The Cardinal V2 build thread (tap tempo + analog schems, layout, & audio)

[midwayfair]

First a little disclaimer: I have not had time to build another one of the analog versions (someone bought my first build while I was in the middle of finalizing it), so I haven't been able to shoot a video demo of the new version. It's coming, though, I promise. You can't buy a PCB of that version yet anyway. I'm okay with sharing the schematic at this point with an audio demo of the tap tempo version, because they sound incredibly similar on the sine and square wave settings.

Schematics:

Analog version:


Tap Tempo Version:


How It Works:

The build document for the tap tempo version (I'm doing a small group buy on Madbean for that one) has all sorts of notes, modifications, substitutes, and a more comprehensive "How It Works": https://docs.google.com/document/d/1HPqDLkMM6WATcSJJT87YK4OvpYAgYb0QhC9qYiV3mB4/edit#

The harmonic tremolo effect works by modulating two bands, one high-passed and one low-passed, 180^o out of phase. The effect is fairly unique, though it sounds similar to a mix between a very mild phaser, vibe, and tremolo. Unlike a phaser or vibe, however, there is no phase cancellation (and thus no frequency notching or pitch shifting) because the bands are in phase with each other.

The dry path is a simple trio of FET amplifiers, with a low-pass filter on Q2 in harmonic mode and a high-pass filter on Q3. Q3's input is grounded in full-range mode, which makes it stay nice and quiet, and the capacitor for the low-pass filter on Q2 is LIFTED from ground in the same way. (This is the same as the previous version.) The actual tremolo effect is accomplished by increasing the resistance between the output of Q1 and the inputs of Q2 and Q3. The light dependent resistors divide against the 100K gate resistors of Q2 and Q3, with the specified vactrol reaching several megaohms of resistance for a very, very deep waveform. (Despite appearances, this is pretty similar to how the big Brownface amps' grid bias modulation effect works, except there, a voltage is applied to the audio signal that attenuates it before it goes to the grid of the tubes; in other words, both here and in the amp, the signal is attenuated before it hits the next amplification stage. Also, although it's optical in the Cardinal, the arrangement is different from a Blackface amp's. We get our blackface sounds here by manipulating the waveform instead.)

The outputs are mixed directly at the drains, and the sources are connected together and share a biasing resistor. This means that a matched pair of FETs is required. I've speced 2N5457 because they're very common, but there are plenty of FETs that work just fine. If you don't want to match your own, there are some forum members (including JMK) who sell matched sets. I've also talked to Steve at Smallbear Electronics and he's going to see if it's feasible to sell matched pairs of FETs over the summer.

Since the TT build doc doesn't talk about the LFO, I'll fully describe how that works here:
A basic oscillator is set up in IC1A; the input pin is biased to 4.5V with R11 and R12. R13 and R14 provide feedback to the inputs of the op amp. Self-noise from the op amp then triggers oscillation.  The Rate control forms a low-pass filter with C13; as the resistance decreases, the frequency of the feedback in the oscillator goes up, and the rate increases. The total resistance of the Rate pot + R15 and C13 set the LFO frequencies. It goes from very slow (~.15Hz) to slightly faster than an amp (4.8Hz). I chose to limit it a little more than the Lune's LFO, because I never found the highest rate settings useful in any way.

A pulse output (square wave) is at pin 3, while the triangle output is at pin 1. The Wave control pans between those two outputs. The wiper of the Wave control connects our LFO's varying voltage to the input pin of IC1C, which is just a buffer. Our depth control is a simple voltage divider (volume control) that pans the input of IC1C between the output of the buffered LFO voltage (pin and a steady reverence voltage at IC1B (pin 7).

IC1D buffers the depth control from the LEDs. The extra buffering stage also seems to smooth out the waveform. The LEDs are then arranged so that one band's LEDs go from the +9V rail to pin 14, and the other band's LEDs go from pin 14 to ground. At minimum depth, pin 14 settles at about 4.5V (depending on the trimpot setting, which is adjusted to account for variation in the LEDs and current limiting resistors). The upper LEDs see this as a lower voltage on their cathodes and also have a path to ground through the lower LEDs + R17, so they turn on a little more than halfway. The lower LEDs are the opposite: their anodes get a steady supply from the op amp and also have a path to +9V, so they turn on a little better than halfway. The LEDs inside the VLT5C1s are in series with the indicator LEDs (one of which can be switched with your bypass switch if you like).

When we increase the depth, the varying voltage from the LFO output is allowed to pass to IC1D, and the voltage on pin 14 varies. The upper LEDs turn off when the voltage swings up toward 9V (no voltage difference between their anodes and cathodes), and the lower LEDs turn on more during the same voltage swing. Huzzah, they're working out of phase! A really neat consequence of this is that the volume in any particular mode stays relatively constant regardless of depth settings, something that isn't the case with the regular Lune LFO.

What's new in this version:

The design is much more efficient in terms of parts and uses trimpots for the biasing, minimizing variations between FETs. It's optimized for using lower gain FETs for more headroom, which was a common issue with V1. There is greater separation between the bands after the first transistor, and the output mixing is done directly at the drains of the FETs, instead of through capacitors like in the original version, avoiding problems with treble loss. The circuit is now far, far more accurate to the sound of the amp. The cutoffs roughly match the Brownface Pro, although the values used to achieve it are different. Also, I think the full depth range is now very useful, whereas before the harmonic mode could sound obnoxious at high depth settings.

V2 also uses a much better LFO so that both frequency bands can now work exactly the same, as opposed to the original version, which had the treble band bottom out as the depth was turned down (resulting in a significant loss of treble in Harmonic mode at low depth settings and compromises made to the full-range side to compensate for that).

The TAPLFO version includes basically all the bells and whistles available from that chip but still fits in a 125B.

There are a couple drawbacks to building this version:
1. You need a matched set of FETs. This was more than just for saving a few parts; it ensures a consistent effect.
2. The VTL5C1 vactrol is required. In the previous version it was just "highly recommended" but this time around I'm going to insist that it be used for a useable effect. It's worth the extra cost.
3. Because the analog version uses a quad op amp, you're a bit more limited in terms of what you can use there for low current consumption without spending a lot of money; but the TL072 or TL062 both work just fine.
4. There are a few weirder sounds missing. There was a neat setting on the old version that sounded incredibly close to an organ through a Leslie (I especially liked using it to play the rhythm part on "Green Onions" at the blues jam I go to sometimes) that I can't quite cop in this version.

However, what you gain is an effect that's easier to easier overall to get right while building, and is, I think, pretty much the most versatile tremolo you can make: The analog version alone can get extremely close to every Fender amp's tremolo and does a full stutter like the repeat percussion.

Audio demo of the Tap Tempo version. I stick to waveforms, modes, and tempos that are easily achieved with the analog version. The only difference is that the sine wave setting is a little smoother in the digital version, whereas the analog version has a tiny bit more maximum depth in the "sine" wave setting. (The analog version also doesn't have any ticking at all. I minimized it as much as possible in the TT version and it only appears in square wave mode, but I've never built anything with the TAPLFO that didn't tick a little without completely nerfing the effect, so it's just sort of something that I have to accept, like white noise from a PT2399.)

https://dl.dropboxusercontent.com/u/9878279/Jon%20Patton%27s%20layouts/Cardinal%20Tremolo/Tap%20Tempo%20Cardinal%202%20audio%20demo.mp3

Perfboard Layout (this is my layout, so I omit the optional 22pF caps on it):

(https://dl.dropboxusercontent.com/u/9878279/Jon%20Patton%27s%20mini%20layouts/Cardinal%20Tremolo/Cardinal%20V2%2021x14.pdf)

There is an etch mask for the analog version:
https://dl.dropboxusercontent.com/u/9878279/Jon%20Patton%27s%20mini%20layouts/Cardinal%20Tremolo/Cardinal%20V2%20etch%20mask.pdf

Finally, if anyone's curious, despite having all those controls on the Tap Tempo version, I rarely deviate from the following setting: 8th or 16th notes, harmonic mode, depth at half, balance in the center, volume just above unity.

[mth5044]

Nice, thanks for posting. Will need to do a lot of reading  Are R17 and 18 supposed to be different values in the TT version?

[midwayfair]

Are R17 and 18 supposed to be different values in the TT version?

Yeah. R17 needs to be above at least 4k7 to minimize whining, whereas R18 doesn't seem to make a difference for whatever reason (I didn't hear any whine even when it's a jumper). I actually just made it 1K just so there was a spot for the resistor and I didn't feel right connecting the base right to the chip. A larger value just lowers the amount of voltage variance available from the PWM, so I figured I'd just make it pretty small and go from there.

[jubal81]

Sounds really, really sweet. Great work, bud!

[italianguy63]

Yes.  Amazing!  MC

[Kipper4]

Good work Jon. Thanks for sharing. Another for the List. 

[GGBB]

Very nice Jon.  Amazed that you found ways to improve an already amazing circuit.

Design questions (not questioning your design - just want to understand  ):

In the doc you describe the following: "On the Q2 side, R4 and C5 form our low pass filter for the "bass" band in harmonic mode; in normal tremolo mode, this is an all-pass stage. On the Q3 side, C8 and R7 form the high pass filter for the "treble" side of the effect." How do you get the 68Hz low-pass and 636Hz high-pass cut-off values from there? There is surely more going on which I can't figure out.

Why both Q2 and Q3? Could you not simply join the vactrol outputs and send them into one JFET?

Also, is the "vibe" mod for V1 doable on V2, or is that one of the weird settings that is no longer possible?

[alfafalfa]

Sounds great , thanks for all your hard work.

Tremelo is one of my favourite sounds .

Alf

[duck_arse]

a diferential pair? very interesting.

[midwayfair]

In the doc you describe the following: "On the Q2 side, R4 and C5 form our low pass filter for the "bass" band in harmonic mode; in normal tremolo mode, this is an all-pass stage. On the Q3 side, C8 and R7 form the high pass filter for the "treble" side of the effect." How do you get the 68Hz low-pass and 636Hz high-pass cut-off values from there? There is surely more going on which I can't figure out.

Bass side low-pass: R5 (47K) + ~2.5K for trimpot 1 and C4 (47nF to ground) = 68.4Hz. Treble side high-pass: C7 (2.2nF) and R10 (100K) = 723Hz. I thought I mentioned something about that one being close enough to the amp (250pF and 1M), but that statement must have been lost at some point in the revisions. I clarified it in the doc. Thanks for the catch!

Why both Q2 and Q3? Could you not simply join the vactrol outputs and send them into one JFET?

I don't think if it ever occurred to me to try just using the resistors as the mixer. It never would have worked in the amp for different reasons. My gut says that it works as long as the effect is oscillating but something funky might happen when the vactrols come to rest at low depth and their resistance drops to about 20K. I'll make a note and try it ... it would certainly be neat to save the trouble of making a pair.

EDIT: On the tap tempo board, it would be pretty trivial to experiment with this when omitting the 22pF caps. You can solder a wire to the gate side of C5, socket R8 and Q3, and plug the wire into the gate's spot of the transistor socket.

Also, is the "vibe" mod for V1 doable on V2, or is that one of the weird settings that is no longer possible?

I can't get it to sound like that mod. A big part of that was having a really big filter sweep effect created by the variable high-pass filter in version 1. There might be some way of implementing something vaguely similar, but I haven't been able to find it. That mod had the biggest issue with treble loss at low depth settings, though.

[GGBB]

Thanks for the clarification. You might want to go over the doc - in parts where you are talking about the filters, you write C5 when you mean C4 and C9 when you mean C8 (I think).

Sounds great.

[midwayfair]

Thanks for the clarification. You might want to go over the doc - in parts where you are talking about the filters, you write C5 when you mean C4 and C9 when you mean C8 (I think).

Sounds great.

Gah. Thought I got them all. I'll go over it again tonight. I had to fix all the component values after deleting a few parts between the prototype and the final board. Thanks for the catch.

[midwayfair]

I added something to the demo, starting at about 7:45 -- I wanted to show what happens when you increase the cutoff of the treble cut in the full range side. The answer is: Not a whole lot! But there are some subtle changes. The smaller cap (I stopped at the "Twin" cutoff, 144Hz, twice as high as the Pro's) has a bit more "bubbling" to it, and a more obvious effect. There's also, unless I'm hearing things, a tiny bit of pitch shifting, probably just from the inherent phase shift of a low-pass filter.

The cap sizes are, in order, 47nF, 39nF, 33nF, 27nF, and 22nF.

p.s. The TT build doc has been updated in general to fix the capacitor numbers in the circuit discussion.

[midwayfair]

1776 Effects released the PCB for the Cardinal V2 today:

http://1776effects.com/

Josh is also stocking the VTL5C1s AND FET sets necessary for this project, with a great price on the matched FETs. If you'd like to purchase different sets of matched FETs, the 2N5952 sets that Smallbear sells will also work for Q2 and Q3; the pinout is backwards but the 2N5952 will, on average, have slightly more headroom and slightly less gain/output at normal biasing settings.

Also, I just realized that I never updated this thread with the video demo:

[rankot]

Jon, is there any recommendation for JFET Vgs/Idss values? I've built this, and my first stage is already overdriven a lot, although I removed 22uF source cap. I have to roll down guitar volume a lot, to get rid of distortion.

[midwayfair]

Jon, is there any recommendation for JFET Vgs/Idss values? I've built this, and my first stage is already overdriven a lot, although I removed 22uF source cap. I have to roll down guitar volume a lot, to get rid of distortion.

I usually leave off the source caps -- I think the only time I included both was when I used a completely different FET that was really low gain. I've also noticed that some instances of 2N5457 since I made the version 2 seemed much higher gain.

Are you above unity gain? And it's definitely the first stage you're getting distortion from?

If removing the source cap still doesn't clean up the signal, you can try increasing the input resistor size. Unfortunately Josh's site is down and I don't have a copy of the PDF, so I have to kind of go by memory on this stuff, but there should be a 33K on the input,. You can knock down your signal by upping that to 68K or even 100K. If you have to get really big with that, you may need to take some other steps.

Unfortunately (or fortunately as the case may be), FETs aren't the cleanest component, and there is supposed to be some measure of amp sim to this, because a little distortion helps enhance the effect.

Good luck

[rankot]

Thanks Jon! I'm sure distortion is present after the first stage, because I monitored it with an oscilloscope. I'll try to adjust Rs and Rd as in Fetzer Valve page from ROG. If it doesn't help, I'll add Gain pot in front of the 1st stage.

EDIT> I noticed that I have a lots of gain over unity, so I put my JFET's params into their calculator and it gave me 14 times voltage gain, which is huge. I tried to increase Rs to 1k, but still need some gain reduction, so I will set Rs/Rd accordingly, and also add A1M Gain pot instead of R2.


I will see if there's a need to adjust tremolo section after that.

I also used Chinese VTL ST vactrols for my build. They seem to work fine (I can't find VTL5C1 at the moment). Their light/dark resistance is fine, but maybe their release is too slow, so I will also try to build this with rolled vactrols using GL55xx series photoresistors.

[rankot]

Just for others to know - I built it using GL5528 LDR and bright green LED (2.51 Vf) and it works fine. I also put a 1M audio pot instead of R2, so I can control gain and thus distortion. Works like charm.