Synthesizer Build part-66: ROLAND TB-303 VCF.

The famous acid house filter from the Roland TB-303. A Eurorack friendly project and a ladder-filter that sounds amazing. 

This is the 14th filter on this website and this is one with a very specific sound so I thought let's make a project out of this because I think that this filter in particular will be of great interest to many people because of it's unique sound. I based my layout on a layout that Jake Jakaan made from a schematic he found online in combination with the original service manual schematic of the TB-303 (TB 303 stands for Transistor Base 303). 

I have to warn you, the filter sounds great in itself but to get that Acid-House sound out of it requires more than just this filter. That specific sound is a delicate balance between filter settings and envelope input and maybe some LFO or offset voltage added. When I first tried this filter I didn't get anything near that classic sound. However, I found some tips and instructions online which helped a lot. I posted a very helpful short video at the bottom of this article below the video demo's that tels you how to get that sound.

I got close in the end though as you can see in the demo video below. My mate Jake Jakaan built a few of these filters and he can really make them sound like a 303 should sound but then he's a professional musician. 

This is really the first filter I ever built that you have to learn how to use. I'm getting there tho ;)

A LITTLE HISTORY:

The TB303 was a bass synthesizer made by Roland and released in 1981. It was supposed to simulate bass guitars but it sounded nothing like a bass guitar and it became a commercial flop. It was taken out of production in 1984 after a run of 10.000 units. These were sold off cheaply by Roland. (If only we knew then what we know now @___@)  However, cheap second hand 303's were picked up by electronic musicians and the twirping, squelching sound became a main stay of electronic dance music genres like Acid House, Chicago house and Techno. There are now numerous clones on the market and original units fetch prices of over $3000,- on the second hand market. Originals were also modified in the 80's, adding distortion and external inputs (Nova mod).

The TB-303 was designed by Tadao Kikumoto who also designed the TR-909 drum machine. It has a single oscillator which produces a sawtooth wave or a squarewave. This goes into a 24dB/Oct lowpass ladderfilter which is manipulated by an envelope generator. 

I have read that it's actually an 18dB/Oct lowpass filter instead of 24dB but I don't know if that's true.

SCHEMATIC:

Here is the schematic. It's a bit low resolution because this was originally a file with a black background and bright green lines. I took it into Photoshop and inverted the image and brightened it up and made it more legible. I also included the transistor pinouts. All transistors on the schematic are NPN 2SC945's except for the two at the bottom marked 733. Those are two 2SA733 PNP transistors.

The capacitors in the schematic are not marked as polarized but the 10µF electrolytic caps are obviously polarized and for the 1µF you can use either type. Polarized or non-polarized. I used polarized caps in the layouts below so that you can see where the minus pole goes if you choose to use polarized caps.

Here's the filter part of the service manual schematic for reference. It has two 2K2 resistors from +12V to T1 and T2 but I think that's a misprint. It needs to be 22K:

The filter does not use any negative voltage. It is powered by +12V and it also needs a +5V powerrails which is provided by the onboard voltage regulator. The +12V goes through a 100 Ohm resistor. I wondered whether or not to include that but I wanted to see how much voltage that resistor takes off from the original 12V and it's only 0.2V so I left it in.

Staying true to the original includes using 2SC945 transistors for the NPN trannies and 2SA733 for the PNP transistors. You can however use other transistors like the BC547 and BC557 but beware when you do because you'll have to redo the layouts. The 2SC945 and 2SA733 have an unusual pinout. It's emitter to the left, collector in the middle and base to the right. I had to constantly keep this in mind when designing the layouts and it wasn't easy but I managed it in a day.

The transistor pairs at the top and bottom of the ladder and the transistors next to it with the common emitter connection need to be matched pairs!! Very important with this filter.

I came to the conclusion that my usual method of matching on Hfe didn't meet the case here so I did it with setting up a differential amplifier on a small breadboard. The method is shown below.

I ordered a hundred of the 2SC945's and made 10 matched pairs and I used those transistors in this project even with the middle trannies in the ladder. I thought I might aswell use all matched transistors but you don't have to do that. You can use other transistor types like the BC547 and BC557 which are used in the Doepfer A-103 VCF6 filter but you'll have to redesign the layout because their pinout is different from the 2SC and 2SA transistors I used.

Two resistors in the circuit have been replaced by trimmer potmeters so we can tune them in to our liking. These are in the Cutoff and Resonance control so they are important to the sound and they do make quite a difference. The way I set them was almost fully open for both (max. resistance).

MATCHING TRANSISTORS.

For this filter I didn't want to rely on just measuring Hfe and matching the transistors on that value. I used the Ian Fritz methode. I took a small piece of stripboard and set up a simple differential amplifier with two transistors. If the transistors are matched then the voltage measured between the two emitters should be zero. Make sure you let the transistors cool down after handling them with your fingers.

For D1 any silicon diode will do. The voltage drop over this diode ensures both transistors get exactly the same Collector Base voltage. Beware this setup requires a dual voltage source of +/-12V. You also need to make sure the two 100K resistors have exactly the same value.

You then need to switch the transistor positions and measure again. I didn't bother with that though. An other method is to leave one transistor in place and change the second one. If you find two transistors that display the same voltage difference from the fixed transistor, those two will be matched.

This method worked very well. I used matched pairs throughout the ladder filter and also for the differential amplifier made up of T1 and T2.

Please read the full article on transistor matching by downloading the article by Ian Fritz. 

Below is a picture of my transistor matching stripboard. I can get them matched to within 1/10,000th of a Volt or 0.1 milliVolt. I cut a DIP8 IC socket in half and connected the top and bottom pin together. I use that as socket for the transistors under test and with this setup I can measure NPN transistors with different pinouts because I have an emitter contact at the top and the bottom. I placed the sockets away from eachother to make it easier to change transistors without influencing the other transistor. I usually accept transistors that measure a difference within 0.3 milliVolt or lower. If you go to extremes with accuracy you'll be measuring until doomsday before you find a match.

LAYOUTS:

Below are the layouts for this project which are verified as ever. I used them to build my filter.

Wiring:

I numbered the transistors that are not part of the ladder, using the same order as in the schematic so you can easily understand which transistor is which when you compare it with the schematic. The light grey transistors are the 2SA733's. I included an extra CV input with the same level control as the Envelope input. 

The transistors in the ladder have the base and collector connected together so they actually function as diodes.

As you can see the envelope and CV input level potmeters have pin 3 connected to a 10K resistor and not straight to ground as is usual with input level potmeters. This is done so that the Envelope input is never fully closed. This is also the case in the original Roland TB-303 because the envelope input is very important for the characteristic sound of this filter. In my own build I did connect pin 3 of the CV potmeter straight to ground instead of the 10K resistor because I wanted to be able to fully close that input. So I leave it up to you how you want to wire that up.

If you look closely at the audio input you can see a 220K resistor on the stripboard that isn't used. I have the audio going straight into the filter through the 1µF cap. Originally that 220K should be in series with that cap but I think the value is too high. I later experimented with a lower value but you can also leave the resistor out.

Stripboard only view:

Cuts and wirebridges seen from the component side. As always, mark the cuts on the component side with a Sharpie or Edding marker and then stick a pin through the marked holes and mark them again on the copper side. Then cut the strips at the marked positions with a sharp hand held 6- or 7mm drill bit.

Bill of materials:

PICTURES:

Here are some pictures of the build proces:

Cuts and wirebridges done:

Everything is soldered in.

Here's the design I made for the panel. Feel free to us it if you want.

And here's how the panel came out:

You can see the colours don't come out as strong with clear waterslide paper as opposed to using white waterslide paper. But I like this effect. The design shouldn't be too overpowering I think.

Here's a look at the finished module:

Side/rear view. I had built a version before this one but it didn't work but I re-used the panel I made so the mounting holes are not positioned where they need to be so that's why the M3 bolt is bent sideways.

VIDEO DEMO:

This filter has that typical ladder filter quirck where if you turn the resonance up the volume goes down and you get less bass. Most ladder filters have this characteristic. The Moog ladderfilter does it and even the Doepfer A-103 VCF-6 does it. 

Here's demo, trying to get that Acid sound using some of the tips from the YouTube short video below. I came close but it's not quite there. I had a slowly rising sinewave on the CV input and a short pulsing envelope, with just some decay and all the other parameters of the ADSR closed. Instead of an LFO I think an offset voltage alone would be better. You can hear it reaches that "eeeuuurrrghhh" sound as the LFO rises in voltage but then it gets too high and it starts to whistle more. I'm going to do more experiments, using the voltage processor and see where that gets me.

Here's an other short test. Beside the envelope input I also had an offset voltage going into the CV input. That offset voltage came from the dual voltage processor to which I also had a sawtooth LFO connected. I set the processor in such a way that I always fed an offset voltage to the filter but the voltage would swing between about +2V to +4V. You can do this by raising the offset and then using the attenuverter to limit the maximum voltage. A very useful module to have in combination with this filter.

Here's a YouTube short explaining how to get the characteristic 303 sound:

Documentation:

Here's the webpage of Tim Stinchcombe about the TB303 ladder filter.

Here's Ian Fritz's original article on transistor matching in PDF form

That's all for this one.

If you have any questions or remarks about this project then please put them in the comments below or post them in the special facebook group for this website.

If you enjoy this content and would like to support the projects on this website, you can buy me a coffee. There's a button for that underneath the main menu if you're on a PC or Mac. You can also use this PAYPAL.ME link and cut out the middle man. All donations go straight back into new projects and the upkeep of the website. Thank you!