Jump to content

Sustainer Ideas


psw

Recommended Posts

A modification to reduce the power consumption at rest would make the amp more suited for the sustainer.

So how would you go about that ?

cheers

Col

Welcome in Babylonia!

You mean how I was supposed to reduce the power consumption at rest.

The oscillation frequency depends on R3 and C4 (and R2 and the output impedance of whatever is connected at the input I reckon).

To lower the oscillation frequency and quiescent current increase C4 and run the test again.

From what I understood from the article the power consumption has to do with the real dead time setting and switching frequency. I wouldn't be surprised if the voltage was the determinant factor, but that's not what the article says. Only one way to find out. Swap capacitor and see what happens.

Cheers

Fizz

Link to comment
Share on other sites

From what I understood from the article the power consumption has to do with the real dead time setting and switching frequency. I wouldn't be surprised if the voltage was the determinant factor, but that's not what the article says. Only one way to find out. Swap capacitor and see what happens.

I suppose that most of the losses occur during the crossover.. caused by dead time, linear mode operation.. whatever... (my class-d knowledge is minimal)

If you reduce the carrier frequency, there are fewer crossovers, so efficiency should be improve.

If you reduce the carrier frequency, in order to avoid an increase in switching noise, the low pass filters cutoff needs to be lowered, or the Q increased or both.

I think this would be a good option for us though. For normal audio, you would want a good response up to at least 18kHz, for our sustainer we probably only need to get to 1.5kHz - 2kHz max, so plenty of scope for reducing the carrier frequency.

cheers

Col

Link to comment
Share on other sites

fresh fizz....I agree about the biasing of that FET.... as battery voltage starts decreasing, then so does the biasing. Not good.

Whilst all very interesting, I'm not quite understanding the pursuit of 'power modulation' ideas, when perfectly good Class D power amp chips exist that cost pennies....

http://uk.farnell.com/jsp/search/productde...jsp?sku=1648680 (ok, so at 3mm long, it's small, but this isn't insurmountable & is surely going to far exceed the results hobbyists could achieve switching the rail on/off quickly!)

Edited by Hank McSpank
Link to comment
Share on other sites

ok, so at 3mm long, it's small, but this isn't insurmountable

Pretty damn tricky though. How would you do this? Dead bug style ?

Maybe not so bad if you have cnc, or all the kit for etching your own pcbs.

Any suggestions for the rest of us?

Link to comment
Share on other sites

ok, so at 3mm long, it's small, but this isn't insurmountable

Pretty damn tricky though. How would you do this? Dead bug style ?

Maybe not so bad if you have cnc, or all the kit for etching your own pcbs.

Any suggestions for the rest of us?

Sure.....an adapter board...

http://cgi.ebay.co.uk/Surface-Mount-SMT-SM...A1%7C240%3A1318

re soldering pins that are only 0.65mm apart ....well, for a start if you've shaky hands, just flood all the pins with solder, then use copper braid to get the excess off thereby removing 'bridges' (the capillary action of the braid will lift the solder off...except the solder that's underneath the pins) - there are plenty of tutorial youtube videos about, you can get the idea here...

(from about 3m46sec in).

Another method is to buy some solder paste (eg http://cgi.ebay.co.uk/SOLDERPLUS-LEAD-FREE...A1%7C240%3A1318 - you normally need a syringe, but you could get away with dabbing a needle into the container & then dabbing it on to the pin pad) .. place a small amount on the pin pads, mount the chip (the paste holds it in position) & whack it in your oven (got to use the right temp!) - the paste morphs into solder - no bridges with this method.

A lot of folks are put off by SMT, but all the best & latest variants of chips are often only available in SMT nowadays.....going this route sure beats the hell out of rolling your own class D amp! (ie 90p for the Class D chip & £2.00 for the adapter - cheaper than a Marks & Spencer sandwich!)

By the way, a CNC ain't much use when it comes to SMT (unless you throw £££££££s at the CNC build - mine was made from scrap!)....at best all I can expect from it, is to make a larger 'experimental' board & things like coil bobbin cutouts.

Edit: I've been using a TDA7052A as a poweramp chip ....I feel it's worth posting up details of a TDA7052A 'quirk' (just in case anyone else uses this poweramp chip). This particular chip performs admirably when the peak to peak signal feeding into it is under 3V (the exact level I can't be sure as I noted it down at home ...i'll update this post when I get home)...ie no visible distortion. However, once it get's above a certain level...the output signal seen across the driver goes really whacky. It's as if the frequency has doubled (a bit like an unsmoothed full wave rectifier output)...it's not a clean representation of the input at this double frequency - no clipping though. This has proved to be a real bummer for me, as I really needed the full 5V swing heading into the power amp chip (for the most flexible AGC)- I guess what I really need is a unity gain voltage to current converter as the output stage.

Incidentally, last night was the first chance I've had to try & integrate my 'digital PIC AGC' with my analogue sustainer electronics - as a first run, it went *very* well.

I have full digital control over...

1. How quickly the sustainer kicks in after a string has been plucked (ie how quickly the sustainer takes up the slack as the string fades out.... the idea being towards a seamless 'marriage')

2. The 'starting level' gain for the AGC (else each time the 'sustainer start' string threshold is breached, it has to start getting to the correct 'gain region' from scratch, which can pose a problem wrt hitting the 'time deadline to wrestle control of the string before the string fade below the 'point of no return'!)

3. How quickly the AGC reacts to incoming signal variation (ie how quickly it adjusts the JFET ...which in turn changes the gain), both...

i) Initial 'coarse' AGC (ie to get to the approximate gain 'region' real fast)

& once it gets 'in the zone' ', then...

ii) Final 'fine' AGC (this takes over from above coarse AGC, to stop it 'hunting'/pumping' ..this is where the sustain 'drive' level ramps up/down as the AGC coarsely tries to get 'hit' the preset optimum gain setting)

4. Sustainer 'enable' threshold (I can cease the sustainer's 'hold' on the strings by slightly damping them - once the string level falls below a certain level...the AGC gets disabled, and therefore the sustainer is curtailed)

5. Release period - I can either have my sustainer 'hold' a note forever at the same level, or fade it out over a preset-able period (eg...slowly fade over 60 seconds)

6. Intensity level.

...I'm chuffed to bits! The mind boggles though as to how the commercial manafacturers got their's working using discreet components!

Some difficult choices lay ahead soon, do I...

1. Stay focused on a six channel sustainer (Hex...which is how I started out this journey!)

2. Have a sustainer for six inputs (hex), six AGCs ...but then sum them into one driver (much lower component /driver count...less power)

3. Have one standard mono guitar input & tweak what I've done to suit! (the lowest component count of them all...but not likely to be as 'controlled' due to the wider & combined frequency band)

Edited by Hank McSpank
Link to comment
Share on other sites

Incidentally, last night was the first chance I've had to try & integrate my 'digital PIC AGC' with my analogue sustainer electronics - as a first run, it went *very* well. I have full control over how quickly the sustainer kicks in after a string has been plucked (ie how quickly the sustainer takes up the slack as the string fades out.... the idea being towards a seamless 'marriage'), what preamp gain level the AGC should start at, how quickly it reacts to incoming signal variation (ie how quickly it adjusts the gain) - both 'initially' (ie to get to the approximate gain 'region' real fast) & then 'final' (a 'finer AGC' takes over ...this is needed to stop AGC 'hunting'/pumping, where the sustain level ramps up/down as it tries to get to the optimum gain setting), ...also it's easy to set the string threshold to cease the sustainer (plus other cool stuff, like 'initial full sustain'...changing to a setable very slow sustainer release/fade out) ...I'm chuffed to bits. The mind boggles as to how the commercial units have got their's working in discreet!

Thats good news.

A couple of things:

There will be a limit on how much control you have at the point where the sustainer takes control of the string. This is caused by the way the natural vibrations of the string change after the pluck - they go through a weird combination of horizontal, vertical, eliptical and figure 8. The sooner the sustainer kicks in, the more damage it will do to the character of the sound.. leave it too long and there will be a dip in output. It's possible to get a good balance, but you won't have a lot of leeway there, unless you're going for a quite synthetic sustain sound.

As far as how the commercial units have got it all to work. Its not an particularly difficult project as long as you can use big boards and/or smd components. There are ways to control the various features of the system without resorting to software. Having said that, software is a nice way to reduce the parts count.

e.g. theres a nice little circuit snippet in one of the THAT corp app notes that shows a 'non-linear capacitor' (I think thats what they call it). It uses an op amp and a few discretes to create a simulated cap that allows a very fast reaction to an initial signal, then a much slower reaction after the signal has begun. This does what you discribe - preventing pumping and distortion due to the attack time being set too low.

Of course, these extras cost money, board space and a little power.

I would still prefer to go with a full digital signal path if I was going down the PIC route. This would allow the fine control your describing over the phase as well as the amplitude. No fet would be required. It may also make it easier to use some of the 'digital' class-d chips out there in order to further reduce the parts count.

I could experiment with delay/reverb effects to better mimic the 'loud guitar in a room' type of feedback. If the chip is powerfull enough, multiband control so phase and amplitude could be tweaked depending on frequency. Hehe, with enough processing grunt, you could have lots of fun.

I still want to get the basic analog version set up as good as I think it can be.

I've been playing around with lenearising resistors around the fet. These do reduce the distortion , but they also increase the response time somewhat.

So I have some experiments to do on the breadboard as soon as I have time.

I also really need a scope :D

cheers

Col

Link to comment
Share on other sites

There will be a limit on how much control you have at the point where the sustainer takes control of the string. This is caused by the way the natural vibrations of the string change after the pluck - they go through a weird combination of horizontal, vertical, eliptical and figure 8. The sooner the sustainer kicks in, the more damage it will do to the character of the sound.. leave it too long and there will be a dip in output. It's possible to get a good balance, but you won't have a lot of leeway there, unless you're going for a quite synthetic sustain sound.

I agree that the initial guitar signal is *very* rich in harmonics...but you'd be surprised how quickly these settle, resulting in something more akin to a sine wave. There is no sustainer in existence that can do anything other than stimulate a string & *not* yield a sine wave type string output (this is why sustainers have a slighty different tone in the tail to the guitar's initial attack (it's not much of a problem in reality though ....Roland worked out that us humans decide what it is we're hearing mainly based on the 'attack' portion of a sound - Roland released LA synthesis to take advantage of this aural illusion! http://en.wikipedia.org/wiki/Linear_Arithmetic_synthesis). The only way around this sinewave-esque tail, would be to have the sustainer circuit 'sample' the string sound just after the attack (as it's initially *too* rich in harmonics then) & then loop that sample back to the string driver...that's an awful lot of processing & specialist h/w (& we're probably some way off that level of sustainer detail!)

As far as how the commercial units have got it all to work. Its not an particularly difficult project as long as you can use big boards and/or smd components.

That's the point...we're all - in the main - restricted by parts count (even SMT to an extent)...that's what makes this a difficult project!

It'd be a doddle if we we're all upright double bass players! (plenty of room in one of those puppies!)

I would still prefer to go with a full digital signal path if I was going down the PIC route. This would allow the fine control your describing over the phase as well as the amplitude. No fet would be required. It may also make it easier to use some of the 'digital' class-d chips out there in order to further reduce the parts count.

I don't think that'll be possible with PICs for a good while. the only option then is something akin to a Variax board - take your guitar signal, AtoD it using dedicated AtoD chips, a chunky motorola DSP & do it all in software ....end result? A board that's fairly big - oh, yeah...it'll cost you £200+ to make!

re the Scope - yes...certainly if you want to come up with your own 'design', I'd say you definitely need some visual feedback wrt what's going on with your breadboarded circuit ...it needn't cost a lot (anything?) - there are plenty of soundcard based 'software oscilloscopes' on the net - more than up to the job of monitoring guitar frequencies. BTW: I finally got round to using MultiSim last night...very cool, but it can only take you so far - for example, that TDA7052A quirk i mentioned earlier...I doubt a simulator would ever be so 'customised/detailed' to simulate that one...& without a scope, how would I have known why my guitar output suddenly sounded awful?

I genuinely find it amazing that so many DIYists have spent as long on their sustainer journey without one. :D

Edited by Hank McSpank
Link to comment
Share on other sites

I agree that the initial guitar signal is *very* rich in harmonics...but you'd be surprised how quickly these settle, resulting in something more akin to a sine wave.

I'm not talking about the harmonic burst during the attack. I'm talking about how the string vibration changes through the horizontal and vertical axis. The sustainer can only drive the string in the vertical plane.

I reckon (yep pure conjecture, but bourn out by listening and fiddling) that the sooner the sustainer grabs the string and yanks it out of its natural vibration, the more likely that that transition will be noticable/audible. If the sustainer comes in smoothly as the string naturaly settles into more or less a sine wave, then it will sound more natural. It's hars to get this to happen without either a small dip in volume on some string/frets, or a more aggressive 'audible' sustainer sound on other string/frets.

In the end, it doesn't really matter. I was just making the point that the natural physics of the guitar will limit the advantage that fine control over the system gives you with a PIC controlled AGC.

I would still prefer to go with a full digital signal path if I was going down the PIC route. This would allow the fine control your describing over the phase as well as the amplitude. No fet would be required. It may also make it easier to use some of the 'digital' class-d chips out there in order to further reduce the parts count.

I don't think that'll be possible with PICs for a good while. the only option then is something akin to a Variax board - take your guitar signal, AtoD it using dedicated AtoD chips, a chunky motorola DSP & do it all in software ....end result? A board that's fairly big - oh, yeah...it'll cost you £200+ to make!

look here

The PICS in this chart have features like

40MIPS cpu, hardware 17bitx17bit multiplication, fast barel shifter, 16k RAM, 16bit architecture, A/D and D/A converters, 128k program memory.

Programming wise, I can certainly do software AGC, filtering and delay/reverb effects with this. probably some sort of multi-band processing although there might not be enough grunt to have enough bands with tight enough crossover between them.

It would be a lot of work though. And not much fun.

cheers

Colin

Link to comment
Share on other sites

Ok, I'm just about there ....but this last hurdle is proving to be a significant one.

What I've got ticked in the box thus far...

1. A Single string driver design (my inductance meter arrived from China yesterday ...allowing me to finally measure the string driver that I've having the most success with ...a 0.15mm variant - I was surprised to see that it has a surprisingly high inductance ....2.95mH. I will address this using with less turns, taking the DC resistance down a little, but for now it works)

2. A reasonable ballpark preamp circuit (it may not be totally optimum, but it's good enough for now)

3. Digital AGC (this one is proving to be a blinder - I'm getting visual feedback as to what's going on with incoming digitized string level readings as my digital AGC adjusts preamp gain to suit in in real time)

4. A poweramp chip that is 'ok-ish' (for now)**

My 'test guitar is a particularly bad strat 'catalogue' copy with high action ( my logic being if I can get it to sustain well, then that should cover most bases! Accordingly I'm using .009 gauge strings)

The Problem? - the variation in sheer amount of power needed at the driver, towards getting the highest fretted notes vs the open strings to sustain.

At the moment, for convenience of access, my driver is held in place *above* the stings, therefore open strings are closer to the driver then those fretted higher up the fretboard. Having tried driver placement everywhere, I'm sure those that have gone before me know that there really is only one good solid location that yields a constant, nice, clean sustain - that happens to be around the neck pickup region (as I move the driver closer to the bridge end, too many 'odd' harmonics are introduced, resulting in the loss of that bell like quality).

The problem with driver placement at the neck end, is that that's where the difference in action is at it's greatest! (& like I say this test guitar is a particularly bad actioned guitar - something in the order of 8mm-10mm action!)

As I recall, there's some equation the goes something along the lines that power needs to be applied logarithmically with distance (or perhaps what I recall is more something along the lines o f 'that sound pressure drops off logarithmically with distance...I digress!).

Anyway...how are others dealing with the seemingly *huge* variation in output requirements to 'excite' those string that are close to the driver & those that are far away? I'm struggling here to rationalise how this will even be possible & *not* have a current hog of a poweramp 'beast on board'.

I can only think of one solution at this early stage - have the AGC to become more 'intelligent'. At the moment, I simply set the optimum string level & if after plucking a string, the incoming level is too high , it starts adjusting a PWM duty cycle to suit (this is essentailly an altering DC output that I use to control a JFET)...like I say, within bounds it works extremely well...but when the action is too high, it can't get enough 'grunt' dispensed quickly enough. therefore perhaps what I need to do next is have some form of decision making based on how well the AGC is 'wrestling control' of the string.

Something like this....

(Deviation = a 'number' respresenting the difference between incoming level & a user preset optimum level')

1. String gets plucked

2. String gets AtoD'ed

3. PIC assesses this incoming string level (sample 1) to establish the 'deviation' number.

4. PIC Adjusts preamp gain circuit up

5. Again PIC assesses this incoming string level (sample 2) but then calculates how quickly this deviation is narrowing in percentage terms

"If the deviation hasn't narrowed by at least X%...then start bringing in some major artillery at the power amp end"

Whereas, at the moment, I have this...

1. String gets plucked

2. String gets AtoD'ed

3. PIC assesses this incoming string level to establish if it's above/below optimum (as it goes, it'll always be above just after plucking, but I've put the 'iff too high' condition first so it always acts on this aspect first)

4. PIC Adjusts preamp gain circuit down

5. Again PIC assesses this incoming string level - establishes if too high or too low

6. Adjusts preamp gain again.

& continue

(it's a little more refined than I've explained, as there's the concept of initial coarse adjustment then finer adjustment as the incoming string gets closer to the optimum level)

So, things still to do...

1. Solve this big power 'window' problem.

2. Make something that'll hold a few of my smaller drivers (I'm thinking here that probably three drivers might be a good balance between component count, power drain & yet still give very good control)

Step 2 is the only fly in the ointment ...& the subsequent test after I have the 'holder' it's likely to be a Biggie - ie inevitable ineraction between adjacent drivers. If it's too bad...hey ho, nothing ventured etc...I'll just then hone in on a six string driver! (watch out Pete...I comming at ya! :D )

My new Class D Power amp chip arrived in the post this morning (MAX9700) ...it won't need a whole lot of solder (& probably best I stay off the coffee for a day or two before I try!)

max9700.th.jpg

Edited by Hank McSpank
Link to comment
Share on other sites

Anyway...how are others dealing with the seemingly *huge* variation in output requirements to 'excite' those string that are close to the driver & those that are far away? I'm struggling here to rationalise how this will even be possible & *not* have a current hog of a poweramp 'beast on board'.

My Idea of "dynamic range inversion" thought up yonks ago was intended to deal with this problem and it WORKS.

Crap name, but thats what it does (over the desired range) high inputs map to low outputs and low inputs map to high outputs. There is a flat area where maximum output is retained over the first half or so of the desired input amplitude range, then the output ramps down as the input is increased.

The only trouble I have is that its hard using analog components to have enough control over thresholds and slope of response curve. You shouldn't have any trouble with this. I want the size of the flat area of maximum output to be longer.

Basically, the idea is that once the input is loud enough to register, but is still a low level, you should be giving it maximum drive.

As the input level increases, the drive is reduced. At the point where the input is as big as you want it, the drive should be minimal - just enough to keep the string going, but not enough to accelerate it.

This way, the weaker areas of response get more drive for longer, and when they settle, they will be constantly driven harder.

Power efficiency is also improved using this approach. You need much more power to 'swell' a low level string pluck up to full sustain than you do to keep a very responsive note going at your desired sustain level. This approach shoudl supply just the right amount if its set up correctly.

This does work pretty will in evening out the response of various string/fret combos. There is still some variation in response, but not enough to be annoying.

The problem I had was that if I have my markII circuit set up sensitive enough to boost the low levels, its reducing the gain to early on the sustaining string. The sustain is good, and even, but not quite lively enough for me. This is the problem with setting the thresholds using analog components. If I could add another few op-amps, it would be easy, but the parts count is already too high.

PIC wise, probably the way I would go if I were you, rather than using calculations of 'deviation' and decision branches in the code (that seems to be what you were describing?), I would go with a look up table and some linear interpolation. That way its VERY easy to change the response. You could even have different modes by just switching look up tables. You can scale and bias the tables as well. So you just have table lookup with some very simple arithmetic.

Depending on your programming chops, you could write a little app on the PC to create the tables. Or maybe cobble something together using Excel.

cheers

Col

Link to comment
Share on other sites

BTW, I'm being a bit slow...I know why this is causing me so much grief - it's because my present AGC has to be set with an 'optimum' string level (& it does that wonderfully) The problem of course...the optimum level for one part of the fretboard is *not* the optimum level for another part! Doh. (I should be forgiven this very simple lack of foresight on the basis, that I've only just brought all my sustainer 'pieces together' ....you kind of have to make that journey to realise!) A bit like painting the floor then realising your in the corner with your back to the wall & can't get out of it without walking over the freshly painted floor!)

Basically, the idea is that once the input is loud enough to register, but is still a low level, you should be giving it maximum drive.

What you describes, sounds to me like a bog standard AGC approach with threshold ...ie cranking the drive when the incoming note is low & vice versa (or am I missing the salient point?!).

My problem isn't getting low notes up...but getting the sustainer to 'get a grip' on a normal strength plucked note (but one where where it was plucked where the action is high). In this scenario, the string comes in as a healthy enough signal (& my PIC AGC just thinks "Aha...here's an incoming note...& it looks like the incoming signal is healthy...I'll therefore go through my normal gain adjustment routine"...my program is slick, because the PIC can talk to itself :D ) , but because the action is so high & therefore the string that much further away, what it really needs is for the drive seen at the coil to ramp up *very* quickly, in order to get a 'grip' on that string before it fades past the point of no return.

This is why I was pondering the concept of monitoring how quickly the sustainer is 'getting a hold' of the string (ie the deviation I spoke of...if the deviation between 'optimum' level & present incoming level' is narrowing fast, then nothing to worry about...if it isn't than that must mean the action is high - and the string further away - therefore more firepower needed at the coil pretty darn quick!)

So, would it be fair to say that the biggest challenge facing DIY/Hobby sustainer designers...is perhaps not phase, not frequency, not the type of output chip or whether it's 0.15mm, 01.9mm, 0.27mm gauge wire, etc - but making sure the overall solution can cater for the chunky range of power requirements needed for the wildy varying scenarios between a high Open E (a lot of drive needed) vs low E string fretted towards the top of the fretboard (not much 'drive' needed at all)

I guess what the perfect sustainer needs, is a fretboard location & string dependent 'drive' circuit!!! (but how can we give it this info?!!)

I have to admit...for now I'm genuinely stumped as to the immediate solution. I reckon a look up table will only work if there's a 'unique' condition coming in at the input (eg if I could tell the PIC...this particular incoming note means the string has been plucked at a where the action is high - go get 'em floyd)....but the PIC doesn't have enough grunt to analyse so many note in real time to derive this info (eg was the action high or low action when the string was plucked)

I guess with a one driver per sting approach (Hex solution), I could have the preamp use frequency based gain - eg for an 'A' string driver, a 110Hz note *must* be an open string, therefore this is where the action is known to be high, therefore ensure more more gain is applied, right up to 440Hz at the top of the fretboard for the same string - less gain required). Then use an output stage correlating voltage to current. This could be done with discreet components.

Hey ho...needs a bit of thought...ideas welcome!

Edited by Hank McSpank
Link to comment
Share on other sites

Basically, the idea is that once the input is loud enough to register, but is still a low level, you should be giving it maximum drive.

What you describes, sounds to me like a bog standard AGC approach with threshold ...ie cranking the drive when the incoming note is low & vice versa (or am I missing the salient point?!).

You might be missing the point, I'm not sure.

With a 'normal' AGC - e.g. a compressor or a limiter, the gain is reduced for higher input levels (and/or increased for lower input levels).

The difference is really one of scale.

A normal compressor/limiter tries to ensure that the output stays at a similar level over a certain range, or that the dynamic range is reduced - so low level inputs produce output that is still lower than high level inputs just not as much lower...

What I'm suggesting for the sustainer is that the outputs for a high level input signal should be lower than the outputs for a low level input signal - the dynamic range is inverted.

Another way to look at this is that the driver and pickup become part of the feedback loop for a standard compressor limiter - where it is trying to achieve an even output level, but this signal should be even at the pickup rather than some where on the circuit board.

To achieve this, there are two important factors:

You need a powerful amp & driver so that unresponsive strings can get loads of drive. (More responsive strings will quickly have their drive reduced)

You need the response curve of the AGC tweaked so you get just enough sensitivity to low level signals, but at the same time, the threshold where the output drops off rapidly for higher inputs is in the correct place.

If you can get this set up correctly, everything else just works.

You get a sustain with a very even level over most of the fretboard.

There will be some variation - weaker areas will have sustain that takes slightly longer to kick in.

Some parts of the neck will tend to bloom harmonically while others don't...

But generally, this approach to AGC will automatically account for differences in amplitude response due to: action, string gauge, phase difference, pickup frequency response etc.

...but because the action is so high & therefore the string that much further away, what it really needs is for the drive seen at the coil to ramp up *very* quickly, in order to get a 'grip' on that string before it fades past the point of no return.

With the AGC set up as I've been describing, as the string fades, the driver output will increase to give as much power as the amp can supply.

If it still can't 'grab' the string, then there's no kind of AGC that will help you. You need either a better power amp. Or, for that particular note on that string, your systems phase response is getting far enough away from optimal that it's not providing a constructive drive.

cheers

Col

Edited by col
Link to comment
Share on other sites

Hello, I made this circuit with tda1015, works like tda1011 but its consumption is lower, I'll try the 2 CAG,

to test which one works best, using its own preamplifier.

sch176.gif

2CAG.jpg

Hank McSpank,

D-class poweramp that I tested did not go well with the coils, due to its instability.

tda1015 needs no more to works.

cheers

Zfrittz6

Link to comment
Share on other sites

Ok, I've sussed my problem (though haven't been able to implement the fix yet)....sometimes you can get so close to the wood, you can't see the trees!

It's my mistake...due to my fixation on tweaking the AGC so much, I overlooked the fact that I'd tapped off the AGC feed from the wrong point in the overall signal flow chain!

This messy sketch of my block diagramexplains it all...

blockc.th.jpg

Essentially where my AGC is (soon to be 'was'!), it's not monitoring incoming the incoming input signal level...it was purely focusing on making sure the output level of my 2nd stage preamp always remaning the same (& a damn fine job it does of that too!). I now need to move AGC's the tap off point to be just after the first preamp stage ...the AGC's control voltage still feeds to the second stage to alter the gain to suit. It's been a bad day in McSpankland!!!

zfrittz6...that's not an obvious choice in chip - it's fully integrated ...ie a preamp & output amp are combined...its input impedance is way too low to attach to a magnetic guitar pickup directly...it'll suck come life out of the natural guitar signal. You'd need to front end that chip with a buffer....& hey, if you're going to need to make a buffer, then why not just go the whole hog & make it a preamp (dispensing with the integrated monolithic solution.. It also looks a little over spec'ed power wise (ok, so if the power isn't needed, it won't be used...but that's a little akin to going to Wal-Mart in a Formula One car!) & the external component count is quite high too.

Point taken about the Class D - there must be a way of getting it to work becuase the Sustainiac uses class D (if not, no matter... I've received about 10 different Class D samples....so they're all free anyway!)

Edited by Hank McSpank
Link to comment
Share on other sites

Some thoughts about fizz, squeal, grunge etc.

Guitar pickups are designed to sound good to our ears, not to function well in close proximity to a sustainer driver.

Most guitar pickups sound the way they do partly due to their resonance properties. The resonance is caused by the relationship between the resistance, the inductance and the capacitance of the pickup.

Most guitar pickups have a resonant frequency between 2kHz and 5kHz (Humbuckers closer to 2k single coils closer to 5k)

This next bit is based on intuition, so may be completely bogus, but I'm interested to hear if anyone has any thoughts on this.

Most of us have experienced squealing feedback with our sustianer experiments. I reckon that the squeal is likely to be at the resonant frequency of the pickup.

I also think that the pickup is going to be MUCH more sensitive to EMI at frequencies near its resonant peak. If so, this is bad news for us because those frequencies are not required for good sustainer performance. However, any atonal harmonics and noise caused by clipping or other distortion that fall near this frequency are going to be magnified by the pickup, and will be heard as fizz etc.

I've been thinking about solutions, and the most sensible and likely to succeed IMO at this stage is a passive notch filter placed between the circuit and the driver. Not sure yet what that will do to the phase response - I'll have to go and work it out. A low pass would be simpler, but might not be good enough at rejecting the target frequency.

Edit - changed my mind about the notch filter...

Any ideas about a solution for this ?

Maybe a bandpass in the LM386 feedback, although that might be unstable.

(FWIW, The resonance caused by a driver as used in the designs on this thread combined with a 100u - 220u cap falls just where we would want it. at or near about 300Hz - yet another reason why Pete ended up where he did after his many many iterations)

So any thoughts on this?

cheers

Col

Edited by col
Link to comment
Share on other sites

re soldering pins that are only 0.65mm apart ....well, for a start if you've shaky hands, just flood all the pins with solder, then use copper braid to get the excess off thereby removing 'bridges' (the capillary action of the braid will lift the solder off...except the solder that's underneath the pins) - there are plenty of tutorial youtube videos about, you can get the idea here...
(from about 3m46sec in).

Another method is to buy some solder paste (eg http://cgi.ebay.co.uk/SOLDERPLUS-LEAD-FREE...A1%7C240%3A1318 - you normally need a syringe, but you could get away with dabbing a needle into the container & then dabbing it on to the pin pad) .. place a small amount on the pin pads, mount the chip (the paste holds it in position) & whack it in your oven (got to use the right temp!) - the paste morphs into solder - no bridges with this method.

I've been in electronics manufacture for the last 10+ years and currently function as electornic assembly technical expert and trainer. Flooding and wicking might get the job done, but I would advise against it and use the paste and hot air technique or look into getting a tip designed for drag soldering (it holds liquid solder in a reservoir that dispenses a small amount of solder upon contact to each pin) as every time you heat the component and the solder joint, reliability of both component and PCB are reduced. One, because you increase the thickness of the intermatallic layer (junction of the solder joint and the PCB land) which means embrittlement that leads to fracturing and two because you're applying another round of extreme heat.

Once you start using some of these more elaborate devices, you guys should also start considering an ESD-safe workstation to minimize the chance of rendering your devices nonfunctional due to electrostatic discharge. The basic setup would consist of a dissipative mat with a path to ground which is also common to any tools being used and the individual via a bracelt and ground strap.

In general, the device datasheets should indicate an ESD sensitivity level. If it references ANY level, you should take heed in order to not waste your devices. The industry spends a lot of money on protection from electrostatic discharge prevention and this increases avery year as devices get smaller and more complex (weaker with regard to static shock). Nothing worse than trying to troubleshoot something that comes down to internal device failure due to an ESD.

Checkout www.esda.org and give S20.20 a read if interested. I can also recommend very cheap ways of setting up the workstation for this.

Link to comment
Share on other sites

I have 1 test

- CAG bs170, tda1015 6V. single coil 8 Ohm in neck

1,2,string good sustain, not works AGC.

3,4,5,string good sustain, works very well AGC.

2 test

- CAG bs170, tda1015 6V. single coil 8 Ohm in position medium

not works.

3 test

- CAG bs170, tda1015 6V, double coil 8 ohm in neck

1,2,3,4,5,6, string good sustain, not works AGC.

4 test

CAG bs170, tda1015 6V, double coil 8 ohm in position medium

1,2,3,4,5,6, string good sustain, not works AGC.

this one CAG works only when there is more power, tomorrow ,testing the other CAG

Hank McSpank, tda1015 input impedance is 100k,enough to EMG81.

tda1015 6V.= -1W

datasheet

The TDA1015 is a monolithic integrated audio amplifier circuit in a 9-lead single in-line (SIL) plastic package. The device

is especially designed for portable radio and recorder applications and delivers up to 4 W in a 4 W load impedance. The

very low applicable supply voltage of 3,6 V permits 6 V applications.

Special features are:

· single in-line (SIL) construction for easy mounting

· separated preamplifier and power amplifier

· high output power

· thermal protection

· high input impedance

· low current drain

· limited noise behaviour at radio frequencies

Supply voltage range VP 3,6 to 18 V

VP = 12 V; RL = 4 ohm Po typ. 4,2 W

VP = 9 V; RL = 4 ohm Po typ. 2,3 W

VP = 6 V; RL = 4 ohm Po typ. 1,0 W

Input impedance

preamplifier (pin 8) |Zi| > 100 k

power amplifier (pin 6) |Zi| typ. 20 k

Saludos.

Edited by zfrittz6
Link to comment
Share on other sites

Ok, having failed to get my sustainer to yield a constant, predicatable, grip of 'distant' strings - I went back to basics tonight - I took the AGC temporarily out of circuit, just a simple two stage preamp (on non inverting, into an inverting)...feeding into the TDA7052A & a single string driver coil at 8 ohms 1.2mH ....I used a sine wave into the preamp (& tuned the guitar until it was totally resonant with the sig gen) ....then tuned for maximum smoke :D the bottom line is, my particular power stage combination is not up to the job of getting enough grunt out to the thinner strings when they're more than 4mm away from the driver.

by this I mean if I place the driver 5mm+ away from, say the the G string, the TDA7052A does not have enough firepower to get the string moving sufficiently with an 8 ohnm coil (4mm is its 'distance' limit ....I'm using different sized drill bits to establish the gap between the string & the driver coil!). I'm summising here that this is because I'm only using a 5V single supply in combination with an 8 Ohm Coil. I scoped the signal across the coil...it was 3V max (if I put any more into this chip, the output gets ugly)...no matter what I tried, I couldn't get anymore out of this combo....if my TDA is limited to 3V peak to peak output @5V rail, then to get more grunt, I either I need to take the supply voltage up, or the impedance of the coil down.

My preference is to take the coil's resistance down to 4 ohms (but then doubling up the wire gauge to maintain the winding count at about 150 turns), so that's what I'll be doing in the next night or two. (it's becoming clearer why most Sustainers use 9V supply & have a discreet push pull output stage!!)

It really is going to be a very tricky task to design a sustainer circuit that can cater for the (relatively large) extremeties of power needed to excite from say a 'top fretted' bottom E string thru an Open Top E string (at least certainly for a guitar where the action is high!!)

Edited by Hank McSpank
Link to comment
Share on other sites

make the coil, works very well, wire 0.13 to 0.18, a coil as close as possible to the other

DISPOSICIONDELASBOBINAS-1.jpg

Do you already done so?

tda7052-circuit.png

tda7052-1watt.gif

(Circuit designed around a single TDA7052 IC. Ideal part for larger projects. Drives 8 Ohm speaker (not supplied). Ideal for battery powered circuits. Operates on 3-15VDC )

Link to comment
Share on other sites

zfrittz6 - my driver coil is different to everyone else's - I'm using a small single string driver...I find this is extremely useful for flushing out specific issues, because I can hone in on a string at a time. What I've now established, is that with a single ended 5V single supply combined with 1x 8 ohm driver coil, it's not possible to 'excite' a string sufficiently when the string is more than 4mm distance away from the driver.

Having slept on this, I have a few plans of attacks...

1. Increase the VCC of the sustainer cct (I'll take this up to 7V...as then we're into 2 x Li-ion batteries in series or midi guitar +7V supply territory)

2. Reduce the impedance of the coil (probably split the difference & go for 6 ohms))

3. A chunkier output IC (2W?)

& if those two don't work, then

4. Start thinking about having two coils 'rowing together' in parallel

I'm therefore thinking of resurrecting my TDA7053A, if steps 1, 2 & 3 don't work...then feed the TDA7053A the just one mono signal into both strereo channels .....& then out to two separate coils. This ought to put a *lot* more firepower at the business end on demand

A big win here, is that the TDA7053A has a more flexible logarithmic DC volume control (vs the TDA7052A) - so rather than mess about with JFETs - which, even though work admirably with my PIC ...they're not linear within their ohmic region (you can linearize them, but then you lose a lot of dynamic range - http://freespace.virgin.net/ljmayes.mal/comp/vcr.htm )...which makes for somewhat fierce fluctuation in gain. It makes sense to try & take advantage of the hard work done by the chip manafacturer!

BTW In a moment of madness/frustration/puzzlement I did at one stage last night revisit the LM386 ...it's a truly awful little piece of junk. Bearing in mind there's very little cost difference between it & other poweramp chips...based on what I saw on my scope, I'd urge everyone to avoid it like the plague!

Edited by Hank McSpank
Link to comment
Share on other sites

3. Start thinking about having two coils 'rowing together' in parallel

I'm therefore thinking of resurrecting my TDA7053A, if steps 1 & 2 don't work...then feed the TDA7053A the just one mono signal into both strereo channels .....& then out to two separate coils. This ought to put a *lot* more firepower at the business end on demand

I think what zfrittz is suggesting here is that running two coils connected in parallel from the same amp will give you more drive for the same DC resistance and effective inductance - two 2.4 mH (in || = 1.2mH) coils taking half the current each will give you more drive than a single 1.2mH coil

(half the current through 4 times as many coil turns = double the drive, although, in reality, I think the gain will be less due to coil coupling effects etc. I'm not so sure that his suggestion to have the coils as close together as possible is correct)

A big win here, is that the TDA7053A has a more flexible logarithmic DC volume control (vs the TDA7052A) - so rather than mess about with JFETs - which, even though work admirably with my PIC ...they're not linear within their ohmic region (you can linearize them, but then you lose a lot of dynamic range - http://freespace.virgin.net/ljmayes.mal/comp/vcr.htm )...which makes for somewhat fierce fluctuation in gain.

It makes sense to try & take advantage of the hard work done by the chip manafacturer!

BTW In a moment of madness/frustration/puzzlement I did at one stage revisit the LM386 last night...it's a truly awful little piece of junk. Bearing in mind there's very little cost difference between it & other poweramp chips...based on what I saw on my scope, I'd urge everyone to avoid it like the plague.

I think the LM386 is very sensitive to the load its driving. Did you use a nice big output cap and a zobel network in your test to keep the impedance it sees at its output more constant?

My MarkII circuit produces clean sustain on all strings, using an LM386. Petes system produces clean sustain using an LM386.

EDIT looking at the datasheet, you can see in the charts that increasing the impedance to 16ohm will increase the distortion levels dramatically, even at lower output levels. A 1.2mH driver is going to present an impedance of over 16ohm for any signal over about 2kHz. Without a zobel or some other way of keeping the output impedance down at around 8ohm for higher frequencies, there will be lots of distortion.

A similar problem occurs if your output cap is too small, but the distortion appears on low frequencies rather than high ones.

As far as action, you are correct, its a big problem. I like to play with a high action and really play hard - it sounds way better - but that is a problem for sustainers. It's always going to be a problem though, no matter how powerful your amp is. The magnetic field drops off with the square of the distance so the difference between 4mm and 5mm is significant.

The ideal sustainer guitar would have a short scale length, no trem, heavy gauge strings, low action and crap pickups (that have no 'character' due to a very poor resonance.)

Don't forget the final purpose is to play the guitar. You don't need or even want exactly the same response on all strings, as long as there is some type of useful response on most parts of the neck, it is a worthwhile addition. Surely no-one wants or needs sustain on the 20th fret of the low E string. Poor sustain on the open high e string is also not really a major issue, just get that note elsewhere if you want it to sustain.

I'm more concerned about how it sounds in the places where it does work, and how it responds to playing.

e.g. does it choke the natural sound? will it kill natural and pinch harmonics? does it produce a lively sound or a boring sinewave? what happens when I play chords? how fast does it respond to different types of playing?

here's an example of the kind of thing that might increase playablility:

using your PIC, what about sensing the strength of the initial attack, then producing a strength of sustain depending on that ?

So if the player plucks gently, the resulting sustain will be a soft low level one, while if they hit the string hard initially, they get a full blooded drive sustain.

This is the kind of thing thats doable with an analog approach, but not practical.

cheers

Col

Edited by col
Link to comment
Share on other sites

Hi Col,

I'm not suggesting the LM386 is not capable of producing clean sustain, but what I'm seeing on a scope across the driver coil is downright ugly (since you ask, I was using a 330uf cap in series with the coil - that was the largest I had to hand). What my line of thought is....is that *if* the poweramp stage is adding distortion (not meant in the more common 'grunge/clipping' sense, but more just 'changing the signal' vs the original)...then overall, it's going to be a struggle - because most JFET based AGC circuits are also adding their own little bit of JFET based distortion into the mix ...therefore adding these two 'distortions' together results in a final signal appearing across the driver that's different by some margin to the the original incoming string signal. If it's not an exact copy of the original signal, then this can only mean the sustainer have to work that much harder at 'exciting' the string (meaning less efficiency = more battery drain)

Re the PIC....what you say is do-able...but the problem is the initial signal transient varies wildly (I can see the transient levels in real time on my PC screen as reported by the PIC) - in my opinion, AGC's really needs a less volatile input signal (else you'll only end up with more 'hunting'/pumping)...also, the transient 'fade time' varies greatly between strings - a low E's transient doesn't appear to fade as fast as a high E - therefore it'd be hard to base an AGC 'gain needed' decision on these initial transient levels .....far better to wait a portion of time (say 200ms), for the string to establish a more meaningful 'average level find its 'average level - & then have the AGC act...but the problem then of course is that there's much less time for the sustainer to wrestle control of the string before it fades past the point of no return! The 'dynamic' sustainyou're proposing...is a nice touch, but I'll lleave that for my MKII version (I'm not even at MK 0.5 yet!)...that aspect is probably just best left to an 'intensity control' pot for now!

A predicatable, controlled AGC for high action'ed guitars really is one outrageously tough balancing act!

I'm now searching for a chunkier output poweramp - & the TPA0252 - http://focus.ti.com/lit/ds/symlink/tpa0252.pdf - now looks like the latest contender (now winging its way to me)

Benefits include...

Digital Volume Control (controlled up/down via pulses ....this will dovetail well with my PIC signal level analyser...therefore, I'm hoping this digital volume control will make for a good, lazy man's AGC!)

A much chunkier 2W Stereo capability (therefore potentially driving two separate coils).

A low 5V supply (PIC friendly)

A Low current draw @8mA (& only 150uA in shutdown!

Muxed inputs (therefore I can feed both inverted & non inverted guitar signals into the chip & have either the PIC or a simple SPST switch select between the two - that ought to make for a very & easy harmonic mode switching solution!)

Very low external supporting component count.

It remembers where the Vol control was last set after coming out of shutdown mode (useful for making sure the volume level is 'in the ballpark' right out the starting gate)

A low THD @0.3%-ish.

Very Small!! (though you might view this as a problem if you're put off by SMT!)

Cons include...

Quite costly in comparison to other poweramp chips.

Very small! :-)

Edited by Hank McSpank
Link to comment
Share on other sites

Cons include...

Quite costly in comparison to other poweramp chips.

Very small! :-)

-40 to +20dB in 31 discrete steps ? it this going to be enough control granularity? yeah, probably.

But you're right about the size putting me off - and many others I would guess.

I would also like to get a more powerful output stage, however, I really want to avoid SMDs.

A class-d output stage might convince me, but it would also mean going back to the drawing board on the AGC.

A THAT corp based AGC with a nice filterless class-d on the output could be really nice (and expensive)

Anyhow, I'm looking forward to hearing how you get on with the new IC, so good luck with that.

cheers

Col

Link to comment
Share on other sites

A THAT corp based AGC with a nice filterless class-d on the output could be really nice (and expensive)

Col

I'd considered a THAT VCA...but I've not quite wrapped my head round how it'll integrate with the Virtual Earths we're using?!! (also these THAT VCAs will definitely need a 9V supply, as they need a +ve...earth...-ve & the minimum spec for these is 4.5V...not sure what happens to the VCA when the battery life starts fading away?!)

I've all but written off controlling the JFET with my PIC...sure, it works, but the JFET is *not* linear...meaning if my PIC increases the gate voltage by set 'predictable' steps in voltage, I don't get predicatable 'steps' of JFET resistance change - this plays havoc with the gain structure of my variable gain 2nd stage preamp. For example, if the signal coming in is below 'optimum', the PIC increases the gate voltage by a certain amount.... the preamp gain then increases by a certain amount, this feeds into the power amp & I get an increased signal across the coil by the required amount. However, after few more steps of this (ie analysing the input, increasing the gate voltage), the JFET resistance then leaves it's linear range' & therefore the its Source-Drain resistance is not proportional to the predictable 'voltage steps' I've applied. The gain of the opamp then whacks up...this can result in a whacking great increase in signal fed into the poweramp stage, ultimately vibrating the string stronger in almost 'step up' fashion, the resulting larger guitar signal then feeds straight back into the sustainer circuit, where the the AGC thinks "Whoah, DECREASE THAT GAIN SHARPISH!!!)...so this all results in it taking longer to 'lock' the AGC gain at optimum (also hunting/pumping enters the fray). Sure, there are articles which show how the JFET can be linearized, but it seems you then lose a large portion of the controlable JFET range....meaning ultimately the gain range of the variable gain opamp (ie dynamic range) is hampered ...I reckon we need to control of almost the full rail-rail range of incoming signal.

So, (stating the obvious) unless we have a totally, predictable linear AGC, then it'll be a real struggle to get our Sustainer 'tuned' well.

Also, I've not quite wrapped my head whether we need a logarithmic gain structure or linear for the circuit driving the coil. I suspect we need logarithmic (on accountwe hear logarithmicly)....if so, this is even more reason to use the THAT type VCAs (or an output stage with a logarithmic volume control).

Does anyone have anything more definitive on this last aspect?

Edited by Hank McSpank
Link to comment
Share on other sites

Also, I've not quite wrapped my head whether we need a logarithmic gain structure or linear for the circuit driving the coil. I suspect we need logarithmic (on accountwe hear logarithmicly)....if so, this is even more reason to use the THAT type VCAs (or an output stage with a logarithmic volume control).

Does anyone have anything more definitive on this last aspect?

We hear logarithmically, but the magnetic pull of the driver varies linearly with current (I'm pretty sure this is correct) so the string 'hears' the driver linearly (and the pickup hears the string linearly). That meant that for 'grabbing' the string, linear control would be better IMO

As far as linearising the jfet, its possible to fully linearize it an lose much dynamic range, or linearise it a little and lose a little of the range. Getting the compromise right between linearisation, dynamic range, response time and low frequency distortion is the real trick, but assuming the output amp is good enough, I think that a jfet can do the job. It might take two jfets to get closer to an ideal system, but its certainly not an unnatainable target.

cheers

Col

Link to comment
Share on other sites

Guest
This topic is now closed to further replies.
×
×
  • Create New...