The Quackmire: A simpler vactrol LP envelope filter for 1590A

[Fancy Lime]

Hi everyone,

continuing with my recent obsession I spent most of my Saturday designing the simplest (within reason) and smallest true-lowpass envelope controlled filter I could think of. It uses a single NSL-32 vactrol to control a Sallen-Key lowpass. The other resistor of the Sallen-Key is contains a range pot. This pot sets the total range from wher to where the filter goes and also at what frequency the maximum Q occurs. Since only one resistor varies with the envelope, the Q factor changes along the sweep and reaches its highest value when both resistors are equal. The maximum Q of 21.7 can be shifted between 245 and 1468 Hz using the Range pot, providing some variability in the sound. The changing Q gives the effect a somewhat vocal timbre, especially at lower Range settings.

The rest is fairly standard as seen before in many other envelope filters. The Volume control is a set-and-forget trim pot since the other settings hardly affect the output volume. This is sort of the little brother to the Funkkontakt:
http://www.diystompboxes.com/smfforum/index.php?topic=118966.msg1109620#msg1109620

The schematic:


The sound:
http://www.aronnelson.com/DIYFiles/up/Quackmire.mp3

I also drafted a 12x15 strip-board layout, which should just about fit into a 1590A enclosure. I will post this as soon as I have checked and verified it.

I called it The Quackmire because the objective to have a true resonant lowpass with only one control element that fits into a 1590A enclosure but is flexible enough to allow some individuality in setting the sounds proved to be quite the quagmire.

Mod suggestions:
For a less aggressive sound I suggest increasing C5 to 500pF or 1nF. For a more aggressive sound increase C4 to 1µF. The Range pot can be a linear type if you do not have a rev log, although rev log has a slightly better control characteristic in this position. The Range pot can be made a trim pot or fixed resistor (combined with R7) to further reduce complexity. Change the charge cap C10 for slower or faster attack and decay. Low values (1µF-2.2µF) will giver a more synth-like sound. Q1 can probably be any high-gain NPN BJT, like 2N5088 or 2N5089, although I did not check. Just try whatever you have, this is not a demanding role for any BJT.

As always: Let me know if you find errors or bad design choices or if you have suggestions for improvements or questions.

Have fun,
Andy

[GibsonGM]

Very nice, Andy - quite a bit of quack for a simple one, too!!   Thanks for posting it    

[bean]

That sounds super. Can you explain a bit the necessity of D1 and D3?

[Kipper4]

Sweet. Nice sound . Is the bass active or passive?
Nice low parts count too.

You might be able to save a part if you if you make R6 a pot.
What size is D3 please.

Great job.

[Fancy Lime]

Hey guys, thanks for the kind words!

Can you explain a bit the necessity of D1 and D3?

D1 and D3 are used for biasing:
D1 together with R2 and R3 tries to keep the Base of Q1 at one LED drop above (supply voltage - one LED drop)/2, i.e. at 5.2V (for a 9V supply, assuming the LED drops 1.4 V). I did that because the Emitter of Q1 is DC coupled to the input of the filter op-amp IC1b. Since we want to keep the op-amp around 1/2 supply voltage and the Emitter of Q1 sits 1 Si-diode drop (0.7 V) below the Base the LED provides just the right shift.
D3 is a bit simpler: It biases the envelope follower op-amp IC1a to 1 LED drop above ground, together with R10. This way the envelope starts just below the conducting threshold of the yellow diode in LDR1, for best sensitivity. Props for this idea go to Jack Orman, who used it in his Dr. Quack for the same reason. You may find that the envelope section of The Quackmire bears an uncanny resemblance to that of the Dr. Quack / Doctor Q.

Is the bass active or passive?

Cort B4 fretless with stock pickups (Bartolini license MK1) but with the original active preamp yanked out. Best modification ever. Sounded like utter crap with the preamp and now sounds absolutely fantastic purely passive. I only used the neck pickup for this (as I mostly do).

You might be able to save a part if you if you make R6 a pot.

Which one?

What size is D3 please.

All LED's are 3mm. But that should not matter (much) to their functionality, should it?

Cheers,
Andy

[Mark Hammer]

Giggity!

[bean]

D1 and D3 are used for biasing:
D1 together with R2 and R3 tries to keep the Base of Q1 at one LED drop above (supply voltage - one LED drop)/2, i.e. at 5.2V (for a 9V supply, assuming the LED drops 1.4 V). I did that because the Emitter of Q1 is DC coupled to the input of the filter op-amp IC1b. Since we want to keep the op-amp around 1/2 supply voltage and the Emitter of Q1 sits 1 Si-diode drop (0.7 V) below the Base the LED provides just the right shift.
D3 is a bit simpler: It biases the envelope follower op-amp IC1a to 1 LED drop above ground, together with R10. This way the envelope starts just below the conducting threshold of the yellow diode in LDR1, for best sensitivity. Props for this idea go to Jack Orman, who used it in his Dr. Quack for the same reason. You may find that the envelope section of The Quackmire bears an uncanny resemblance to that of the Dr. Quack / Doctor Q.

Awesome. Thanks for taking the time to explain!

[Kipper4]

Same bass as I have.

I also note theres no decoupling cap between the input buffer and the filter.
I guess theres no need.

[thermionix]

I know all of jack squat about vactrols.  I have one labeled VTL5C1, is this an equivalent?

How do you think this circuit would sound with guitar?

[Fancy Lime]

Same bass as I have.

I also note theres no decoupling cap between the input buffer and the filter.
I guess theres no need.

's'right. DC coupling the two stages saves some parts and for some reason seems to have more headroom and sound better than with separate biasing. It also means that Q1 works in a double role as an audio buffer as well as a bias buffer and therefore must itself be biased 1 Si-diode drop above V 1/2 (so that the Emitter is at V1/2), which is what D1 does.

I know all of jack squat about vactrols.  I have one labeled VTL5C1, is this an equivalent?

How do you think this circuit would sound with guitar?

The filter is designed for the characteristics of the NSL-32's I have, which have an Off Resistance of >10MΩ and an On Resistance of 1kΩ. The VTL5C1 seems to go to even lower On Resistance, which means that the filter sweep will go higher. So, yeah, you can use it, it would actually be better than the NSL-32. You might want to adjust R6, R7, C4 and C5 to adjust the sound to the vactrol and your personal preference but I don't think it is necessary. Here is the datasheet:
http://logosfoundation.org/instrum_gwr/playerpiano/Optor_VTL5C1_87223.pdf

To calculate the response of the filter to component changes I like this tool:
http://sim.okawa-denshi.jp/en/OPseikiLowkeisan.htm

For guitar use it should be fine too. It may sound a tad aggressive as it is now because the pick attack frequencies are amplified a lot. Increasing C5 to 500p or 1n would make it a bit rounder and less peaky. Also I personally like fast attack and decay for bass but slower attack and decay for the typical funky rhythm guitar autowah sound. Try increasing the charge cap C10 from 10 to 22µF, I think that may sound better with guitar. You can socket this cap for extra flexibility.

Cheers,
Andy

[StephenGiles]

I get "internal server error" when I try to download the circuit???

[Rob Strand]

Sounds pretty good, especially while you keep riding the envelope.  It's got a clean wide bandwidth sound.
I'd try to clean-up the sound right at the end of the tail of the envelope, it cuts off/out a little quick.  (Maybe decrease R6 (1M)?)

Many moons ago I played around with envelope filters in the DSP domain.   I started by writing a VST plug-in for the MXR envelope filter.  What I found is second order filters for the envelope detection filter sounded better.   (If there was a secret that would be it.)  If you look at the MXR envelope filter it actually has a second order filter for the envelope.  The vactrols kind of give you a second time constant by nature.

For the circuit,
-  I'd put in a larger R3 and remove D1.
- Also D4 doesn't do anything.    You normally put this diode in when there is a cap between the opamp and D4.  This gives you a voltage doubler; IIRC the MXR has this.

[Fancy Lime]

Hi Rob,

thanks for the input!

Sounds pretty good, especially while you keep riding the envelope.  It's got a clean wide bandwidth sound.
I'd try to clean-up the sound right at the end of the tail of the envelope, it cuts off/out a little quick.  (Maybe decrease R6 (1M)?)

Do you mean that the release/decay is too fast? That is actually a feature, not a bug. I added R13 after originally leaving it out because I found the decay too long. Matter of taste. I like when I can return the filter to zero by quick muting and have to willfully stay on top of the envelope rather than the envelope making me cruise along the crest automatically. Increasing R13 to 100k or 330k will give longer decay times. For guitar use I would recommend longer decay times. One may even want to change the charge cap C10 to 22µF for guitar. Decreasing R6 would set the starting point of the filter higher. Also an option that probably makes more sense for guitar than for base. I should probably write a more extensive Mod Guide for this thing one of these days.

Many moons ago I played around with envelope filters in the DSP domain.   I started by writing a VST plug-in for the MXR envelope filter.  What I found is second order filters for the envelope detection filter sounded better.   (If there was a secret that would be it.)  If you look at the MXR envelope filter it actually has a second order filter for the envelope.  The vactrols kind of give you a second time constant by nature.

By "second order filter for the envelope detection" I assume you mean double the R12 C10 arrangement? I did this sort of thing on OTA based envelope filters, where it does indeed help smooth out ripples in the envelope. I also tried it with vactrol based filters. However, I found that when using vactrols and relatively "large" charge caps (C10=10µF), I could not hear a difference so I left it out. But for faster control elements you are definitely right. The more flexible solution if you want a smooth sound even at very short attack and release times is to use a precision full wave rectifier with an over-sized de-glitching cap (which effectively works like the first pole of the seconfd order filter you mentioned) instead of a half wave rectifier.

For the circuit,
-  I'd put in a larger R3 and remove D1.
- Also D4 doesn't do anything.    You normally put this diode in when there is a cap between the opamp and D4.  This gives you a voltage doubler; IIRC the MXR has this.

The R2-D1-R3 biasing is a result of me being to lazy to get my multimeter out. But you are right of course, removing D1 and adjusting the R2/R3 ratio until the Emitter of Q1 is at V 1/2 works exactly the same with one less part.
And finally: Yepp, D4 is indeed pointless and shall be removed for the next version. Can't believe I missed that. Rookie mistake. Might be a good idea to reroute it to the op-amp feedback to keep reverse leakage from eating away charge on the negative swing, although the effect will probably be minimal. We'll see.

Thanks and cheers,
Andy

[Rob Strand]

Do you mean that the release/decay is too fast?

No, it's not the general fast attack.   That's fine.

I added R13 after originally leaving it out because I found the decay too long. Matter of taste. I like when I can return the filter to zero by quick muting and have to willfully stay on top of the envelope rather than the envelope making me cruise along the crest automatically.

Yep,  I understand you here 100%.  (I do the same thing.)

Increasing R13 to 100k or 330k will give longer decay times.

Agreed, and it goes against what you want.

So the general fast decay isn't the issue.   As you play, the envelope filter will shape the envelope/fiter, as normal. However when you let the note ring out (eg. on your sample at say 0:14, 0:32, 0:51) you get to hear the full sweep of the filter.  The "issue" is when the filter gets near the end of its "trajectory".  To me the rate at that point is a tad too fast (not bad).   I think the reason is because R33 is connected to ground, this makes the discharge of the cap aim towards 0V.  However, the circuit stops at 1.5 to 1.7 due to the the Vactrol LED.  So the cap doesn't get a chance to discharge naturally to zero and the filter sweep just stops.   The time slope of the cap voltage at 1.7V discharging to 0V is higher than a cap at 1.7V discharging to say 0.6V.  If the ground side R13 (not the cap) is moved to D3 then the cap will discharge towards 1.7V and it will slow down the discharge;  this will probably be too slow.  Maybe 0.6V will help, basically a silicon diode connected like you have D3.

The other thing that's going on is as the vactrol gets near cut off the resistance goes up.   Eventually it will rise to a point where the parallel resistance R6 will dominate and the filter cut-off will settle to it's lowest point.  The presence of R6 also slows the filter sweep rate at the end of the sweep.   So I was thinking as a quick hack you could see what decreasing R6 sounded like.   It's not the same as playing with the discharge slope as changing R6 will raise the final cut-off frequency.  Both aspects affect the final sound.

By "second order filter for the envelope detection" I assume you mean double the R12 C10 arrangement? I did this sort of thing on OTA based envelope filters, where it does indeed help smooth out ripples in the envelope. I also tried it with vactrol based filters. However, I found that when using vactrols and relatively "large" charge caps (C10=10µF), I could not hear a difference so I left it out. But for faster control elements you are definitely right.

Yes, the vactols give the extra filtering for free.  It's also hidden as it's part of the vactrols behaviour.  When you take it away by using an OTA something is missing!

[Fancy Lime]

Hi Rob,

ah, now I see. You want to make it "land more gently". I'll try move the R13 bottom from ground to the R10-D3 junction and see what happens. It should as you say slow the draining shortly before touch down.

R6 is near the useful limit at 1M. Making it smaller means that the filter idles above the fundamental frequencies of the lower register on bass. Which also sounds good, but takes away some of the "flashy quack" so I decided to go with 1M. Changing R6 to a 1M or 2M5 pot would make an additional "floor" control very similar in effect (but very different in execution) to what I implemented in the Funkkontakt envelope filter.

Andy

[Rob Strand]

You want to make it "land more gently".

Yes, that's it.

R6 is near the useful limit at 1M. Making it smaller means that the filter idles above the fundamental frequencies of the lower register on bass. Which also sounds good, but takes away some of the "flashy quack" so I decided to go with 1M.

Yes it's a fine line playing with these things and you don't want to upset the goodness in what you already have.

[Fancy Lime]

Hi Rob,

I had a minute to try to get the gentler touch down. Adding a diode works but as it turns out, so does turning up the Sensitivity a smidge. Then I listened to the sound sample again (later realizing it should have done that earlier) at the points you mentioned (0:14, 0:32, 0:51) and realized that you probably thought that I let the notes ring out when in fact I muted them. When letting notes ring out the filter follows the envelope nicely to zero without any abrupt drop.

Once at it I also played with the charge cap C10 as Rich (Kipper4) suggested elsewhere. Turns out 1µ, 4µ7 and 22µ all provide very cool and quite different sounds. That in itself is not surprising but I was mildly surprised that they all sounded so good without any additional changes. Although the original goal was to keep it as simple as possible, I think an additional Mode switch between the 3 caps (synth, fast, slow) is warranted, considering the benefits. And it should still fit into a 1590A, although it's getting tight now.

Cheers,
Andy

[StephenGiles]

I get "internal server error" when I try to download the circuit???

Downloaded OK today - how strange!!

[Fancy Lime]

Alrighty,

removed the unnecessary D4 and added a Mode switch, making it much more versatile (but you can just as well leave it as well as C10 and C12 out and only use your preferred value for C11). It is a bit of a challenge to fit it all in a 1590A enclosure with the extra switch but it is doable. I'll post pictures of my build one of these days, promise. I checked and double checked the layout and I am fairly certain it is correct now. As you can see, I am no da Vinci when it comes to layouts but I hope you can make sense of it. If not, feel free to ask and/or complain.




Please let me know if you find errors.
Giggity,
Andy

[Rob Strand]

had a minute to try to get the gentler touch down. Adding a diode works but as it turns out, so does turning up the Sensitivity a smidge. Then I listened to the sound sample again (later realizing it should have done that earlier) at the points you mentioned (0:14, 0:32, 0:51) and realized that you probably thought that I let the notes ring out when in fact I muted them. When letting notes ring out the filter follows the envelope nicely to zero without any abrupt drop.

Interesting result.