduces gain as the gain control input current falls (I haven't tried fiddling too much with the circuit yet)? So we want to generate a control voltage that gets progressivly bigger as the input signal exceeds the threshold further, yeah? The easiset way I can see to do it is to have a static DC voltage applied to the gain control pin, and subtract the control voltage from it - that's just an opamp stage with a reference voltage applied to the non-inverting input (call it "VCA normal gain" control - could even be a variable resistor) and the control voltage applied to the inverting input. The output of that opamp becomes the control voltage for the VCA. Hehe...some people would argue that trying to build a sustainer and writing 132 pages of concepts and ideas is madness BTW Col, what software are you using for your circuit simulations? Cheers, Curtis

I think you're getting it... As Curtis pointed out, there are two ways to set up the control circuit in a compressor - one is to get the compressor to try to make the signal louder or quieter depending on its own output, the other is to use its input signal as the deciding factor... (I was re-inventing the wheel describing the second version as an approach for the sustainer). In the sustainer system, the input to the circuit comes from the ...driver->string->pickup, so if the input is used to decide whether to add more or less drive, it will automatically adjust for any variations caused by string, fret, pickup etc. If a normal 'stompbox' guitar compressor that uses its own output signal as a control is used, the level is evened out _before_ the variations caused by strings and frets are added to the mixture... so they remain :-| I think there's still some confusion as to how exactly a compressor works and the difference between a comp and a limiter. A compressor does not amplify anything. Signals below the threshold do not get extra gain. All a compressor can do is turn the volume down once the input signal gets too loud. By how much the compressor turns the volume down by determines whether it behaves as a compressor or limiter. Here's a really poxy analogy Imagine the input signal as a row of trees and the compressor as a dude with a chainsaw. The chainsaw dude wants to take the tops off the trees to make them shorter so he lines all the trees up and lops the top off them - some of them are still a little taller than others, but they've all been trimmed a bit. Some of the shorter trees didn't need any trimming and he's left them alone. So we're left with a row of trees that are all a wee bit shorter, roughly the same height, but there's still some height variance in them. That's all a compressor is doing. However, all the trees are now shorter and we now want them brought back to the same height they used to be because we can see too much of the next door neighbours property (ya, it's a **** analogy). We then put the trees on an elevated platform and jack them all back up to the height they used to be at. That's the function of the makeup gain control, and it happens after the compressor. It can give the impression that there's some special amplification happening below the threshold, but it's entirely separate and unrelated to the act of compressing. Limiting is just taking the row of trees and lopping the tops off all of them so that they're all exactly equal with little to no deviation in height. Here's my thoughts - the feed forward compressor that Col's putting forward is the way to go. The trick is going to be making the compressor more or less responsive to particular frequencies and amplitudes (string size does play a part, but in the absence of being able to make a driver per string we have to ignore it and focus on frequency and amplitude - the sustainer is a monophonic device being asked to work on 6 different signal sources at once). I think we need to get a reliable simple feed forward compressor going, and then we need to tweak the behaviour of it somehow. My thinking is that we want a low threshold, but high(ish) gain before the compressor. The sidechain signal needs to be treated in some way so that it generates a smaller control voltage for higher frequencies so that the compressor "ignores" the higher strings and focuses more on turning down the drive to the thicker strings (it's the skinny bits of steel that need more drive afterall). In order to do this I reckon we need some frequency shaping network in the sidechain that removes a lot of high frequency content before it gets a chance to affect the compressor. We may also need some additional post-compressor frequency carving going on so that the lower strings are naturally driven less anyway. Here's the (image-free, sadly) signal flow I'm proposing: Guitar pickup -> Buffer (with extra gain perhaps) - > VCA -> High pass filter -> Fixed output volume (or trimmer - set and forget) -> driver amp (LM386 or other?) -> Driver. At the point where we tap off the sidechain signal, between the buffer and VCA above: Buffer -> Low pass filter -> Precision rectifier and threshold control (similar, but simpler than the THATcorp appnote) -> DC smooting -> VCA gain control pin. You could probably realise the above with a quad-opamp chip, VCA element (be it FET, LM13700 or otherwise), an LM386, and a handfull of resistors, caps and diodes. Should be buildable on a 2" square PCB (or a couple of stacked boards if space is at a premium). In your diagram, the gain control pin is normally tied to the 9V supply via that series resistor, which sets a fixed gain as determined by the current flow into the input pin. You've placed the transistor in parallel with the gain control pin. So with the transistor switched off the LM13700 gets all the gain current, and as the transistor switches on more it steers more current away from the LM13700 and into the collector of the transistor. You should only need to feed the DC control voltage directly into the LM13700 gain pin via a resistor to get it to behave properly - no need for fancy transistor switching I wouldn't think. That's probably more to do with the capacitor you have in there (1uF). You've formed a lowpass filter between the 1uF cap and the 2K base resistor. By changing the base resistor you've changed the cutoff freq of the filter, making it more or less responsive to high freq's, thus making the circuit quicker or slower to respond to changes at it's input. Also changing the base resistor changes how much current gain is reflected through to the transistor's collector - very roughly, current in the collector is base current Ib x hfe (the transistor current gain). Making the resistor smaller makes Ib bigger, which makes collector current bigger, but only to a point at which you reach saturation. Delving into the dim, dark recesses of my memory, having a bigger IB will probably give you a faster switching speed too. Cheers, Curtis.

Ok, first glance assessment is all I can give it as I don't have much time to spend on it today: So you've got your virtual earth buffer in the top left which gives our psuedo 0V supply - cool. Input buffer at bottom left has variable gain which feeds both the side chain and VCA. Input impedance is a little low for connection to a pickup (1M and 8K2 in parallel). Ditch the 8K2 series resistor on pin 5 of the LM358 and run direct into the non-inverting input. Next stage is your normal/harmonic phase inverter, feeding the VCA which is wired up pretty much like the app note, but with the gainset and linearising diode resistors scaled to reflect the lower supply voltage? In the side chain is another buffer, but it doesn't look quite right to my thinking - you've got the diodes connected outside the feedback loop of the opamp. It won't rectify small signals (less than the forward voltage drop of the 1N4148 diodes) properly like that. I'm not sure the biasing of the BC337 trasistor is right - are you trying to steer current away from the LM13700 gain control pin when the input signal exceeds the threshold? I think the way you have the transistor biasing set up it will always be conducting (base is 0.6V higner than the emitter) and conducting even more when the control signal appears - probably why you're experiencing such high power consumption? It's also possible that the transistor is being run saturated meaning that no matter how much extra control signal you apply it will never reduce the gain any further - the transistor is already conducting as much as it can - possibly why you're getting such bad compression performance for high input signals. Also, I'm assuming the variable resistor "R" is your threshold control, but it controls the strength of the signal entering the VCA and the sidechain - as you wind down the threshold you're also reducing the signal that the VCA sees. Threshold should only affect the sidechain signal - the signal entering the VCA needs to be uneffected by this control. Have another look at the THATcorp appnote to see how they implement the Threshold control into the side chain, and also how they rectify the input signal for the control voltage. That's just about as simple and effective as it gets. Cheers, Curtis.

Got a diagram of your alterations so we can debug it? Again, got a schematic? The sidechain of the THATcorp app note I mentioned earlier (here if you can't be bothered trying to find it again) is a pretty standard setup for opamp based feed-forward designs. I've used it myself on at least one other compressor that worked fine, and I know of at least one other high-end compressor that uses a very similar setup. Referring to page 2 figure 1, if you omit the 2252 RMS detector and just use the components based around OA2 and OA3 you should be able to easily generate a constantly varying DC voltage based on the AC input signal that is dependent on the Threshold, Compression and Makeup Gain controls (omitting the RMS detector will make the compressor a peak-detecting type rather than RMS detecting - differences are possibly not too important in our application). I simulated the above circuit earlier today and I can get it to work quite well for a variety of input voltages. Based on past experience, the real-life performance is quite good aswell. Implementing a circuit like this into the LM13700 should at least be a good starting point, though I would expect a bit of tweaking is required to get the correct operation. Yes, the detector circuit of the 1176 is waaaaay too complicated for our use, but that gain reduction element is ridiculously simple and effective - a few resistors and one FET. It's possible that you could marry the THATcorp sidechain with the 1176 gain reduction element (with a bit of fiddling of course) and have a highly effective compressor that's smaller than the LM13700 chip alone (and probably a lot kinder on battery consumption too!). Cheers, Curtis

Ahhh, yes, of course! I keep on forgetting that we're not compressing blindly to set an output - the string and driver form the signal source that gets fed back into the system via the pickup. The whole thing relies on a lot of nested feedback loops to function properly. Yes, I have a few. Sadly they're quite overly complex for our application though. And looking through the datasheets for some of those VCA's I mentioned earlier on, most of them require at least +/-4V to run, which pretty much excludes 9V battery operation. What we're probably after is as simplified version of a FET based compressor, like the 1176. Note: Yeah, I know it's a feedback design, but you could still conceivably design a FET compressor using feed forward arrangements. In order to keep this cost-effective and keep the components in the easy-to-get basket, we're probably going to have to design (or adapt) a descrete compressor. The MXR Dynacomp might be another alternative circuit? The LM13700 will work as feed forward mind you. Figure 2, page 8 of the LM13700 app note shows you how to connect the VCA up. All you need to do is provide a precision-recitfied signal derived from the input signal, and use it to control the gain on pin 1. Setting it up "just so" for the right threshold and compression ratio settings will be somewhat tricky though. Cheers, Curtis

Errr...I think you've just described exactly how a compressor works. The compressor needs to use the input signal to control the AGC, otherwise it won't compress! Have a look at this appnote by THATcorp: http://www.thatcorp.com/datashts/an100a.pdf Ignoring all the unnecessary detail, the way the compressor is constructed is exactly the way you're proposing - The input signal (Vin) is split and goes through the VCA chip (218X), and also into the sidechain (2252, OA2, OA3). The output of the sidechain feeds the gain control pin of the VCA (pin 3) and the output signal is buffered by OA4. I think the only difference between what you're proposing and what you've used in the past (the LM13700) is that we're changing the compressor topology from feed back to feed forward control. The thinner/tauter strings will require more drive and the thicker/looser strings require less drive. The comrepessor, by it's very nature should naturally even things out anyway, regardless of what compression topology is used. Thinner strings will naturally generate a smaller signal (due to mass, vibration amount etc), which causes a smaller control voltage to be generated for the compressor, resulting in less compression (perceptibly louder), whereas the thicker strings will generate a higher control voltage and more gain reduction (perceptibly softer) for a given compression threshold. Cheers, Curtis.

Yeah, that's what I figured too - all the extra components make it less attractive, but I guess we're after repeatable, reliable performance..? I'll have to have a look, thanks for the tipoff Nah, it's just really poorly drawn. Remove the earths at the top right (just below the 110K feedback resistor), and the output of the opamp (labelled "Vs/2") and it'd work fine. The supply bypassing isn't quite right either - the two caps on either side of the opamp supplies should either be drawn with their "earthy" ends sitting off the Vs/2 output, or straight across the battery terminals (omitting the negative side of course, as it's not doing anything bypassing itself). I can redraw it for you if that sounds a bit confusing The battery terminals become +Vs and -Vs, and the Vs/2 becomes the new ground for the opamps that require it. Special IC's (like the LM386) that require no ground are designed to look after themselves. Connecting one circuit up with it's Vs/2 ground to another with it's -ve battery ground will present no problems provided you capactively couple the inputs/outputs. All we're doing with the Vs/2 output is providing a reference voltage around which our audio signals "wiggle" around on. The same thing happens with dedicated split supplies (eg, +15V, 0V, -15V) - you could argue that the whole thing is one giant 30V supply, but the difference is that we already have our midpoint ready to go thanks to the more elaborate power supply/voltage regulators, and that we can source/sink a lot more current than the opamp vs/2 supply. Righto, off to have a fiddle with this sustainer thingy! Cheers, Curtis

My thinking is that high value resistor (250K) is used to prevent the signal current flowing into the LM13700 from the previous stage from throwing off the gain reduction function of the linearising diodes. You'd probably get better performance if you used the LM13700 in the configuration shown in the app note on page 8, figure 2. However you'll need more support circuitry to use it in this way - the control signal is completely different to the one you're currently using. You'd need a precision rectifier to properly generate the control voltage to apply to pin 1 (labelled "gain control" in the app note). There are other smaller ones yet. Thing is they cost even more. I'm waiting on some free samples of the THAT2181 which is an 8 pin SIP package VCA. I reckon the buffered voltage divider is the better way to go (page 88, figure 7). Seeing as we're looking at running the sustainer from battery supplies we want the "mid point" of the battery to remain at exactly half the supply as the battery drains. With the zener earth reference, as the battery starts draining the earth starts falling relative to the positive side of the supply, which will bugger up the earthing of the system - you might start out with a 9V battery with 4.5V in the middle (+4.5V, 0V, -4.5V, but after a couple of hours of playing your battery has sagged to 7V and you're left with a supply sitting at +2.5V, 0V, -4.5V. Not so good. At least with the circuit shown in Figure 7, if your battery discharges to 7V, your supplies are still divided equally around earth. Current shouldn't be too much of a problem, there are plenty of opamps capable of sinking/sourcing enough current to work a few stages - NE5532, NE5534, LM833, even the good old LM741 would probably work OK. We're only supplying the earth for the preamp section afterall (light duty stuff), the LM386 will look after itself with just a single 9V supply. Cheers, Curtis.

Certainly does. Was using it myself a couple of days ago during my tests, works well Cheers, Curtis.

Hehe, check this out Also from the Nat Semi Audio Handbook (what a brilliant publication!), SPDT switch, no extra gain and phase inversion. With the switch in the "up" position there is no phase inversion, with the switch in the "down" position there is phase inversion. The input impedance is 100K/2, so you might have to increase the resistors to prevent loading of guitar pickups (or feed the input from another high impedance buffer), but make sure you keep all resistor values equal. Cheers, Curtis.

Of course, if we're considering external power sources as an option (phantom powering etc) then bridged amping is still do-able, or indeed any other non-battery-friendly system. BTW, just in case I'm not getting the full story and missing large swags of conversation, bridged amping is not the same as class D. Oh, and this is the easiest way I can think of to flip the phase of the driver signal without swapping the driver wires: All the extra circuitry is left out for clarity. The same DPDT switch, but all you're doing is swapping the inverting and non-inverting inputs of the LM386 (pins 2 and 3). Cheers, Curtis.

Bridged operation is a simple way to get more watts into a load, usually double the power. If you're going to stick with the LM386 as a driving source you'll need to add a second one to do it, or swap the LM386's out for one of those stereo chips. Unfortunately doubling the output power using the same amp topologies in the LM386 also results an a doubling of the power requirements - your battery will drain twice as fast! The load might not be referenced to earth, but you're still pushing twice as much power through it. Still, it might be interesting to see if the unearthed load used in bridged amp topologies has any impact on the amout of garbage injected into the guitar's earth. If it does, we could still make it work by running the amp in bridged mode but limiting the amount of power that it can put into the driver, thus maximising battery longevity.