What Has To Be Adapted In Practice

CV Input

CV Filtering
Overall Noise Considerations
Testing And Verifying The Side Chain
Power Supply

Housing And PCBs

Some Paradoxical Things...

Using The Unit

Reference

Vintage LDR or OTA based compressor pedals with their countless iterations are well known, mostly by their property of being noise generators. More modern ones seen are seemingly derivatives of application notes published by THAT Corporation and follow very simple fixed attack and decay schemes that are usually a compromise.

Acting too obvious, such units never gained any ground for live music playing due to the compression artifacts they are prone to produce.  What may work for a specific situation, can be counterproductive for a different setting. Usually those application notes rely entirely on the IC’s on-board dynamic processors for side-chain processing, driven by cost considerations.

(click on the picture to load larger image)

SOMA (SOM-2 revisited) Nick Boscorelli’s spares no effort to introduce program-dependent attack and decay timings into his SOM-2[1], which can be switched from hard knee to soft knee for a very unobtrusive compression experience maintaining very high signal quality. Indeed, the unit comes out exceptionally quiet and versatile, while still fitting into a fairly small package. Interestingly, none of the usual muffling is noticeable. Used with discretion, I never heard any pumping or similar compressor plagues. But of course, you can weaponize it and make a tool of destruction out of it... A similarly powerful unit in pedal-form cannot easily to be found. Please excuse the mediocre photographic expertship. My camera is over-whelmed with reflective parts. Or me... Take the numbers on the ratio control with a grain of salt. It appears that ratio can have negative values. The setting depicted is called the "sag" setting - probably emulation of a tube amp output stage sag... For the front plate layout, I had to make concessions to the internal layout in order to fit the boards into the limited space.

"Auto variable attack and decay are easy to accept once the player learns to trust them. Their convenience tends to spoil one for manual modes."    - Nick Boscorelli, The Stomp Box Cookbook, p.241

Yes, it is addictive - like Soma, the mysterious drug. Unfortunately the SSM2120, its heart, is gone. But fret not - there is a replacement that brings it to life again.

SOM-2  uses the dynamic range processor’s building bricks (such as the RMS unit and the VCA) bare-bone in order to get all the mundane work done, and does all the side-chain processing by himself using discrete components, thus having full control over all parameters without being hindered by the quirks of the processor chip itself.

Unfortunately, the dynamic range processor / VCA circuit used (SSM2120) is unobtainable, a fact which does not promise to change in the foreseeable future. Fortunately, THAT’s range of processors are quite similar in that they have a rectified and log-ed signal accessible and dual-control VCA’s. With small adjustments a work-alike unit can be built.

There is an application note from THAT called Replacing the SSM2120 Level Detector (DN126), which is their response to customers'  demand for a replacement. It is however by no means a straight pin-for-pin replacement, but this app note outlines the procedure in general.
THAT’s series of all-in-one compressor ICs are very similar to the SSM2120, and they also have several detector-only and VCA-only ICs that could have been used.
THAT4305 was chosen since it combines both. It has several simplifications of features we don´t need anyway. It also has only one VCA and RMS unit where the SSM has two, which is just perfect since only one is needed anyway.

Signal board: (click on the picture to load larger image) Here are the schematic and PCB that incorporate THAT's dynamic processor into the signal part of SOM-2.  The large IC in the center is representative for the DIL-16 to SMD adapter.

• The whole side chain processing is almost identical to the original, except for the decay network around the transistor. Read more further down at the testing section.

• Note that D1 has to be selected as described in the original document.

• The SSM asks for a different drive impedance. Description further down.

• The unused control input is tied to ground directly.

• The snubber network has to be adapted according to the needs of the 4305.

• Non polarized electrolytics have been used, because THAT is not consistent in the way they want the polarity.

• The current limiting resistors into VCA and RMS were optimized by trial and error to avoid clipping the VCA, but are in the same ballpark.

• There must be no resistor on the LOG output of the RMS unit to ground, where the SSM needs one. SSM uses a large resistor for establishing a timing current (10µA), whereas THAT provides a 7.5µA internal current source.

• THAT recommends a small capacitor on this output. 1nF has been found working, 10 nF filtered too much.

• No overall bias current needs to be set for the VCA (R14 – 1.5Meg on the SSM)

• The input op-amp has been chosen as single unit in order to be able to pick a low noise specimen. See later on.
  

Control board (side chain): (click on the picture to load larger image) Here are schematic and PCB for the side chain. This deviates only slightly from the original.

Back To Index

---

CV Input

The way the control inputs on the SSM are made, they ask for an impedance not higher than 200 Ohms, driven by a circuit using a suitable series limiting resistor. In other words, a voltage divider.

THAT chips won’t work like that. They ask for a driving impedance < 50 Ohms, in other words, being driven directly by an op-amp’s output. Luckily, I had one spare on the prototype. The necessary CV range is guaranteed by using approximately the same division ratio of 8.2k > 220 Ohm beforehand. This works a breeze.

Both unit’s inputs are current driven inputs, hence the series resistors are current limiting resistors.

Back To Index

---

CV Filtering

I did not see it mentioned by THAT explicitly, but it is pointed out on the SSM datasheet (and well documented otherwise): control voltage must be noise free. Since CV directly modulates the gain, any noise appearing on the control input(s) directly affects the output, so it should be as pristine and noise-free as possible.

For their own circuit suggestions, THAT applies some severe high end tailoring in almost every feedback loop of the driving OPA, with a corner frequency of something like 10k into 100 or 47nF. This cuts off at 2-300 Hz with a rise time of less than
1 ms.

A suitable capacitor has been added directly to the final voltage divider, which is not present in the original.

In any circuit, no doubt the first amplifier has a major impact on overall noise. Assuming the input is driven by a 10k input impedance, with the circuit given, noise calculations suggest using a bi-fet OPA such as the LF356, or an OP27. The OP27 is marginally better with lower impedances, with higher ones the LF356 or similar, but you will be pushed hard to hear that. Since a modern sound chain is usually low impedance, the OP27 has been finally used.

The working gain established on the input is counteracted by the output for unity gain. This quasi-companding helps to keep the noise at an absolute minimum.

Indeed the unit is exceptionally quiet for a compressor,  caused by the fact that the voltage controlled „amplifier“ is not really amplifying, but rather attenuating for most of the time.You hear no breathing and pumping.

You need an oscilloscope and a frequency generator.
Note: any component designators mentioned are in reference to the original drawing by Boscorelli.

Use op-amps with j-fet inputs for the side chain. Any of the TL0xx series will work. LM324 and other bipolar types likely will not function as expected as Boscorelli points out on several places in the appendices. I used TL074 throughout because I had them and was not pushed for current saving (they need a little extra supply current).

• remove the dynamic processor IC (THAT4305) for the test.
• Feed a 100mVp-p sinewave into the „LOG“ signal.

• Verify the presence of the sine wave on pin12 with an oscilloscope. Note the amplitude.
• Monitor pin14. The gain should be ca. 100.
• assuming all works, look at pin1. It should be at 0V
• increase signal amplitude and swiftly decrease it again. A negative going pulse should appear on that pin.

• Verify the presence of the sine wave on pin12. Note the amplitude.
• Monitor pin14. The gain should be ca. 40.
• For checking the peak detector (D7/C14) do NOT look here with an oscilloscope probe. The latter has a resistance of 1M which will significantly falsify the measurement. Look at the output of the subsequent OPA instead (pin8), which is a plain follower.
• Produce a signal step as above. Monitor the voltage at the follower to see C14 charge up.
• Decrease input signal slowly (to not trigger the fast decay network) and observe a slow discharge of C14 over seconds.
• Produce a fast on-off step as above and observe a rapid discharge of C14.

Note: Boscorelli planned for a ca. 3s discharge time according to the accompanying text. However, the only path to discharge C14 is either through capacitor leakage (not to be expected) or through the transistor. Boscorelli uses the ubiquitous 2N3904, and since this is so plain that it was replaced by a generic transistor. But this assumption was wrong. The description says that when IC1-D is at zero, Q1 would discharge C14 via R24. This did not happen. When the emitter and the base of a transistor are at the same potential (ground), it is off.

Indeed, this was happening. The transistor used (BC549C) was too good. The peak detector would charge up and then stay there indefinitely. You can use LED D1 further down the line as an indicator to see if it lights up and dims with the decaying voltage.
A fast decay would indeed discharge C14 as expected (by actively turning the transistor on), but it would not trickle discharge in normal mode. Only a leaky transistor can accomplish that. God knows what old transistor Boscorelli used for this circuit, but a reverse wired 1N4148 between C-E provided all the leakage ever needed. Note that while other diodes may work, only the 1N4148 has a foreseeable reverse leakage. In any case, verify the discharge time. At this point in time, it is not clear how this worked for him.

The value recommended for C14 (5n6) yields a decay time of 10s, which comes out too much for the targeted  decay of 3s from 5V to zero (according to his specs). Reducing C14 to 3n3 produces exactly the envisioned behavior. Adjust C14 if necessary.

►	The falling edge detector is explained on p. 177f. in the PDF.

The fast attack circuit cannot easily be verified by using an oscilloscope, since attack is naturally fast. Again, D1 can be used as an indicator. Turn up the voltage slowly until you measure about 3V on the output of IC2-c (pin7). D1 will start to glow dimly and then turn on progressively more.
Similarly, the signals at the „hard“ and „soft“ pins are not easy to judge by looking at the scope.

Back To Index

---

Power Supply

Boscorelli designed the circuit to use two 9V batteries for a symmetric supply. Those can be expected to deliver no more than +/- 8V in practice. Having built some of Boscorelli’s other projects earlier, I had a pre-made circuit ready using 9Vdc in conjunction with an LT1054 charge pump for a negative rail.

►	Boscorelli deviated with this supply substantially from his usual +/-15V supply. He normally uses a switching converter for the negative rail for all his projects with no adverse effects on signal quality.

With a total current consumption of about 50mA, a well filtered negative rail (using a passive L/C filter) delivers 7.5Vdc, so total voltage is about 16.5V. This works perfectly, and not artifacts are audible coming from the switching supply.

Symmetric PSU (+/-9Vdc) (partial view): (click on image to download complete pdf) The application is standard according to the datasheet. Use low ESR caps such as the recommended OSCON caps for low loss. An LC filter has been added to provide further attenuation of the 40kHz switching frequency. An oversized specimen (current-wise) has been chosen for L1 to achieve low wire resistance (and less loss).

PSU and bypass switching PCB: (click on the picture to load larger image) I combined power conditioning, LED and true bypass switching all on one board, which fits into a Hammond 1590N1 case. This is a very space saving approach. I use it for all circuits by Boscorelli and beyond. The "oil barrel" shaped thing is the choke.

Back To Index

---

Housing And PCBs

The prototype fits into a Hammond 1590BB2 case. All components are THT, resistors are standing. This allows for a very dense population. I never encountered adverse effects by using this technology.

THAT4305 is mounted onto a DIL-16 adapter.

Individual PCBs had been made matching the schematics. If any of the newly developed PCBs had failures beyond recovery, the damage would have been local to that PCB. The added indicator board, albeit minute, naturally resides on a separate PCB for similar reasons. A final PCB could be shrunk down considerably, but I don´t have any plans to do so.

Somebody might find it an attractive idea to make a production unit. If they do, I expect honorary mention at least.

In a follow-up article I present a true gain reduction indicator circuit that uses a single LED to inform you about the compressor's activity such as

• depth of compression
• attack and release speed
• frequency of activation vs. static compression (= static attenuation),
  

which is a huge improvement over simple below-threshold-indicators that just tell you that the compressor has been triggered.
All that at a minimal additional effort.

Back To Index

---

Some Paradoxical Things...

As fate has it, I happened to look into THE PDF again.
There is Project #7, Squeeze-O-Matic (SOM), which in retrospect, would probably have been called SOM-1, had been anticipated that there is a successor. I had not payed much attention to it, since all projects are self-contained. But that maybe was a mistake. This is the "manual" sibling of SOM-2. It has a few interesting features that wait to be explored.

Interestingly, here too Boscorelli uses the term "parallel compression" interchangeably  with "feed-forward compression", which, as we mentioned above, which in my ears and in retrospect was a very unhappy choice of words.

Also, he mentions paradoxical dynamics or paradoxical compression. He used the same term for SOM-2, achievable with ration settings bigger than two third of the range. The term paradoxical compression is used in data compression algorithms for a technically (theoretically) not achievable amount of compression. It seems not to be currently used in musicians lingo.

"...with the ratio pot set past halfway, the device exhibits paradoxical dynamics: the louder the input, the softer the output."    - Nick Boscorelli, The Stomp Box Cookbook, p.28 "Paradoxical compression results if the ratio is taken past 85%: gain falls as the signal intensifies"    - Nick Boscorelli, The Stomp Box Cookbook, p.241

However, what he refers to is currently known as negative ratio, which some DBX compressors ( incidentally now THAT) and others can do. Hence the labeling on the ratio pot.

"Turn the ratio up, and a compressor becomes a limiter. Go even further and you enter the world of reverse dynamics — and unique sonic effects."    - Mike Senior, Negative Ratio Compression

In the chapter of Dynamic Effects (p. 154ff) Boscorelli mentions on sustain:

"Downward sustain compresses the signal at high ratio and relatively low threshold. (...) [Using a] Feedback control [path] leans to subtlety, parallel  [=feed forward] control to a more forward sound, even to the point of sag."    - Nick Boscorelli, The Stomp Box Cookbook, 2nd ed., p.169

If his perception of a tube amp sag emulation hits the nail, I cannot judge.
 
Back To Index

---

Using The Unit

Boscorelli has used his brain when designing that unit. He also seems to have a very thorough understanding of the matter.
In the chapter of Dynamic Effects (p. 154ff)  he gives several hints on using this unit:

• low ratio and low threshold (p.158)
  
• medium ratio and medium threshold
• high ratio and high threshold (most audible)
• low threshold and high ratio (traditional sustain)
• hard- vs soft-knee
  

"Soft-knee compression sounds subtle enough that the player cranks the ratio so high as to defeat the point. Players seeking audible compression should stick to hard-knee transfer. Soft-knee settings are best fixed by meter, rather than boosting the ratio to the point of blatant compression."  - Nick Boscorelli, The Stomp Box Cookbook, 2nd ed., p.164-165

Read about a suitable gain reduction indicator in the follow-up article.

Back To Index

• Feb. 25, 2026: chapter "Some Paradoxical Things" and "Using The Unit"
  
• Feb. 23, 2026: first release