Transformer Ring Modulator

[dthurstan]

Hello

For the past 4-5 years I have been trying to design a transformer based ring modulator. I think like a lot of people I saw the simple 6 component design & thought that looks easy to build! There are probably easier ways to build a ring modulator, but I liked the look of a transformer type and stuck with it. There aren't really any schematics out there except of course the Ring Stinger, but that's too big for what I wanted. The schematic below is pretty much the Ring Modu-Matic circuit from Stompboxology Vol 8 No 1. That article was a great source of information about ring modulation. I chopped and changed things round to get it to work with the transformer implementation instead of using an OTA chip. Here is a flow diagram of the circuit.



The i/p goes to a buffer which goes to the balance control and it also goes to an amp with a gain of 5.7. This goes to the modulator i/p and to the gate (marked comp) which switches the oscillator off when there's no signal at the i/p(this is set with the trim). I spent a long time looking at sine wave generators but couldn't get any to work well enough. So I went with a triangle wave generator, which has a good range of freqs. The o/p of ring modulator is attenuated and sent to the other side of the balance control. I decided to put in an octave switch as it is easy to implement.



Here is a pic of the breadboarded circuit which I've been tweaking;


It sounds good, there is a lot of bleed thru but I didn't match the ring diodes as I have the gate to stop the osc. Im not sure about the attenuator after the ring mod I think I havent set up the roll off points correctly? Also I think I should have a LPF before the i/p signal goes into the modulator to roll off the highs, I was thinking somewhere around 1 KHz as that still has the fundamentals of all the notes on the fretboard?

Let me know what you guys think.

[senko]

The 6 component ring modulator (2 transformers and 4 diodes) is one of the simplest ways to create a ring modulator.  However, when I built one, the bleed through of the carrier signal oscillator was fairly noticeable.  Like the Stompboxology article explained, gating of the carrier signal is important enough to expand the circuit component count.  I also believe that the article concluded that the signal generated contained an unusable amount of harmonic distortion. 

I'm currently working on a small PCB layout for the two transformers and the four diodes so that I can experiment with this particular module or substitute the OTA circuit into the mix. 

Although it catches some flak and isn't really a standard stompbox component, the ICL8038 and other waveform generators are perfect for experimenting with ring modulation.  Get a control voltage into the function generator and you might actually have something going.  That's my plan when I get back to my workstation.

I will definitely report back with my findings.  Ring mod is cool enough to pursue.  My conclusions thus far are that you'll definitely need more than 6 components to use with a guitar and there's lots to explore (better than the ol' fuzz box, eh? )

[PRR]

It has been decades since I messed with ring modulators.

However I think you want more than 0.1uFd driving an audio transformer.

And I'm pretty sure you need the voltage across the diodes to _reverse_. The way you have it, both ends of the diode bridge go up-and-down together. It may "work" only by diode mis-match and secondary effects.



While you may want a low-pass, I'm suspecting that what you have now is "all highs" from too-small driving caps and stray capacitance overwhelming poor diode switching.

Which transformers are you running?

[Jazznoise]

I'm currently working on a FET based ring modulator, the simulations would currently show carrier levels in the main audio path to be pretty low without any deliberate carrier supression. I'm breadboarding some of it Friday and I'll let people know how it's sounding!

Transformer method seemed very high noise to me, it seems like having a differential amplifier is the only way to reduce carrier levels drasticly without resorting to things like gating. The only commercial units I can think of that don't utilize gating use Balanced Modulators, which exist both as a chip but are generaly just good broadcast practice. Read up on some telecoms, we steal more from them then we'd like to admit!

[amptramp]

The U2 pins 8/9/10 section seems to be a bit strange.  It buffers Vref and provides an output to a pot also returned to Vref.  If it is supposed to act as a variable voltage source, you go all the way from Vref to Vref.  I'm not following this.

[slacker]

I think that's an error on the schematic, looks like it's supposed to be the comparator part of an LFO so the right hand side of the 10k in the feedback loop should go to pin 14 of U2 not pin 8.

[dthurstan]

...I also believe that the article concluded that the signal generated contained an unusable amount of harmonic distortion.  

It is noisy and there is some harmonic distortion (it's hard to tell whats ring modulation and whats distortion) but I kinda like that it's not quite right ( some would say it's just working ) I'm considering building a ring mod based on the AD633 after this to get a more "pure" RM sound (hopefully wouldn't take 4 years), but for now I like this messed up noisy beast.

@ Jazznoise I've seen a bit here and there from telecoms I should read some more I've got The Art of Electronics which as some stuff in there.

@PPR & amptramp cheers for spotting those mistakes in the schematic (the circuit is wired correctly). Here is the corrected;



yeah I thought there where problems with RC networks. I find it difficult working out if it's high passing or low passing.

So the modulator (or X) input sees R=5k7 (1+4.7) C=10u fc=2.8 Hz
carrier (or Y) input sees R=1M (1.1M at the moment as its set) C=10u fc=0 Hz
output sees C = 1u R=48k (10+38) fc = 3.3 Hz

I haven't tried to balance the diodes at all and maybe I should, I'm thinking I might use 1N914's instead of the 1N456's Im using at the moment. I did have 1N914's in at 1 point and I now have about 16 so I have more to match. I read something about diode array, but can't find any in the correct config here in the UK.

Cheers for all your replies.

[PRR]

> Transformer method seemed very high noise to me

No, that's fine. Transformer rings are a standard in RF work where noise is bad and gain at base frequency is not possible. Oh, sure, you can build a hissy transformer RM.

> commercial units ...... use Balanced Modulators

Audio transformers are lumpy, costly (or cheap), attract hum problems. Balance tolerance can be good, but when it isn't there's no trim. For other reasons, chip modulators are (or were!) very available and sometimes very-very good. And they come with full datasheets for modulation work with suggested hook-ups.

_I_ think you want to have two good sine generators and an oscilloscope when learning ring modulators. (A spectrum display would be real nice but last time I did RMs a SA cost more than a house.)

[amptramp]

I prefer Gilbert cell modulators like the MC1496 and others of that ilk because diode modulators introduce harmonics that need not occur with semiconductor mixers.  With RF mixers, this isn't usually a problem - a mixer providing a 100 MHz signal will have harmonics that can be filtered easily.  This is not possible in audio, since many harmonics are also in the audio band.  In addition to the harmonics caused by diode mixing, there will be small amounts of crossover distortion where nothing conducts due to the forward drop of the diodes.  You should be able to get decent results from this design because harmonics are usually welcome in stompboxes, but it will not behave as an ideal modulator.

[dthurstan]

I've played through it a bit more tonight and the you can hear a lot of carrier bleed thru. I play it tuned, so it's like a drone. I think it I match 4 1N914's as close as I can that should tame it a bit. Looking at arrays all I found was UC3611N which is 4 Independent Schottky diodes in an IC. Not sure if that would work and I can't get hold of 1 here anyway without paying almost £20!

I'm using 42TM018's for the transformers btw, I found it sounds better with 0.1u instead of 10u cap from the osc to the carrier i/p. Looking at geofex splitter;
http://www.geofex.com/FX_images/splitter.gif
I see that the o/p of the transformers have LPFs with fc= 15 kHz, R=10k C=0.001u.

[DavenPaget]

I've played through it a bit more tonight and the you can hear a lot of carrier bleed thru. I play it tuned, so it's like a drone. I think it I match 4 1N914's as close as I can that should tame it a bit. Looking at arrays all I found was UC3611N which is 4 Independent Schottky diodes in an IC. Not sure if that would work and I can't get hold of 1 here anyway without paying almost £20!

I'm using 42TM018's for the transformers btw, I found it sounds better with 0.1u instead of 10u cap from the osc to the carrier i/p. Looking at geofex splitter;
http://www.geofex.com/FX_images/splitter.gif
I see that the o/p of the transformers have LPFs with fc= 15 kHz, R=10k C=0.001u.

Farnell doesn't sell it that expensive .

[dthurstan]

Farnell is where I found it. It's not in stock so to get it from the US is an extra £15!

[Taylor]

When I was playing with these, the best I could do was with a uc3611 quad schottky diode array. That's the best matching you'll get, and less gated audio because of the lower diode drop.

[PRR]

http://www.acoustics.hut.fi/dafx08/papers/dafx08_29.pdf

[dthurstan]

@ slacker
Yes cheers my dodgy drawing skills. sorted now.

@ PRR

Thats an interesting read, cheers. Some points I noted.

• So I should take the transformers induction into count when working out the frequency response of filters/amps around the ring mod.
• Ge diodes are used for there soft slopes and fast turn on, si diodes create harsh distortion in the o/p. In the model the bias point is shifted above 0 using si diodes with a constant, I take it this can't be done in a real ring mod. However other people have used si diodes with good results.
• The modulator & carrier should be within +/- 200mV so 400mV RMS, I'll have to measure what is going into my ring mod.

@ Taylor
I think it was a post from you where I pick up the IC idea. I might have to try and get hold of 1, it may just mean waiting till there in stock in the UK b4 buying.

Thanx

[dthurstan]

The oscillator is outputting 800mV. I need to reduce it, I'm not sure what is the best way to do it. Should I increase the 10k resistor, or should I use a potential divider on the output of the oscillator?

[brett]

Hi
try a pot to cut the modulation voltage.
One of the main problems that I found with transformer-based designs was that the centre tap might be 10 or 20% off-centre. Maybe you should apply an offset voltage?
It is certainly possible to obtain much stronger modulation than distortion. Distortion should be only a few percent.

I've had success with a 386 opamp driving a 8:1k transformer, ring and 1K:8 transformer. The idea is to boosts the guitar signal (nominally 0.1V) with the 386 to about 1 V, then step it up 11x in the secondary of the transformer (to about 11V). The high voltage in the secondary and diode ring minimises losses and balance issues. The second transformer (1k:8ohm) presents a high impedance and steps the signal back down to a volt or so.
cheers

[dthurstan]

Here's further alterations I've done;



Cheers Brett. In the end I increased the gain of the guitar signal from 5-6 to about 11 (increasing the attenuation at the Ring Mod o/p).

I have now replaced the 1N456s with 1N914s. I tried matching the diodes using a DDM, but there wasn't much between them and they change according to temp etc. I have a circuit somewhere which shows a better way. Using the 1N914s has changed the sound of the circuit but I'm not sure how, it's not a massive difference.

I also have added the 100k trim at secondary of the 1st transformer as in the ring stinger.
http://homepage.ntlworld.com/paul.nelson666/StringRingerv12r.pdf
This doesn't do anything, I assume its supposed to balance the secondary i/p. The Stinger does have the guitar going to the secondary though?

The bleed thru isn't too bad however mixing in more dry makes it more pronounced.

[kristopher612]

i'm working on a transformer based RM myself right now.  i believe that the trimmer on the input transformer is to cancel bleedthrough from the oscillator.  if you look really close though, and this is about where i am in the project after having to put it aside when this semester started, the oscillator is going into the transformers primary, and guitar signal is going into the center tap.  i haven't had a chance to rewire this portion and confirm bleedthrough cancellation, but it looks like it should work.  if you get a chance to try that soon, let me know.  if not, i may be able to work on this some next weekend, and i'll give a heads up about it.

[dthurstan]

kristopher612 just swapped the modulator & the carrier around ala the Ring Stinger and it doesn't change the carrier bleed thru at all. From what I can gather si diodes from the same batch will be pretty close anyway so I'm not sure if it's even worth getting the UC3611N as it probably won't make much of a difference?? I might try Schottkey diodes (1N4001) see if that helps.