That corp.'s "One Knob Squeezer"

Started by Processaurus, January 08, 2007, 04:11:52 AM

Previous topic - Next topic


Might make a quality, simple to control stompbox compressor.  About the design they say "The circuit allows for the continuous adjustment of gain, threshold and compression ratio with a single potentiometer, forming the basis for compressor designs which are more intuitively accessible to a wider range of users."   

mostly built around a single THAT 4301P chip, which Small Bear sells.  To guitar adapt I'm guessing it might benefit from a high impedance FET input stage (and perhaps some gain to get it up to a level the circuit expects) and a volume control. 


Very cool. I downloaded it for a later, more "in depth" read.



Was looking through the PDF this am while on a flight (nice chunk of uninterupted reading time :)

It looks like it requires a bipolar supply. Also, I'm not quite sure what those MD... transistors are, haven't seen that symbol in a schematic before. Any ideas?

Finally, I think you are right about a buffer in front for guitar use.


Analogue Signal Processing


Quote from: A.S.P. on January 08, 2007, 04:25:40 PM
woodwork squeekz

I agree. I think.

Appears that the MD984 parts are obsolete. I wonder if That's 4315, being a newer chip has more up to date app notes.


Quote from: A.S.P. on January 08, 2007, 04:25:40 PM
woodwork squeekz

... and out come the freekz.

An email to That folks and they might recommend suitable substitutions.

~ Charlie
"I tend to like anything that I think sounds good."


I emailed them about substitutions for the weird transistors.  The only datasheet I could find on them says "special order from factory only".  Hmm.. OK... :icon_rolleyes:


That corp got back to me and reccomended the MBT3906DW1T1-D for the pnp dual and MBT3904DW1T1-D for the npn dual.  When I asked if they needed to be dual packages for matching purposes (though the datasheet said nothing about these being matched), or if the design will work fine with single transistors, they wrote back:

Dear Ben,
You'll get better matching with the parts packaged together.  Though
they don't state it, and they don't make any warranties, the fact is
that they choose devices from adjacent locations on the wafer, and the
matching is almost always pretty good. My former boss (and I) at a
different company went through this exercise with a similar part;
Manufacturers do not take step to randomize the selection of devices.
They always select adjacent devices, unless the adjacent device does not
work, and only occasionally is there a step from wafer to wafer (in the
same lot), and only rarely is there a jump from lot to lot.  You'll
almost always get good matching with the devices I mentioned. 
As to whether all of this is necessary, that's up to you.  You'll get
better performance with these fairly well matched (though not specified)
devices than you will with totally unmatched devices (The best
performance will be with matched devices like the THAT320).  I don't
know if the performance hit will be an issue to you (and your customers)
or not.  However, I know that using discrete, unmatched transistors
will work.
Best regards,
Frank Thomas

So thats a solid "maybe".  Its like a psychologist asking: "But the question is, do you feel its important the transistors need to be matched, Processaurus?"  Hmm... I guess not, doc... its just a guitar compressor...


However, I know that using discrete, unmatched transistors will work.

Sounds like a gray area to me.

Mouser shows the parts mentioned as non-stocked, and Jameco has never heard of them.

I wonder if the THAT part has advantages over the the SSM2166 chip that has been mentioned before.... I'm working up a layout for a SSM2166 with a JFET buffer in front, hope to have it up and running soon.


I'm intrigued by this.  I need to DIY a pedal compressor and was wondering if anyone took this very far. 

wayne kirkwood

THAT Corporation commissioned me to do a modern update to the classic dbx 163 "One Knob Squeezer" using the new low-voltage, low-current THAT4316 Analog Engine.
This is a preview to the DIY community is in advance of its publication as a Design Note.

The article, which includes a circuit description can be found here

Battery-Powered One Knob Squeezer Guitar Compressor Using the THAT4316 Analog Engine



You might want to increase the input resistance if you keep the bypass as drawn.  Passive guitar and bass have an interaction with the cable and input of what follows. 

wayne kirkwood

Thanks Gus for the input.
That's easily done. I agree that it is a bit on the low side.
The input R1 could easily be 10M; R2 is limited by the bias current of the 33178 but could also be made larger than shown.

We actually thought about showing a full bypass with LED but decided to keep that part simple figuring the OEM had their own ways of doing it already.
I see that the forum's store has a deal on 3PDT switches.


Thanks for posting your design here, Wayne! 

Regarding the bypass, the "True Bypass" used by stompbox people is pretty universal currently, I would recommend either showing it with a 3PDT true bypass scheme with LED, or no bypass shown at all.  The currently shown bypass is known as the "tone sucking" bypass, because old pedals with low input impedance (old wah wah pedals especially) would load down the guitar pickups, even when the pedal was bypassed, causing the guitar to not sound as good, when that pedal was in the chain.  Yours and other opamp based designs have a higher impedance, but the tone sucking bypass is still avoided to this day, because it is convenient and helpful to guitarists to know that when a pedal is bypassed, it is 100% out of the circuit, and can be trusted it is not doing anything to the sound of the guitar when it is bypassed.

Beginners might appreciate everything they need for a finished pedal being shown in the schematic, and pros would be able to easily, mentally erase the bypass, if they have a different bypass method in mind.  If you do decide to show a true bypass, the version that grounds the input would be good for your circuit, to avoid any oscillations from the input being left floating, which can happen with high gain circuits.  An example can be seen here, on the GGG big muff schematic.

Have you published any other designs for THAT? Interesting company!

wayne kirkwood

Thanks Ben for the feedback.

Your comments regarding 3PDT bypass echo those exactly of THAT's CTO when I submitted the design for review. It was the "sticker shock" of the $30 Carling (Mouser X-wing price) that made me go for the "mental erase" strategy since the original article was targeted to OEMs. I didn't want to suggest a switch that might be the most expensive thing on the BOM.

Are they serious?

I agree completely however that it's probably best to show a true DPDT bypass with a third LED pole. I like the idea of back-grounding the de-selected input you referenced and will see if I can get THAT to edit the drawing to show a 3PDT switch.

I've written a number of things for THAT over the years as well as done proof-of-concepts using their product for both them and myself.

I admit to not doing much work on guitar effects at all since the 70s when Bill Hall and I did the Dynaflanger which was really intended as a studio box and became a guitar effect by happy coincidence. I'm not a player but occasionally come up with circuit ideas I know you guys might like. I have two EFX projects in the pipeline one of which is about to beta test. The layout for that one uses a 3PDT which along with the case appear to be the two most expensive things on the BOM.


> "sticker shock" of the $30 Carling (Mouser X-wing price)

Mouser is another planet. Most pedal-builders know that their market demand has driven 3PDT stomp switch prices down down down in the last decade. Even StewMac will sell you a brand-name 3PDT for $12. This forum's STORE has them for around $6. This offensively named and MSIE-unfriendly site has $2-$3. eBay prices run $2 ea in 10-lots. asks a buck-fifty in lots of 1,000.

But 100-1000 DIY sales won't keep THAT Corp's ovens warm. THAT would love to do a deal for 100,000 chips. Which means someone who can sell into Guitar Shack and other mega-retailers. The people who build pedals on that scale do, as you say, have their own ideas about how to do a bypass: Same way they did it on a dozen other products already. And with robot PCB stuffers, it really does become cheaper to use a $0.12 tact-switch, a $0.20 chrome button, and a half-dozen other bits to bypass electronically, than to rely on specialty switches of varying price and reliability. (And size, which really foils PCB production.)


Why does J1 IN say "-39mV"? (Or is that a squiggle?) (Then why a squiggle? Test condition 40mV exact. Working condition up to 1V input without gross fuzz.)

Could save a part if the 4316 could eat 12V directly (sloppy 9V) instead of through a 5V regulator. Is it really on a 7V process? Could it be on a 15V process without cost or sonic harm?

Eight opamps is a LOT for a pedal. I can see how to lose opamp D, maybe G too, but it's still a lot of pins to stuff. Also MC33178 is unusual (in stomp-land) and may not be minimum-cost? TL072/TL074 is an old friend. Of course any mega-builder will have his own jellybean opamp. Might be good to list vital specs (all must thrive on 7V power, opamp A wants low Ibias, etc).

wayne kirkwood

Thanks PRR.

It's a squiggle because -26 dBu is not 39 mV exactly.
40 mV at the input would result in 80 mV at TP1 which is not -20 dBu.
I prefer to use dBu so I used a squiggle for the mV readings...

The 4316 has a 6V absolute maximum Vcc. I suspect it uses internal CMOS op amps.
I think a HV process would likely increase IC cost more than the added 5V sub-regulator.
I note for the readership that although the 4316 operates at 5V its audio ports are current in/current out and can scale to higher operating levels with resistor changes.
Because they are current in/out the VCA and RMS detectors are not limited to 5V signal levels.
The control ports and RMS scale at -6.1mV/dB and are also not supply voltage-limited.

A TL072 should not be used to drive the VCA  Ec pins due to rising output impedance vs. frequency.
TL072's driving VCA pins tend to cause the VCA gain cell to oscillate.
Otherwise the op amp choice is pretty flexible...

The 33178 is fairly low cost, low power, low Vos, will drive 600 Ohms, relatively low noise and maintains a low output impedance at high frequencies to drive the Ec pins. It's an MC33078 but low power. Again OEMS have their own ideas about what part they want to use.

Losing op amp D completely, the I-V converter, is not as easy as it appears: Though a resistor could be used for I-V conversion the output current mirrors cannot swing the needed current with AC voltage swing also appearing on the output. Out of curiosity I tried it and the output swing is limited to -15 dBu before clipping. THAT's VCA design engineer confirmed this.

Op amp B exists because it was a left-over and a polarity flip was needed due to the non-inverting nature of the 4316's internal gain cell. (Other THAT VCA gain cells such as the 218X and 2162 invert polarity but with I-V conversion overall polarity is non-inverting.) In a minimalist very low part count design B, C and G could be eliminated.

I believe a sentence in a closing paragraph reads: "The reader is encouraged to tweak this design to suit their requirements."

I also  notice that Mouser's catalog price for that Carling Switch is about half the online price. Still they're expensive.  I read however that people have to test the "Blue" ones. I wonder if the Alphas are OK. Not that it's reality but the Carling are rated 50,000 operations versus 30K for the Alpha.


> Losing op amp D completely............

Ak!! My brain fart! I meant opamp B!

I know D brings us out of current mode and up to the voltages we like to see. That the VCA raw output is 0.6V max(*) and much less for low error, preferably zero voltage (a mV or so). One of D or B is not "essential".

Polarity flip? E-H LPB inverts, FuzzFace does not, nobody seems to care.

I do appreciate opamp B's lovely gain control, in contrast to a simple reostat across opamp D's NFb.

> A TL072 should not be used to drive the VCA Ec pins due to rising output impedance vs. frequency.

Rising Zout, sure, granted. Cheap and cheerful chip. I did not know that about driving VCA inputs. (In higher-voltage work I'd tend to pad-down, but I see here there isn't much room.)

(*) Later I saw "output swing is limited to -15 dBu" or 0.138Vrms, say 0.2V peak. I suspect I have not kept-up with modern VCA topology.

wayne kirkwood

Op amp B (or one of them) was a left-over dual. I decide if I was going to use it I might as well have it flip polarity as well as actively attenuate.

I do realize a lot of devices flip polarity relatively unnoticed. I thought about Eric Johnson in particular when I made up "up" again. There will always be someone who will hear it or claim they do. On some highly asymmetric material I admit to hearing it hear it - guitar isn't always one of them.

You can run your own polarity audibility test here: