Build an Analog Vocoder

Intro: Build an Analog Vocoder

This is my first Instructable and my first attempt at such a large electronics project. I designed the circuit from scratch and I am making it available to all so anyone interested can build their own. I have called it the Morphatron. I will not say it is a piece of cake, but I have done most of the hard work for you, and either way it is very much worth it, because this machine can accomplish amazing things.... too bad I am so close to the deadline, I was unable to make a decent demo video.

STEP 1: Design and Schematics

The first thing we need is a circuit design, which I am posting as a series of images. I designed it from scratch, although I did investigate as many vocoder schematics I could get my hands on. I focused on using easily obtainable components that are inexpensive, and I tried to keep the circuit as simple as possible without sacrificing functionality.

A vocoder basically has two inputs and one output. The first input is the program (usually a connected to a microphone) and the second input receives a carrier signal (usually a keyboard). The program signal is then fed to an analysis section, which extracts the spectral information from the sound and applies it to the carrier signal. This vocoder will analise the signal on 14 bands, but the design can easily be modified to include more channels, or actually fewer (if you are in a hurry).

First, we must have input amplifiers for each of the two inputs. Then, we must build each channel, and finally mix all the channels. These are the three main blocks of the circuit. All the 14 analysis channels are identical, except for the values of the capacitors (I purposely designed it this way to make it easier to build).

Each channel in the analysis section consists of two identical band pass filters. The first receives the program signal which then goes into an envelope follower. The output of the envelope follower then controlls an VCA (voltage controlled amplifier) which amplifies the carrier signal coming in through the other band pass filter.

The next step contains a parts list, however, I will tell you right away that ALL the op amps in the circuit are TL074. So all the components you will need are the TL074's, resistors, condensers (capacitors), and NPN transistors (2N2222). I am trying to keep it simple. Also, you will obviously need wires, connectors, pots, jacks, and some material to put the whole thing in (I used plywood, MDF, and acrylic for the casing.

This circuit also requires a bipolar power supply (+/- 10-15 volts). I pulled out the power supply from a damaged old computer, which is bipolar, I used the +12v, ground, and -12v from this power supply (they are standardly marked yellow, black, and blue respectively in a computer power supply). Using an existing power supply will save you a lot of time and money.

In this step you should download and study the schematics just to get acquainted with the project.

STEP 2: Buy the Components

Besides the condensers listed in the previous step, I used the following components. You may want to buy a little more than the quantities I list here just in case you damage some while working.

Condensers:
17 x 4.7u
2 x 10u
14 x 150n
14 x 0.1u

Variable resistors:
16 x 50K
1 ganged 50K

Small calibration pots:
14 x 50K
14 x 1K

Jacks:
27 mono phone jacks (because each channel is patchable)
1 stereo phone jack

Cable:
3 meters of 10 wire ribbon cable
30 male 10 wire ribbon cable connector
30 female 10 wire ribbon cable connector

A bunch of wire to make all the connections (I destroyed an Ethernet cable to get this copper cable, its great because it is color coded and all).

Resistors:
28 x 1.2K
28 x 2K
2 x 1K
123 x 10K
28 x 20K
56 x 33K
28 x 120K
28 x 100K
42 x 200K
28 x 220

Transistors:
42 NPN transistors (2n2222 or similar)

Op Amps:
30 TL074

LEDs:
15 RGB

Diodes: 42 signal diodes

STEP 3: Test Build

Do not try and solder the circuit before testing the individual parts on a breadboard!

STEP 4: Solder It Up

Since I am self-taught I used standard circuit boards instead of printing the boards myself. These photos show the process of soldering up the filter bank.

Someone asked for images of the underside... so I have posted 3 more images to this step, the images are of one completed bank board (which actually has 2 filter banks on it). The first image is the top, the corresponding is the underside (flipped vertically, not horizontally), the third image is a copy of the underside with some visual cues.

STEP 5: Make the Casing

Depending on your style and the way you decided to wire the circuit up, I recommend you make your own design. I routed channels into two plywood boards to hold the filters vertically and hold the side panels. The front and side panels were cut from 4mm acrylic, and the back is cut from an 4mm MDF board. Then bolted the top and bottom lids together.

STEP 6: Enjoy

Now just get down and mess around with your new vocoder. I had never played around with one of these and I am just about scratching the surface... This vocoder (with patch cables) will allow you to invert the spectrum or transpose it, (even use external CV from analog gear like sequencers).

Credits:
The design borrows ideas from several vocoder schematics on the net. Many of these vocoders used difficult to find parts and many more components. I used a filter calculator provided on the net by Okawa Electric Design (http://sim.okawa-denshi.jp/en/OPtazyuBakeisan.htm) to design the filter banks. Also, I borrowed the concept of only varying condensers on the filter from the Okita vocoder, this makes it very easy to make all the banks because they are identical except for 8 capacitors in each; making the voltage control patchable was borrowed from the Elekvoc. The voltage controlled amplifier is based on a circuit made available from Rene Schmitz.

Comments

[Victorian DeLorean]

This circuit is just what I was looking for! One modification I think would improve the circuit is to use something like a LM13700 series OTA for the variable gain amplifier, instead of the voltage controlled transistor differential amplifier stage in the schematic. As it stands now, there will probably be significant harmonic distortion coming out of that stage as the large input signal causes the gain to vary around the quiescent gain set by the current through the 200K resistor. The LM13700 gets around this by dividing down the input voltage to a low level, using linearizing diodes to provide negative feedback, and converting the signal to a current which is more immune to noise. It would increase the cost somewhat, as they're each about $1.50 (for a dual unit).

[HeDog]

Hi all !
First of all thanks Planetone for this project, and thanks
Victorian DeLorean for this mod. I aim to try and build it as a kind of
playable instrument version, with faders instead of pots, mute buttons,
filters etc. The CVs inputs are perfect in that matter.
I think I
understand how this modification integrates in the first schematics, but
to be sure, is the "Control" input right after the 50k variable
resistor and the patchbay jack for CV/gate in ?
As for the LM13700
OTA suggestion, would it integrate "easily", as in "just replace parts
of the VCA design by it" or would it imply modifications of the design
that I am absolutely unable to do by myself ? (I am not experienced in
case you didn't already realize ^^)
Also, has anyone measured the
current drawn by the whole circuit ? I found designs for a bipolar power
supply that would be a lot smaller than a pc one, and would make it
more transportable, but I need to know how much current is needed for the circuit.
I know it's an old subject, but if anyone drops by and has answers, that would be awesome !!
Thanks

[GORDONR27]

About 20 years ago I built a 10 band vocoder using VCA's with only 1 transistor + 1 opamp, with very acceptable results, despite probably quite high levels of harmonic distortion.

[planetone]

You are absolutely right, but here in my country OTAs are rare and either really expensive or have to be ordered. This is why I tried to use a single type of op amp. There was an even more elegant solution in the Okita Vocoder... a chip that would handle envelope following and voltage controlled amp called a compander (I imagine this chip to be awesome for analog projects like compressors and the such but had no luck finding it without placing an international order). To your knowledge, is there ways I could reduce the distortion (could you explain the quiecent gain issue a little) would attenuating the signal more help?

[Victorian DeLorean]

Basically, there are 2 possible sources of distortion in the design as it is. First, the 200k resistor is only an "approximation" of a constant current source, so the gain will vary somewhat with the input signal as the voltage at the emitters of the differential pair varies along with it. The more serious issue is that if the signal voltage coming out of the filter stages is higher than about 40mV peak to peak there will be distortion due to the following: Suppose the current through the differential pair is 1 mA at some control voltage level, so each transistor draws 0.5 mA. The Ebers-Moll model says that for each 18 mV increase in base to emitter voltage, the collector current will double. If the input voltage increases 18mV, then the input transistor is carrying 2/3rds of the current, and the other is carrying 1/3rd instead of 1/2 and 1/2. At 36 mV input, the ratio will be 4/5ths and 1/5th. Pretty soon the input transistor is basically drawing all the current it can through the "current source" and it saturates, or cuts off if the signal goes in the other direction. That's the cause of the major part of the distortion and why you want to use small signal levels when going into a current-controlled VCA. Fortunately both issues should be pretty easy to address. You can divide down the signal just by putting a resistor in series with the output of the filter and making a voltage divider with the base resistor you already have to get the millivolt-range input level you want. Then you would modify the output differential amplifier to make up the gain. The tradeoff of course would be extra noise on the output, but hopefully the signal to noise ratio wouldn't be too terrible. You could modify the control current circuit to make it a true constant current source just by adding a transistor and a resistor and changing a few component values. I'll see if I can draw something up over the weekend. I think (hope!) that making these changes could improve the vocoder's performance quite a bit - I'm self taught as well so I'm just learning along with everyone else. :)

[planetone]

Thank you for this response, the information is quite helpful. I would appreciate your offer to draw up the constant current part as I am not sure how to go about this (if you find the time of course). I think it would be good to update the schematics in the Instructable to reflect these improvements. Thanx

[Victorian DeLorean]

Hi again, Planetone. I got some assistance from a more experienced engineer on another forum, and what we came up with is in the image below. Unfortunately making a good constant current source VCA without using an OTA requires a few more parts than I thought, but it's not too outrageous! Basically in the schematic, Q9, Q10, and U2 form a voltage controlled current source. This current is mirrored by Q6 and Q7 and used to control the gain of the differential amplifier, which also uses a current mirror as its load. Using this arrangement increases the transconductance of the stage and greatly improves the common mode rejection, such that the trimmer resistors probably won't be necessary. However, Q1 and Q2 should be matched and be in close thermal contact. Resistors R4 and R7 may need to be tweaked - since I don't know the voltages coming out of your carrier filter stage and modulator envelope follower stage. Basically you want to set it up so the circuit has unity gain for a carrier signal of the correct frequency when the amplitude coming out of the envelope follower stage is at its maximum value. Decreasing the value of R7 will increase the gain for a certain input control voltage. Adjusting R4 will adjust the "make up" gain, since the input signal has to be divided down to stay within the 50 mv peak to peak linear region. So for example if you use a voltage divider to divide down the input signal 100 to 1, 10k for R4 will boost the gain back up by 100. If the carrier input signal coming out of the filter stage is hot and needs to be divided down more than that, R4 should be increased.

[planetone]

Hey, I know this is real old, but could you post a larger image for your solution here... just so that others have access to this fix.

[Victorian DeLorean]

I'm sorry I didn't do this sooner! Stuff happened and I completely forgot...:(

The original circuit diagram wasn't done by me, and in looking over my larger copy I actually noticed a pretty significant error that would probably prevent the original circuit from working! I've made a few corrections and now have a circuit that should work properly (note I haven't actually tested this circuit in the project).

[Victorian DeLorean]

Here is a direct link to a large version:

http://i227.photobucket.com/albums/dd240/bitrex2007/VCA.jpg

[Victorian DeLorean]

I'll have to make a new drawing of this, unfortuantely I just noticed that the one posted here is for some reason showing up just too small.

[Aud1073cH]

curious if a larger img was coming?  thanks

[planetone]

We have to ask bitrex to help us out with that... this part is his improvement (to reduce total harmonic noise)

[dsauer]

I am sorry to hear about your troubles with getting LM13700s. The die size is so small that today the package probably now costs more than the silicon. A die photo of the LM13700 can be found at my web site at www.idea2ic.com, and a die this size should cost only a few cents to produce. But designing ICs in silly-con valley has always been like living in a Dilbert cartoon. Things make semi-sense at best.

[PierreB71]

Do you have thé pcb layoutof this vocoder,so I can etch the pcb's my self
Kind regards
Pierre bladt
Bergdal 55
B-8830 Gits
Belgium

[planetone]

Hello Pierre, I never made PCB layouts for this, just soldered it on generic prototyping PCB's. Sorry.

[fernackleberry]

WTF is a condensor in the real world?

[planetone]

A capacitor

[MatthieuR]

Hey, thanks for the inscrutable! Could you provide the schematics other that how it is today on the site? They are shown on the page as videos, and when I download the intructable's PDF they are low resolution. Thanks a lot.

[coolduder1001]

Stupid question but what is the minimum amount of potentiometers that can be used with some result?