The Weak Joe Optical Compressor
A Tweeker's Version of David B. Thomas'
Reverse-Engineered Joe Meek Compressor
By now many audio DIYers would be familiar with "What Compressor?", a project that apparently had its origins in one man's dissatisfaction with Alesis' 3630 compressor and his fascination with a friend's Joe Meek device.
DeeT, as he calls himself, generously shared his story and schematic but not his printed circuit board design. That's a shame because the complexity of the circuit precludes easy wiring on a Veroboard. Still, not many people would have enough skill or patience to deal with surface mount devices that he apparently used in the 10 prototypes he mentions on his page. To say little about gaining access to these SMDs in the first place.
I'm aware of one other site that deals with this intriguing design and has a PCB layout but was somewhat dismayed with that author's choice of opamps for his version. DeeT specifically mentions that substitutions to the MC33178 opamp should be a low noise device "able to drive 600 ohm(s)" loads.
The obvious choice is the bipolar NE5532 with its extremely low noise and very low distortion at 600 ohms. Unfortunately 1176neve site's printed circuit board is designed around JFET TL074 quad devices which, aside from not fulfilling the above requirement as well as the NE5532, is not pin for pin compatible with any device meeting DeeT's basic specification (there is the MC33179, a quad version and even harder-to-find part than the MC33178 though). Therefore if you want to use dual opamps even the MC33178 device originally specified (which can be obtained in a ordinary DIP as well as SO-8 surface mount package) the layout is useless without extensive modification.
Curiously, according to data sheets from Motorola and On Semi, the distortion from the MC3317* family at 600 ohms isn't all that great despite their first bullet point being "600 Ohm Output Drive Capability". The venerable NE5532 has less than half the distortion at 1kHz and greater than ten times less at 10kHz! DeeT is right, however, about selecting a good opamp the optoisolator can easily present these loads when driven hard.
A Few Small Repairs
DeeT's reasoning for the lack of controls on his "What Compressor?" is understandable he wanted to build a dedicated vocal compressor. But his earlier effort, the Joe Cheep [sic], does include an attack control. Not only that, it does not contain the "April resistor" that mixes a preset ratio of straight (unprocessed) signal into the compressed one. Arguably, this makes his earlier semi-passive design more versatile than his later one.
My printed circuit includes the return of the attack control from the Joe Cheep and a variable release control. I've eliminated both the "April resistor" and the mix control of my earlier version; if you want to mix a dry signal into your compressed you should divide your signal and use a mixer. I've changed the value of the gain control to give you more range and relabled the "COMPRESSION" control to "THRESHOLD", a more correct terminology (it would make more sense to call it the compression control if this knob controlled the compression ratio, which it does not). There is a side chain input/output which could be connected to a normalled TRS 1/4" jack so you can easily access the side chain (pre 3dB/Oct filter). You should jumper these points together if you're not going to use this option.
DeeT's specification shows the compression ratio of his "What Compressor?" to be 1.5:1 a very small amount indeed. Perhaps this accounts for his unit's 'transparency' there isn't much going on in the first place. I've added a ratio control that allows the circuit to go from 1:1 to negative reduction. With the addition of the attack/release control and the ratio knob the unit has become something quite unlike the original "What Compressor?" - something that can go from subtle to extreme at the twist of a knob. I find the attack control particularly useful for allowing the transients to sneak through giving percussive sounds the most amazing "pop" when used at higher compression settings. The best thing about optical compression is that no matter how much you apply it never sounds truly bad interesting, unique or different but never bad.
Vactrol VTL5C9 and Homemade Optoisolator (foreground)
An entire study could be devoted to substituting different optoisolators into the circuit. Currently I'm testing Silonex NSL-32SR3 devices in the circuit.
The way I have implemented the ratio knob is by varying the ratio of the resistances at the voltage divider rather than changing the gain of the control voltage circuit. Originally this was done to overcome the difference between the Vactrol in the "What Compressor?" and a homemade device I made for the prototype which had a much greater 'on' resistance. This causes the input impedance of the output circuit to vary making this part of the circuit susceptible to EMI/RFI. Therefore, the PCB has been redesigned to make the leads at the LDR as short as possible. I also suggest using shielded wire at the ratio pot to pin 5 of U5-B. It may help to electrically connect the chassis of the ratio pot to the ground plane as well.
I haven't seen another schematic that implements ratio in this manner yet in fact many optical compressors have no variable ratio controls at all. It is an elegant (some might say simplistic) approach but consider that it allows the LDR to operate in the same range regardless of the ratio setting and is a true ratio control, not something which alters the gain of the control voltage circuit feeding the control device, but an element that alters the ratio of the voltage divider.
I've eliminated the extra circuitry DeeT used to generate "dual" release times. Not only would its effect be negated by the release control without extra (and possibly superfluous) circuitry but data from Silonex, Eddie Ciletti, et al. suggests that LDR recovery time changes with respect to exposure to light over time. That is, the photocell recovers more slowly when illuminated for a longer period. "Dual release" or rather constantly variable release is something which is apparently inherent in the optical gain control element itself! Still, keep in mind that the effect will occur only at the lower settings of the release knob. After a point the envelope generator will create a release time far longer than the recovery time of the LDR no matter how saturated it becomes.
All of this saves one IC and means that a complete single channel now fits on a single board no longer is there a shared IC between two seperate channels. It's not hard to see where another IC might be eliminated as well perhaps in another iteration...
The high pass filter (U1-B on the "What Compressor?" schematic) in the sidechain is an idea apparently patented by ATI at least a couple of years before "What Compressor?" appeared. I don't agree with DeeT's assertion that it makes the compressor into a de-esser. If de-essing is the goal it would be better to target the actually ess frequency rather than give the whole sidechain a 3dB/oct slope, though the filter does give the compressor a nice response to many different instruments and indeed entire mixes due to its even response to program input - never sucking out with a bass note or kick drum. It's a great idea but the fact that someone can take a high pass filter, stick it in the sidechain and patent it seems ridiculous. Luckily if you're building one for your own use (or EQing the sidechain of another compressor) you don't have to worry about paying ATI.
Finally, my Weak Joe uses 8 pin DIP dual devices which allows you to use either the DIP versions of DeeT's originally specified MC33178 and TL062 opamps or substitute any of the range of ancient to modern devices available in the dual package. You can even use a TL072 in the audio path if you really want to (though I'd suggest you use it only in the sidechain)! If someone ever builds a opamp actually better than the NE5532 you can drop that into the circuit. Use sockets for the ICs.
The ground plane ensures low noise. Since I make the boards myself and do not make plated-through holes, any grounded pad that would normally be covered by the component on the top side is connected to a trace and a ground pad to the plane that allows you to solder a wire from the bottom to top side of the board. Components that allow you to solder from the top have square pads that can be directly soldered to the component on the top ground plane. Connect all ground plane pads to their corresponding solder-side pads. If you are making a single-sided board, be sure to connect all the ground points together. These are indicated by the square pad with a cross on the top layer.
I've designed the board for direct installation of 9mm PCB mount potentiometers. You can simply connect hook up wire to these points if you are using different pots. The bias trimmers sit neatly at the back where they can be easily accessed to calibrate the optoisolators. All other connection points are on 2/10" (5.08mm) spacing for easy hook-up wire or PCB mount screw terminal connections. A single channel compressor fits neatly on a board approximately 4 inches square.
This printed circuit board has been verified as 100% functional.
Each board is 3.775" wide and 3.9" deep. Reduce the artwork as required.
Some preliminary information on operating weak joe and a photo of the prototype can be found on the weak users page
To hear Weak Joe in action, go to my Audio Page.
- 27 March 2004 Added schematic
- 14 August 2003 Fixed error in side chain bottom layer
- 25 July 2003 Minor corrections to ground plane layer.
- 15 July 2003 First Weak Joe Compressor.
- 30 June 2003 Last "What Compressor?" version D.
- 26 June 2003 Release control added, mix knob refined. Version C.
- 15 June 2003 Mix knob added. Version B.
- 13 June 2003 Original PCB of plain "What Compressor?" Unreleased.