Boss CS-3 Design Choices

[Incubusguy]

Around 6 years ago I bought a Boss CS-3, then a couple of years after that I decided to do the Monte Allums Opto Mod.

Cut to now and I've finally decided to address some of the issues I've always had with this pedal. Since switching to Seymour Duncan P-Rails pickups and installing an active buffer into my guitar a couple of years ago, even at minimum Sustain there is far too much compression for my tastes when using the bridge pickup. With that in mind, I increased the top ranging resistor for the Sustain control to 10k (I never go beyond halfway anyway), meaning I can now get reasonably subtle compression at minimum Sustain.

Anyway, I've been exploring the design and there are some things I just don't get:

- The input stage appears to be some sort of non-inverting log amp, yet the THAT2159 has an internal log amp input anyway. Why have a duplicate log amp input stage (unless it's not a log amp)?

- The negative supply is used as a current source for the THAT2159. 2.4mA is the recommended max current for 15V bipolar supplies and good headroom. The negative supply terminal sits 4 diode drops below ground, so for our 9V supplies it sits around 1.7V (verified by measurement). Boss prescribed a 2.7k resistor, meaning the current available is around 630uA, far below the 2.4mA recommendation, even considering the 40% supply voltage difference. The datasheet says the core current is 430uA, meaning there is around 200uA of headroom. Swapping the resistor down to 1k for 1.7mA actually causes the device to cease output for some reason, so I need to look into that a bit more (I'm pretty sure it's not my power supply current-starving).

- The datasheet shows the THAT2159 output connected to an active low pass filter with a cutoff in the RF range. Boss, however, chose an 8.6kHz cutoff, and this output is used for the VCA feedback and is where the pedal output comes from. This means that output is always band-limited to 8.6kHz. While this might not be as noticeable for clean guitar signal input, this will surely have a big effect on the character of a distorted input? I get that filtering transients will give better VCA response (and helps cut some noise) but why not just filter in the control path instead?

- Finally, I find that I have to have the Level control at maximum in order to get unity gain, even with minimum Sustain and slowest attack. I'm pretty sure that's not normal, so if anyone has had a similar experience I'd really appreciate their input if it can save some painstaking debugging!

Here is the THAT2159 datasheet (http://www.thatcorp.com/datashts/THAT_2150-Series_Datasheet.pdf) and the compressor schematic (created by forumites Incubus and ayayay!) (http://farm4.static.flickr.com/3277/2371510004_63132d9b22_b.jpg)

So, are there some gaps in my understanding or did Boss engineers really make some strange design decisions with this one?

Thanks in advance!

[cctsim]

The input stage is a voltage follower. The diodes D2-D3 are for input protection. They are effectively connected between the opamp inputs pins 2 and 3.

The "low pass filter" at the output of the VCA is used for stability rather than filtering. From the datasheet:

"A small feedback capacitor around the output op
amp is necessary to cancel the output capacitance of
the VCA. Without it, this capacitance will destabilize
most op amps. The capacitance at pin 8 is typically
30 pf."

There is definitely something wrong with your CS3. Mine has plenty of output and unity gain is around 11 o'clock.

[Incubusguy]

The input stage is a voltage follower. The diodes D2-D3 are for input protection. They are effectively connected between the opamp inputs pins 2 and 3.

The "low pass filter" at the output of the VCA is used for stability rather than filtering. From the datasheet:

"A small feedback capacitor around the output op
amp is necessary to cancel the output capacitance of
the VCA. Without it, this capacitance will destabilize
most op amps. The capacitance at pin 8 is typically
30 pf."

There is definitely something wrong with your CS3. Mine has plenty of output and unity gain is around 11 o'clock.

Ah, that makes more sense. The 100% negative feedback looked like a follower but the positioning of the diodes made it look like a log amp, so I guessed some sort of hybrid. So it's clamping the maximum voltage swing to +/- 0.7V. Given that my bridge pickup can put out almost 2Vpeak with hard picking, I imagine I'm getting some strong clamping action going on. It's a shame the CS-3 doesn't have an input gain/attentuation pot as well.

Yeah, I noticed the feedback capacitance paragraph. Figure 14 on page 9 shows a typical application using a 47pF cap, so why did Boss choose one over a factor of 10 larger?

Hmm, I feared as much regarding the low output. Looks like I have some fun debugging ahead of me with that one!

[slacker]

The diodes are to protect the opamp's inputs, they don't clamp the signal in normal use. There's  an explanation here or google for "op amp input protection".
http://www.analog.com/library/analogDialogue/archives/41-05/input_protection.html

[ayayay!]

Here's another schem from the original.  

The first thing you may want to do is reverse the Opto Mod, since about 1/4 of it is nowhere in the signal chain, and he makes claims like "Smooth opto-like vibe" and "adds clarity and body."  It has you modify C10, C16, C18 and D1 for "better tone."  Seriously.  Read the doc.  

With that in mind, I increased the top ranging resistor for the Sustain control to 10k

- Finally, I find that I have to have the Level control at maximum in order to get unity gain, even with minimum Sustain and slowest attack.

Are we talking about R3 here?  If so, you're on the right track and answered your own question.  I think you went too low.  I wouldn't drop that below 18k.  20k is probably doable.  

On another note, I'm not one to slay every "crappy component" dragon out there.  However on compressors I have to agree that upgrading caps and resistors can help noticeably.  Metal films caps and resistors do help in this circuit.  As petemoore says, "Electrons have a habit of coming out sounding like what they were passed through."  

Hope that helps!

[Incubusguy]

The diodes are to protect the opamp's inputs, they don't clamp the signal in normal use. There's  an explanation here or google for "op amp input protection".
http://www.analog.com/library/analogDialogue/archives/41-05/input_protection.html

Thanks, that makes sense now.

Here's another schem from the original.  

The first thing you may want to do is reverse the Opto Mod, since about 1/4 of it is nowhere in the signal chain, and he makes claims like "Smooth opto-like vibe" and "adds clarity and body."  It has you modify C10, C16, C18 and D1 for "better tone."  Seriously.  Read the doc.  

With that in mind, I increased the top ranging resistor for the Sustain control to 10k

- Finally, I find that I have to have the Level control at maximum in order to get unity gain, even with minimum Sustain and slowest attack.

Are we talking about R3 here?  If so, you're on the right track and answered your own question.  I think you went too low.  I wouldn't drop that below 18k.  20k is probably doable.  

On another note, I'm not one to slay every "crappy component" dragon out there.  However on compressors I have to agree that upgrading caps and resistors can help noticeably.  Metal films caps and resistors do help in this circuit.  As petemoore says, "Electrons have a habit of coming out sounding like what they were passed through."  

Hope that helps!

Yeah, since revisiting it in the last few days, I came to the same conclusions about the Opto mod and ended up reverting everything. Unfortunately, 6 years ago I knew a lot less about electronics than I do now!

Actually, according to the circuit diagram you posted, it was R4 that I modified, increasing it from 4.7k up to 10k. I was getting strong compression at minimum Sustain and also never turned it past halfway with lower output pickups; therefore, I decided to shift down the range of the whole control by increasing the upper resistor of the divider, R4. I measured the Sustain pot output at around 2 o'clock rotation and calculated the 10k value from that.

Incidentally, I have been methodically swapping out some resistors which I considered key in reducing noise (i.e. any high value ones and any used in gain stages) to metal film equivalents since the noise level is quite obtrusive with distortion on. In all honesty, I can't say I've noticed much difference so far but maybe if I swap some more it will become noticeable.

I understand that changing the voltage output of the sustain control will directly affect the VCA gain, but the low output issue was there before I changed R4. Since the Opto mod only involved swapping some caps (mainly just little things like input cap if I remember rightly), the input protection diodes and the reversed-PSU diode, I can't see how any of those changes would cause such a huge drop in volume. No resistors were changed (I double-checked every carbon film one I pulled with my DMM and made sure the metal film replacement also measured the same value) so gain ratios haven't changed, meaning it must be something like a bad solder joint or broken trace/pad? I suppose I should be thankful that it's not a huge parts count so tracking down the problem won't take too long... Famous last words, perhaps?

Thanks all for the help so far  

[PRR]

Gitar speakers have SHARP roll-off above 5KHz. For good musical reason. The 8KHz rolloff is not a big isuue. You can change C5 and try.

IC1a is (as cctsim says) not a log amp and not a clipper. It is a perfect follower up to within a volt of the rails. Because you never know who/what will use/abuse it, because the input could force the follower outside the rails, D2 D3 clamp the input when follower action fails to keep up.

The bias current, nominal peak voltage, and VCA input resistor R9 33K are related. 0.6mA idle current may be switched to cancel R9 current. Clipping will happen at 33K*0.6mA= 20 Volts peak. Actual input won't be over 4V peak (with +/-15V rails it could be higher). It is nowhere near clipping through the VCA.

It is a limiter with a bypass. Dry signal should be similar to the sidechain threshold. Here the sidechain threshold is Q8 Q6, Si transistors, 0.6V peak.

> I increased the top ranging resistor for the Sustain control to 10k

That sets the no-action static gain. Your symptoms seem to be about limiting action.

> I have to have the Level control at maximum in order to get unity gain

Suggests your dry signal is hotter than 0.6V peak.

> pickup can put out almost 2Vpeak

That's extremely hot. Turn your guitar down, your amp up.

[Incubusguy]

Gitar speakers have SHARP roll-off above 5KHz. For good musical reason. The 8KHz rolloff is not a big isuue. You can change C5 and try.

IC1a is (as cctsim says) not a log amp and not a clipper. It is a perfect follower up to within a volt of the rails. Because you never know who/what will use/abuse it, because the input could force the follower outside the rails, D2 D3 clamp the input when follower action fails to keep up.

The bias current, nominal peak voltage, and VCA input resistor R9 33K are related. 0.6mA idle current may be switched to cancel R9 current. Clipping will happen at 33K*0.6mA= 20 Volts peak. Actual input won't be over 4V peak (with +/-15V rails it could be higher). It is nowhere near clipping through the VCA.

It is a limiter with a bypass. Dry signal should be similar to the sidechain threshold. Here the sidechain threshold is Q8 Q6, Si transistors, 0.6V peak.

> I increased the top ranging resistor for the Sustain control to 10k

That sets the no-action static gain. Your symptoms seem to be about limiting action.

> I have to have the Level control at maximum in order to get unity gain

Suggests your dry signal is hotter than 0.6V peak.

> pickup can put out almost 2Vpeak

That's extremely hot. Turn your guitar down, your amp up.

Ah, I guess I oversimplified things a bit when looking at the bias current! Thanks for clearing that up.

Before now, I thought the circuit looked like this:

- Input hybrid log amp feeds VCA input
- VCA output is passed through 8.6kHz low pass filter, into cut/boost tone control, level control then out through a follower
- Low pass filter output is tapped off; both signals fed through Q8 and Q6 set up as common emitter amps, but the input to Q6 is first inverted, then the two signals meet again
- If the signal is stronger than 0.7V, a Sustain bias is added; any less and the diode looks like an open circuit
- Attack adds some extra variable current supply and sets Q6/Q8 gain, and C8 determines release
- This goes through another follower and into the VCA sidechain input

Seeing that my signal was always stronger than 0.7V and, therefore, that a Sustain bias would always be added, I thought lowering that bias would help things.

Hmm, I'm curious/confused about how the Sustain upper ranging resistor, R4, sets the no-action static gain, partly because of the above sentence.
I guessed that D4/D5 were just Baker clamps for Q6/Q8. Way off on that?

Since it seems clear to me that my understanding of this sidechain control path is incredibly sketchy at best, would it be possible for someone to give a quick walkthrough, please? I'm sure that, given a rough idea, I could fill in the blanks with my existing electronics knowledge and I'd really appreciate it 

Ah, that's one problem I suppose I'll have to solve with some padding perhaps: I like to switch between single-coil and series humbucker with the P-Rails pickups, meaning that at any point my signal can be between 0.5V and 2V, depending on the pickup configuration used. So, it seems that the best solution is to engage a pad when using series coils to bring it down to a signal my CS-3 will play nicely with.

[PRR]

> If the signal is stronger than 0.7V, a Sustain bias is added

Not sure what you are saying.

Using the factory schematic posted by Jono:

In the NO-signal case, VR3 pulls-up C8. C8 voltage is buffered by Q7 to the VCA, offsetting the internal current-split amp. High voltage is HIGH gain.

We do not want HIGH gain at idle. It brings-up the universal hiss. D6 clamps C8 voltage to a player-selected level. "Sustain" is not a technical description, but does make sense to a guitarist because a higher base gain brings-up more string-decay time.

When signal peaks at IC2b output exceed 0.7V, Q6 Q8 conduct and pull C8 further down, reducing gain. This is not directly adjustable.

Evidently they expect output level to be 0.6V or some lower level set by "Level" pot. Since guitar level is "usually" 0.2V to 0.5V peak, this is reasonable.

I do not see any "curable" hiss problems. High-value resistors shunted by low impedances (such as R3 shunted by guitar; R25 shunted by C14+IC1a output, R27 shunted by VR1 ~~15K) give little hiss. IC2a should have good hiss specs, both voltage and current noise, also low bias current and ample stability. R6 should not be a 1930s cracked-carbon resistor; carbon-film should be fine though one metal-film here won't break the bank. Leaky/dry electroltyics can cause hiss in odd ways. Hard-worked limiters always seem "hissy" because the loud parts can't get louder so we bring-up the soft parts (and the background hiss).

[PRR]

If IC2a (factory plan) M5218 is a variant of 1458 or 4558, you can do a bit better.

THAT Corp suggests LF351, an old single BiFET. TL072 is nearly the same in a dual and super-common. TI is suggesting their TLE2072 has a trace lower noise-voltage than '072.

I just noted the factory-plan posted by Jono has 10K in and out of the VCA, opposed to 33K  on the plan posted at Flickr. That makes more use of the VCA idle current: 0.62mA*10K= 6.2V overload. We can't have 4V signals; however pushing closer to max current makes the THD rise from point-something toward 1%. I don't think that matters for a guitar limiter; indeed some rise of THD seems to be musically useful in strong limiting.

[Incubusguy]

> If the signal is stronger than 0.7V, a Sustain bias is added

Not sure what you are saying.

Using the factory schematic posted by Jono:

In the NO-signal case, VR3 pulls-up C8. C8 voltage is buffered by Q7 to the VCA, offsetting the internal current-split amp. High voltage is HIGH gain.

We do not want HIGH gain at idle. It brings-up the universal hiss. D6 clamps C8 voltage to a player-selected level. "Sustain" is not a technical description, but does make sense to a guitarist because a higher base gain brings-up more string-decay time.

When signal peaks at IC2b output exceed 0.7V, Q6 Q8 conduct and pull C8 further down, reducing gain. This is not directly adjustable.

Evidently they expect output level to be 0.6V or some lower level set by "Level" pot. Since guitar level is "usually" 0.2V to 0.5V peak, this is reasonable.

I do not see any "curable" hiss problems. High-value resistors shunted by low impedances (such as R3 shunted by guitar; R25 shunted by C14+IC1a output, R27 shunted by VR1 ~~15K) give little hiss. IC2a should have good hiss specs, both voltage and current noise, also low bias current and ample stability. R6 should not be a 1930s cracked-carbon resistor; carbon-film should be fine though one metal-film here won't break the bank. Leaky/dry electroltyics can cause hiss in odd ways. Hard-worked limiters always seem "hissy" because the loud parts can't get louder so we bring-up the soft parts (and the background hiss).

I was looking at it from the perspective of D6 as being either open-circuit or a short; therefore, I reasoned that D6 would only become short-circuit when the guitar signal was greater than 0.7V - I thought that the output of Q6 + Q8 was still an AC guitar signal (then again, if Q6+Q8 are essentially acting as variable resistors to ground with resistance controlled by the guitar signal at the base, wouldn't you still have a gain-adjusted version of the guitar signal at their collectors?).
IC4b buffers a bias provided by the Sustain control, so I thought that only when the diode was a short would this bias be added to the AC signal on the Q6+Q8 output (if the diode is open-circuit, it effectively disconnects IC4b's output). As a result, I thought that, at the points where the signal is greater than 0.7V, you'd get the AC guitar signal riding on this bias voltage.
Hmm, looking back at your description, is this basically what you were saying, but for the case of an applied signal? I'm just not sure I see the reason for D6; why not have a fixed resistor to constantly apply the bias from IC4b?

Since Q6 provides an inverted form of the guitar signal and Q8 provides the normal, then these are summed, is this a way of rectifying the guitar signal for an ~DC control signal to the VCA? Considering that Q6+Q8 are trying to pull down 9V, is it the case that, when the guitar signal is strong enough to make them pull C8 down to 0.7V, D6 is now reverse-biased and disconnects IC4b's output? If so, why disconnect the threshold control when the signal is strongest?

Hmm, I do apologise for the question bombardment; this is all just rolling of the top of my head as I'm trying to understand something which looks like it should be simple 

By the way, in your sentence regarding signal peaks at IC2b's output exceeding 0.7V, do you actually mean IC2a? I'm not trying to be overly-pedantic; just want to make sure I understand what's going on!

According to the M5218 datasheet (http://www.datasheetcatalog.org/datasheet/MitsubishiElectricCorporation/mXtqqut.pdf), it is a 4558 variant. I have loads of TL072s lying around so I may try substituting. Interestingly, the THAT2181 (a drop-in replacement which supercedes the discontinued THAT2150 series) datasheet now specifies an OP275, so I assume their recommendation is basically any high-quality, modern low-noise op amp.

Thanks a lot for the added explanation of the control path; I really appreciate it   I was having one of those frustrating moments where you think you have a circuit nailed, only to find out you were way off. Then again, that's usually the point where you stand to learn the most!

[PRR]

> do you actually mean IC2a?

Yes. Sharp catch.

> if Q6+Q8 are essentially acting as variable resistors

Not very variable. Assume the bases are driven hard. At 0.5V the transistor is "off", at 0.7V the transistor is "ON!". It's pretty brutal. If peaks happen, the collectors PULL C8 voltage DOWN hard.

The 47K in series with the base does soften this slam.

[Tightpants]

Sorry to bump a very old thread but all the info schmatics are already here...Soo a question for you compressor experts: I have a sadly defunct Boss ME5 and I was wondering if swapping out the VCA chip in my CS3 (DBX 1252 from the ME5 for That 2159 in the CS3) would make my CS3 sound any better? I note from the schematics that the older (and more revered) Japanese CS3 has the DBX chip and the swap could be achieved using the same board, just swapping out a couple of resistors and a cap. I should explain that I love the comp on the ME5 but the CS3 not so much, is it worth bothering (I know there are other differences in the circuit that could account for much of the different sound)?
I suppose what I'm asking is has anybody directly compared the MIJ CS3 with the DBX chip with a later model using a That chip and what were your thoughts on the sound? Thanks in advance