This Week on the Breadboard - The Engineer's Middle Finger Compressor

[Chuck D. Bones]

LoL. I'm glad you guys are deriving some benefit from my ramblings.

 

[music6000]

Final update:
I changed C7 (tant), C8 (film), R18, R21, VR2, VR3, added R23. 78L05s can be a little noisy and R23 kills the noise. C4 can be 22nF or 33nF. 33nF adds a little more midrange when BRIGHT is turned up. With compressors, it's a trade between speed and distortion. Fast ones can make distortion and low distortion ones have a slower attack. Setting RELEASE below noon will produce some distortion on the lower strings, but the initial pop when SUSTAIN is dimed has been minimized. Not a bad compressor, but not as good as the Delegate IMO. Next up: one more look at the Thumbsucker.

To Sum it up, Do you think it is worth building or not???

 

[Chuck D. Bones]

I'm not going to take it beyond the breadboard stage. So, in summary, no. It's given me some ideas on tweaking the Thumbsucker and also caused me to reconsider optical compressors.

 

[fig]

I'm not going to take it beyond the breadboard stage. So, in summary, no. It's given me some ideas on tweaking the Thumbsucker and also caused me to reconsider optical compressors.

Well, at least I don't feel as bad about not finishing it

 

[Chuck D. Bones]

Back to the Engineer's Thumb...



This version is sounding pretty good. Some of it is easy to kludge onto the Thumbsucker board, other parts not so much.

The ATTACK control was in the wrong place. When it was inside U2A's feedback loop, it did almost nothing because U2A would just drive harder and longer to overcome the delay caused by the ATTACK control. Located outside the feedback loop, the ATTACK control can actually slow down the attack response. I wanted a solid Vref, so I used a 78L05. Vref2 is very clean, Vref1 is not quite so clean but can source more current.
This freed-up U1B, so I configured it to drive U3 differentially. Makes for lower distortion, although the difference is small. R202 biases the linearizing diodes for very low distortion. R200 & R201 balance the currents into pins 1 & 16 of U3. I consider the THRESHOLD control to be useless at best, so I replaced it with R207 & R208. C200 was added to prevent oscillation in U2A. The BRIGHT switch was replaced with a BRIGHT control. It can add up to 8dB treble boost. C201 was added to improve DC bias stability. D6 & D7 absorb some of the initial overshoot when RATIO is maxed. R206 was added to smooth out the bottom-end of the RATIO control's rotation. R210, Q200 & D200 provide an option level indicator. Use a superbright LED for D200 and adjust R210 to vary the maximum brightness.

As Valve Wizard says in his write-up, this is a compressor, not a limiter. It reduces the dynamic range but does not flatten it. The RATIO control sets how much the dynamic range is reduced by limiting the maximum amount of gain available.

This works very well as drawn. I'll see if I can simplify it a bit to make modding the Thumbsucker easier.

One more thing: C5 & C6 must be film for minimum leakage. Aluminum electro is too leaky and even tantalum is borderline.

 

[Dan M]

Pros and cons vs The Delegate?

(Up above on the previous version you commented “not as good as the delegate”)

 

[Chuck D. Bones]

Now that I have the EMF working the way I want, I'll run another comparison with the Delegate and report back here.

 

[aquataur]

This thread is not too old, and what I have to say belongs here, and nowhere else.
There is some observations I have made independently of what Chuck writes here, and many are overlapping. Remarks in no particular order.

• The threshold pot does not control threshold in the traditional sense. Threshold is a DC value which is set somewhere in the action range of the input signal (or outside, as the case may be).
• Low threshold means a lot of the signal is affected, high threshold means little or nothing of the signal is affected. In this sense, the increase sense of the threshold pot is wrong. Conversely, if naming it an input gain control, the rotational sense makes, ahem, sense. I personally view it like that.
• But I agree, the average guitar player will not know about the inner workings of a compressor, so choosing this nomenclature probably served as simplification. In this case clearly a knob termed compression is superior. And indeed, a two-knob unit was the original goal.
  This compressor uses an adjustable version of the input signal (AC) as an input to the envelope generator, while the threshold remains implicit. It thus resembles more of a traditional compressor like the 1176 that boosts or cuts the input signal for the same purpose.
• Minimum attack settings easily lead to audible clicks when the compressor engages. It also changes the timbre of the played guitar tone. This is very noticeable on short staccato picks. A medium setting lets some of the transit through, which improves this vastly. Medium is still very short on this compressor. Produce some palm muted clicks with a fast attack, behold the unnatural tone, and slowly increase this until it sounds natural.
• In my opinion this is not a design flaw, the signal overshoot is inevitable even with the shortest of attack times, unless one incorporated look-ahead, which is non nonsensical. Using such short attack times for guitars and the like is also not musical, and applying this rather simple unit universally for all sorts of instruments is neither in the intention of the designer nor possible due to oversimplifications – as we will see later. Whoever wanted to make a 5 knob unit out of this at all cost: it is not.
• The minimum settings for release as possible on the NOMAD variety are smaller than the originally recommended minimum values. Indeed this leads to some breathing. Again, a medium setting works well for guitar, which it is optimized for. If a studio unit had such long release times, it would be unusable for material like drums, because the gain cell will never release before the next beat comes along. If a compressor never returns from compression in time, it becomes static, which is called static attenuation. More on this later. One would think this gets in the way, but it works eminently well for gently and unobtrusively ironing out the picking strength variations of a stringed instrument.
• The unit (with paralleled OTA’s) is exceptionally quiet. I can see no need for noise gates and the like. Any unit with (potentially) high gain will expose the predecessor’s noise.
• The make-shift reduction indicator LED (in the rail splitter’s loop) yields little additional information, which very likely is why it has not been included in the NOMAD rendering, and indeed has been dropped on Chuck’s later versions. However, it reveals the really long release times. The other circuitry suggested in the DIYstompboxes thread with the secondary current mirror is not much better in this respect and introduces other dangers, such as corruption of the Vref potential by heavy current swings.
• Setting the threshold pot to low (which would be akin to a high threshold setting on a studio unit) consequently requires to increase the ratio for a noteworthy compression. Doing this, something gets overloaded eventually and some heavy signal distortion appears. This is likely to be attributed to the OTA cell. This effect can be observed on a scope and was reported by some users. Low settings also creates some gremlins that make strange noises that you have observed, too. I thus recommend to stay away from low input gain settings.
  I finally settled on a center position and use ratio as a compress knob.
• People notice a dulling of the signal. This is a normal compressor action, particularly noticeable with fast acting compressors (short attacks), because the initial peak gets flattened, making a more sinusoidal signal out of what has been a fast rising slope. This in fact removes the higher frequency content of the transit. So this effect is not explicitly a characteristic of the ET, but of all (fast acting) compressors alike.
• I noticed a perceived reduction of bass. I say perceived, because there is the yin-yang of tone. An increase of bass will be perceived as a reduction of treble and vice versa. Also, a louder signal will be perceived as sounding better. And compressors fudge with our sense of loudness. So what frequency portion of the signal has changed in loudness? This is to be taken with a substantial grain of salt.
• Merlin says that the ratio setting effectively acts as a blend control, meaning the output signal is a mixture of the uncompressed and the compressed signal. I have a hard time to imagine the sonic extents of this in this design at low and high ratio settings.
• The above observations make me believe, that the parallel component of this compressor, which a true blend design would introduce, is rather small. A true blend resp. true parallel processor would not create such a pronounced shift in yin-yang.
• While it appears valid what he says about feed-forward designs in general and their independence of the controls, the ratio control is NOT independent. Ratio scaling would need another separate OPA stage to be independent, in my humble opinion. In this design, the clue for ratio is taken from an already compressed signal, from the output.
• Merlin says that the side-chain is input referenced, making it a feed-forward design. Such a design gets its clues from the uncompressed signal prior to the gain cell. In my view, this is not entirely true for the ET, because the ratio is derived from the already compressed output, which would make it a mixed-mode compressor.
• Ratio and threshold are two values that are normally tweaked together – setting both high resp. low for a given signal and purpose. If this sounds confusing, remember that a high threshold setting means little compression - or, to be more precise, only the highest peaks of the signal are affected by the compression action. The gravity thereof, or how much the chosen region is affected, is determined by ratio. The result of setting both controls in unison can be very different depending on the desired purpose. For example, a high threshold and high ratio setting may be used to balance the peaks on a vocal take, while a low threshold, low ratio setting may be used for a mastering compressor. Totally different effect. For the ET and its peculiar intertwined ratio control, this may not be so dramatic. It determines the side chain signal from a mixture of the input level (note: a scaled version of the whole input signal, not a region thereof set by threshold) versus an already compressed output signal.
• A genuine blending function in studio terms is called N.Y. compression, or parallel compression. This enables the user to compress the crud out of a signal without noticeable artifacts. Settings, that would probably sound horrendous on a normal serial compression (I don´t know if this term exists, but is used to distinguish it from parallel compression…) can go unnoticed because the original signal is left intact. It is also called upwards compression.
• A pronounced soft knee, or better: a large knee span, usually conceals much of the compressor’s action. On parallel compression however, this is mostly unwanted due to the aforementioned effect. Heavy compression is wanted in this case.
• Summing all above criteria up, the unit is best viewed as a two-knob device. Indeed, that is what Merlin originally suggested. And it works terrific for that. Expecting something like this, I made attack, release and threshold internal trim pots on my PCB. I concede that a tweak might be in order if used for a different instrument. Bass may be used as is due to the natural relationship.
• Merlin has originally designed the ET to be a guitar effect for stomp-box usage. It was never meant to compete with a studio unit, whose purpose generally is to work on a diversity of instruments. It may not work (well) on such. I reckon somebody smelled an advantage in selling it as such, but it excels at being a two-knob stomp box compressor for a live performance, one termed compression and one level (such as Merlin has recommended it). As a matter of fact, any such crude unit would very likely interfere with a studio recording, which is why on serious recording sessions, the signal is split early and an untreated signal is sent to the board.
• I have seen people saying that this compressor sounds inferior. Please note that a pedal-board compressor is rather a levelling tool than a sound shaping device, and is not meant to sound any particular way. Studio units can be used differently.
• So if such a unit is deemed bad sounding, then, excuse me, it is either the wrong tool or applied wrongly (leaving obvious design flaws aside), or the expectations are wrong. Compressors are a supreme discipline to master.
• My Verdict: the ET works well for ironing out a guitar’s inconsistent playing if used as a two-knob unit – compress and level. It does this unobtrusively and is very quiet. It must not be mixed up with the performance of a (digital) studio unit or a DAW’s plug-in, but it was not planned to be either, despite sales arguments that want you to make believe it can do this. And it is not necessary either. It works well as a serial compressor, but in order to be a genuine parallel processor for that N.Y. fattening effect, it would probably need an additional equal power panning stage, which again raises the question: is this needed for my Gilmourish tones that are highly processed anyway, besides drowning in the wall (pun intended) of other bombastic instruments?

 

[Chuck D. Bones]

Good write-up. That's a lot to digest.

The RATIO knob is better named SUSTAIN since it controls the maximum available gain.

I think the main reason compressors can sound "dull" or lack treble is that the high-freq content on a guitar note decays faster than the low-freq content. The compressor adjusts the gain on all frequencies equally, but because the treble/bass balance shifts going into the compressor as the note decays, the treble/bass balance shifts the same amount coming out. This assumes no feedback from the speakers to the guitar.

In the end, I decided that this compressor is not to my liking. The peak detector follows the peak signal (hence the name) which is dominated by low-freq content. With chords, there is a natural phasing/chorusing between the strings which causes the envelope to modulate up and down at a slow rate. This modulation goes thru the peak detector and modulates the gain, which creates a breathing effect that I don't like. This is a natural consequence of a simple feed-forward compressor. Some signal processing of the envelope signal could minimize the breathing, but is beyond the scope of this design.

On the plus side, OTAs respond much more quickly and precisely than an LDR. Putting the OTAs in the feedback loop minimizes noise and distortion.

 

[aquataur]

The RATIO knob is better named SUSTAIN since it controls the maximum available gain.

Yes, and because there is a feedback into the side chain, you run out of gain eventually, because ratio is derived from a signal that is already compressed. The more you wind up ratio, the more compression until this game stops eventually - you run out of gain. This means at some point the ratio does not increase any more. It has been proven.

I think the main reason compressors can sound "dull" or lack treble is that the high-freq content on a guitar note decays faster than the low-freq content.

This may well be. Roey Itzhaki in his great studio book explains how the transit gets mangled in a way that it loses its shape towards a shape that contains less high frequencies. But indeed, this does not explain the loss on a strummed guitar for example. How much of tone is determined by the initial transit, how much by the sustain part?

The peak detector follows the peak signal (hence the name) which is dominated by low-freq content.

That's the crux with all compressors. The low frequencies are more powerful. Which is why for broad-band program material you must employ some high-passing, but this addresses bass and bass drum in a mix. I agree this is beyond the scope of this unit.

One thing that is mostly ignored is the fact, that if you play on your own, like in your bedroom "studio", you want your guitar to sound full bodied. This is comprehendible, but this is most certainly not what is wanted in a band context - you have to dial some bass out. If you don't do it, the man at the mixing board will do it - if you are aware of that or not. You are competing with other instruments. But you are not liable to notice that, because the musical "space" is filled with other instruments. The mid space is reserved for guitar. So dial out this bass right from the start and live with subsequent effects in peace. Fuzz face will thank you, maybe compressors too. Those are the things that normally slip our attention when we try to emulate some big shot's sound chain.

As an aside, the revered Big Muff is mid deficient. And what does Gilmour do? He sticks a transparent overdrive after it with, guess what? Bass and Treble controls to carve out the mids...

I never grew friends with compressors for live instrument usage, but with my recent interest in Gilmour style playing and forever lasting ambient music I can see the merits of such a device. This just keeps your sloppy picking even. And since even such great musicians employ one, there must be some need for them. To an extent, this takes the life out of your playing, but sometimes it is about rhythmic patterns that sound better compressed. I thus do not have experience with such devices, neither do I plan to gain such. I can however be convinced ;-)

It has a purpose and it serves it, and the problems you describe are not apparent for what I need it. As I postulated, this is a very specific tool for a specific purpose and thus should maybe not be an always-on device.

I invested a lot of time into compressor plug-ins for DAW work, and this is a can of worms. I believe there are not many people who can truly harness them. But none of them uses compressors indiscriminately.

OTAs respond much more quickly and precisely than an LDR.

This is not always a merit. It depends entirely on the purpose. Opto compressors are the tool of choice for vocals, but some have used it successfully for other material for fresh and unusual results. Opto's are good for leveling, but some have achieved excellent results with very fast compressors.

The most forgiving method appears to be parallel processing, or several compressors in series, both dialled in with discretion.
But this is going to far. I can only recommend to spend some time with plug-ins.

 

[aquataur]

Re: compression indicator LED:

this is the circuit I am addressing in my first post. It has been suggested in the other forum. The OP suggested that it may draw its supply from Vref, which I felt was placing a very high load onto the OPA stage driving Vref. Indeed this issue has been addressed here, which in turn goes back to an application note by THAT.

Their concern was that the supply rails were corroded by the momentary current shocks. Indeed, in the above mentioned wiring this is a valid concern. Since you use a linear regulator and a post-regulator, this is mitigated somewhat.

Unfortunately, the overall benefit of such a LED is dramatically small in this unit, as different to a full-blown gain reduction meter in a plug-in.
On software compressors, such an indicator is an invaluable tool to see a) how deep your reduction is, and b) how frequently you come out of compression. A visual clue is particularly useful for all, who are not studio masters, a criterion that will be met by most of us.

FWIW, I prefer digital read-outs over needle meters, because they have their own ballistics, and you must understand, what they say, letting alone their volatile nature.

The LED circuit must turn off completely in moments of idling, which it does not, plus it must release timely, which it does not, plus it must reflect the depth of compression by a comprehensible and linear change of brightness, which LEDs are notoriously bad for. As it is, it just tells you that the compressor is doing something, which is not exactly an indispensable gain in knowledge.

THAT's comparator circuit is more aptly called an "above/below threshold indicator", which reflects, when the compressor kicks in. In our case, this will be on for most of the time, and we still do not know, how deep the compression is...

But again this discussion may be futile, since we are not talking about studio processors. In comparison, a compressor pedal appears a relatively blunt one-trick pony, which brings us back to the two-knob unit.

Re: linearization diodes

I tried this one, but for some reason it produces a shift that presents DC to the ratio pot. It crackles. Removed it again, since nothing obvious changed.