Wednesday, February 11, 2015


This little project came to me while i was cleaning out the basement. Heidi and i are moving soon, and i have been trying to liquidate as much as i can before we go. we have lived in this apartment for 10 years now, so our basement was full of years of postponed projects and toys that i have collected. i have actually done pretty well at cutting my collection down to bare necessities. i did however end up filling one box with things that i thought i should maybe sleep on before deciding to get rid of them. in the box was a fisher price guitar from the late 80's. it had been so long ago that i had added it to my collection, that i wasn't even sure if i could remember what it sounded like. i do remember having the keyboard version that was sold around the same time. i remember that keyboard having particularly terrible sound. for whatever reason, i ended up scrapping the keyboard, but kept the mainboard over all those years. i'm not sure why, but now i'm glad i did. when it was finally time to make a decision on whether or not to keep the guitar though, it occurred to me to maybe plug it in and have a listen first. i did, and the sound was terrible! so terrible in fact that i decided to open it up and see why. well, one thing led to another, and i decided that i probably had enough time before we move for one last project. 

Upon opening the guitar, i quickly found that the primary reason the sound quality was so bad was because of the way the voice was generated. the guitar's voice chip sent 8 digital parallel outputs to a shift-register/ resistor array, in some sort of external DAC(digital-to-analog converter) configuration. the digital outputs were also being sent to the keyer/programmer chip before being sent to the shift registers to be mixed. kind of seemed weird to me, but i guess it kind of made enough sense to move on. i tried to find some kind of envelope control for the guitar voice, but it seems that that was being generated digitally too. so that's why it sounds so terrible! well, there would be no 'cleaning-up' the sound here. instead i focused on that DAC business. mixing in the raw digital signals to the audio path made for some great distortion. i also thought it might be cool to send the digital signal through a PWM(pulse-width-modulator) circuit, as kind of a highpass filter. the PWM sounded great, but none of the combinations of the digital outputs and the analog output sounded very good together, so instead of making a mixer section for the 9 different output options, i just picked out the 6 most harmonic sounding digital signals and put them on a six-position selector switch. each of the six 'timbres' have unique harmonics. especially through the PWM circuit. on one end of the spectrum, the voice is loud and has rich harmonics that carry. on the other end of the spectrum the voice is total 'dial-up' noise. then there are four varying levels in between. 

The guitar originally had three strummer paddle switches. to play the guitar, you would hold a note or chord on the fretboard, and strum the paddle switches like strings. when playing a single note, the three paddles all strum that same note. when playing a chord, the paddle switches play one note in the chord individually. i wanted to replace the strummer paddles with an LFO(low-frequency-oscillator), but i also wanted to maintain the chord strumming option, so i built a 3-step looper with a CD4017 chip to cycle through the notes sequentially. i added individual switches to the outputs of the CD4017, so that any or all triggers from the looper to the strummers could be turned on or off in any combination. i also added a big arcade switch to act as a manual trigger for the first string, or the root note of the chord. i may end up going back in and changing the manual trigger to the second string though. i later found out that string two is responsible for advancing the melody recorder in playback mode. more on that later.

at this point the guitar was sounding really a noise-loving kind of way. the sound still needed something though. i didn't want to go too crazy on this project since there wasn't much time, so i decided to add a PT2399 echo circuit to the mix. it did a great job at taming those harsh digital timbres. after being totally satisfied with the sounds i was getting from the circuits i had built, for some reason i really wanted to hear those digital noises through a lowpass filter. i ended up stripping out three buffer/filter stages from the original signal path, since i wouldn't be needing them, but there was one of the four op-amps of an LM324 that was still needed for the weird power switch circuit in the guitar. this meant that i had 3 left over op-amps to use, so i went with a simple 2-pole vactrol based lowpass filter. this is where things got stupid... while i was reverse engineering the mainboard of the guitar, i was constantly flipping the PCB(printed circuit board) around to see the tracings on the bottom, and the components on the top to draw out the circuit to better understand it. i did a great job of drawing everything out, but for some reason i reversed the power pins to the LM324 in my drawing... you would think that after all of these years of using LM324's, i would know by now which pin is positive, and which one is negative. thanks dyslexia! anyway, i built up the filter exactly how i had drawn it into my guitar schematic. it wouldn't work??? no, why would it? i tried a million things to try and figure out why every time i turned the power on to the filter, my voltage would drop. i tried different filters, different power supplies, different voltages... at one point i thought that the regulator to the guitar circuit just wasn't capable of sending any extra current to the filter, so i decided to give the filter 12 volts to see what it would do. as i was pulling a jumper carrying 12 volts over to my breadboard, the  jumper sprung out of my hand and landed directly on one of the pins to the guitar's voice chip. from then on it didn't make a sound. all of the other pins were still fine. they all did what they were supposed to, but this pin gave me nothing. it was the pin that sent the 'enable' signal to the DAC's shift register, so no sound would come out without it. DAMN! this is when i remembered that i had kept the mainboard to the keyboard version. i dug around and found it, and thankfully the voice chip was a match. i de-soldered the old one and soldered in the new one and it worked again. it was then that it occurred to me to check the polarity of the op-amp. it was backwards. i swapped the polarity and the filter worked perfectly.

Once the circuit was finally working, i quickly drew up some faceplates for the guitar, and started tooling the enclosure to make room for all the new hardware. after that, the enclosure would need a substantial cleaning. with the exception of some coloring-crayon stains here and there, it came out pretty clean. i then drew up a little circuit board design for all of the additional circuitry, and engraved it with my CNC. i did have a little bit of trouble with the PCB though. i forgot a few components in my 'bill of materials'. when i used all of the components that i had listed, i just assumed it was done... oops, forgot a couple. eventually it all worked out great, and assembly was pretty smooth. after everything was built, and the guitar was all closed up, i started playing with it. not only did everything work exactly how it should, but i also realised that this guitar has a pretty BAD-ASS SEQUENCER! i knew that it had a record and playback function, but i didn't end up trying it out until everything was said and done. the guitar has a 'record' button that puts the guitar in record mode. in record mode, the fretboard is used to record up to 39 notes without the need for the strummer switch. in playback mode, the second string plays the sequence one step at a time on each strum. since i wired up the strummers to an LFO, the rate of the sequence can be played back independent of the guitar's internal clock(pitch). at the end of the recorded sequence, it restarts at the first step and plays through continuously, and there is no noticeable gap between the first and last step! very cool. i am very glad i didn't just give this little guitar back to goodwill. 

This time around, i didn't film a video.. i tried, but i wasn't able to get good footage and play cool stuff at the same time. my camera skills are just too horrible. instead, i just recorded some audio and put it to a slideshow. please let me know if this bothers anyone. just makes more sense to me.

Friday, January 23, 2015


Ahhh... internet at last. been having trouble with my connection for the past week or so. haven't had the opportunity to post. the NT03 is the latest of my NT(noystoise) series. i initially started developing the NT03 in early November of 2014. the first of the six was completed at the beginning of December, but a commission came my way, so i put the rest on hold for a month before finally finishing them all. unfortunately the commission fell through. thankfully it was only one month wasted. hoping to quickly recoup my losses, i cranked out the final five NT03's in record time(two weeks), only to find that my 3 month old CenturyLink DSL router was all dried out. one after another, i finally got a replacement that works(for now). now to catch up!

The NT03 consists of two square wave voltage controlled oscillators, a holtek voice modulator circuit, a voltage controlled filter(VCF), and a low frequency oscillator(LFO) that can either/or modulate the filter cutoff, gate the voice modulator transpose sequencer.

The two VCOs are made up of one 74LS124 dual VCO chip. i was able to get a pretty good deal on these chips some years ago, but i haven't really put them to much use. this felt like a good opportunity to use them. i also had a few holtek HT8950 chips that needed to be used. the HT8950 is basically just what you would find in one of those toy megaphones that turns your voice into a robot voice(all circuit benders should know what i'm talking about). the difference in the HT8950 is that in addition to being able to incrementally select the pitch setting of the modulator, there is also a three bit parallel input that can be used to select the chips' eight pitch settings; Robot, x2, x1.6, x1.3, x1, x.9, x.8, x.6. the sample rate(BIT RATE) is resistor set. initially i tried clocking the UP/DOWN inputs on the 8950 with an external oscillator, but found that there is a pretty significant debounce protocol built in. this makes it difficult to achieve faster transitions at lower sample rates, so instead, i built an eight step parallel sequencer from the CD4029 chip. this way the eight pitch modes can be accessed quickly with no debounce(latency/stutter). the sequencer can also advance in either direction.

After the two VCOs are mixed and sent through the voice modulator, they are then sent to a resonant 12db lowpass filter(VCF). the filter's cutoff can be modulated with an envelope derived from the sequencer clock, but the attack and decay of the envelope are preset. the filter cutoff, pitch of VCOs 1&2, and the sequencer clock rate are all controlled by one of four joystick axis. the 3D joysticks are a recurring theme in my NT series. this time i decided to use all white thumbsticks, just because they seem to go with everything. they do tend to get dirty though..


Each of the four joystick parameters can also be controlled externally, although the joystick still controls the maximum threshold of the incoming control voltages(CV). the frequency of the sequencer clock cannot be controlled with a CV, but it can accept incoming clock, or negative going gates. the sequencer clock can be triggered internally and/or externally. to advance the sequence internally, there is an 'ENABLE' switch, and a 'DIRECTION' switch. the DIRECTION switch controls the direction of the sequence with either internal or external clock advance input.

The NT03 also includes BIT-RATE control, resonance control(Q), VCF MOD switch(on/off), POWER switch, VOLUME control, LINE OUT, clock rate LED, and center positive 9VDC supply input jack. the NT03 draws about 60ma max, and can also be powered by internal 9v battery.

The NT03 enclosures are made from laser-cut plywood and laser-cut acrylic faceplates. everything is made inhouse(in my house) from scratch. when i finished the first NT03, i had originally intended on painting them all primary or secondary 'candy' colors. hence the yellow one. after some time away from the project though, i thought it might be nice to come up with some colors of my own. i ended up mixing all of the colors for the final five by eye. some are a clean wash, and some ended up with some inconsistencies. namely the olive green one that has my big fat finger print in it. i suppose i could fix it or change the color, but despite the imperfection, it still looks really clean. if no one buys it, i'll just have to keep this one for myself.


This project began in early December of 2014. it was originally a commission, but the deal ended up falling through, and i never got paid for it. hopefully i can find someone out there to buy it from me to make up for my losses.

The original concept was based around the Synare3 circuit. the Synare3 consisted of two CMOS oscillators, a white noise generator, a four-pole resonant lowpass filter, a voltage controlled amplifier, and two envelope generators that modulated the filter cutoff and VCA amplitude.

My task was to build a modern variation on the circuit that would enable CV/EXP. PEDAL inputs to the oscillators, filter, and VCA. this would mean replacing the CMOS oscillators with voltage controlled oscillators. i was also asked to add the ability to 'transpose' the oscillators frequency range, create multiple waveforms, add a 'crossfade' between one oscillator and the noise generator, add a line-in mix, and a pushbutton switch that would fully trigger the envelope generators.

with all of this on the list, i began designing a circuit that would achieve it all. in the end, the circuit was quite a bit different than the original Synare3 schematics i was able to find. for the VCOs, i used a basic dual op-amp design that produces both square and triangle waveforms, and has a pretty decent frequency range. transposing the VCOs would just be a matter of switching in additional timing capacitors. 

the white noise generator in the original schematic was not giving me very good results, so i ended up using the two transistor type that i have used in the past with the NT02 series. it works well, but the circuit requires at least 10vdc. since the buyer asked that the finished product be powered by a standard 'center-positive' 9vdc supply, i would also have to build a 'charge-pump' circuit to create the necessary voltage for the noise generator circuit. like the NT02 rev.b, i used a mirrored signal from 'oscillator 2' to drive the charge-pump voltage multiplier.

the filter in the Synare3 originally consisted of four separate 'OTA' 6db lowpass filters in series to create the four-pole filter. the circuit originally used the CA3080. in the interest of space and cost, i decided to use the dual package LM13700 instead. i felt that the fourth filter in the chain was not really necessary in this circuit, so i cut the filter down to three poles, and used the fourth OTA for the VCA. i was able to cut down my OTA needs from five single amplifier chips to two dual chips. also, i didn't really see the need for two separate envelope generators, and the buyer only wanted one decay control knob, so i only built one to control both the VCA and VCF. in the end, the only thing that stayed original to the Synare3 was the drum-sensor/envelope-generator circuit. the seller also insisted that the sensor be a 3 inch loudspeaker like the original Synare3. the speaker did have a more natural response(maybe), but when the deal fell through, i ended up switching it out for a piezo instead. the speaker was just not sensitive enough for finger tapping. i also ended up adding a lot of different modulation routing to the LFO, VCO, VCF, VCA, etc. the finished product does well as a drum synth, but excelles as a drone synth. plugging other signals in to the trigger input is especially fun. it can take LINE, MIC, INST, and other sensor inputs safely.

the buyer had also asked that the VCOs, VCF, and VCA have standard 0~5vdc CV inputs that doubled as expression pedal inputs. this was new territory for me. i didn't understand how you could send 5 volts to a jack in a TRS situation, and then just short that 5 volts to ground when you plug in a TS jack, but the buyer knew what he wanted, and it is common for newer moog and roland gear to have this option. after a couple of days trying to come up with some sort of current sensing switch that would disconnect the 5 volt supply when shorted, i ended up just asking to borrow someones moog pedal to see what they do to overcome this. ironically, the instructions on the bottom of the moog sent me to their website, where the manual with the description of the circuit is. all it really ended up consisting of was a dedicated 5 volt supply that was limited to 5ma per output. if one of the four was shorted to ground, the rest would not be able to power an expression pedal. if all four are connected to TS CV inputs, the circuit will draw no more that 20ma. since this thing isn't running off of batteries, it's not a big deal. so i was able to do it the way moog does. i just added a 7805 regulator with four separate 1k resistors coming off of it to the four CV/EXPR inputs.

the LINE-IN MIX jack was pretty basic. the buyer really just wanted to have the option of mixing another LINE-IN without the need of an external mixer. the LINE-IN has a little bit of gain too, so the signal can be summed from the external device's volume control.

the body of the synth is made from laser-cut wood, and the faceplates are laser-etched hand-cut sheet metal. working with sheet metal can be a pretty stressful when you don't have the right tools. my second try turned out much better than the first. i wasted a lot of hours, but it was a good learning experience.