Designing a 500-Series Optical Compressor Pt. 1 March 30, 2017 13:33

It’s time to let folks in on the next collaboration between JC-diy and DIYRE. Our first effort resulted in the EQP5 passive equalizer. This time around we’re doing a classic-style optical limiter.

I have always been a fan of the classic optical leveling amplifiers, and for good reason — their musical processing and simple functionality (just two knobs, one for reduction and the other for gain make-up) make it hard to produce bad results. The most famous leveling amps are the LA-2A and LA-3A from Teletronix (later Urei) which are lovely not only in the character of their dynamic processing but in the inherent tone of their signal paths. It has long been my desire to create a DIY-friendly project whose behavior evoked the spirit of these classics while being accessible in both cost and build, so now we’re making it happen! 

We are tentatively calling it the OLA5 (for Optical Leveling Amplifier 500 series ). 

So What is a “Leveling Amplifier” Anyway? 

“Leveling Amplifier” is a name given to early optical compressor/limiters, as their intention was to level (average) a signal’s dynamic range to make it easier to balance in a mix. The gain reduction element in these designs used a voltage divider featuring a light-dependent resistor (LDR) driven by a light source (an electroluminescent panel or an LED) to drop the level of the signal being processed. As the intensity of the light source (driven by its sidechain amplifier) increases, the resistance of the LDR drops, causing a reduction in level. Because the reduction is produced by a passive device (a resistor) there tends to be less obvious distortion in optical compressors compared to other designs with active gain components.

Additionally, LDRs possess an inherently non-linear release characteristic which changes based on how hard they’re driven. Subtler processing yields soft, forgiving time constants, but wind them up a bit, and you’ll have plenty of grab followed by a fast initial release that slows as it gets closer to full release. This non-linear response gives these processors a unique and musical ‘leveling’ characteristic (as opposed to a more linear, VCA-based compression most of us are already quite familiar with). This characteristic is highly prized for its ability to flatter vocals, bass, and guitars in particular.

The LA-2A was a tube unit, and the LA-3A was all discrete solid state, and both had input and output transformers. By contrast, the later LA-4 was all IC-based with an electronically balanced input and a transformer balanced output. There are already existing DIY projects based on the LA-2A and LA-3A which are wonderful. If you build one of these you won’t be disappointed. However, my goal for this project was to create a simpler and more cost-effective build that still retained the classic leveling characteristic that we love in those old designs. So I chose to start with the LA-4 circuit, and then tweak from there.

Where’s the Love for the LA-4?

Despite the almost universal love for its forebears, Urei’s LA-4 hasn’t enjoyed quite the same enthusiasm from the pro audio community. In fact, I have more than once heard it referred to as the “red headed stepchild” of the LA-2A and/or LA-3A. 

Why the disrespect?

Well, the overwhelming gripe people have with the LA-4 seems to be with the signal path of a stock LA-4, not so much its dynamic behavior. Audio passing through a stock LA-4 has a tendency to sound a bit dark or ‘veiled’ while being a bit abrasive in the midrange, particularly when working with stronger signals. Both of these characteristics are a result not so much of poor design, but of the proliferation of the less-than-stellar RC4136 quad opamps that make up almost the entire signal path. 

The lackluster 4136 lowered production cost and simplified the build, but it also limited the speed of the audio (called ‘slew rate limiting’) making signals dull. The remedy for this is a switch to better opamps, which is a challenge given the unique pinout of the 4136. But we’re talking an entirely new, ground up DIY project here, so opamp selection isn’t a problem for us... we can use whatever IC’s we want!

Upgrading the opamps reveals the gain reduction circuit's refined, vintage dynamic character for which classic optos are coveted. My first prototype coupled my own electronically balanced input stage with the LA-4’s original output circuit (with modern substitutes for the actives), and it sounded great. I have since swapped the output stage for the proven output arrangement I used in the EQP5 Passive Equalizer. So this unit’s signal path offers the same, exceptional performance as that in the EQP5 including the option for the vintage discrete opamp/transformer output bundle to give it more old-school love!

Expanded Feature Set and Functionality

We've added or improved features compared to the LA-4 design, including sidechain HPF, better stereo linking, a legit, discrete opto cell (instead of the off-the-shelf vactrols that everyone else is using). I will cover all of that and the path we took to get here in a future post, but our immediate need is for feedback from you regarding the possible inclusion of one specific feature.

Question for you: Wet/Dry Mix?

I must admit this feature, included on the prototypes, has become a personal favorite of mine, but one that some may feel doesn’t really belong on a compressor of this heritage—the ability to do parallel processing. It seems more and more hardware and software dynamics processors these days offer on-board parallel functionality, but is this because people find it really useful, or just because it gives marketing something to boast about? 

The beauty of the LA-4 sidechain is that it excels at leveling audio signals in a musical, but not overly heavy-handed way. Yet I find that this musical sidechain, when slammed hard, does, indeed, have an attitude of its own that blends well with many unprocessed signals to affect both the perception of size as well as really sticking a signal dynamically “in your face." Under mild-to-moderate processing (which is not uncommon for such a true leveler such as this) the usefulness of parallel processing is less obvious, but when you really wind it up the process can be quite magical.

So this, dear readers, is the primary question we are struggling with regarding the design of the OLA5: should we include parallel functionality, or should we simply respect its classic heritage and stick to purely serial processing? 

What do you folks think? We've included some samples and clarifying info below. And please feel free to comment or make suggestions on any other aspect of its design too. It’s not too late to change things, and we wanna know your thoughts!

Rock Mix Samples

Guitar and Vocal Samples

Why Two Knobs for Wet/Dry? In the original prototype I used a single ‘blend’ knob that adjusted the balance between the unprocessed and processed signals, but I found that I couldn’t quite tweak its ballistics to my liking — everything from about nine o’clock to three o’clock on the pot sounded remarkably similar with only the outer throws of the pot yielding much perceived control over the balance. This really bugged me in use, so I switched gears and chose a separate fader for the “dry” input signal just like the processed signal has. While this is a different approach to a blend knob, I find it quite simple to make the desired adjustments, and both knobs provide a proper feel. Plus, those who aren’t interested in parallel processing can simply turn the dry signal all the way off.