Lowering the noise/hiss of an Ibanez CP9 compressor/limiter?

[Speeddemon]

Hey guys, I was modding my CP9 a little, so that it wouldn't cut off so much low-end (replace the input cap of 0.047uF with a 0.1uF or 0.15uF) and replaced the crappy red LED (red LEDs on a red enclosure are not smart, IMO) with a clearlense green one (had to raise the 3k3 ohm LED resistor to 22k!), but I was wondering if there are any simple and effective ways to clean up the signal path, other than replacing ALL the resistors with metalfilm ones.

I've noticed some tantalum caps, some electrolytics, maybe they could be replaced? I replaced the 3 knobs too, so it looks kinda custom now. 

And for those of you that can use the schematic, my buddy D_H has a schematic (CONTAINING ERRORS) on his webpage:
http://www.dirk-hendrik.com/images/Ibanez_CP9.gif

I don't know what the errors are, but he put the warning on his site.

Would the input stage benefit from replacing the 510k resistor with a higer one (1M Ohm)?

[Mark Hammer]

Feed it a slightly hotter input signal.  Feed it a cleaner input signal.

This is one of THE most common sources of complaint about compressors, but 95% of them simply can't help it.  A very large proportion are designed around transconductance amps that can't comfortably handle hot signals.  To avoid distortion, the input signal has to be attenuated before going to the OTA (whether LM13600/13700, CA3094, BA662, BA6110, or CA3080).  Unfortunately, the compressor works by boosting the daylights out of whatever is at its input, unless that input signal increases, at which point gain is reduced.  The input signal provides some hiss.  The compressor itself provides some hiss.  The input is attenuated to avoid distortion, but not always enough that the residual hiss, amplified many times, is hard to hear.  As a result, compressors always "hiss out" when you stop playing.

The ideal is to either:
a) use a much better OTA/VCA that can handle hot input signals, providing a better signal to noise ratio on the output
b) use an LDR or something else that is equally immune to level-related distortion, again, for the purposes of maximizing S/N

If you boost the signal a wee bit (I wouldn't go more than x4, personally) with a clean, clean, clean low-noise JFET buffer, the compressor may be able to handle that increased input level without clipping.  Because the input is hotter, you compensate by turning the compression amount down a bit, and turning the pedal volume down a bit.

Ultimately, though, having a guitar signal which is as hum and hiss free as can humanly be provided is the best solution.  This is why a recommended practice is to stick a compressor at the very start of your signal chain, so that no additional pedal-related hiss has an opportunity to be added to the signal ultimately fed to the compressor.  It's also a reason why the very first cable from guitar to pedal needs to be as short and decent as you can manage.  The one from last pedal to amp can afford to be lower quality, but NOT the first one.

[Speeddemon]

Hey Mark, I think I got the first part covered;
I'm using a 4,5m (15') Monster Rock cable and the only pedal before the CP9 is a true bypass wah (Tech 21 Killer Wail).

When used clean, I don't find it that hissy, but I use it a lot to add a bit boost, tightness and sustain to my H&K Tubefactor (medium/high gain setting on Factor 2) and then ofcourse it's hiss-city.
I've had good results in the past with modding my Ibanez TS9 RI, especially replacing 3 electrolytics (1uF) by film caps, 'cleaned' up the sound, but I don't know if something similar can work for the CP9.

It has a LM13600, is that one perhaps replaceable with something better/low-noise? Or some of the JFET's perhaps?

[Joe Kramer]

Hey Speeddemon,

You might try replacing the 2SC2458 xstrs with something quieter, like 2N5089/MPSA18.  It's especially important for the input xstr, since the compressor is going to amplify any noise from there.  The 2SC2458 doesn't look too bad in the specs, but I wouldn't be surprised if the 2N5089 is a little quieter in actual practice.  Here's the datasheets just for comparison:
 
http://www.datasheetcatalog.com/datasheets_pdf/2/S/C/2/2SC2458.shtml

http://www.datasheetcatalog.com/datasheets_pdf/2/N/5/0/2N5089.shtml

Also, it may help a little to add small tantalums (1-2 uF) to bypass the xstr collectors and the PS pins on the OTA chip.   Good luck!

Joe

[Mark Hammer]

The Fairchilds (5089) claim to be 3db noise figure, compared to 10db in the Toshibas (2458), so you might be onto something there.

[Speeddemon]

Hmm, they have different pinouts. I wonder if I can bend the legs in such a way, that they don't hit eachother.
I have a few 5089's around (from a Ibanez SM9 cloning project, I still haven't gotten around. That's one tricky distortion!), but I don't know whether they're from Fairchild. Do the others have similar low noise, or is this brand-dependant?

Would changing just the one near the input help, or the one at the output as well?

[Joe Kramer]

Hey SD,

Might as well do both 5089s.  If one helps at all, two will be better.  Just use a piece of insulation stripped off some wire on one of the xstr legs to prevent a short.

Looking at the schemo again, it's interesting they use the integrated Darlington output buffers of the 13600 on the "bypassed" signal.  No telling how noisy those might be.  In fact, you might try replacing it with the LM13700 instead.  Here's an excerpt from that datasheet:

The LM13700 series consists of two current controlled transconductance amplifiers, each with differential inputs and a push-pull output. The two amplifiers share common supplies but otherwise operate independently. Linearizing diodes are provided at the inputs to reduce distortion and allow higher input levels. The result is a 10 dB signal-to-noise improvement referenced to 0.5 percent THD. High impedance buffers are provided which are especially designed to complement the dynamic range of the amplifiers. The output buffers of the LM13700 differ from those of the LM13600 in that their input bias currents (and hence their output DC levels) are independent of IABC. This may result in performance superior to that of the LM13600 in audio applications.

If you try this, better be sure about your de-soldering skills first.  Put in a socket too.

Joe

[Speeddemon]

Hey Joe, nothing wrong with my desoldering skills. 
I've done the low-noise mod to one of my Boss GE-7's and replaced all 4 of the op-amps, as well as numerous TS9 mods.

So, to recoup, changing the input and output transistor + changing the op-amp would give a lower noise floor?

[Joe Kramer]

Hey SD,

Mouser, 513-NJM13700D.  Keep in mind, I'm not certain changing from the 13600 to the 13700 would be an improvement, but if I were able to swing the parts, I'd probably find it a worthwhile experiment.  As always, YMMV.  You could start with the xstrs that you have on hand, and go from there if that doesn't make big enough improvement.


Joe

[Speeddemon]

Hey Joe, I just noticed that Banzai does have the LM13700 for sale, but they were under 'misc. op-amps'.
They have a JRC13600 as well. Both are about $1.15 each.

The more I use my CP9, the more I'm convinced that THIS is a compressor worth cloning, next to all those Ross and Dyna-clones. I have an original (no LED, block logo) Dyna Comp, and while it's a great comp on singlecoil equipped guitars, it's too dark with humbuckers/ Les Pauls.
The CP9 adds a bit high-end, that makes humbuckers cut through better. Also, it's fairly transparent (after the input cap mod), not as much country chick'n pick'n squash goin' on. But then again, if you're a Nashville cat, playin' on Tele's, the CP9 ain't your thang. 

Now I wonder whether all the stuff in the schematic is necessary for a clone, seeing that it comprises so much more parts than a Ross/Dyna layout.

[Jay Doyle]

[Edit: I don't think going to a 13700 is a good idea, it will require some design changes just to get in there and without biasing the diodes, you aren't going to get much bang for your buck, just slightly better buffers (which you don't have to use anyway and can replace with any buffer type you want) but if you do linearize the diodes, the function of the OTA blocks change and with it, the operability of the circuit requiring circuit changes that would probably be close to a 25% redesign (the whole FWR). I think it would be easier to work with what you've got, there really isn't all that much to this circuit in the audio path, that you can control at least.]

Try a search through my old posts on the Ross Comp. This is essentially the same circuit with the addition of the Darlington buffer in the 13600 before the phase splitter and the 'Attack' control. Without the linerarization diodes biased, the 13600 (and the 13700 BTW, good catch Joe, but the diodes have to be biased, which also changes other important aspects) is essentially a 3080, though I don't know why the Iabc port on the unused OTA is driven by the output before the Darlington, but I'd guess it has something to do with the supply current/voltage and the 13600, that may be a good place to experiment (actually, now that I've read Joe's quote it is because they use the output buffer of the 'unused' OTA for the clean signal, reading between the lines in the quote, the 13600's buffers ARE operationally dependant on Iabc current). The whole mess on the collectors in the 'Attack' control is like the 150k resistor on the Ross.

In sum, you basically need to follow the signal path, the control path can be ignored (Meaning no need to change anything after, and including, the phase splitter transistor). Which, in this case doesn't leave much because the Darlington in the 13600 is used. You definitely want to change the 150k before the buffer to metal film. That will make a difference, I promise, and is probably the easiest way to improve either the Ross or this circuit in one quick step. The output of an OTA is a current so any resistance here is going to have a lot of current noise across it, making metal film the obvious choice. You could not use the Darlington, which may not be a bad breadboard experiment, it doesn't matter what the buffer is made from as long as that 150k stays 150k, the cap is there to limit the 13600's speed, it is extremely fast, and you can play with the cap to try to cut some of the high hiss out. In total there are 8 caps and 14 resistors I see that would be worth experimenting with, along with the input and output buffers (I'd suggest MPSA18s), just follow the signal path...

[Joe Kramer]

Hey Jay,

I was under the impression that the 13700 was a drop-in replacement for the 13600.  I didn't realize it required different biasing--thanks for setting me straight. 

Semi-on-topic: last week I built a Ross/Dyna comp with an electret mic hardwired to the input, for use as an "auto-leveler" for a cassette recorder.  I went pretty much with the stock schemo (more Dyna than Ross actually), the only exceptions being 2N5088 xstrs, and 1/2 watt MF resistors throughout.  FWIW, it's remarkably quiet and tonally neutral.  I plan to build one for proper guitar use eventually, but I'll have a look at your Ross schemo before I do. . . .

SD, you might consider just doing a clone after all if you really want the cleanest possible compressor.  That way you won't have anything in the signal path you don't want, like all those switching FETs and so forth. . . .

Joe

[Jay Doyle]

Hey Jay,

I was under the impression that the 13700 was a drop-in replacement for the 13600.  I didn't realize it required different biasing--thanks for setting me straight.

What sucks right now is that everything you've said is right. 

It isn't the chip that is the issue but the CIRCUIT. Now that I look at it closer, it wouldn't be a big deal to do. Just remove the 100k resistor from the output of the 'used' OTA to the Iabc port of the 'unused' buffer. But even doing so only gets you slightly better buffers (with DC offset being the biggest advantage, something we really don't have to worry about a whole lot here) because you haven't biased the diodes and if you were to bias the diodes...

You see what I'm getting at? It isn't that you are wrong, just that I don't think the quality to price ratio of switching chips is worth the effort. If I were building from scratch, I'd follow your advice to the T. And if I had this thing on the breadboard, I'd definitely fiddle with biasing the diodes but if I'm remembering right, doing so changes the Iabc/gm curve pretty drastically so the whole FWR will have to be tweaked (hence the breadboard).

Semi-on-topic: last week I built a Ross/Dyna comp with an electret mic hardwired to the input, for use as an "auto-leveler" for a cassette recorder.  I went pretty much with the stock schemo (more Dyna than Ross actually), the only exceptions being 2N5088 xstrs, and 1/2 watt MF resistors throughout.  FWIW, it's remarkably quiet and tonally neutral.  I plan to build one for proper guitar use eventually, but I'll have a look at your Ross schemo before I do. . . .

This is definitely a case where metal film REALLY helps a circuit. Unfortunately it is now past the point where OTA chips are cheap, but swapping out a couple of 3080s is a good idea. Other than that, it sounds like you know exactly what to do.

SD, you might consider just doing a clone after all if you really want the cleanest possible compressor.  That way you won't have anything in the signal path you don't want, like all those switching FETs and so forth. . . .

Agreed.

[Joe Kramer]

Hey Jay and SD,

Here's the earlier Ross post (written by Jay, slightly edited by Joe).  Lots of good info here for future reference!   


Some suggestions about noise and the transistors involved in the Ross Comp.
-----------------
1/.Lose the 0.0022uF cap at the input, I put a 22pF there to eat any RF that might have ideas about getting into my sound.
-- The original Ross has a 220 pF cap for C2, the .0022 will cut too much treble.
2/. I'd increase C15 (RG's schematic diagram) to get good bass response, (1uF mkt cap would be small in size, and would do the trick).
-- The low freq -3 dB roll off point is about 64 Hz, that should be low enough for everything but bass, I think the lowest note on a guitar is 80 Hz. A .1 uF will make the roll off 26.5 Hz.
3/.I'd either remove C11, or replace it with a value like 1000pF or less. Omitting the cap will result in boosted high frequency response.
-- It definitely will boost the highs, but way too much. You don't want to take it out completely, the 3080 has a current output and can reach into the MegaHz so you will want to limit that with this cap.
4/.I'd also fool with C10, first try removing it to see how it sounds with out it. Try replacing the 0.01uF cap with a 0.1uF cap, the important thing here is to see if the compressor distorts. Another different option would be to reduce the value of R11 (for about the same output as the stock setup), and bypass it with a small cap, again 1000pF or less. This could give some of the sparkle you are after.
-- I'd leave this as is and switch C2 to 220 pF to see how that improves the high freq. response. If that doesn't work for you, lower C10, but definitely do not leave it out. Most of those caps in the Ross are there to smooth out the high frequency spikes of the input signal to minimize false triggering of the compressor mechanism.
5/. Experiment with the values of C13, and C14. Larger values I would think should result in more compression of bass frequencies, under the right conditions of course.
-- The rolloff points of C13 and C14 with the 1 Meg resistors following them are about 16 Hz, way below what a guitar will put out. No need to change them.
6/. I did notice while faultfinding, that it took a couple of seconds for C16 to charge to 8VDC-ish. To get faster responses you could try reducing the size of the cap. You could also try experimenting with the value of R18, though this would possibly limit the sustain (gain) of the compressor.
-- What you are seeing as the cap charges is the decay. The attack happens as the cap discharges through the collector-emitter paths. Raising R18 will increase the decay time, lowering it will decrease the decay time. Lowering the value of the cap will lower both the attack and decay times, raising it will have the opposite effect.
---------------------------
Personally, I would forget the 2SC1849s and just go with the 2N5088s. IMO matching isn't much of an issue but the noise will be. I can see how matching the two transistors in the level detector/full wave rectifier MAY have some effect but they are both running at such high gain (no emitter resistors) that I doubt it would be an issue for any transistors that are over 100 hFE.
The gain and noise will definitely play a part with the other transistors. The input transistor is running as an emitter/voltage follower. The lower the noise here the better as any noise that is injected here will be boosted by the CA3080. The higher the gain (OF THE TRANSISTOR), the closer the output gain of the emitter follower will be to one. Though I must mention that the difference will be negligible in regards to hearing any difference.
Of the other transistors, you also will want a low noise transistor for the phase inverter that drives the full wave rectifier as the output signal is taken from this transistors emitter.
The transistor that drives the Iabc port of the 3080 OTA may make the circuit sensitive to it's gain, but I doubt that a big difference would be noticed.

Quote from: Darren
: CA3080 vs. CA13600
: I've also read that the CA13600 is a low-noise version on the CA3080. Has
: anybody ever built a Ross Compressor clone using these parts or have any
: experience using the CA13600 IC at all that can attest to it's quality?

Here they are definitely NOT easily swapable. The 13600 is a dual OTA and is 16 pins, the 3080 is a single. You would have to design around the difference. The 13600 may be quieter, I personally don't know.
If I were going to design for lowest noise in this circuit I would:
A - Pick the lowest noise transistor I could find for the input emitter follower.
B - Change the 150k resistor directly after the OTA to a metal film.
C - Audition a couple of CA3080 chips to find the lowest noise one.
If you still find a good bit of noise, start changing more of the resistors around the OTA to metal film. (You may want to try and reduce the values of the 1Meg resistors before the OTA, anywhere down to 220k or even 100k, though I haven't tried this). Though keep in mind that any upward compressor is going to be noisy without any signal as it boosts whatever is at it's input regardless of the source. Nothing that has gain is noiseless.

[Speeddemon]

Hey Jay, you're starting to lose me there...

When I look at the schem, I see 2 150k resistors; 1 parallel to a 3n3F cap and the other attached to a 10k (trim) pot.

Which one do you mean as the output buffer?

Also regarding cloning; I wonder if this could all be fitted in a 1590B enclosure, with battery option AND TB switching... 

[Jay Doyle]

Hey Jay, you're starting to lose me there...

When I look at the schem, I see 2 150k resistors; 1 parallel to a 3n3F cap and the other attached to a 10k (trim) pot.

Which one do you mean as the output buffer?

Also regarding cloning; I wonder if this could all be fitted in a 1590B enclosure, with battery option AND TB switching... 

The one right under pin 11 in the schem, in parallel with the 3n3.

I don't see it fitting in a B easily but it could be done it would all be in the PCB design; when you take out the FET switching (which also removes any need to use the other half of the 13600) there isn't a whole lot left.

[Jay Doyle]

Hey Jay and SD,

Here's the earlier Ross post (written by Jay, slightly edited by Joe).  Lots of good info here for future reference!   

Joe, I appreciate your quoting me but I would have rather you had not edited it any, there are some things that you edited that I don't agree with, please see below...

Some suggestions about noise and the transistors involved in the Ross Comp.
-----------------
JOE: 1/.Lose the 0.0022uF cap at the input, I put a 22pF there to eat any RF that might have ideas about getting into my sound.
JAY: -- The original Ross has a 220 pF cap for C2, the .0022 will cut too much treble.

I found that in the Ross, lowering this too much can cause false firing of the FWR because of high energy, high frequency components that occur in a guitar signal, especially after a distortion.

JOE: 2/. I'd increase C15 (RG's schematic diagram) to get good bass response, (1uF mkt cap would be small in size, and would do the trick).
JAY: -- The low freq -3 dB roll off point is about 64 Hz, that should be low enough for everything but bass, I think the lowest note on a guitar is 80 Hz. A .1 uF will make the roll off 26.5 Hz.

The -3dB point is below the lowest frequency on guitar, I don't see any need to change this cap unless you want to CUT bass response by reducing it, or are going to be using this compressor for bass.

JOE: 3/.I'd either remove C11, or replace it with a value like 1000pF or less. Omitting the cap will result in boosted high frequency response.
JAY: -- It definitely will boost the highs, but way too much. You don't want to take it out completely, the 3080 has a current output and can reach into the MegaHz so you will want to limit that with this cap.

Do not remove C11, the cap after the OTA, altogether and do not lower it too much. This is THE source of injected noise in the circuit so a good, high quality cap is in order. Also, OTAs have incredibly fast slew rates, because they are current output, so they can easily reach into the MHz. Without a cap there, or one of too small a value, the OTA could be oscillating and you wouldn't be able to hear it but it wouldn't sound good. I'd stick with a metal film resistor and a good poly-type cap of near the original value and fiddle with response elsewhere.

JOE: 4/.I'd also fool with C10, first try removing it to see how it sounds with out it. Try replacing the 0.01uF cap with a 0.1uF cap, the important thing here is to see if the compressor distorts. Another different option would be to reduce the value of R11 (for about the same output as the stock setup), and bypass it with a small cap, again 1000pF or less. This could give some of the sparkle you are after.
JAY: -- I'd leave this as is and switch C2 to 220 pF to see how that improves the high freq. response. If that doesn't work for you, lower C10, but definitely do not leave it out. Most of those caps in the Ross are there to smooth out the high frequency spikes of the input signal to minimize false triggering of the compressor mechanism.

I found that playing with those caps altered the sound a tad but not so much as to make it all that worthwhile. They are there mostly to filter the DC signal being fed to the + input on the OTA for balance and feedthrough noise cancelation.

JOE: 5/. Experiment with the values of C13, and C14. Larger values I would think should result in more compression of bass frequencies, under the right conditions of course.
JAY: -- The rolloff points of C13 and C14 with the 1 Meg resistors following them are about 16 Hz, way below what a guitar will put out. No need to change them.

At a roll off of 16Hz, all frequencies down to the sub-sonic are passed to the FWR. The only reason to change these is if you want to DECREASE compression on bass notes by lowering them.

6/. I did notice while faultfinding, that it took a couple of seconds for C16 to charge to 8VDC-ish. To get faster responses you could try reducing the size of the cap. You could also try experimenting with the value of R18, though this would possibly limit the sustain (gain) of the compressor.

Without a signal present, the transistors are completely off and the cap charges to V+ through the 150k resistor (R18 on R.G.'s Dyna/Ross comparison schem, or the combination of resistances around the 'attack' control on the Ibanez); when the cap is at the V+ voltage the OTA is running at max gain, when it is at ground, the OTA is running at min gain. When a signal comes in, it turns the transistors in the FWR and turns them on HARD. The cap discharges through the parallel resistance of the two emmitters in the FWR. This happens extremely fast because that resistance is on the order of a couple dozen ohms. THIS is actually the 'attack' of the unit. When the signal is removed again the transistors shut off and the only way the cap can charge back up to V+ is through the 150k resistor, this is the 'decay'.  Raising the value of the cap (C16) will increase both the attack and decay, lowering the cap value will reduce both attack and decay.

Mark Hammer has come up with a good mod for the decay portion that involves a switch to adjust the 150k resistance, a search should yield it. The Ibanex calls it attack, but it's decay.

-------------------------
Also,

I looked into it last night and while biasing the forward diodes on the 13600/13700 series does allow for a significant input headroom increase before it overloads, the gain is significantly less requiring 3-4x the amount of resistance on the output of the OTA to get the same circuit function, which means injecting 3-4x the amount of current noise from the output of the OTA. The problem is you do not get 3-4x the headroom so you don't get a chance to cancel out the noise by increasing the S/N ratio with a larger input because you are adding 3-4x the noise on the output before even feeding it a signal. In essense, by forward biasing the diodes, you get more noise and less input related distortion.

Regards,

Jay Doyle

[Joe Kramer]

Hey Jay,

The only thing I edited from your post was the first sentence, prior to the words "Some suggestions about noise. . . ."  The rest of the post I left intact, hence the "slightly edited."  However, I did fail to notice that there was a dialogue going on within the post between you and another stompmember, which you understandably mistook for my opinions.  Sorry for the confusion!       I probably should have just linked to the page in the first place.  Here's the actual page, with your post being #10, near the bottom:

http://www.diystompboxes.com/smfforum/index.php?topic=8871.0

Regarding, your advice on the Ross/Dyna, I have yet to actually breadboard this circuit and experiment with any of the values, but what you're saying about the stock values and their functions points towards leaving a good thing be.  I was very pleased with the Dyna/mic circuit I built, and it's more-or-less stock.  Not only that, but I happen to have an original grey Ross too, and there's really nothing I would particularly care to "fix" about it.

Thanks again for your insights.

Joe

[Jay Doyle]

Regarding, your advice on the Ross/Dyna, I have yet to actually breadboard this circuit and experiment with any of the values, but what you're saying about the stock values and their functions points towards leaving a good thing be.  I was very pleased with the Dyna/mic circuit I built, and it's more-or-less stock.  Not only that, but I happen to have an original grey Ross too, and there's really nothing I would particularly care to "fix" about it.

No big deal Joe, I just thought that I would expand and clarify what was in there a bit...

I owned an original Ross too and managed to buy it for $50 before they exploded. I loved it but wanted to improve on it so I dug into it...

I do think that the Ross circuit especially has been designed and tweaked very well. In my mind, if you like the compression, the only thing to do is improve the noise which can be done by swapping in a few metal fims, two transistors and auditioning a couple of 3080s. The caps can be upgraded and adjusted as needed. This circuit obviously does what it does very well because it is the basic compressor circuit for nearly every major manufacturer...

Jay