MUFF WIGGLER Forum Index
 FAQ & Terms of UseFAQ & Terms Of Use   Wiggler RadioMW Radio   Muff Wiggler TwitterTwitter   Support the site @ PatreonPatreon 
 SearchSearch   RegisterSign up   Log inLog in 
WIGGLING 'LITE' IN GUEST MODE

For your review: a dual function Ringmod / VCA circuit
MUFF WIGGLER Forum Index -> Music Tech DIY  
Author For your review: a dual function Ringmod / VCA circuit
drmarble
Presenting the Multi-Mode Multiplier, a dual function ring modulator and VCA module, based around the AD633 chip. I'll be making it into a DIY 4hp Eurorack module.

I've had this idea for a while, and only recently got around to breadboarding it and testing it alongside other Eurorack modules. Looking for ideas for improvement to this circuit (in terms of both electrical function and being user friendly).

As-is, the front panel would have 5 components: 3 jacks, 1 toggle switch, and 1 knob.

It's largely taken from the AD633 datasheet, so nothing groundbreaking there. But the idea I was tweaking the most was implementing a "VCA" mode. One of the inputs to the multiplier gets halfwave rectified, so that negative values of input signal have no effect.

Initially, I tried the ideas of using a "superdiode" for that rectification, and also a BAT43 diode with a super low forward voltage. However, in practice I found it useful to keep a “deadband” of about 0-100mV CV input, where the output of the diode is still at 0v. This turns out to be a desired effect, since some envelope generators don’t return to exactly 0v, and tend to float a bit above that. If my EG floats up to 5 or 10 mV at idle, I don't want the VCA letting sound through.

So, that's why you see a 1N4148 used in this circuit. After some testing, I am not hearing any sudden truncation of the tail of an envelope. Even though the part is spec'd to about a 0.65 forward voltage, in practice, in the context of this circuit, it does not exhibit a sharp cutoff.

One technical concern is whether I should hang 10V Zeners off the AD633 inputs, to protect it from anything going higher than its spec'd 10V maximum inputs. I don't know if higher voltages pose an actual danger to the chip...?

If wanted, it would also be simple to sub a 100k pot in place of R1, to have an attenuator on the Y input signal as well. However, it doesn't seem all that important to have attenuators on *both* inputs to a multiplier, so it seems like it would just be extra cost and front panel space for an unhelpful control. Thoughts?
cygmu
I think the AD633 can withstand input up to the supply rails so you don’t need the Zeners. +/-10V is the recommended range but the rails are the max. A TL07x powered on 12V won’t get much above 10V anyway.
guest
you probably dont need the opamps at all. the X input has a large impedance, so buffering this isnt a bad idea, but could be solved other ways. the main issue is the high bias current of the AD633 (1uA). so you want to keep the input impedances balanced at the inputs. so if you stick with the opamps (honestly they are cheap, and add another layer of protection to the AD633) then id reduce the value of R5/R4 so the impedance seen there is closer to the very low output impedance of the opamp. maybe 100k/100ohm.
robin87
I don't think that you'll need any sort of input protection. It'll work just fine without the opamps. I wired up something similar on a piece of perfboard the other day, since I'm working on a circuit that will use one or maybe two AD633s.

Keep in mind that the AD633 uses a scaling voltage of 10V, so if you multiply two signals with a +/-5V output swing (as in a typical ringmod patch with two VCOs), the output will be half the amplitude. (5V*5V/10V=2.5V) Also, you'd need a 10V envelope to reach unity gain if you're using it for VCA tasks. Take a look at Fig.17 in the datasheet for how to bring the level back up.

The AD633 is great fun and really versatile! If only those 2$ AD633s on chinese ebay weren't fake..
drmarble
Thanks everyone for the comments!

To address the inclusion of opamps: I initially breadboarded the circuit without them. However, if I would trim the X input in ringmod mode for good feedthrough rejection, and then switch the diode into the circuit (for "VCA" mode), the feedthrough jumped up a lot.

I'm not great at detailed circuit analysis, but my sense was that I needed to buffer the drastic impedance change between diode/no diode, and so that's when I introduced the buffer for the X input.

Then, once you've got one opamp onboard, might as well find a use for a dual opamp in the same real estate, so that's why I also buffered the Y input. As others have pointed out, the opamp stage provides input protection, which for a $10 IC it seems prudent.


guest wrote:
the main issue is the high bias current of the AD633 (1uA). so you want to keep the input impedances balanced at the inputs. so if you stick with the opamps (honestly they are cheap, and add another layer of protection to the AD633) then id reduce the value of R5/R4 so the impedance seen there is closer to the very low output impedance of the opamp. maybe 100k/100ohm.


Ah, I didn't realize it had a high bias current... it lists a 10MΩ input impedance, so I started thinking of it as "just worry about voltage, it's not going to draw current".

So, changing R5 to 100Ω would give the + and - pins of input X a better-matched impedance? I'd probably keep R4 at 1M, since I'm barely moving from the center point of the trimpot, and could use more resolution rather than more range (but yes there is a minor but audible improvement in feedthrough when using the trim vs. just connecting pin 2 of the AD633 to ground).
drmarble
robin87 wrote:
Keep in mind that the AD633 uses a scaling voltage of 10V, so if you multiply two signals with a +/-5V output swing (as in a typical ringmod patch with two VCOs), the output will be half the amplitude. (5V*5V/10V=2.5V) Also, you'd need a 10V envelope to reach unity gain if you're using it for VCA tasks. Take a look at Fig.17 in the datasheet for how to bring the level back up.


Hmm, good point... I'll look into that too.
drmarble
robin87 wrote:
Keep in mind that the AD633 uses a scaling voltage of 10V, so if you multiply two signals with a +/-5V output swing (as in a typical ringmod patch with two VCOs), the output will be half the amplitude. (5V*5V/10V=2.5V) Also, you'd need a 10V envelope to reach unity gain if you're using it for VCA tasks. Take a look at Fig.17 in the datasheet for how to bring the level back up.


I thought on this issue a bit more, and looked at the voltage levels in my system in more detail. Most of the VCOs do swing +/- 5V. The control signals, however, vary - some swinging 0-5V and others 0-10V.

So, my thought is to include a second switch, to allow a doubling of the output level of the AD633 for when using a 5V control signal. Here's that variable scale factor method you mentioned from the datasheet:



I'm thinking 10k for both R1 and R2, to give a 2x gain (while keeping the additional load on the output low). Using a pot for R1 (R1 in that schematic, not my original) would be possible as well, but you don't get a unity gain option out of that (since gain is always more than 1).

I'm wondering if for this gain switch, I should just use a SPST toggle to connect or disconnect R1 from pin 6, or if I'll want a DPST to also bypass R2 and connect pin 6 directly to ground. I mean, it'll work either way, but it's a noise issue, right, whether pin 6 sees ground directly or through a 1k resistor?


Or... maybe the simplest option is to just hardwire the 2x gain increase, and use the module's attenuator to knock 10V control signals down to 5V?
neil.johnson
Nice little project, although note that what you have designed is a switchable 2/4-quadrant product multiplier and not a ring modulator.

Neil
drmarble
neil.johnson wrote:
Nice little project, although note that what you have designed is a switchable 2/4-quadrant product multiplier and not a ring modulator.


You mean, by virtue of it not containing an actual ring of 4 diodes?
neil.johnson
drmarble wrote:
neil.johnson wrote:
Nice little project, although note that what you have designed is a switchable 2/4-quadrant product multiplier and not a ring modulator.


You mean, by virtue of it not containing an actual ring of 4 diodes?

Two reasons really:
1. No ring of diodes
2. You won't get the carrier harmonics that you get with a ring (either driving it correctly per the original design, or incorrectly as is done in the world of electronic music)

Neil
drmarble
neil.johnson wrote:
2. You won't get the carrier harmonics that you get with a ring (either driving it correctly per the original design, or incorrectly as is done in the world of electronic music)
(emphasis added)

As you no doubt are aware, the prototypical carrier signal used in ring modulator effects is a sine wave, which does not have any harmonics.

So, are you pointing out that the four-quadrant multiplication performed by an AD633 chip lacks (significant) distortion, and therefore doesn't add in extra harmonics the way a traditional diode circuit would?

If so, I get it that distortion is an important part of sound design. The multiplication of two signals as performed by an AD633 won't fool anyone into thinking it's a vintage ring modulator circuit. It lacks a certain roughness around the edges, and so it doesn't make a convincing Dalek voice.

That said, it does occur to me that I haven't explored the effects of overdriving the AD633, and possibly that's another option to try out in the circuit before I build it.
neil.johnson
This topic was discussed some time ago...

https://www.muffwiggler.com/forum/viewtopic.php?p=2078438#2078438

Neil
drmarble
neil.johnson wrote:
This topic was discussed some time ago...

https://www.muffwiggler.com/forum/viewtopic.php?p=2078438#2078438

Neil


Ha! By you and I, no less. That's half the reason I write things down, to share the knowledge with my future self.
drmarble
OK, after some further experimentation, here's an updated schematic.

I tried the effect of overdriving the AD633, but it's not very interesting... it just clips hard like opamp clipping. So I'm not attempting to exploit that here.

The value of R4 could come up higher for better trim resolution, but I did find that 1M was too high (after changing R5 to 100Ω) and did not give me enough range. The AD633 datasheet shows an offset trim circuit for ±50mV, whereas mine only has a 4mV range (they need to cover their worst case, max. offset scenarios, whereas a DIY'er can test for the actual unit they get). The tradeoff is range vs resolution, of course. For the record, the difference in feedthrough level between using a trim circuit and just connecting pin 2 directly to ground was slight. Maybe a 6dB difference, best I can tell on a spectrum analyzer. But it's so far down, right above the noise floor, that it's likely to be inaudible in most cases.

Also, I discovered that, even with the X input at 0 volts, that if the Y input was hit with a loud enough signal then the peaks would splatter through quite audibly. In the context of strictly Eurorack this is unlikely to happen. I only heard this when I was hitting the Y input with the output of a external preamp running on 15V rails, so it was peaking at around ±13. So, this would be one reason to add 10V zener diodes to the inputs of the AD633, to clip any incoming signals to ±10V max. But it's a fringe case, so I may or may not bother.

Thanks again to robin87 for pointing out how I might want to adjust the scale factor. I'm going with hardwiring resistors for an overall scale factor of 5 instead of the default of 10. Any situation where that would cause clipping at the output can be easily compensated for with the X/CV attenuator knob.
drmarble
I've been working on the front panel for this module, which as I mentioned I wanted to do as a 4hp.

This schematic shown needs 5 front panel components, but if I use 1/2" spacing between components, I can put 8 components on a 4hp panel.

So there's space for 3 more, which means I can put a second channel on it, with just X, Y, and output jacks (no unipolar switch, no attenutor).


Brainstorming question - what would be more useful for a two channel module:

1. Normalize channel 1's X and Y inputs to ch 2's X and Y inputs, respectively?

2. Or, normalize channel 1 OUTPUT into channel 2 Y/signal input? (Channel 2's X/CV input would normalize to ground.)
degeneratedsines
Or leave some space for the attenuator knob?
drmarble
degeneratedsines wrote:
Or leave some space for the attenuator knob?


I was planning on using a Davies 1900 knob, so that works with 1/2" spacing, provided it's next to a jack and not another knob.

This is the spacing used by the STG .mix layout:
https://muffwigglers.myshopify.com/products/mix

It can be a bit cramped, yes, but doable - and in this case, with only one knob, I would put it at the top of the panel for easiest access.
MUFF WIGGLER Forum Index -> Music Tech DIY  
Page 1 of 1
Powered by phpBB © phpBB Group