synthgeek's blog

I originally wanted to call it “Chimera”, but then I remembered there’s already a synth company using that name, so I went with Cerberus to avoid confusion, and because “Tricephalic Filter” just doesn’t have the same ring to it. I didn’t bother renaming the mp3′s below. But I digress…

Like its namesake, this filter has three “heads”, or in this case, filter modes. The main influences here were René Schmitz’ Late MS-20 Filter, Osamu Hoshuyama’s VCF 1984B and fake SSM2040 VCF, and Ray Wilson’s VCSVF (this was another reason I wanted to use Chimera, to mark it as a composite beast- oh well). Like all of these filters (as well as many others), the Cerberus is based on OTA’s (Operational Transconductance Amplifiers)- in this case, the LM13700.

The  LM13700 datasheet contains a schematic for a single-ended voltage controlled resistor, which is how it’s being used here- in fact, the circuit used in the filter(s) is not much different from the application circuit shown in the datasheet. Each filter stage (of which there are four here) consists of one such VC resistor as part of an R/C filter network, with each stage buffered before going on to the next. This is a handy thing about the LM13700- it includes a buffer for each OTA on the chip. These four stages arranged in a serial configuration give us the “core” of a 4-pole lowpass filter. That’s cool, but what if we want other filter types?

This is where Ray Wilson’s filter comes in. Nice guy that he is, he explains how his VCSVF works quite well at his site. In a nutshell, the highpass filter is achieved by subtracting the lowpass signal from the input, and the bandpass is the highpass signal through the first filter stage. This actually creates a BPF which is steeper on the highpass side than the lowpass side. Since the Cerberus filter has four stages instead of two, we’ll use the second filter stage instead of the first for this.

On to the resonance. This is, in my opinion, what makes a filter. I noticed that the René Schmitz filter, as well as other MS-20 clones and similar filters, uses the trick of using diodes (green LEDs here, I also tried red) in a soft-clipping configuration, like many distortion stompboxes. I decided to give it a try here as well (with the amplifier part configured a bit differently)- works great. I added a potentiometer to the feedback path so you can increase the gain for more distortion. There is another interesting thing about the MS-20-style filter resonance, which is that it is returned through the first stage capacitor, which means the feedback is, in effect, bandpass filtered. Ray Wilson’s filter goes about this in a slightly different way, but the resonance is still bandpassed. I went with a configuration like the MS-20-style filters, taking the feedback signal from the output of the final filter stage. I also tried it with the feedback being taken from the bandpass output, but I preferred the former. This means that the bandpass filter’s resonance is filtered differently from the filter itself, but it sounds good to me, so I’m going with it.

Here’s an early shot of the breadboard (tweaked since then, stuff added also):

… and here’s the schematic:

This may get tweaked a little more from here, but probably only as far as changing the input stage (U1a)- I’m thinking of making the gain adjustable, and adding a pair of diodes which can be switched out (I liked 1N914′s here instead of LEDs), to add another possible “color”. Also, you may notice that the resonance amp and input stage share a dual op amp- this is done so that you can swap out different chips, and in doing so, perhaps alter the sound of the distortion in both places.

You may wish to replace R41 with a 1k resistor, and add an inline 1k trim pot to make the CV response adjustable. I don’t care about 1V/oct compatibility or anything, and I liked the response when I tried the 2k, so there it is.

Here’s some audio… note that I haven’t added a cutoff knob yet.

First up, the lowpass output with a saw wave input, modulation from a triangle LFO and the Semi-Random Source. Later, the lowpass is turned down and the bandpass comes in (using the Phoenix audio mixer), then input is turned down to nothing, and you hear just the filter’s self-oscillation (first BP mode, later LP). The resonance gain is varied throughout. Sorry for the pops, still on breadboard, so there’s occasionally some handling noise when using the pots.

Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.

Next, the bandpass output with white noise as the input:

Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.

This project was started as pure experimentation, toying with op amp distortion. Although it takes inspiration from many other designs (nothing entirely new here, just overdriven op amps), it is not a clone or mod of any specific device. Also, though it certainly might work with guitar (and sounds pretty good on my bass I think), it wasn’t designed for it. Its purpose will be as a distortion channel-type preamp for the modular amp project I mentioned in the ReMock+ post.

For the distortion circuit itself, I referenced the article “Cook Your Own Distortion” from the General Guitar Gadgets site, as well as designs by Runoffgroove and Mark Hammer. There are, of course, several things that could be modified- some of them are mentioned in the schematic notes, those and a few others will be discussed here. This is not to say that this post will cover the entire range of things that can be done with an op amp-based distortion unit, just several ideas for tailoring this one to your own purposes. Perhaps the most obvious would be trying different op amps- I tried a few, and preferred the MC1458 for this device- but, as they say, YMMV.

Getting into the rest of the circuitry, it may seem like overkill to have two gain knobs and a drive control- and it probably is- but they are actually all useful.  However, the second gain knob could certainly be left out or changed, while still retaining a good range of different sounds. The configuration shown gives a variable gain factor from about 5.5 (5.45 repeating actually) to 48 in the second gain stage. As mentioned in the above-linked article, the formula for the gain factor is (R1 + R2)/R2, where R1 is the resistance in the feedback path, and R2 is the resistance to ground.  Just as an example, to keep it at the low side of the current configuration, you could change the feedback resistor to 47k, and just remove the 100k pot (leaving the 10k resistor where it is) for a gain factor of 5.7. This would still give a bit of clipping in the second stage.

The capacitors C2, C3, C7 and C8 (along with associated resistors) create simple filters in the feedback loops. Changing these will alter the sound quite a bit. In the current configuration, they act as bandpass filters which act mostly on the lows- the first stage keeps most of the lows intact, while the second stage cuts the lows for an edgier sound. C3 & C8 remove the very high frequencies- around roughly 33kHz.

Getting back to the the drive control, this could probably be left out of you’re building this as a guitar effect, since your guitar’s volume knob would perform the same function. I’ll be going the other way, and leaving out the output volume control, since in my setup it will always be plugged into something with an input volume control- probably a mixer most of the time.

Another thing you could do to simplify things is to remove the LED’s and/or switch from the feedback path of the first stage. You could go the other way and get more complicated here as well, but I personally chose to save that for another project. As mentioned in the schematic notes, one LED I used is some strange multi-color thing that came out of a computer. I only tried it because it was here, it turned out I liked the sound, so I went with it. I also tried two red LED’s, and it was much more subtle.

On to the tone stack: I chose to place it between the gain stages in order to make it like having two distortions in series, with an EQ between them (which is essentially what it is). As a starting point, I used the “Fender” setting in Duncan’s Tone Stack Calculator- first, changing some component values in the simulation to get started, then tweaking values further on the breadboard until I was happy with the sound. My main goal here was to get rid of as much of the mid scoop as possible, while still retaining a decent range of control. The current configuration still gives a slight dip around 400Hz, also known as the “mud zone”, so we can live with that. As is often the case with passive tone controls, there is quite a bit of interaction, though in this case, I consider that a feature. Specifically, what would have been the treble control is now more of a spectrum tilter-turning it one way simultaneously boosts the highs and cuts the lows, and vice versa. The bass control is still a bass control. I replaced the mid control with a fixed resistor- lowering the value here will cut the highs, but if you make the value too low, it will affect the entire frequency range, and become a volume cut instead of a tone control. I would suggest going no lower than 33k, but again, YMMV. Below is the simulated frequency response plot from TSC. The red line corresponds to flat tilt and bass knob settings, the green and pink lines are the two extremes- bass all the way down, tilt all the way to the treble side, and vice versa. The white line is both controls all the way up:

The schematic includes a bypass switch for the tone stack, which could also be left out. Speaking of the schematic, here it is:

With the tone stack between two distorting gain stages like this, it acts more as a way to adjust the character of the second stage’s distortion than an equalizer for the actual sound coming out of the effect. While this is done by design, you could certainly move it to the end of the circuit if you prefer. I had actually considered including a simple lowpass tone control at the end- but there’s already so many knobs, and as part of the larger project, I plan to build at least one EQ-only module.

Here’s a recording of a sine wave from the K2000 being processed, with various controls being swept/switched:

Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.

We’ll follow up with a few bass riffs… first, both stages cranked, with the LED disengaged:

Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.

Now, with the second stage backed off (minimum gain), first one still maxxed out:

Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.

This time, both stages backed off, but not quite at minimum:

Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.

The next two have the LED engaged, first lowish gain on both stages:

Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.

Now with both stages cranked:

Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.

I recently came across this article: http://www.fairchildsemi.com/an/AN/AN-88.pdf , which details the use of CMOS devices in linear mode. One of the techniques shown uses inverters as amplifier blocks, turning them into something like op amps. This inspired me to try using them as an input buffer/preamp for a 386-based mini amp.

As noted in the linked article, the response is far from linear as the signal swings close to the rails, so anything other than squarewaves will get distorted to the point of hard clipping as the input is driven harder. That means it functions as a distortion device as well as a mixer/amp- in fact, I would almost say it’s more a distortion device that also happens to have an input mixer and speaker output.

Here’s the schematic as it stands so far:

As noted in the red-boxed portion of the schematic, more inputs can be added simply. Per-channel mutes would also be a simple addition, just add SPST switches between the pots and input capacitors. The tone control could also be copied per-input if desired. The next section (U1a-U1c and associated components) is the input buffer/preamp, which is implemented via 1/2 of a CD4069 hex inverter. Note the 47pF cap in the feedback loop- though not specified in the Fairchild article, it is necessary to prevent oscillation and unwanted noise. Higher or lower values did not work.  The 1M pot, however, you can mess with (lower values will give you less drive)- just make sure you leave the 10k resistor there to provide a minimum resistance. To simplify this section, you could find the resistance that gives you the gain you need, and replace VR2 & R3 with a single resistor of that value. Note that removing inverters from the signal path will also decrease gain. Consult the Fairchild article for further information about this.

I did try adding a 1M resistor to ground on the input and plugging in my bass, but wasn’t impressed with the results. If you would like to be able to use this device with hi-Z instrument-level inputs, you could try messing around with that some more. I no longer have a guitar available to try it with, so I didn’t go any further down that road myself.

The next part is the power amp section of the amplifier, provided by an LM386. This section is somewhat similar to, and was inspired by, the main amplifier part of the Beavis Audio Noisy Cricket design. If you wanted to make this section simpler, there are a couple of things you could do. The easiest way would be to remove the 1k audio taper pot between pins 1 and 8, and the 100nF capacitor from pin 7 (Bypass) to ground (this capacitor is only needed when using higher gains), leaving all 3 of those pins unconnected. This would remove the option for added gain/distortion from this section, but it’s a bit OTT anyway, so that’s no big deal. Another option would be to leave the Bypass cap in place, but replace the pot with a SPST switch and a 1k resistor inline, thus having a low gain/high gain switch. It may seem counter-intuitive that adding resistance here increases gain, but it’s because the 386 has an internal 1.35k resistor between pins 1 and 8, so adding a resistor or pot there actually decreases the resistance, due to parallel resistors.

The final sections are the outputs. The 10 Ohm resistor (R5) sets the output impedance for the speaker output. The other branch provides a buffered line output- R7 and R8 form a voltage divider to tame the level. Volume controls could be added to both outputs by replacing R7 and R8 with a voltage divider via potentiometer, like the inputs- pin 3 would connect to the output caps, pin 2 to the output, pin 1 to ground- and adding the same thing between the speaker output and its capacitor.

You may notice there are two unused inverters, I have some ideas for making use of them, so I didn’t delete them from the schematic in the editor yet. As per proper CMOS practice, these unused inputs should be tied to ground if you build this as it stands.

It can be powered from a 9V battery, though I don’t know how long it would last. I would guess battery life should be decent, comparable to the Noisy Cricket and similar designs, but that’s just a guess. I’m currently using a 9V wall-wart supply. It should be possible to run it at anything between 4 and 12 volts, as both chips are capable of that operating range. However, the information in the Fairchild article suggests better performance at higher voltages, as that gives more headroom for the CMOS amplifiers, keeping the input from “hitting the rails” and going into hard clipping. I chose 9V simply because I have an adapter handy that fits the bill, and it makes for easy portability. Also, it fits with the original spirit of the mini-amp devices which inspired it.

Here’s a shot of the breadboard build:

…and here’s some audio- first, a sine wave from the K2000 (first dry into a different mixer, then into the mossifier at lowish gain, then at high gain):

Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.

This is what the high gain part looks like on a waveform display- as you can see, the clipping is not symmetrical:

The next one starts with a 40106 osc, starting at lowish gain, then sweeping into higher gain, and back down around the mid point. Then it switches to an actual patch from Loid being played through it, first with just one input, then a second is brought in later:

Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.

The speaker output is actually pretty loud- definitely louder than the Dirty Cow amp. I successfully drove both a 6″ bookshelf-type speaker, and a 12″ Peavey PA speaker with good results.

Before calling this finished, I plan to try to improve the tone stack. I have ideas for using the two unused inverters there, for an active filter setup. Failing that, I may attempt an inverter-based version of the passive buffered BPF from Loid. More to come…

Complete with examples, graphs, reference, and the ability to save your output as an image or audio file.

http://madtealab.com/

No actual video, only plays audio & shows the cover image (loaded with all 22 tracks from IDE: Interface Vol 1):