By Jack Schoonover
August 2020, Portland, OR
It’s long been a dream of mine to own a Roland TB-303 Bassline synthesizer. That legendary sound has been in my head since long before I knew what it was, so I decided to go the extra mile in quarantine and build the awesome DinSync RE-303 (https://shop.re-303.com/product/space-cadet-bundle/). While the earlier x0xb0x had intrigued me, this one is a true replica of the original 303–same dimensions, part locations, and most importantly vibe. It was a hustle to track down choice rare parts, and I learned a lot about transistors in the process. For anyone with a decent soldering iron and a love of acid, I can’t recommend this project highly enough.
In the eight or so years since I first learned to covet the 303, I’ve done quite a lot of thinking about how my favorite acid masters (Roy of the Ravers, UR, Donato Dozzy, and RDJ of course) got their sounds. While the programming is a language that will take some time to master, much of the wilder acid I’ve loved relied on extensions of the 303 that the original absolutely fell short of. I got to spend some time with a Devilfish some time ago as well, and knew that any 303 of mine would need a little something extra.
Needless to say, before doing these mods I made really, truly 100% sure that my RE-303 was working as intended (i.e. like a “factory” unit), and calibrated it as per Dinsync’s instructions before performing any mods. When attempting any of these, make sure your unit is powered off and unplugged. Each mod you attempt, power the 303 back on and make sure everything still works–none of these should prevent you from accessing the entire original range of the instrument. Lastly, feel free to get in touch with any questions about these instructions, but I can’t be held responsible if you break your RE-303 or TB-303 attempting them.
I gathered my information from (https://www.ladyada.net/wiki/x0x/mods) and (https://www.firstpr.com.au/rwi/dfish/303-mods/), and apart from finding more convenient access points I deviated little from their findings (just implemented them to my personal taste). Enjoy!
Shorter VCA Envelope Decay
I made something of an accidental discovery early on that helped me with this one: R121 is the resistor mainly responsible for setting the decay of the VCA envelope. Its stock value is 1.5M, but I went through my whole build, including calibration, with a 220k resistor here. (Must have gotten it mixed up with R123 right above it.) The 303 sounded so good with this value, however, that I didn’t notice my error until I started poring over potential mods. Knowing I would be messing with this value as soon as I could get my hands on some larger-value pots, I took out the 220k and replaced it with the typical 1.5M, and was very surprised to hear that it made very little difference, if any.
Anyway, I got my hands on a B2M pot thinking I would follow the wisdom on the x0x Wiki that larger value = larger decay and that’s just how it is. I put the B2M pot in series with a 1k resistor, ready to get wild. 1k resistance at R121 produces a decay that is so short that the envelope almost doesn’t exist! This did not sound “intelligent,” it just sounded kind of silly. However, about 90% of the pot travel the other way failed to lengthen the decay at all! A B1M improved things slightly but at least the last 50% of that pot’s travel produced no change. Ultimately I’ve settled on a B250k pot in series with a 47k resistor in place of R121–at minimum setting this produces a nice, bouncy staccato effect, while maximum gives the “stock” decay as far as I can tell.
Some talk exists of lengthening the overall decay range by installing switched capacitors in parallel with C62, the 1uF tantalum on the PCB’s lower right edge. Disappointingly, this did not work for me and I found that doing this didn’t change anything.
Variable Accent Decay
A true complement to the regular VCA Decay mod, this one allows you to manipulate the decay of accented notes. In a “stock” 303 or workalike the accented notes have a shorter decay in order to simulate a slap-bass sound, but wiring a B250k pot between the anode of D28 and Pin 4 of IC12 allows it to be set to a maximum of about 3 seconds, same as the regular envelope. Other tutorials say to cut the trace between these two points, but this is unnecessary–simply lift D28 out of its trace on the left/anode side, stand it on its head, and wire the pot between the trace and the lead of the diode. A bit of heatshrink over the diode is definitely a good idea for preventing shorts, but overall I find this way to be much less destructive.
Manual Accent Trigger
This one made the cut on the Devilfish and with good reason: it’s nice to have some performance controls, and very simple to implement. Simply score a nice momentary On-Off pushbutton and wire one leg to Ground and the other to the base of Q35 or the north side of R140 or R141. Pushing this button while the 303 is running will Accent the next sequenced note.
Manual Slide Trigger
Who doesn’t love the sound of a good 303 slide, or many good 303 slides? Wire a momentary pushbutton with a 100k resistor between TP5 (+5.33V source point from the calibration procedure) and the right side of R59 or the base of Q29. Just like the Accent trigger, the 303 will slide to the next note instead of jumping when pressed. The CV will get lazy and won’t always make it to its next pitch when this button is held down, but that’s a cool effect in itself in my opinion. Note this one isn’t on the Devilfish and perhaps some might see it as less important than the Accent.
VCA/Gate On Forever Mod
The growling, slinky sound of a nice sliding 303 pattern can really bring the funk. So why let it decay, ever? Solder a wire from the upper-left leg of TM4 to the center pole of two SPDT switches. Connect the bottom pole of one to the south side of R133 and the bottom pole of the other to the south side of R134. These will hold the VCA and Gate on, respectively; together, the 303 will sustain endlessly which is awesome!
It seems like few if any 303s live their full lives without this one. As it turns out, the 10k R97 holds the key to the amount of resonance in a 303. As such, different values in parallel can give a user wide control over this range. For my purposes, I found a 43k resistor in parallel, giving a resistance of about 8.1k total, just gets over the top enough with the Resonance cranked without being too painful.
Extended envelope range
Simple but crucial modifications here–short out R61 so that envelope modulation of the filter can be turned off entirely, which is devastating in combination with the aforementioned Gate On/VCA On Forever mods. Likewise, a 220k resistor in parallel with R63 produces a very wide yet totally useful range for the envelope modulation, managing to just avoid some strange noisy behavior in the filter that seems to result when the envelope modulates the self-resonating filter too much.
VCO Filter Modulation
This one is just nasty, and highly recommended. From Ladyada: “Take the feed from the square wave (R36/Q8) and the saw wave (R105/Q28) and wire one to pin3 of an SPDT switch and the other to pin1 of the same switch. Use a SPDT switch so that you can select the saw or square wave as the modulation source. Connect pin2 from the switch to the 3rd leg of a 50K log External Linkpot. Wire the first leg of the pot to the junction of R71/Q9/Q11 (it is best to hook it up here instead of to ground so that it follows the same bias strategy as the env mod does – see page 8 in the 303 service manual: http://dl.analoghell.com/index.cfm?id=107). Hook up a 100K resistor to the center lug of the pot, the other end of the resistor goes to the junction of R71/R72/Q10-base, the summing point for the filter frequency control voltage input.” You could also try AC-coupling the VCO signal with a 100nF-1uF capacitor between the wiper leg and 100k resistor.
I didn’t have an A50k pot so wired a B50k pot with a 6.8k resistor between the wiper and the CCW leg going to the junction of J71/Q9/Q11. This is different from the post-VCA FM mod found on the Devilfish but I really like the different sounds that come from being able to select which VCO waveform is modulating the filter since both are running all the time anyway.
Pitch Tracking for Filter
Once you’ve replaced R97 with a lesser value and extended the range of the envelope, your 303 will really start to chirp. (Even moreso once you start FMing the filter, of course!) This is all good stuff of course, but why stop there? I found that JP34 (which is connected to pin 7 of IC11) is an effective pitch CV output, in that connecting stuff to it won’t alter the VCO’s pitch tracking. I originally opted for a linear pot for this one and wired the JP34 wire to CW, wiper to resistor to R71/R72/Q10 node (or just the end of the other 100k resistor connected to the Filter FM pot), and CCW leg to R71/Q9/Q11, so we’re working with the same reference for the voltage divider as the VCO Filter Modulation mod. The size of the pot and resistor seemingly affects how wide the range is; I went with a B250k pot with 47k resistor on the wiper which seems to give plenty of useful range without blowing anything up. (B50k with 100k resistor was ok too, and probably a little safer.) After messing around a bit, I found that the most useful implementation is to wire this pot up for anti-log behavior favoring the CW side.
This mod is a little funny, in that turning up this parameter can reduce the overall cutoff frequency, rather than just adding to it as with the Keytracking parameter on most synths. I assume this is because the VCO CV output dips below the reference for the Filter cutoff input on lower notes. Honestly, though, this doesn’t bother me very much since it’s easy, sounds great, and you can still turn it all the way down. Apparently this is how it is on the Devilfish too.
A defining feature of the 303 is its somewhat anemic sound, caused by many factors: (at the time) cheap parts, a filter seemingly designed just to skirt a patent lawsuit, strange engineering decisions, Roland’s assumption that bass guitars just sound kind of stupid, etc. It turns out that the filter circuit’s input is capable of handling a signal much hotter than the stock R62 value of 220k provides, though, and this can really beef up the sound. A B250k pot in series with a 3.3k resistor at this point creates a wide range, from a little extra oomph at 3-4 to complete filter-oblitering mayhem at maximum. Note that turning up the overdrive with this mod does increase the 303’s overall volume significantly; something like a B100k pot in series with a 130k resistor might make a nice more nuanced version.
Muffler Diode Distortion
If implemented exactly as described above, you may notice that the Pre-Filter Overdrive has a tendency to “blow through” the 303’s whistly resonant filter, especially past 1 o’clock or so, and consequently the 303’s overall volume output raises considerably as well. If left untouched, this doesn’t sound that good, since you essentially just get loud, gated oscillator when turned way up. The Muffler mod is the cure for this–it turns this range into a satisfying Fuzz Bass synthesizer capable of some very tasty tones.
The x0x Mods Wiki offered up a few ideas worth experimenting with: as per the schematic above, you can try a simple 1N4148 diode with its anode connected to the negative pole of C38 (1uF cap to the right of the Accent pot) and its cathode connected to ground. This produces a very subtle distortion at high gains. However, another 1N4148 facing the opposite direction in parallel makes the distortion *very* pronounced, producing a midrange-heavy crunch that sucks the bass away kind of like a Boss DS-1 whenever anything peaks. In my opinion it sounds aggressive and cool.
One LED configured this way, or even two reverse-biased LEDs, was much more subtle and needed a lot of gain to do anything; I suspect this has to do with the greater current needed for them to turn on. This does depend on the LED used, of course, but I didn’t get a lot out of just using LEDs here.
Given the topology of the filter (which uses NPN transistors in a diode configuration, of course), I had to try NPN transistors configured as diodes in this position as well. One transistor going to ground is again quite subtle, while an opposite pair as with the 1N4148 sounded quite a lot more pronounced, with a nice crunchy distortion that acted less on the bass frequencies than with regular diodes. This mode sounded really good in general and I considered keeping it.
However, while messing around on the breadboard I hit upon my personal favorite, which is as follows: NPN configured as diode in the forward position (Emitter and Base tied to C38 negative), LED in the reverse position (i.e. anode to ground, cathode to C38/E+B of transistor). The gain of the transistor makes a difference and my fave was ultimately the BC550C, which typically has twice the hFE of the transistors in a 303 such as the 2SC945P. This configuration has a really interesting character with a distinct slow-attack quality almost like an optical compressor with a low ratio, and indeed the LED will be quite active visually as the circuit clips. Thanks to the high gain transistor, this configuration has a good deal more bass response than the two 1N4148 method, so I’m certain a 303 in this mode would utterly obliterate a room with a good set of subs.
Those are the mods I’ve done to my RE-303 as of August 2020. I’ll try to post detail of more mods if and when I make them. Thanks for reading, and thanks again to the 303 masters who poked around and figured this stuff out.
TD-3 Gang-Style “Rubber Mod” (not recommended in combination with Filter FM mod!)
The debate about the Behringer TD-3’s very existence is too complex a subject to touch here. The good news for us is that we have many, many users getting their hands on a synth with many circuit-wise similarities to a real 303, and they’re so cheap that it’s okay to break one!
Anyway, this has birthed a whole community of modders with some interesting ideas. One that sounded cool in theory was the so-called “Rubber” envelope mod, accomplished on the RE-303 by soldering the positive leg of a 1uF or 2.2uF capacitor to the upper end of R73 where it meets Q10, then adding a switch to ground coming from the cap’s negative leg. This DOES give the filter envelope a neat (if rather subtle) “wub,” especially with the resonance cranked. What they DON’T tell you is that a capacitor here will cause the VCO Filter FM mod to work somewhat regardless of that knob’s setting, so you kind of get both or nothing with this mod engaged. So, this one didn’t make the final cut.