ARP 16 Voice 3303 Electronic Piano

arp-piano-1
arp-piano-promo
arp-piano-under-hood
debugging-arp-piano

Last Update 02-11-2019


Defective Membranes
Defective Noise
Defective Mute
Derivatives
Initial Repairs
Gating Circuit
Noise Sources
Panel Controls
Sounds
Epilogue

The ARP 16 voice piano is one of the first electronic pianos designed to appeal to piano players, with a wooden keybed that was velocity sensitive.  “16 voice” refers to the number of presets, and the piano was fully polyphonic.  It could had been a banner instrument for ARP.  It followed the infamous ARP Avatar guitar synthesizer, which was such a market flop that the paltry sales failed to recover the development expenses and the failure put ARP in a very precarious financial condition.  Desperate for a product that would save the company, ARP rushed the unfinished 16 voice piano into production which soon followed with pianos coming back in for warranty repairs as fast as they were shipping them.  Two major faults doomed the piano: they were very noisy and the key contacts were hopelessly defective.  Today they are not very valuable as you can expect any of these to be broken and although the noise issue can be improved the labor cost will exceed the value of the piano.  Fixing these things is truly a labor of love.  By the time I completed a major restoration on my own unit (new caps, new CMOS, & modern low noise opamps), the sound was still not very good.

These were first introduced in 1978.  In the 1970s the competition for portable pianos was intense (digital/sampled pianos were still a decade away).  Portable keyboards that sounded like a piano were very desirable, and the reed pianos (Wurlitzer) and tine pianos (Rhodes) did not sound like a piano.  Electronic pianos back then were a better approximation although they lacked a velocity sensitive operation like a piano.  Most did not have a weighted action like the piano.  Two notable attempts were the Yamaha CS-80 and the Moog Polymoog, but neither emulated the varying timbre and envelope across the keyboard range.  There were actual pianos repackaged (some would say disguised) as a gigging instrument (IE Baldwin ElectroPiano, Yamaha CP-70B/CP-80, others), but they were far from portable and because they used actual piano strings (and iron frame!) they had to be re-tuned between gigs, which wasn’t fun for the gigging club musician.

Defective Membrane Switches
ARP approached the velocity sensing by using two switch contacts for each key.  There is a contact for key at rest position and another for key at full depression.  A velocity sensing system was derived by using a system to measure the time between both positions.  ARP chose to use a membrane switch system for the key contacts, which wasn’t a bad decision as they were very cost effective in a growing competitive market.  The membrane switches are normally open contact, which had to be converted to normally closed to work with the velocity sensing system.

Their key switch system used a roller wheel on a spring to apply pressure to the contact.  This closed the normally open contacts of the membrane switches, and it was novel enough that they received a patent on it.  Unfortunately when the pianos went into the field and were stored in the back of a car in the summer between gigs, the hot environment would cause the membrane switch in the key at rest position (which the wheel was pushing hard against) to melt in a permanently closed position.  This rendered any affected key to be completely inoperative.

ARP had to replace the key contact system as there was no way to remedy the fault, and the pianos coming in for warranty repairs were quickly eroding their finances.  They designed a palladium leaf switch replacement that had to be retrofitted to existing pianos in the field and in production pianos, although there are still pianos out there with the defective membrane contacts still intact.

Noisy Audio
The noise issue was really bad.  The audio path has multiple filters in series and parallel configurations that converted the raw square waveforms into a piano sound and other presets.  All of these filters were based around the old 4558 opamp.  Those 4558s were basically two ancient 741 opamps on a substrate and they were not low noise devices.  Chain enough of them from end-to-end and the accumulation of noise becomes unacceptable.  This noise was constant in the audio output.  Opamps with better fidelity simply did not exist back when this piano was designed.

If that wasn’t bad enough, ARP designed the audio chain on single rail circuits.  Typical audio circuits use dual power rails, usually +/-15VDC.  Opamps can operate off a single rail, which was common in guitar pedal circuits.  ARP was driven to this concept because of the CMOS ICs used for switching audio signals in and out.  There are compromises with single rail circuits 1) dynamic range is cut in half 2) circuits will be noisier 3) design approach changes with typical circuits 4) not every dual rail circuit will work with single rail.  With the dynamic range halved, slamming big chords can hit the opamps hard, driving them into phase reversal which imparted a nasty distortion on the audio which hopefully a power cycle can clear up.  ARP should had used a compandor to tame the noise and dynamic range to eliminate the phase reversal, a modification I plan to implement in the future.

Defective Mute Circuit
Contributing to the noise problem was a defective mute circuit.  This muted the output between power cycles and when changing presets.  This was standard design as the transients in the audio during these conditions would result in very loud noise, potentially damaging your speakers and/or your amplifier.  Unfortunately the design for power cycle muting was defective.  The rectified raw AC secondary voltage would feed through the circuit, then amplitude modulate the mute signal resulting in a constant 120hz harmonic riding on your audio output, sounding like the food processor in a kitchen.  This made the piano completely unusable.  I found the offending signal on my oscilloscope below:

debugging-mute-signal

I disabled the mute circuit as there was no hope of it ever functioning and it will need a replacement circuit.  Until that is built, I have to be very careful with power cycles because it sends nasty loud noises that can damage my amp/speakers.

Yeah this thing was far from a finished product.  If the Avatar wasn’t bad enough, the piano scared plenty of musicians away from ARP.

Derivatives
ARP later released the 3633 four voice piano as a lower cost product.  After ARP went bust, the four voice piano went on to become the Rhodes 3633 Four Voice piano, which did not fare much better.

arp-4-voice-piano
arp-4-voice-piano-promo
rhodes-4-voice-piano


The word from the Fender factory, which made the Rhodes pianos, was that the 3633 piano was very much the unwanted orphan.  No schematics are known to exist for the 3633 piano, although the keying/gating circuit boards appear to be identical to that used in the 330x.  Judging from the front panel controls and pictures of the circuit boards of the Rhodes Four Voice Piano, it has a much simpler audio chain that significantly reduced the filtering system thus reducing the noise.  It also eliminated the master VCF, the resonator, and the noisy phaser.

The reputation of the 16 voice pianos was bad enough that most players dumped them in the trash, and surviving specimens rarely came up for sale.  Indeed, after almost twenty years of being on the internet I had never seen one for sale.  I never saw one in a store.  Then at AHMW 2015, one of the attendants had brought one.  This member always had brought something interesting and different to the AHMW gathering.  It certainly lived up to its noisy reputation, but that wooden action really appealed to the piano player in me.  It was far more pleasant to play than most of the digital pianos at that time with action that was too stiff, a growing pet peeve of mine.

I left AHMW with no immediate desire to seek out one of them.  There were no known schematics on the internet and at the time I had enough projects for the bench on the back burner.  I was talking to my repair tech about these pianos and he said he used to own one.  He also lamented about their terrible quality.  Although he disposed of the piano long ago, he still had the original service manual complete with schematics which he gave to me as he had not seen one in for repair in a long time.  I scanned the manual and provided the digital PDF to my friend from AHMW (email me if you want the PDF file, it's a very large file).  Shortly after that, the very first 16 voice piano I had ever seen on eBay appeared, and I snatched it up.

Fortunately this one was one of the last pianos off the production line with the far better palladium leaf switch contacts intact.  During disassembly I found a date stamp of manufacture, placing it a few months before ARP was liquidated - the last of the production 16 voice pianos.  If you really are brazen (or maniacal) enough to take on a restoration - a true labor of love - look for the pianos whose model number begin with “3303”.  If you find models “3301” and “3302” then you risk having the flawed membrane contacts for which there is no remedy.  Here's what each contact system looks like:

Defective membrane contacts with wheel
Better leaf switch contacts
defective-contacts
better-contacts

For the contacts shown on the right, note the remnants of the wheel actuator on the hammer ends which indicated that it was a field retrofit.  They simply cut off the wheel extending from the end of the hammer.  With my specimen being one of the last ones made before ARP went bust, it did not have the remnants of the wheel actuator.

Initial Repairs
My specimen had zero audio output.  By the time I got audio to happen I had to replace several defective opamps.  I even found a defective capacitor that was not passing any signal at all.  You can't troubleshoot one of these without an oscilloscope (although the service manual suggests a handmade probe to use with an amp, I don't recommend it).  There are many many places in the audio chain for a defect to happen.  Then I discovered the power cycle mute defect and its not-so-wonderful 120hz ring modulation.  Probing around, I found many missing signals.  With the noise issue of the opamps I knew I was looking at a massive component replacement job on the voicing board.  Justifying this was other faults I had fixed leading me to question whether the entire audio chain for all the presets were working at all.  There were over 50 opamps to replace plus many many polyester capacitors which aren’t optimal for filter circuits (I had already replaced two defective polyester caps).  Today there are opamps available with far better fidelity and I knew capacitors well enough to know which dielectrics were best for filter circuits.  Here's the work before and after:

Voicing Board Before Rework
Voicing Board After Rework
rework-before
rework-after

Suffice to say, this was a Big Job.  Didn't I mention the job was harder as none of the replaced ICs were socketed?  The sole socketed IC is the 256bit OTP ROM (NOT a typo, NOT 256KB or 256MB!!!).  Besides the caps and opamps, I replaced ALL the CMOS switches (originals were RCA which are notorious for failing with age). Astute observers will note a remaining polyester cap; I had ordered the wrong value at the time the picture was taken, and the correct cap will be remedied in short order.  Since some interconnects were HARDWIRED, instead of piecemeal work of rework/assemble/confirm/rework/assemble/confirm/repeat I had to opt for rework of the entire board before assembling it back together.  All those caps, opamps, and CMOS.  Deep breath and... to my good fortune, it all worked when I finished!

I also reworked the resistor matrix board (pesky RCA 4051 ICs), the keyer boards, the power supply board, and the front panel board.  Luckily I had no goofs in my rework.

There was a rusted connector socket on the resistor matrix board.  A search found no direct replacement, so time for a home remedy... I found one of the connector pins to be redundant so I pulled out the socket for that pin and transplanted it to the faulty one - problem solved.

Oh, and application notes for the modern opamps recommend installing bypass caps on the power pins to reduce noise.  See all those tan colored caps I had to install on the solder side of the board?

Voicing Board solder side, Bypass caps added to all opamps
bypass-caps-added

ARP sprinkled "test points" to confirm signal presence at key circuit points.  Since the piano only generates percussive sounds, you have to repeatedly play a key to get a signal.  About 1/3 of the test points are at the virtual grounds of the opamps, which won't show anything!  Test point 24 was nowhere to be found until I turned the circuit board on the solder side - it is blocked on the component side by two resistors!  While I had this board removed, I installed pins in the USEFUL test points to ease placing scope probes on them.  Some of them are under the key contacts and are hard to access.

During the rework I uncovered either schematic errors or factory changes to component values.  Did I mention this was an unfinished product?

Initially there was an intermittent audio fault with a loud buzz, sometimes followed by degraded audio quality.  Since the mass component replacement I have not heard this defect.  It must had been an opamp on the hairy edge of failure that I pulled.  They will bother me no more...

Some keys were sticking, which were fixed when their bushings were cleaned up.  I had a few keys that didn’t trigger and they were a challenge.  One was due to corroded membrane contacts ("flex" cable interconnect, not the key switches).  Another was due to maladjusted mechanical placement of the key contact assembly.  The last was the biggest challenge.  This forced me to study the velocity sensing system; this was a fascinating and interesting circuit.  Its operation is not adequately described in the service manual and I drew out a better clarity description (PDF can be downloaded here).  It also proved challenging to decipher a key trigger fault, which ultimately turned out to be dirty contacts on the switch points.  A final defect was isolated to a bad flex cable contact, and as there are no direct replacement I am looking at replacing the interconnect system - more heavy work.

The mechanical key action is very interesting.  It consists of the key shank throwing a lever into motion with a lead weight at one end, with the other end moving the contact.  This system is the most accurate emulation of a piano action I have felt, and many piano playing techniques can be faithfully replicated here.  The velocity response is the most natural I have encountered in an electronic piano.  These keysets were built by Kimball or Pratt-Read.  Here's what the hammer action looks like:

hammer-system

Sadly, this hammer system would not carry over to the Rhodes Four Voice piano or the Rhodes Chroma.  Dammit!

Trigger/Gating/Rank System
The voices are derived from a top octave system (TOS) that has its roots in electronic organs and string machines that ARP is famous for.  ARP seemed to be stuck in TOS land.  The 73 note keyboard is split in two halves (the piano is not multitimbral however), the trigger system of each “lower” and “upper” half comprising a large circuit board.  The reason for the splits was due to the limited range of the divider ICs, thus the lower pitched tones were driven by the same master clock but with its frequency divided by eight (three cascaded sets of divide-by-two), and to keep the size of the circuit boards reasonable.  Each trigger half generates raw square waves of three fundamental “ranks” whose intervals were an octave apart (8', 4', and 2' fundamentals), plus ANOTHER three “ranks” whose intervals could be VARIED (inharmonic ranks).  This made chorus effects possible (such as honky tonk pianos), plus it made it possible to generate transient inharmonics that could generate bell tones such as vibes or tine pianos (Rhodes).

ARP implemented a 40 pin Mostek key voicing/divider IC normally used in electronic organs for gating the raw signals (curiously they left the 16' rank unused).   They developed a novel approach that made velocity sensing possible; the gate inputs were in fact voltage controlled, which meant the amplitude of the gate outputs could be continuously varied.  ARP designed the key contact system as a “flight time measurement” system, which generated a voltage that corresponded to the time displaced between the “key at rest” contact and the “key at strike point” contact, also known as velocity of a key strike.  This voltage was used to gate the outputs, resulting in velocity response.  The voltage generator circuit was based on the charging action of a capacitor, which developed the envelope for the key gate (whose decay shape happens to be similar to the decay of a vibrating string).  Each key had its own capacitor for envelope generation.  The envelope had fixed near instantaneous attack but the decay rate was scaled along the keyboard IE lowest notes had the longest decay which progressively got shorter as you played higher notes - just like a piano.  The decay rate for each key was set by a fixed resistor which was scaled along the keyboard, yielding the RC product charging basis per key.  The downside is that only AD/zero sustain envelopes are possible, so only percussive presets could be realized.

Each lower/upper half has TWO of these key voicing ICs.  The goal of the pair per half was to implement complex envelopes and transient timbres.  The flight time measurement circuit for each key actually generated TWO envelopes - "long" and "short".  "Long" was the fundamental AD envelope for fundamental ranks, "short" was the secondary AD envelope for the inharmonic (or transient) ranks.  By combining the ranks together with their respective envelopes, the complex timbre of a piano whose timbre varies with time can be emulated.  For the sake of convention with the service manual, the fundamental ranks are "long" while the inharmonic ranks are "short".  In addition, a pulse train could vary the discharge action on the "short" envelope, permitting varying decay times for inharmonics on other presets (IE the inharmonic "ping" of a xylophone has a much shorter decay than the inharmonic of a piano).  On a real piano when you play harder the notes get brighter, but unfortunately this system did not have any way to vary the timbre with velocity.

Diagram of the TOS/keying system
tos-system-arp-piano

The keying system is completely analog with no computer needed at all.  ARP was always famous for its maverick designers - Phil Dodds designed this product.  It's an intriguing trigger/envelope system before embedded systems were economical for keyboard scanning.

Noise Galore
Then we get to the rank mixing and filtering system, which is crazy.  Check out this signal flow diagram of the voice board:

signal-flow-diagram

The left half of the diagram is a matrix of rank mixing and active filters - combinations of ranks with bandpass, highpass, and lowpass that can comprise a preset sound.  But these are FIXED filters, a mix of Sallen-Key and MFB architectures.  Recall that the audio circuits operate on single rails whose noise performance are compromised.  Also consider that every active filter has inherent noise which cannot be eliminated.  Any preset can have SIX fixed filters engaged - three in parallel, one set for "long" and the other for "short".  Cascading three fixed filters accumulates the noise.  The noise is DOUBLED with the two long/short sets summed.

Only one master VCF is employed, and its cutoff is controlled by an envelope follower.  Only three of the 16 presets utilize the VCF, and with a master VCF this is a paraphonic system which isn't very effective for improving the authenticity of its preset sounds.  Then the summed ranks go through a six channel resonator, a bank of six parallel active bandpass filters whose output level could be varied by the resistor matrix circuit.  Sort of a programmable graphic EQ.  Ah, there is more noise accumulated there. 

At last we reach the end of the audio chain where we find the six stage stereo phaser, and a lovely noisy one it is.  And while a front panel switch turns off the phaser, in reality it is STILL in the audio chain with its depth and rate preset to minimum.  The reason for the minimal phaser state was to impart subtle stereo motion in its stereo outputs.

Can it get better?  Oh yeah - the master volume control is BEFORE the phaser.  Turn the volume all the way down, and you still hear the noise of the phaser.  All the time.  Better when it is active... swoosh, swoosh, swoosh... Have another drink!

The accumulation of noise from the chain of a whopping thirteen active filters (fourteen if using the master VCF) consisting of the cascaded fixed filters, the resonators, and the phaser results in a very noisy keyboard, even with no keys pressed and no malfunctions.  ARP was probably adopting the concept of fixed filters from their ProSoloist/ProDGX.  The fatal flaw was that the piano lacks any VCA at the end of the audio chain, rendering the noise CONSTANT.

As an indication to how unfinished this product was, several presets produce very low level output at the top keys - including the piano preset.  They never finished voicing the presets.

After I completed the replacement of the noisy RCA 4558 opamps with modern ultra-low noise opamps - an expensive upgrade - the noise floor was significantly reduced, but still audible. 

Panel Controls
The front panel was not the source of the membrane switch problems - it was actually the key contacts.  Controls include volume, tone (rolls off treble), phaser on/off, speed, resonance (no depth?), master tune, detune (not active for every preset), left pedal function, vibrato on/off, speed, depth.  Forget about modifying the preset sound.  Vibrato is quite useless, although the speed controls the trill effect in the Trill FX preset.  The phaser is the standout here; its noise notwithstanding, it sounds very good and high levels of the resonance control brings in that 70s "vibe".  By then, ARP owned Musitronics which made the well-regarded Mutron phaser pedals so they must had used (stolen?) some of their design expertise.  Detune is only active in certain presets, and is good for chorusing effects and "honky tonk" pianos.  A stereo headphone jack is provided but it doesn't have enough drive for 600ohm headphones.

The rear panel has stereo and mono high/low 1/4" outputs, XLR mono output, pedal switch, and aux inputs.  The aux inputs are RCA jacks and were designed to accept tape players; it does not have any RIAA phono preamps for record players (remember this was the 1970s).  While this is a great idea for learning new music in private, the aux inputs produce such low level (even with the headphone output of my iPod) that they are useless.  While the phaser is stereo, the routing to the 1/4" outputs is weird.  The mono outputs actually produce the weakest phaser output; I found the right output jack to have the best phaser effect when a mono output is desired.  And that XLR jack is NOT a balanced output - one of my major pet peeves of ARP products (only the ARP/Rhodes Chroma had balanced XLR outputs).  The pedal assembly that shipped with the piano (missing from mine) was a dual pedal system; right pedal was sustain, left pedal was assignable.  There is a front panel switch that assigned the left pedal to soft pedal (mutes the "short" ranks) or to vibrato depth (a mod wheel for your clumsy foot, if you will).  Since I lacked the original dual pedals I opted to use one of my single sustain pedals.  Only to find that sustain control was on the ring of the pedal jack on the piano, NOT the tip... I quickly swapped those around.  Fortunately with the ring grounded to sleeve using a TS plug into the TRS pedal jack, the soft pedal is not engaged.  Wouldn't have a use for it anyway...

The piano came with legs but they were missing from mine.  The frame under the piano to mount the legs is still intact, but shows the rush to production.  The frame is very shoddily built with the threaded insert for the legs welded to the corners of the frame, and it's a poor welding job.  The legs from my Polymoog were not in use and I found that they fit the threads.  But without crossbars, the legs are very wobbly.  I added a plate to secure the crossbars from my Polymoog, and it is much more stable.

The piano is not too hard to cart around and is actually lighter than a Rhodes piano.  They included a handle at the center under the keyboard for carrying with one hand.  When a (well compensated) friend pitches in to help, you'll find handles at the ends.  The handles are spring loaded so they lay flat and unobstrusive when not in use.

Sounds
I never found an owner's manual or any reference for this piano that defined the sounds for the presets.  The promo sheet hints at the sounds.  This table represents my best guesses:

Preset #
Sound
Preset #
Sound
"P"
Piano (duh!)
9
Harp
2
Fifths
10
Celeste
3
Vibes
11
Clavichord
4
Xylophone
12
Bright Reed Electric Piano (Wurlitzer)
5
Trill FX
13
Harpsichord
6
Tine Electric Piano (Rhodes)
14*
Wah Clavinet
7
Clavinet
15*
Auto Wah Reed EP (Wurlitzer)
8
Reed Electric Piano (Wurlitzer)
16*
Auto Wah Clavinet

* uses master VCF/envelope follower

These are not terribly authentic emulations, but they're not useless.  Most are mid heavy sounds and the tone control should had rolled off bass, not treble.  The piano sounds, well... electronic.  You won't use this on a classical concert stage.  The most curious one is "Trill FX" in which a harmonic is being pitch modulated with a square wave (the vibrato speed control on the front panel controls the modulation rate).  Keep in mind this was the era of 1970s when goofy sounds were the rage...  Sadly while the piano presets aren't terrible, the top notes are too soft.  My guess is that "Fifths" - a sound with fundamental and fifth at the same time - is intended for doubling rhythm guitar on keyboard (a role I absolutely abhor...).  The reed piano presets are better than the tine piano, illustrating the effectiveness of the dual long/short ranks and the radical filtering.  Traditional keyboards like harpsichord, clavichord, celeste, harp, xylophone, vibes are pretty good.  Clavinet not so great, wah clavinet is better and has a funky sound with real dynamic staccoto playing technique.

The Four Voice piano with its much simpler voicing system is not likely to be very good.  The only preset sounds it has are piano, vibes, tine piano, and harpsichord.  I have yet to read a positive review on the thing.  Expect to be disappointed.

Epilogue
ARP president David Friend - who was also responsible for the infamous Avatar - made the decision to rush the piano into production.  It was a very unfinished product.  Even after a complete restoration, the stock sound leaves very much to be desired.  Frankly this is one product that is not worth the effort of restoration unless you put some effort into going the last mile of voicing the presets and correcting the defective circuits.  I have future plans to build a "prototyping box" to replace the resistor matrix board, which will aid the re-voicing effort.  During troubleshooting some filter circuits and CMOS switches were obviously not working (and the defective connector socket found on the resistor matrix board), but the resulting sound of some "broken" presets was quite useful which I would like to find again.  To tame the noise further, I plan to implement compandors which was a circuit technique that proved effective for noisy BBD-based analog delays.  To replace the defective mute circuit I am planning on adapting a voltage supervisor - a component originally targetted for microprocessors.  Also looking at switching the phaser completely out of the audio chain so when it is off I mean OFF.  But these improvements are low on the priority list for now.  When I get around to experimenting I will update this webpage.

ARP never even bothered to put their logo on the rear panel.  This was during a time when I was installing bogus badges on some of my gear to mess with guitar players (like the "Marshall" logo on my Leslie cabinet, which REALLY freaks out gearspotters).  A tech friend had some surplus badges that he gave to me to further my evil quest to obfuscate gearspotters, so I installed a Peavey logo on the back panel of the ARP piano (couldn't find a Steinway logo...).

If there is any silver lining to this poor product, it resulted in the ouster of David Friend and the appointment of ARP founder Alan Pearlman to the presidential throne, whose prediction of ARP blunders proved to be correct.  While he took the helm of a sinking ship, he was the guiding light behind the development of a wonderful polysynth we might never have seen - the much-loved Chroma was birthed at ARP, later finished to production at CBS/Rhodes after ARP went bust.  Sadly Alan Pearlman would be left out of the direction of synthesizers after the demise of ARP, and he never returned to building musical instruments again.

To me, the piano's sole redeeming feature is the action - it feels excellent to this lifetime piano player, the dynamic control is very natural (better than any MIDI instrument I played), and I can use any piano playing technique on it including rapid repetitive keying with two hands.  Not far from me, another 3303 piano showed up cheap.  I grabbed it for some components that were missing from my first one.  This one is in worse shape and I have no intention of performing the same restoration, but I want to experiment with building an embedded controller using that excellent keyboard action and turn it into a MIDI controller.  There's a knock at my door, the boys with the paddy wagon must be here...

contact info

Home