Oberheim OB-X Analog Polyphonic Synthesizer


Last Update 10-11-2014 #1
#2 #3

Inherent Design Errors
How to restore an OB-X
New Voicecards
"Tom Sawyer" rezz
OB-X Revs
OB Model comparisions

Back in 2000, Cary Roberts brought his eight voice OB-X to AHMW (the midwest gathering of Analogue Heaven members).  I was really impressed by the sound of the OB-X - thick, organic, edgy, different from my Memorymoog.  I used to own a Matrix-6R but it didn't sound near as good as the OB-X.  The Oberheim OB-X is a unique sounding classic analog polyphonic, but with many of them in disrepair they are not easy nor cheap to restore.  This is not a restore job for the novice DIYer.  By the time I was in a position moneywise, fully loaded eight voice OB-X were not to be found anywhere.  After a few years, I did spot two fully loaded eight voice OB-X on the 'bay.  One sold quickly to a BIN but needed new endcaps and a restoration.  Another was a dead one that had been stored in a shed and had been a home to rodents (the telltale sign of rodent urine on the computer PC board was a deal-killer, that stuff is a serious toxin).  Spare voice cards were extremely rare and expensive so I passed on the four and six voice varities that appeared.  I wanted a fully loaded eight voice.

For lack of finding a suitable OB-X, I snagged this basketcase Oberheim OB-SX from the 'bay back in '08.  Some research on AH revealed that it had its roots in the OB-X so I opted for the SX until I found the OB-X I really wanted.  The OB-SX needed a restoration and I had made progress, but then I got very busy at work and it sat on the back burner.  After a while I decided I really wanted an honest-to-goodness OB-X. 

Any OB-X seems to be a rare bird on the 'bay, fully functional ones rarer indeed.  I had read that only 600 of these were made.  So I decided I was going to have to compromise and attempt to fix a handyman's special if I wanted one at all, and I could land a fixer-upper for a lot less $$$.  So when a dead four voice appeared, so did extra voicecards and a motherboard for the extra voices so it was perfect timing.  Now I had a system for full expansion to eight voices.  It wasn't a mint specimen but that wasn't a concern.  The Oberheim sticker on the top right is not original but I decided to leave it on.

Boy can I pick 'em!  This was a real basketcase as there were many blown components, likely due to a very malfunctioning power supply.  It powered on, but the front panel would not respond.  OB-Xs do not have a good history as their components tend to go bad, and there is a reason why many of them that turn up for sale malfunctioning as I later discovered. Unless they have been completely restored, pretty much ALL old Oberheims are malfunctioning.

All OB-X, OB-SX, and OB-Xa synthesizers have a design error

Curiousity stirred me and I wanted to know why these synthesizers are so prone to breaking down.  After studying the schematics,  I uncovered a few design errors that combined to cause the problem of dead OB-Xs:
  1. A design violation in the CMOS 4051s on the CV S&Hs.  The voltage on the input pin of the 4051 must be within VEE-0.5 < Vin < VDD+0.5.  I found that the DAC VREF and VDD are the same value (5.6VDC), and sitting between them was an opamp buffer with +/-15VDC rails.  On power transitions, the output state of any opamp is indeterminant and can swing anywhere between +/-15VDC rails as they are ramping up, way outside the VEE-0.5 < Vin < VDD+0.5 rule.  Even though this could occur in a split second, this will damage the CMOS gate and ultimately blow the 4051s over time.
  2. The 5.6VDC source is a mere diode piggybacked to the 5VDC rail.  This is not a voltage regulator!  The CMOS ICs on the 5.6VDC VDD rail will subject that source to high current transients when the CMOS switches between states.  Under this condition, the 5.6VDC diode circuit cannot maintain its steady voltage, thus the rail voltage will drop during transients.  If this happens while any analog input on those 4051s is greater than VDD, the VEE-0.5 < Vin < VDD+0.5 rule is violated and will blow the 4051.  And on the front panel circuit with the pots, the 5.6VDC rail is supplied by a proper 7805 regulator biased at the common terminal with a diode, which is the correct design (insert facepalm here).
  3. The original CMOS ICs were RCA components.  I have the RCA databook for that era, and discovered that the first generation CMOS made by RCA did not have input overvoltage protection diodes now standard on modern CMOS!  Also these ICs were made on early 6 micron fabrication lines and are prone to outright dying due to chemical interactions which causes a short to ground.  Today's modern CMOS are made on 0.05 micron lines and the chemical interactions were corrected, making them far more robust.  Coupled with the two aforementioned design errors, it is no small wonder that so many OB-Xs turn up dead or seriously malfunctioning.
These design problems also existed in the OB-SX and OB-Xa.  This was finally corrected in the OB-8.

The corrections to these problems are:
  1. Replace the pseudo-regulator diode circuit with a proper regulator such as a 317-based circuit.  It does not draw much current and probably won't even need a heat sink.  Make sure there is a 10uf cap between the adjust terminal and ground for better regulation under the CMOS switching transients.  Setting the rail voltage a little higher than 5.6VDC - higher than the rail on the front panel pots - is the safest approach, so a trimpot would be wise.  This is the new VDD for the CMOS 4051s on the S&H CVs.  If you set VDD too high above the 5VDC rail on the TTL decoders and demux's then the 4051s will not switch properly.  You'll need to build a sub-board with a place to mount it (I used a tall standoff mounted just to the left of the connectors leading to the voicecards, near the top of the control board).
  2. Install two clipper diodes on the output of the opamp following the DAC, one should clip to ground the other to the new 5.6VDC rail.  During power cycles, even if the 5.6VDC doesn't fully ramp up this guarantees that the opamp will never swing outside the CMOS VEE-0.5 < Vin < VDD+0.5 input rule.
  3. Replace all RCA CMOS ICs with new modern equivalents.
If you do not understand circuit design for these corrections, please consult a competent technician or engineer.

I do not offer repair or restoration services so please do not email me because I am very busy and don't have time to address individual queries.  I often get emails from folks who want to know what my replacement parts were, so here ya go:

Restoring a dead OB-X

This section is applicable to other old Oberheims like OB-SX, OB-Xa, and OB-8 not just OB-X.  It is also somewhat applicable to other analog polyphonics like the Memorymoog, Prophet-5, Roland polysynths, etc.

How prepared are you for a job of this magnitude?  Take a look at these images below.  The one on the left shows my Oberheim OB-SX in the middle of restoration.  See that box to my right?  That's the box of NEW components.  My OB-X needed MORE replacement components than the OB-SX in the picture.  See the image on the right?  Those are 3/4 of the components that were discarded due to age or malfunction (not shown are large electrolytic capacitors, disintegrated keyboard bushings, and many other ICs).  Not kidding, folks.  Restoring an old Oberheim polyphonic synthesizer is a labor intensive effort (plan extra $$$ for a competent tech) requiring about $300US in brand new parts.  It's a lot of work, but in the end you will have a reliable synthesizer that stays in tune and autotunes when you are done.  Do not expect shortcuts here.

The patient on the operating table (no, the KEYBOARD silly!)
The pile of replaced components (!!!)

This unit is serial # 7943xx, indicating it was built in the 43rd week of 1979.  The OB-X was introduced in June 1979 thus this is one of the early units.  This one lacks a lot of field upgrades (aka ECOs or Engineering Change Orders) that have been developed over the years.  It had obviously been gigged with its share of scratches on the panel and the wood endcaps were very soiled, which was a good indicator of the abuse this unit had endured.

I don't often buy dead polysynths because of the time/effort in restoration and part availability can be a problem.  The OB-X was a rare exception because its great sound justified the time/effort and with very few exceptions (80C98s, RAM/ROM ICs, CEM3310s, CA3080s) you can STILL get parts for these beasts.  They used a lot of standard opamps, TTL, and CMOS ICs.  CA3080s can fail but can still be found on the net or scavenged from other synths (the oft-scoffed preset-only OB-SX is a popular target for scavenging valuable parts).  Although the CEM3310 is the only custom IC (two on each voicecard) and are very expensive on the market, they are robust and do not fail often.  Voicecards offered for trade are often stripped of valuable ICs like CEM3310s and 3080s, so if you find any be sure to confirm that they still have them!

The very first thing you should check with any synthesizer is the power supply.  Nothing will work unless it is 100% functional.  You should NEVER swap out ICs until you confirm full functionality of the power supply because bad rails will blow brand new ICs.  A multimeter is not enough - you need an oscilloscope which will reveal faults that will not appear on a meter.  Inspection on my unit revealed that there were several bad rails and multiple defective components on this power supply.  That can blow components downstream under the wrong conditions.  The restoration work on the power supply consisted of:
Electrolytics are polarized so attention is required so they are not installed backwards or you will have fireworks when you apply power!  I left the tantalum caps in the audio path intact (usually at the end of the audio path on the voicecard), they are not subject to the large voltages like power rails and I wanted to retain that "dirty" sound that Oberheims are renowned for (tantalum caps are not renowned for their high fidelity, and this is one beast where imperfections can be a good thing).

With the power supply now verified, it was time to move on to the embedded system.  The front panel would light up but nothing would work.  The OB-X has no bootup fault reporting system.  Something was locking up the embedded system.  One problem with troubleshooting this kind of fault is that the schematics are so poorly drawn that I had to re-draw them to get a logical overview of the embedded system.  The embedded system is locking up because one or more rogue components is tying down either the data buss or the address buss, and the troubleshooting process to isolate the rogue components is what separates the men from the boys.  From that I was able to quickly isolate the problem which turned out to be a bad 80C98 shorting out the data buss.  This re-drawn diagram is very large and not capable of being scanned, so a friend is currently putting that diagram into a CAD system so an electronic copy will be available in the future.

With the embedded system now living and breathing, it was obvious this was going to need a major overhaul.  There were a lot of defective circuits - most of the CVs from the demultiplexed system were not present, and several logic control lines to the voicecards were defective.  If I turned up the portamento control, the pitch would shift and throw all the voices out of tune.  I got it to make audio noise, but with many of the control signals defective it wasn't going to make music.  I decided that I was going to do the full monty and replace all the front panel switches, all the trimpots, CMOS ICs, TTL ICs, opamps, tantalum caps, and electrolytic caps.  With the condition that the power supply was in, it's a safe bet that many ICs were either damaged or stressed almost to the point of total failure.  So I ordered a few hundred dollars of replacement components and a new set of keyboard bushings to replace the dried out ones that go "clack clack clack" when you depress keys.

I already had a stash of NOS front panel switches with caps of varying colors so I elected to add a personal touch with black, grey, blue, and red button caps.  I have white button caps but they clashed with the panel so I did not use them.  As I removed the original switches, they revealed an oily substance under the switch casing that reeked of nicotine - cigarette smoke.  Before installing new switches, I cleaned off this substance to prevent future low impedance shorts (and remove a known carcinogen).  I arranged the button cap color coding into logical groups.  I reserved the red button caps for functions you REALLY shouldn't activate accidentally (such as WRITE) or functions you may want in a hurry (such as AUTOTUNE).  EDIT and MANUAL got the blue button caps; the HOLD, RESET, and program buttons got the grey button caps; and the programmable synthesizer switches got black button caps as originally installed.  You can see this arrangement in the picture at the top of this page - they don't look out of place.

Unfortunately this project had to be suspended due to work demands and serious health issues.  It was a few years before I resumed restoration.  The annual AHMW2013 event was coming up and I decided to get this unit in working order.  I didn't have much time so it wouldn't be a complete restore, just enough to get it functioning musically.  And I was pretty anxious to hear what this thing could do...

When you are performing a mass component replacement, you MUST perform them in logical order per strategic section and verify functionality of each section.  This was the order I performed:
  1. Replace trimpots on the control board and recalibrate.  This is important for lining up the DAC output range, the LFO frequency range, and the octave and pitch bend ranges.  If this isn't done first then voicecard calibration will be incorrect.
  2. Correct the 5.6VDC design error.
  3. Replace all CMOS 405x switches in the CV demultiplexer, the LFO waveform switching, and modulation source switching.
  4. Replace all quad buffer opamps on the outputs of the 4051s in the CV demultiplexer circuit.
  5. Replace all the 7442 decoders and 74174 latches that generate control signals to demultiplexer circuits, to ROM/RAM demux, and to voicecards.
  6. Replace the CA3240s in the polyphonic glide circuits with LF353s (ECO recommends TL082s but the LF353 is compatible)
  7. When I verified the CVs and logic controls to the voicecards, I identified a bad pitch CV to voicecard #4 and traced it to a bad 3080 OTA in the polyphonic glide circuit.
For a system this complex, it is best to divide up into separate sessions and check that the computer still works.  After you finish, apply power and verify that the computer still works.  Then replace trimpots and electrolytics on the voicecards.

The best plan of IC replacement is to fan out from the DAC.  Replace the 4051s, then power up and verify.  Then the quad opamps, power up and verify.  Then the 7442s, 74174s.  Rinse, repeat, and recycle.  Divide into small sessions and you will keep your head from spinning.  Then probe CVs and logic signals at the voicecard pins, confirming that they change when you manipulate the front panel controls.  Then replace the TTLs in the computer circuits, the opamps in the CV and PWM summers, the LFO, etc.  Divide and conquer, divide and conquer.  This is why a mass component replacement must be done strategically - the voicecards are dependent on everything upstream so you must verify 100% operation of the control board which is generating the control signals.

If your OB-X doesn't autotune, DON'T attempt to troubleshoot it yet.  The autotune circuit is a complex closed-loop system that is dependent on proper CV generation and calibration of the voicecards.  In other words, it may work just fine when you finish in the end, which was the case with me.

Other known failures:
  1. Bad miniDIP voice select switch.  This allows a tech to disable other voices so that one can be isolated for troubleshooting.  It is an IC-shaped assembly with eight little rocker switches, located under the front panel to the left.  If you have voices that will not trigger then this should be replaced.  Very common failure point.  My OB-X was early enough that it had three large slideswitches instead of the miniDIP switch to select voices.
  2. Cracked ceramic caps.  These tend to disintegrate with age.  They are many in the OB-X.  The leads can completely separate from the cap body rendering it completely broken.
  3. Leaking S&H caps on the processor board.  If you see the telltale sign of residue around the leads, they're leaking.  Even if they aren't, they're old enough to be replaced with WIMA caps.
  4. Cracked solder joints on the voicecard motherboard.  You have to remove the voicecards and remove the motherboard to reflow these joints.  Reflow all of them because I uncovered the cause of these cracked solder joints and you can eliminate a lot of malfunctions with this correction.
  5. Inspect for excess flux.  You can use a small wire brush to scrape it free, just be careful not to damage any board traces.  Aged excess flux can cause malfunctions via low impedance shorts.
  6. Blown LM311 can render pots inoperative.
  7. Erratic intermittent operation could be due to bad 6508 static RAMs.  I found one in my OB-SX.  I had recently read that static RAMs can be restored by shorting all the pins through a few layers of conductive aluminum foil.  Haven't tried it myself, but it does sound possible.
  8. Failing pan trimpots on the voicecard motherboards.  If you hear audio crackling when you rotate them, you're losing signal.  Unfortunately the original Piher trimpots are no longer made and there isn't a trimpot that fits the same footprint, so you need to use miniature pots mounted somewhere.
  9. A thorough inspection revealed a broken solder fillet on one of the polyphonic glide circuits that rendered that pitch CV inoperative.
  10. Check for the presence of insulating fish paper underneath the computer board, right above the big polystyrene caps in the polyphonic glide circuits.  There is close clearance here and the fish paper prevents the leads of the caps from shorting to the traces on the computer board.  The tape on the fish paper can dry out and the paper may be loose or missing.  Replace with fresh tape or use RTV to secure the paper.
It's a REAL GOOD IDEA to do these steps BEFORE probing circuit malfunctions.  You don't want to waste time chasing down a ghost.  I wasn't kidding when I warned that a project like this takes time!
By this point I had verified that manipulating front panel switches and pots correlated to associated control lines to the voicecards - now it was time to restore the voicecards.  This is why a mass component replacement must be done strategically - the voicecards are dependent on everything upstream so you must verify 100% operation of the control board which is generating the control signals.

Remember the design errors?  Sitting between the DAC and the CMOS 4051s is an opamp buffer with +/-15VDC rails.  On power transitions, the output state of any opamp is indeterminant and this innocent little puppy can swing anywhere between +/-15VDC rails as they are ramping up, violating the VEE-0.5 < Vin < VDD+0.5 input rule of any CMOS inputs downstream.  Even though this could occur in a split second, this will damage the CMOS gate and ultimately blow the 4051s over time (poof!).  This is another reason why old Oberheim often malfunction with age.

One other correction is due - install two clipper diodes on the output pin of the opamp following the DAC; I used a 1N34 to ground (anode to ground, cathode to output pin) and a 1N4148 to the new (+)5.6VDC rail (anode to output pin, cathode to (+)5.6VDC).  This forms a hardware limiter that prevents the opamp outputs from exceeding voltages that can damage the CMOS ICs downstream.  I chose the 1N34 because it is a germanium diode which will limit the low voltage to (-)0.3VDC, adding a safety margin to the VEE-0.5 term of the CMOS input rule.   During power cycles, even if the 5.6VDC doesn't fully ramp up the clipper diodes guarantee that the output of the opamp will never swing outside the CMOS VEE-0.5 < Vin < VDD+0.5 input rule.  Even though you will be installing new CMOS ICs with input protection diodes, this is another shield against damage.  I included a pic on how I tack-soldered the diodes on the solder side of the processor board, and don't forget the tubing to insulate the leads of the diodes to prevent shorts.

It's a good idea to confirm that the CVs can swing between zero and 5VDC after you install the diode clippers.

So I closed the hood and tried to play the partially restored OB-X.  I lost an hour playing, it sounded so good.  3 of 4 voicecards were working, the bad one was isolated and I located one from the surplus that would function so that brought it to a respectable four voice system.  Naturally the tuning was way off on the voicecards because there was work to be done on them.  Before I could calibrate any voicecards I had to:
  1. Replace the V/OCT trimpots on each voicecard
  2. Replace the RANGE trimpots
  3. Remove the tantalum capacitors on the power rails, which have a tendency to fail and short out.  I left the tantalums in the audio path intact, as I wanted to preserve that classic Oberheim edge and dirt and the less-than-perfect tantalums are a factor in that sound and are not subject to the high voltage as those on power rails.
This was enough to properly calibrate tuning on each voicecard - the pulse width and high tracking trimpots would be replaced later. 

I was making good time so I proceeded to install the Encore Electronics MIDI retrofit, a comprehensive improvement to the OB-X.  Installation went smoothly with the exception of routing the RESET line - my unit was old enough that the installation instructions did not match my unit and I had to trace the correct PC board trace manually.  Upon firing up, the retrofit worked flawlessly.  The MIDI output would come in handy because I could use a Kurzweil 1000SX as a rock steady digital tuning reference for calibrating the tuning on the voicecards.

Encore's products have been emptying out and they don't appear to be making any more of them.  They still have OB-X retrofits available, but there may come a day when they are gone.

I removed the tape interface jacks - they're long obsolete and I no longer needed them with the Encore retrofit.  I also removed the CV/trigger input/output system due to obsolescence, and I did not want those jacks to wreak any havoc on the critical CVs going to the voicecard.

Calibration went pretty painless and the OB-X now played in tune!  But the big acid test remained... autotune.  If it doesn't work, the system will defeat any voices it cannot autotune.  A press of the button, deep breath, play the keyboard to cycle through the voices - and all four are playing in tune. Autotune is working!  If none of the voices will autotune, the audio summer opamp on the autotune buss on the voicecard motherboard has been known to go bad.  The 2nd version of the OB-X service manual includes a technical description of the autotune circuit for troubleshooting.  I drew up a clearer schematic and description for reference:


And then another problem surfaces: some of the connector pins on the voicecard motherboard were intermittent.  I had read this was a common issue.  But time was short and this would have to wait later.  Luckily that problem didn't surface while at AHMW2013 (knock-knock-knocking on wood). I also couldn't get the filter cutoff frequencies trimmed between voices which is a bias issue to be addressed later.  One of the voicecards had an EG that isn't the same timing as the others, likely caused by the timing cap that is thirty plus years old.  The restoration isn't complete and more work will follow the AHMW2013 event.

After having the instrument in this condition for a few days and a travel 1/3 across the US and back, I cannot stress enough the importance of new trimpots.  The OB-X fired up in tune - every time - and stayed in tune much better than the Memorymoog at the AHMW2013 event.  I didn't even have to autotune when it was powered up cold.  The original trimpots were CTS open frame and cheap quality, which often failed completely with age.  The Oberheim accountants cut a few too many corners in sourcing cheaper components.  The trimpot settings were "locktited" into place when they left the factory but this made no difference after thirty years.  This is another reason why old Oberheim show up for sale malfunctioning.  New replacements are the only solution.  The layout on the circuit board hint that they were originally intended to receive Piher trimpots but I recommend sealed trimpots from Bourns.  I have had good luck with the Bourns trimpots over the years in other instruments.

I now had a 100% functional four voice OB-X ready for the AHMW2013 event.  All I had to do was dial up 120 patches in two days (right!).  I settled for the factory patches that can be recalled from the Encore retrofit with a power-up reset sequence, and made my own bank of some classic Oberheim sounds such as strings, brass, pads, filter sweeps, fat unison leads and basses, special effects... here are some of the sounds it was making. 

A lot of AHMW attendees were looking forward to trying the OB-X.  It got a great reaction although it had to compete with a cool sounding Roland s100M modular that one of the folks brought.  Everybody liked both but the favorites were split between the Roland and the OB-X.  Doesn't break my heart.  But I must admit after being a Moog fanboy for years that I like the unique sounds coming from the OB-X that none of my Moogs could dream of.  They both have their applications!

The night before the AHMW2013 show I was playing the OB-X in my hotel room to dial up some more sounds.  I placed the OB-X on a footstool and got comfortable on a chair.  As I was working, my freshly restored OB-X started malfunctioning.  I noticed if I placed pressure on the case I could make it malfunction.  It dawned on me that the base of the chassis was thin enough that it could flex.  With the footstool supporting the base in the middle, that was exactly what was happening.  This flexing can transmit to the PC boards inside, which will cause malfunctions.  This is a root cause why the connector pins on the voicecard motherboards often have cracked solder joints that have to be reflowed.  I have added angled extruded aluminum to the base of my unit at strategic points to correct this problem.  It is strongly advised that your keyboard stand use arms that support your OB-X fully from front to back (like Standtastic) or as wide as possible and avoid any support point in the middle of the base.  If you use your OB-X on a tabletop with the endcheeks to support it, this is safe.

I set out to expand the voicecards in my unit.  Mine was a four voice, and in order to expand beyond four voices you need some very rare parts.  At the time I acquired my OB-X I was fortunate to find:
  1. Five more voicecards, enough for a full eight voice plus a spare.
  2. Extra voicecard motherboard to accommodate four more voicecards.  These are very hard to find.  Voices 1-4 sit in the lower motherboard while 5-8 are suspended directly above it.
  3. Metal mounting tray for the motherboard, which is even harder to find.  This provides a rigid mounting frame to swing the assembly out of the way for calibration.  Without it, the motherboard will flex and crack with handling.
Fortunately, my good friend Scott "Old Crow" Rider was inspired to replicate the OB-X voicecard using modern components after hearing my unit at the AHMW2013 show.  He is also building the motherboard (but not the mounting tray) for OB-X owners wishing to expand their units to full eight voice.  I have heard the new card in person and can confirm that he nailed the vintage OB-X sonority.  Bare boards are available for DIY assembly, and he is near completion on an SMT version for assembled cards for those who don't have DIY skills.  Full details here.

The components already exist on the processor board for up to eight voices.  You will need to build connectors between the processor board and extra motherboard, and for connector J between upper and lower motherboards.  Connector G is not used on the upper motherboard.  Motherboards were configured for either voices 1-4 or 5-8, there should be a label to the right of the pan trimpots.  The motherboard for voices 5-8 are the same circuit board with unneeded components unpopulated.  If you acquire the motherboard for voices 1-4 with intention for voices 5-8, you must remove the active components in the voice summing circuit and the 7x15 regulators for it to work with voices 5-8.  If you do not know how to interpret schematics then you should refer this task to a competent tech.

It required some effort to getting the extra voicecards restored, they all had faults that had to be corrected.  I found bad 7x15s, opamps, FETs, 3080s.  The CA3086s have been known to fail, mine were fine.  All the trimpots, electrolytic caps, and the aforementioned caps vulnerable to cracking were replaced.  One voicecard even had a broken trace, that took some detective work.  DO NOT HOT SWAP VOICECARDS WITH THE POWER ON!  I knew this precaution but on one occasion I forgot to turn off the power when I installed one of the freshly repaired voicecards.  That fried a CMOS IC on the board POOF!

One complaint I have read often was that in OB-X consisting of greater than four voicecards, VCO calibration was difficult in the lower voicecards.  Each motherboard fitted four voicecards each, voices 1-4 are on the lower motherboard with voices 5-8 directly above it. 

Everyone knows that heat rises and that accumulated heat throws analog VCOs out of tune.  Studying the OB-X framework, there is a large aluminum wall separating the keyboard and voicecards that also supports the front edge of the panel when it is closed.  That wall combined with the standoffs for the upper motherboard combined with a solid metal tray that the board mounts onto (Oberheim failed to put vent holes in the metal tray!) created a closed box that restricts airflow and traps heat in the lower voices.  There is little room at the rear of the OB-X for heat to escape.  This is another reason why OB-X have a reputation for tuning drift.  This is why calibration is difficult on the lower voices - when you raise the upper voices out of the way to calibrate the lower ones, the heat has now escaped.  Close the upper voices back down, and the heat is again trapped and throws off your calibration!  To make matters worse, the VCF exponential converter has zero temperature compensation thus raising/lowering the upper tray makes VCF calibration an empirical job with the heat changes!!!

The easiest solution is raising the standoffs for the upper motherboard by 1/2", providing better ventilation for the lower motherboard beneath it.  There are six standoffs, three at each end.  One set of three have hinges to swing the upper motherboard like a hood, the other set are fixed height.  The hinged set is a two piece assembly consisting of a fixed 1" standoff coupled to a hinge that is another 1/2", total 1-1/2" height.  The other set is fixed height of 1-1/2".  Since we want to raise them 1/2", we can replace the 1" fixed standoffs with the other 1-1/2" ones, when combined with the 1/2" hinges results in a new 2" height.  All you need are three new 2" fixed height standoffs.  I confirmed there is ample clearance for all components when the hood was closed.  After using this modification for a year I can report that the lower voicecards are very stable.

The later OB-Xa and OB-8 replaced the wall between the keyboard and voicecards with a pair of standoffs.  This allowed air to circulate and they are easier to calibrate and stay in tune better than the OB-X. top

That cool opening sound in "Tom Sawyer"

Rush's "Tom Sawyer" opens with this really cool resonance sweep (aka "rezz") that has evaded emulation for YEARS on other synthesizers.  I tried to emulate it myself and never quite got it.  At the time of the making of Moving Pictures, Geddy Lee was using both his old eight-voice SEM polysynth and an OB-X, which can be seen in his picture in the inlet of the album. Rush has a DVD on the making of the Moving Pictures album where Geddy Lee confirmed that they created that sound on "the oberheim" but he doesn't say which one!  Nonetheless, both can get that sound.  These are the secrets to "that" sound.

SEM Multimode filter.  The OB-X uses the same multimode filter as the older SEM-based polyphonic systems.  It is a 12dB/oct state variable filter that can generate simultaneous lowpass, bandpass, and highpass outputs.  No other filter, not even the Moog ladder filter in 12dB/oct mode, will get this sound.  I tried.  The SEM filter has that unique grind at high resonance and growl at low frequency cutoff that contributes to the Tom Sawyer rezz.  Later OB models abandoned the SEM filter.  The CEM3320-based OB-SX and OB-Xa were almost as good, OB-8 and later OBs lost it.  The SEM filter is THE tool for this sound.

Knowing this, I fiddled and fiddled to get THAT sound when I first got my OB-X.  Only when I pressed the UNISON switch was that moment when I said "WOOOOAH, so THAT'S how they did it".  That sound requires multiple voices firing in unison.  A single voice (such as a monophonic) will not work, you need a polyphonic with unison mode firing multiple voices at once.  There's another slight clue in the song that confirms this.  If you're not in unison mode, the OB-X operates in multiple trigger mode which means the EG re-fires with every key.  There is one place in the song where the EG does not re-fire between A and G key.  If you're in unison mode on the OB-X, it is operating in single trigger mode where the EG does not re-fire until you release all the keys.  That one place in the song is only possible with unison mode.

But I had done the same trick on my Memorymoog and Andromeda and they didn't sound like the OB-X.  I'm no slouch with analog programming, so something was missing.

That "something" turned out to be inherent filter tracking error and mismatched EG decay rates between voices.  On newer polyphonics like Memorymoog, Andromeda, et al they all perfected the matching of filter tracking and EG rates between voices.  The OB-X is one of those cases where imperfections can be a good thing.  The SEM multimode filter in the OB-X has its faults (or features depending on your perspective).  The filter tracking of SEM filters is only within a paltry three octave range, outside of that they are not guaranteed to track.  The cheap expo converter in the CV summing circuit for cutoff control contributes to this "fault".  Once outside that three octave range, the error between voicecards will not be the same.  Translated, the cutoff frequency between voices will vary despite best efforts to calibrate them.  Indeed, filter calibration has been a common complaint because you will never get all the voicecards to track the same.  In unison mode with multiple voices firing at onces, that tracking error causes a flanging effect - due to multiple voices with mismatched filter cutoffs - with the "Tom Sawyer" rezz as the cutoff sweeps down.  Also listen closely to the original song - when the filter closes, you're hearing mismatched cutoffs.  They're panning the voicecards across the stereo field.  It's a cool stereo effect that you can't get with a modulated delay like chorus or flanging..

Adding to the filter tracking error is the tolerances of the EG timing between voicecards.  The original timing capacitors used for the EGs on the voicecards were wide enough tolerance that the decay rate varied slightly between voicecards.  That exaggerated the filter sweep flanging effect between voicecards further, especially as the filter sweep falls from high cutoff.  This was further confirmed when I went just a little too far in restoration and replaced those "faulty" timing caps with new WIMAs and "lost" that fault, but I have another modification in the works to get it back.  Another confirmation was playing with an Oberheim Two-Voice as I tried that rezz sound.  I used both SEMs for the sound.  With multiple SEMs, there is no way to get the EG decay perfectly matched.  Indeed, as I varied the decay of one of the SEMs while the other was untouched, the unique "growl" of the Tom Sawyer rezz began to emerge.

So the secrets of the "Tom Sawyer" rezz are the SEM multimode filter and imperfection.  All succeeding polyphonics perfected the "faults" of the OB-X, thus they are not capable of replicating that sound. top


Oberheim's previous polyphonic synthesizers was a bunch of SEM modules (SEM was a compact synthesizer module) ganged together and coupled to a polyphonic keyboard scanning system.  The popular four voice was model FVS (Four Voice System), while the bulky (and ultimately unwieldly) eight voice was model EVS.  Oberheim retrofitted user patch storage to both polyphonics but this programmer did not memorize all the settings on the SEM modules.

In 1978 the Prophet-5 polyphonic synthesizer was released, and its main feature of memorization of all synthesizer parameters attracted legions of customers.  The OB-X was introduced in June 1979 to compete with the Prophet-5.  The OB-X voicecards were near identical to the SEM, only now every parameter was programmable.  They were offered in four, six, and eight voice configurations with list prices of US$4595, $5395, and $5995 respectively.  These were very expensive machines for 1979.  Owners could add more voicecards later, up to eight total.  The OB-X has no split or layer capability later to be introduced on the OB-Xa.  Unofficially they are categorized into four revs with the latest ones sporting the pre-MIDI "Oberheim Interface" in the left hand endcap to communicate with their DSX sequencers and DMX drum machines (which wouldn't be released until AFTER the OB-X was discontinued...!).  None of them ever had a MIDI upgrade from the factory as the OB-X was discontinued in early 1981, two years before the birth of MIDI.

There are 19 control knobs and 21 buttons on the front panel for tweaking the synthesizer patch, all of which can be stored for instant recall.  You have to press the EDIT button to tweak the patch using the knobs and buttons, and the MANUAL button makes a brand new patch using the current position of the pots.  Memory provided storage for 32 user patches (the Encore Electronic retrofit expanded this to 120 patches).  There is a big slideswitch under the hood to disable write capability and protect your patches (not useable with the Encore retrofit).  Left, right, and monophonic outputs with pan trimpots under the hood route the audio to respective jacks.  Volume is global to the instrument, there is no programmable volume per preset.

Usually the serial badge on the rear will specify how many voices are present (unless a unit has been expanded to more voices in the field).  The voicecard architecture is identical to the classic SEM discrete circuit, you could follow the schematics from left to right and it is 90% identical with exceptions of circuit differences in the filter, lowpass only in the OB-X, and omission of PWM/FM via VCF EG.  Each voice featured two VCOs, each capable of providing ramp or variable pulse width waveforms.  VCO2 can be hard synced to VCO1 and you can configure the left hand pitch bend lever to modulate only VCO2 for screaming hard sync effects.  The OB-X was one of the meanest sounding hard sync sounds.  One pulse width control is shared between both VCOs.  There is an option for oscillator cross modulation (XMOD) using VCO2 ramp to VCO1 frequency, but the depth is fixed and VCO2 cannot be put in LFO mode.  There is a pink noise source on the OB-X.  Crude mixer level controls of each oscillator is limited to full/off for VCO1, full/half/off for VCO2, and full/half/off for noise.  You cannot combine waveforms - ramp or pulse, not both.

The OB-X features polyphonic glide ("portamento" in Oberheim speak), which the Memorymoog had but the Prophet-5 did not (P5 glide only worked in unison).  The polyphonic glide is totally analog.  It does also have a fault - or feature depending on how you appreciate it.  Because each voice has its own glide circuit, the key component - the CA3080 OTA - isn't exactly matched between voices.  This means the glide rate is different between voices.  When used with unison with all voices firing, the unequal glide rate results in a unique fat detuned sound as you transition between notes.  Using the stereo outputs with voices panned across this was a dramatic effect.  Later OB-Xs used matched OTAs - marked with a red dot - in the glide circuit for better but not perfect matching between voicing.  The mismatch was pretty pronounced on my unit, enough that I swapped out the unmatched OTAs for better ones.  Digital glide was first introduced on the 56-preset OB-SX (which emulated the unequal glide rates in the OB-X!!!) while the OB-Xa (produced at the same time of the 56-preset OB-SX) still had analog glide.  Oberheim finally progressed to digital glide in the OB-8 that was generated by software and guaranteed matching glide rates between voices.  That eliminated the cost of the high parts count and the labor intensive matching for the OTAs that were thorns of analog polyphonic glide.

Modulation is pretty basic on the OB-X.  There is no voice modulation like that offered on the SEM module, Prophet-5, or Memorymoog.  There is no provision to route filter EG to VCO frequency (later added to OB-Xa) or pulse width (later added to OB-8), nor can you impart audio FM on the filter cutoff with a VCO (later added to OB-8).  There is a single LFO with triangle (panel is incorrectly labeled "sine"), square, and S&H waveforms with two independent depth controls for frequency or pulse width modulation.  The mod wheel (OK I know they're pinball flippers not wheels, but I'm retaining the convention used in the rest of the world, especially MIDI) always routes the triangle LFO waveform to both VCOs.  A control pedal via the rear panel jack can also be used to blend in vibrato like the mod wheel.

The VCF is the same 12dB/oct filter found in the SEM, but only with lowpass filtering.  A switch turns on keyboard tracking although the service manual says that tracking is accurate only over a three octave range.  The filter will not self-oscillate, but it has a wonderful resonance that is colorful and unique.  The only modulation option is from the LFO or the EG.  There is a rear pedal jack for filter control which affects all voicecards simultaneously.

The VCA is the same discrete circuit found on the SEM and is largely responsible for the dirty edgy sound to the OB-X.  The VCA is comprised of a CA3080 OTA followed by a TL081 to buffer the weak current output of the OTA.  The 3080 OTA is not exactly a high fidelity component and some distortion is introduced here, which lends the OB-X its dirty edgy sound.  Definitely not clean but there is a pleasant mild creamy overdrive here that is pleasing to the ears.  At the time the cheap-and-dirty CA3080 was the most cost effective that technology had to offer.  The better ones weren't better enough to justify the higher cost.  This same circuit is also used in the voice summing circuit, further distorting the signal (especially in unison).  Interesting note: the SEM polysynths had the OTA VCA, but no OTA in the voice summing circuit!  In fact you can correlate the loss of this dirty edgey sound with the progression from OB-X->OB-Xa->OB-8 as changes were made to cleaner VCAa and voice summing circuits (see the comparision table).  No doubt when Tom Oberheim lamented that the OB-8 was "too clean" and that something was lost along the way...

On the left hand panel, a switch selects the range of pitch bend to either a full octave or two semitones.  Frankly the pitch bend is a little abrupt using the "pinball flipper".  Another switch routes pitch bend to VCO2 only, handy for hard sync sweeps.  A third switch changes the range of the keyboard up or down one octave.

CV/Trigger inputs on the rear panel only work on voicecard one and the jacks are wired such that when used the computer will skip voicecard one.  This lets you use a sequencer on one voicecard with the remaining voicecards are controlled by the keyboard, but because the OB-X does not feature multitimbrality you can't get an individual patch on that sequenced voicecard.  There are CV/Trigger outputs provided but they are the signals from voicecard one only.  Because it preceded MIDI, the OB-X had a cassette interface system for archiving and restoring patches.  Since I was going to use the Encore retrofit which feature MIDI control and sysex patch I/O, the cassette interface and rear panel CV/trigger jacks were no longer needed and I removed them.

One of the coolest features is UNISON.  This turns the OB-X into a monophonic instrument with all voices firing together.  Because the calibration between voicecards isn't precisely matched, unison yields some very thick sounds.  A large part of this is due to free-running VCOs in which none of them are oscillating in synchrony.  Very few analog polysynths have this feature.  Combined with polyphonic glide and the HOLD/RESET buttons, you can arrange one key to play a chord and get end-of-the-world effects.

Of the OB line, the OB-X has the least features but the most organic sound.  It has that colorful alien resonance in the filters (especially in filter sweeps), that discrete girth and bite (especially in unison), that subtle pleasant distortion in the VCA and voice summers which is a classic Oberheim trait.

The OB-Xa and OB-SX voicecard is neither interchangeable or compatible with OB-X.  They used different signals and the connector pins do not match.  top


There are unofficial "revs" of the OB-X scattered about the net that help define how old an OB-X is.  Some are obvious from the outside, some aren't.  Through significant circuit changes from reading ECOs (Oberheim "Engineering Change Orders") and cosmetic changes in the panel, I narrowed down this list of unofficial "revs".

Rev #
"OB-X" block typeface on top middle of panel, no Oberheim name on front panel, voicecard enable three large slideswitches (like TEST1 and TEST2), only two slideswitches on LH panel (OSC 2 ONLY and TRANSPOSE), rev A voicecards, no solder mask on any boards, mounting screws visible on LH panel slideswitches, no bezzle in hole for mod/pitch levers.  Polyphonic glide circuit used CA3240 which fail with age.
"OB-X" block typeface and Oberheim logo above keyboard, three slideswitches on LH panel (OSC 2 ONLY, BROAD/NARROW, OCTAVE)
"OB-X" fancy typeface top left of panel, "Oberheim" brand above keyboard on right, "HOLD" "RESET" labels right-aligned, voicecards have solder mask
"HOLD" "RESET" labels center-aligned, voicecard enable mini-DIP switch (common failure point). Noise filter added to AC power inlet, processor & main board has solder mask, mounting bolts for LH slideswitches now integrated into LH panel with only visible screws on mod/bend levers.  CA3240 replaced with TL082 (LF353 is suitable replacement) in polyphonic glide circuit.
Power switch moved to top of panel, "RESET" renamed "CHORD" and center-aligned, rev B voicecards (gate LED added)
Bezzle added in hole for mod/pitch levers.  Gray outline changed to darker scheme with hint of blue.  OTAs matched in polyphonic glide circuit for better (but never perfect) tracking of glide rates between voices.
LH panel slideswitches has gray border, rev B processor board (adds "Oberheim interface" DB37 connector on LH endblock), panel boards have solder mask.

So which is the the best "rev" OB-X to look for?  Soundwise, they're the same.  If you want one with minimal maintenance/reliability, there is no better rev - they all have the design faults, flakey CMOS ICs, failing solder joints, and bad trimpots.  If you want one with minimal restore effort, the later ones are not much safer.   top

OB Model Comparision

There's a lot of confusion between the different Oberheim OB models so I built this table to compare them

4, 6, 8 (even only)
2x discrete
2x CEM3340
2x CEM3340
2x CEM3340
expanded in software matrix
VCF (lowpass only)
Discrete cascaded OTAs, 12dB
CEM3320, 12dB
2x CEM3320, 12dB or 24dB
1x CEM3320, 12dB or 24dB
3080 OTA + TL081 buffer
3080 OTA + TL081 buffer 2x 3080 OTA + TL081 buffer 2x 3080 OTA + TL081 buffer or CEM3360
Voice volume control
2x CEM3310 VCF & VCA
2x CEM3310 VCF & VCA 2x CEM3310 VCF & VCA 2x CEM3310 VCF & VCA
two (for dual timbrality split/layer)
two (software)
Noise Source

# of patches
24/48/56 depending on version
32/120 depending on version
User Storage
Polyphonic glide
56-preset version only, digital
Audio Field
Voicecard Audio Summing
3080 OTA + TL081 buffer 3080 OTA + TL081 buffer CEM3360
DAC Resolution
MIDI Retrofit
Encore Electronics
Highly Liquid UMR2 Encore Electronics Encore Electronics

Fully functional Oberheim OB-X units rarely come up for sale and when they do they command some serious bucks.  Most casual owners don't have the resources to keep them running and resort to "parting out" valuable stuff like the voicecards.  If you can't afford a functional unit, by all means don't buy a broken on until you arrange the services of a competent tech for restoration work (and enough $$$ for the labor and parts) - unless you're lucky enough to possess an EE degree for DIY jobs.  Luckily you can still get replacement components for the OB-X, with the exception of the 80C98, CA3080, and CEM3310.  CEM ICs are getting expensive and hard to source.  Unlike the CEM-heavy OB-Xa and OB-8, the only CEM in the OB-X is the CEM3310 EG and those usually are quite robust.

Most people would be crazy to own both an OB-X and Memorymoog in that both are service nightmares, but I have been restoring and maintaining these things for years and can keep them running in top condition.  I have loved my versatile Memorymoog for years, but the OB-X compliments it nicely, fills in some gaps, and produces unique sounds only the OB-X can do. The filter is very unique and colorful, and despite appearances of basic architecture the OB-X can produce some wild and crazy sounds.  Most people identify the OB-X as the sound of the 80s for good reason.  Many classic songs featured this beast, and something was "lost" when the OB-X was replaced with the OB-Xa, OB-8, Xpander... Oberheim never reclaimed that "sound".  There aren't many of these units left, and people who have good units in their possession are not giving them up.    top

contact info