Must FETs in the OS be matched or checked?

[lowvolt]

I've read seemingly confliciting information addressing matching the jfets used in the common Orange Squeazer compressor.  I've been messing with the GGG OSQ. jsyk.  I've found some mods for use with bass guitar, and some other little bits of dope here and there.  Some have said that the 5457s MUST be matched (a process I am as yet ignorant of), others have said that it's not necessary.  The GGG instructions mention nothing of this process as well.

I have noticed that sometimes when I build one, it seems to sound a little different (better?) than a previous attempt.  Is this due to the inconsistancies of our beloved fets?  Sometimes it seems as though there is a bit longer attack time.  I always use pre-tinned 24guage wire, 50v 105c 2k-hr load-life Nichicon caps, brand-name ICs from Mouser, 1% metal film resistors, and the fets are supposed to be Fairchilds.  Although I can't tell if they truly are ... I mean the little "F" is there, but sometimes it's a double-outlined "F", other times it's a solid "F" .. so who knows?

I am not a commercial builder, I'm just a hobbyist with a sickness.  I've built about a half dozen of them that end up in various projects or on friends' pedalboards.  And as I said there seems to be some inconsistancies.  Any ideas?  Is it the fet-matching thing?  Would it be prudent to measure/record the values when I happen to build a "good" one?

Thank you.

[midwayfair]

There's no reason for the FETs to be matched, and anyone who told you they needed to be misunderstood the circuit. One FET is a constant current source, the other (the one with the bias trimpot) is the variable resistance element.

The variance in attack times very likely boils down to the 20% tolerance on the 4.7uF cap after the diode, the forward voltage of the diode, the total gain of the op amp (which will vary according to the pretty wide tolerances of a handful of caps), the exactness of the biasing, and any number of other things. You can't step in the same river twice.

[R O Tiree]

I'd concur with Jon completely on these issues.

The only things I'd add are as follows:

1.  The trimpot can really be considered to be adjusting the voltage at the junction between the Drain of the current-source FET and the Source of the resistive FET (even though it is directly connected to both).  I've played around with the values around this area of the circuit at some length to try to alter the "cross-over" voltage that I've mentioned on a few occasions in threads about the OSQ (more on that later...) but only 2 things that I've tried work, and it isn't the trimpot/2k4 resistor combo. I've even tried dropping the 2k4 to 220R... that gives a trim-pot value somewhere in the region of 2k or so. Put the values back "to stock" and the trimpot will be somewhere around 5k or so with a JFET with a V_GSoff of around 2.5V. At 100Hz input frequency (almost as low as your low "E" string) the reactance of the 4.7µF cap in parallel with the trimpot is only 338 ohms and this clearly drops with increasing frequency, which makes this cap the driving factor in passing the "waste" signal to GND from the Source of the resistive JFET of this circuit, not the trimpot.

2.  The other big variable in this circuit is the resistive JFET itself. The 2N5457 that the OP is using has V_GSoff values from -0.5V to -6.0V from the datasheet.  Together with the R_DS characteristic, and as far as I've been able to prove, these are two of only four factors that affect the "cross-over" voltage that I keep banging on about with this circuit. By this I mean (in order to avoid the OP having to look up other threads) the point at which V_In = V_Out. Below this value of input signal, the output is amplified. Above it, the signal is attenuated. With a V_GSoff figure at the low end (-0.5V), the cross-over will be around 200mV or so - great for someone with single coils, but for someone with humbuckers pushing out +/- 1V or so, it will be totally squished and with a comparatively quiet output. With a figure of -3V or so, that will probably put the cross-over at around 1V. The humbucker guy will be very happy with this, leaving the single coil chap a lot less happy - his 200mV pickup output will be well into the amplification region of the compression curve AND IT WILL BE TOO DAMN LOUD, so he'll turn it down... given that he's already looking at a gain of about 4 or 5 at that input signal level, he'll turn the output down to about 20-25% of max, so the maximum gain he'll ever see in these circumstances will be around 4... the result will be compression, sure, very subtle indeed and quite "transparent" (picking dynamics well preserved) but with little sustain. The chap trying to run EMGs into either of these compressors will be massively unimpressed. Even if we take a JFET right at the top of the range (-6V) he's looking at a cross-over voltage of around +/-2V, which is still pitiful compared with his max output of around 4.5V, so he'll still wonder what all the fuss was about with this otherwise elegant circuit.

3.  The third factor that affects the cross-over voltage is the resistor connected from the input cap/resistor to the Drain of the resistive JFET (stock value 82k).  Dropping this value raises the cross-over voltage reasonably easily (1/4 of the original value pretty much doubles the cross-over point).  Sadly, it doesn't work very well the other way... an 820k (10 times the original) only seems to drop the cross-over point by about 30% or so in the circuit I'm playing with (YMMV).

4.  The fourth factor is obviously the resistor connected from the opamp output pin to the negative input pin... Halving the value (from the stock value of 220k) lowers the cross-over point again by only about 30%, but with the unwelcome result that the max gain is halved... not what we're looking for, then. Similarly, raising this value raises the cross-over point, but max gain is also raised and, with it, any noise in the signal - again quite unwelcome results.

5.  Unless I've missed something, all this would appear to leave the resistive JFET as the primary governor of this aspect of the compressor's behaviour.

Edit for typo

[bluebunny]

^^^ yep, completely agree with Jon and Mike.  Built two so far and not given a second thought to matching JFETs.  Both work perfectly.

[lowvolt]

EXCELLENT information, folks.  Thanks so much.  As I stated in another thread I'm just getting started with getting to better understand semiconductors.  I have an excellent grasp of the basic diode and rectifier, but the fet family and opamps are new, as well as the transistor.  I'm just beginning to get into fuzz circuits as well.  I still need to make up an HFE tester (or buy one).

...... The variance in attack times very likely boils down to the 20% tolerance on the 4.7uF cap after the diode, the forward voltage of the diode, the total gain of the op amp (which will vary according to the pretty wide tolerances of a handful of caps), the exactness of the biasing, and any number of other things. You can't step in the same river twice.

So is there a trimpot that can be installed or perhaps replacing a fixed resistor with a trimmer or pot that could produce a variable attack time?

[R O Tiree]

The circuit is designed to react very fast.  As soon as the output from the opamp exceeds +0.3V the Ge diode conducts and the current has 2 choices... try to get to GND via the 100k resistor or charge up the 4µ7F cap.  No contest really.  So the cap charges rapidly and raises the voltage at the Gate of the resistive FET which lowers its R_DS, allowing some of the AC signal to escape through the FET and onward to GND through the other 4µ7F cap attached to the FET's Source pin.  It MUST do this fast for one very simple reason... with no input signal, the gain is pre-set at 23.  In fact the gain of the opamp is always 23 - by allowing a significant proportion of a large input signal to escape to GND before the opamp, then whatever is left gets amplified 23 times, but we've limited the output by limiting the input.  So, let's hit it with a big chord from a humbucker equipped guitar, pushing out +/-1V.  The opamp is going to try to produce +/-23V (impossible with 9V supply, clearly) and it will therefore clip horrendously until the output signal can be brought under control. Even a single-coil guitar, pushing out +/-200mV will develop +/-4.6V out of the opamp during the first cycle of the input signal.  If you try to slow down the Attack, you'll get a discernible period of farty, blatty distortion which will sound utterly terrible, instead of just one or two cycles which is extremely unlikely to be noticed.

[lowvolt]

Hmmm ... Like I said I've had some of them put together with a noticably longer attack time, and it doesn't distort, at least not audibly.  I don't know how to explain what I'm hearing vs what you've detailed out here.  So I have no viable argument.  All I know is what I'm hearing. 

Going by what member "midwayfair" said about the variations I'm hearing are partly due to tolerances in the "cap after the diode" it makes me think that there are some possibilities to have some control over attack time.

Ok, so if I'm understanding this properly, what if you change the value of the 4.7uf cap after D1 diode (like putting a 1.0uf in parallel with the 4.7uf to create a total of 5.7uf or more) ... wouldn't adding capacitance increase attack time, and vice-verse?  Or increasing/decreasing the value of the resistor before D1 diode (R11)?

One other gnewb question, why must it be a Ge diode?

[midwayfair]

Hmmm ... Like I said I've had some of them put together with a noticably longer attack time, and it doesn't distort, at least not audibly.  I don't know how to explain what I'm hearing vs what you've detailed out here.  So I have no viable argument.  All I know is what I'm hearing. 

Going by what member "midwayfair" said about the variations I'm hearing are partly due to tolerances in the "cap after the diode" it makes me think that there are some possibilities to have some control over attack time.

Ok, so if I'm understanding this properly, what if you change the value of the 4.7uf cap after D1 diode (like putting a 1.0uf in parallel with the 4.7uf to create a total of 5.7uf or more) ... wouldn't adding capacitance increase attack time, and vice-verse?  Or increasing/decreasing the value of the resistor before D1 diode (R11)?

One other gnewb question, why must it be a Ge diode?

1. The diode does not need to be a Ge. 1N60P, or any other Schottky, really, is just as good and cheaper. It's very close to the Fv of the 1n100 in the original, which is harder to get.

2. Increasing that 4.7uF cap CAN give you more sustain in parts where the compressor is actually working. You won't notice it much if you aren't playing lots of notes together. Basically, the cap has more ... "room" ... to charge up before it's fully charged and the resistor starts siphoning off the excess to ground. Decreasing it makes the sustain shorter. I used 2.2uF in one of my favorites.

Playing with the cap is not very productive in practice. Changing the resistor in series with the diode is also not terribly prodcutive. I find it much more useful to do one of the following if you want to alter the compression characteristics:
1) Create a "perfect" bias control. Put a B1K pot in series with the 10K bias trim, and put a 2.7K across the outside lugs. This creates a ~700Ohm pot, with a large portion of its travel (thanks to the reverse log taper we've created) in the "best" range, the absolute minimum setting in the "boost" range, and the maximum around the top of the
still makes noise" range.

2) Create a gain control. Put a 10K pot in series with a [1.5K-4.7K] resistor in place of the 10K on pin 2 of the op amp. This creates variable gain, and it's the only "real" compression knob you can add. It'll go from "less" to "more" compression than the stock OS. Very good especially in cases where the guitar pickups might not actually trigger the compression much.

3) Swap the 100K in parallel with the 4.7uF hold cap for a 500k pot. This is okay; it makes a variable decay control much easier than playing with the cap.

I'm not sure there's really a way to vary the "attack" in the OS -- or really any other stompbox compressor. The "attack" is tied to whether the compressor is actually "on" at the time the note is played. This seems to be a common misconception. You can make it harder for the compressor to turn on with the size of any given signal, but that's really part of the threshold; and you can make it turn off quicker so your next note will have some attack, but that's really part of the decay. The actual speed at which the compressor turns on, however, will be set by the variable resistance element, which can be very fast (a FET or other transistor) or slower (a vactrol -- or really, how fast the LED inside the vactrol lights up). In studio compressors, especially with digital compression, you might be able to set up different conditions.

[lowvolt]

Yea, in analog synth language that is known as "single trigger" .. meaning that the gates won't reset until there is no gate rec'd.  So the envelopes won't retrigger until you completely lift from playing (for even just s split second) only then will the envelopes retrigger when you commence playing.

I liken the "attack effect" of the comp to that situation.  You're not going to hear another "attack" until the comp totally resets, and that only happens when there is a no-input situation (or darned near it).  The caps have to discharge before another ~attack~ is heard.  This can be good, or bad, depending on what one expects.

Thanks for the help.    I've taken notes, which gives me lots to think about. 

[R O Tiree]

When "tuning" an OSQ by ear, there's a fairly wide band of trimpot values where this thing sounds OK to good. That said, there can be some pretty weird dynamic effects at pick-strike and for a little while afterwards as the voltages sort themselves out if the biasing is not spot-on. Could this be the variation you're experiencing?

[lowvolt]

When "tuning" an OSQ by ear, there's a fairly wide band of trimpot values where this thing sounds OK to good. That said, there can be some pretty weird dynamic effects at pick-strike and for a little while afterwards as the voltages sort themselves out if the biasing is not spot-on. Could this be the variation you're experiencing?

Sure .... yea, sure it could.  Perception is 100% of reality.  I do know for certain that some of them have turned out noticeably different than others.  But as has been pointed out, it's most likely due to spec tolerances.  What I did NOT do was record the actual observed values of the caps I installed in each construct.  So .... sky's the limit on what it is that caused the variations.

I'm realizing that things weren't the way I thought they were as far as how each one sounded.