Real World Interfaces logo

Engineering Change Orders for TB-303 Devil Fishes

Last update 2016-05-18  Robin Whittle
Update history below.  To the main TB-303 Devil Fish page: ../.


In February 2016 there are 297 TB-303 Devil Fishes to date - SN -001 to 016 versions 1.x and 017 to 295 versions 2.0 to 4.2B.  I created this page to list technician-installable hardware changes which correct problems in Devil Fishes which I discovered after shipping the machines to customers.

These ECOs do not concern the version 1.x machines, some of which have a Sweep Speed switch which was not part of the original design.

Some ECOs have a PDF file instructing a technician on the fault which can be fixed by a hardware change, and the details of these hardware changes.  Filenames are of the form ECO-DFTB-xx-yy.pdf where xx is the ECO number and the yy is the revision number of that document, starting at 00.

This page is intended to be read by technicians and Devil Fish owners in order to determine whether any ECOs are applicable to their particular Devil Fish. 

These ECOs do not cover every change in the Devil Fish circuitry.  For instance they do not cover the earlier arrangements for Filter Frequency CV In which were not approximately 1 volt per octave.  Nor do they cover some machines with 32 Banks of Memory which may behave erratically if there are fresh (and therefore high voltage) alkaline C-cell batteries installed.  The workaround for this is to use somewhat used batteries.  See the Devil Fish manual's Known Reliability Problems for a discussion of potential poor contact in the CV In socket leading to wavering pitch, for serial numbers up to 083 (April 2000).

The ECOs

Number and PDF links
Applicable modifications, versions, serial numbers and dates
V2.0; SN 017 (perhaps) 019 and 020; July and September 1996. 
A problem with filter frequency oscillations can be solved by installing a 22uF axial capacitor in place of the one currently installed, between the original C14  positive pad and pin 6 of the Resonance pot.  Please email me for details.

V2.0, 2.1 and 2.1A; SN 017 (perhaps), 019, 020, 023, 024, 025, 028 to 032, 034 to 038, 040, 041, 042, 044, 046 to 051; July 1996 to August 1998.
There may be excessive clicks at the start of notes when the Soft Attack pot is set towards the clockwise end of its scale.  This is due to my use of a Schottky diode instead of an ordinary 1N4148 diode, and perhaps some other aspects of the circuitry.  Please email me for details.

Please also see ECO-07 below.

V2.0 to 4.2; SN 017 to 259 (2012-11-10), except those I subsequently worked on after 2012-01-07.
There is a slight buzz in the audio output due to front panel LED drive currents.  This is worse if the original red LEDs are still installed.  With new Blue LEDs, the current is reduced and so is the noise.  The fix is simple - on the back of the TB-303 main PCB (which is easily accessible by removing the rear cover) solder a reasonably thick wire between the left terminal of the Volume pot and the ground terminal of the Audio Out socket.

MIDI In (not MIDI In and Out) systems installed between  December 2004 and  February 2013 and which I have not subsequently  worked on in  February 2013 or later; SN 165 to 186, 191 to 195, 197, 198, 200 to 207, 209, 210, 214 to 221, 223 to 235, 240, 241, 243, 246, 248 to 255, 257, 259, 261, 264, 265, 267 and 271.
The firmware in the microcontroller of these MIDI In systems is version 1.0.0, 1.0.1, 1.0.2 or 1.0.3.  All these versions have some bugs which are listed at the start of the user manual: ../DF-MIDI-In-1.0.3-Manual.pdf.  This manual replaces the one I sent with the Devil Fishes.  It contains workarounds for these bugs. 

If your technician is prepared to dismantle and re-assemble the Devil Fish, which is tricky, then please email me and I will post you a microcontroller with Version 1.0.4 firmware, and a new manual.  The microcontroller is in a 28 pin DIP package and can easily be plugged in, replacing the original microcontroller.

In addition, these machines may power up with the Accent flip flop set to Accent On, which will cause MIDI In notes to play with Accent.  The workaround is to play and stop an Internal Sequencer pattern which has no Accents.  The hardware fix for this is relatively easy for a technician - remove the rear cover, isolate (by cutting a thick track) pin 10 of IC 13 (active-high SET which resets the bar-Q output of pin 12), and connect this pin 10 to pin 7 of the CPU, which is an active-high power-on reset signal.

V2.0 to 4.2; 017 to 274, except machines I worked on after 2013-12-03.
These machines use a 10k ohm pull-up to +6 volts on the Accent Out socket.  This is probably fine for most applications. 

Starting with SN 275, I changed this to a 3.3k, so that the signal is strong enough to drive the Slide CV Input socket to turn on the Gate signal, even if the machine is running from batteries which have lost some of their capacity

The fix requires completely dismantling and reassembling the Devil Fish: Install a 3.3k 0.125 watt (or at least a small resistor which might have a higher wattage rating) in place of the 10k resistor just to the right of the Accent Out socket.


MIDI In systems - see ECO 04 above.
As explained in the 4.2A sub-section of the Version History section of the Devil Fish manual, machines with MIDI In or MIDI In and Out are now installed with capacitors and resistors on the Sync socket.  This is not absolutely essential, but it may reduce problems due to capacitive coupling in 5 wire MIDI leads where the driving device is not connected to the same ground as the Devil Fish and one or the other is driven to a high AC voltage with respect to the other, most likely due to being driven by an AC power adaptor with no other form of grounding, especially when the mains voltage is 230 volts or so, rather than 120 or so.

The fix is relatively simple - your technician can remove the rear cover and install between the centre pin and each of the two outside pins of the Sync socket (ground and the Run/Stop and Clock pins), a 0.01uf capacitor in parallel with a 33k 0.125 (or at least small) axial resistor.

All Devil Fishes except those I initially modified, or subsequently worked on, in February 2016 or later.
This relatively simple hardware change significantly reduces a remaining problem with quiet but sometimes annoying clicks in the output signal at the start of notes when  there is no signal coming out of the Filter.  This problem remains after the fixes mentioned in ECO 02 above.

There are naturally clicks at the start of some notes due to the fast rise of the volume envelope when there is a low frequency signal coming out of the Filter, and so into the VCA. 

The problem fixed by this ECO concerns a small mid-to-high frequency click which may be audible at the start of notes, when either there is no signal going into the Filter and the Filter is not self-resonating, or any signal going into the Filter is not coming out of it into the VCA, due to the Filter frequency being lower than the lowest frequency component of the input signal.  The nature of the click changes considerably according to the setting of the Soft Attack pot, when Accent is not active.  In these circumstances we expect the VCA to open up (except when the Decay pot is fully anticlockwise - ACW - and the Soft Attack pot is fully CW), and to the extent the the VCA opens up, we expect to hear only its internal noise and the internal noise of the Filter.  We do not want to hear a click, especially a high one, since our ears are very sensitive to this.

The problem occurs inside the VCA chip when there is a sudden positive change in its control signal on pin 1.  To reduce this, and so to significantly reduce the clicking problem, the hardware fix involves capacitive damping and raising its lowest voltage from 0 to about 0.258 volts:
  1. Remove the rear cover and install these three components on the bottom surface of the main TB-303 board.  This does not involve disassembly of the TB-303 or Devil Fish boards.

  2. Install a 10M resistor (0.25 or 0.125 watt) between +12 volts (pin 4 of the VCA chip, IC 15) and pin 1 of the VCA, which is the right-most pin.

  3. From this pin 1, install two components in parallel: A 220k resistor (a small 0.125 watt size fits best) and a 0.1uF monolithic ceramic 50V capacitor.  (The actual value of the capacitor is likely to be somewhat greater than 0.1uF, which is fine.)
A photo of the component installation: ECO-DFTB-07.jpg


MIDI In and Out systems installed before February 2016; SN 283  (September 2014), 284, 285, 286, 287, 288, 289, 292 and 294 (January 2016).
When the Internal Sequencer plays notes longer than 1/16th note (that is, the first 1/16th note is followed by one or more 1/16th notes which continue the first note, to make, for instance, a 1/4 note from four 1/16th notes) the pitch of the subsequent 1/16 bar sections of the one note is slightly raised.  The CV can rise by about 20mV, which is about 1/4 of a semitone.

This only affects notes played by the Internal Sequencer, not those played from MIDI In or when an external CV signal is plugged into the CV In socket.

The cause is quite complex, involving leakage currents on some input pins of the the PIC microcontroller due to one or more other input pins being raised more than about 0.3 volts above the power supply voltage, which is normally about 5.0 volts.  See the PDF for details.

The degree of the problem will vary somewhat from one machine to the next, and it will generally only occur if the machine is running from a power adaptor or perhaps from very fresh alkaline C cell batteries.

The workaround is to run the machine from batteries which are not entirely fresh.

The hardware fix is relatively straightforward for a technician who can remove the rear cover of the machine.  It is not necessary to dismantle the TB-303 or Devil Fish circuit boards.  A single additional resistor soldered on the back of the main TB-303 circuit board will fix most of the problem.  Remaining pitch fluctuations can be fixed with six more resistors.  See the PDF for details.

It is also possible that pitch fluctuations would occur if an external Roland/DIN Sync source is driving the Devil Fish, where the Sync signals are driven with greater than 5 volts.  This is not the case for the TR-606 or TR-808, or for the TB-303's internal tempo system or with the Devil Fish MIDI In or MIDI In and Out drive of the Sync socket, which can be used with the Sync lead to drive other machines .  However, it is possible that some other sync devices output voltages to +12 or +15 volts.  The workaround is to avoid such sync sources, or have a technician construct a cable with resistors and Zener diodes to limit the voltage to 4.7 volts or so.

Update history:

© Robin Whittle 2016 First Principles and Real World Interfaces
Return to the main Devil Fish page.