posted 06 November 2010 10:32            

'I'm currently experimenting with Ring Modulators and am extremely unhappy with the pure mathematic multiplication happening in the digital realm. It sounds harsh and the dynamics are out of control. So I tried real analog diode/transformer ring modulators and they do it just right.'

I know this is completely out of my range/depth/abilities, however I am fascinated to gain a handle on the concepts involved.

I use and love the FXpansion DCAM synths. http://www.fxpansion.com/index.php?page=62

they make a free mini synth to showcase the technology http://www.fxpansion.com/index.php?page=12&tab=60

These synths where created by measuring the components on similar circuits and represented in DSP code. (perhaps using http://www.mathworks.com/)

One assumes that given enough time/motivation/dsp any analogue circuit/components could be constructed.

Rather than building a specific circuit, (in Mathis case ) would it be possible to build a library of electrical component 'measurements/algs' and then find a way to re-construct a circuit of approximate non-linearity within kyma somehow? This way we can build a library of common components, (one at a time) (uumm yeah!)

in the fantasy in my head, we would then be able to re-construct simple *analogue*ish filters/ringmods/distortions complete with non-linearity's! (perhaps from commonly found circuit diagrams, or indeed taking technology apart.)

Well obviously I have no idea about the scales of effort science and technology to achieve something like this......

Charlie

(*some days later.... * - well I have done some reading on a few of the topics involved, and (obviously now) really its a total 'pie in the sky' concept. Not without an army of quite dedicated people and a bucket full of oversampling to make it worthwhile)

posted 08 November 2010 15:32            

DTRingMod.kym

Attached is a kyma emulation of the old Diode/Transformer ring modulator I did about five years ago.

hope it helps

posted 08 November 2010 21:28            

PetesDTRingModVS.kym

I wanted to hear the differences between the 'stock' ring mod and your funky circuit.

I created a crossfade so you can blend and switch.

Fascinating stuff!

posted 09 November 2010 06:43            

Solving this challenge is not simple and much of the work in creating a top DSP emulation comes down to managing the aliasing. The most straight-forward approach is usually some combination of working at a much higher sampling rate and properly band-limiting the two source signals.

posted 09 November 2010 16:36            

PetesDTRingModVSvr2.kym

I have tried to implement the Single Side RM into the equation but have been unsuccessful.
The prototype only has the one input, I was expecting two, I tried mapping the same fader the generate the frequency, yet I am not sure how to manage the frequency scale parameter to ensure parity in the experiment.

anyway....

I am currently distracted by the waveshaper, I am sure this holds the key to unlocking some analogue-ness. I will leave the ring mod conundrum to Mathis for the moment.

Thanks again.

Charlie

posted 18 November 2010 14:51            

this is very interesting because admittedly I didn't think about the mathematical differences between the diode ring and the multiplication. And I have to admit I still don't really get it, could you help me a bit.
I always thought the main difference is the diode voltage drop and especially the transformers saturation, but obviously there's a fundamental difference how the multiplication is rendered.

posted 20 November 2010 10:30            

SSDTRingMod.kym

When I looked at the circuit (many years ago) I worked out what would happen if the signals were 100V P to P. In this case the 0.6 v voltage drop and the log law of the diodes would be irrelevant. This is what I emulated in the Kyma sound.

You will notice that there are no multiplication modules in the sound only mixers. It is all done by addition, subtraction and clipping. The module is offsetting the wave form to make use of the kymas -1 to +1 limit and causing clipping or half wave rectifying to emulate what perfect diodes would do.

I've attached another version called slow sine set up so that you can see how the the module doesn't multiply at all but instead offsets and clips a 50 hz sine under the control of another sine at 0.1 hz. You can see that what this module produces is chopped up bits of sine wave which has the general feel of ring modulation but is well and truly distorted with loads of extra harmonics due to the clipping.

I had not even considered how the log law of the diode could help to make something that was closer to the almost perfect ring modulation you get with a simple DSP multiplier.

So maybe adding some of the log law of the diodes (when not fully conducting) to my modules but remembering that the log law is not perfect for the whole range of currents will give something closer to the imperfect analogue DT ring modulator.

I suspect that the imperfections of the transformers have hardly any influence on the sound as compared to the diodes loose log law-ish-ness.

In my module I offset the signal by -1 to create a clip (rectification) and then offset it back by +1 to center it up again. If instead you used some InputOutputCharacteristic modules you could emulate both the clipping and the log-ish law of the diode without having to offset to -1 and back, and you may well get the sound you are looking for.

Hope it helps

Pete

posted 21 November 2010 12:35            

I've now looked at the DT circuit and found out how it works again so ignore what I said before as 100 V would simply blow the diodes to bits as soon as the input voltage got above 2 times the forward voltage.

Anyway the circuit is quite difficult to understand what's going on unless you break it down and think about it a bit at a time.

The first thing to do is to move the ground point (Ov ref) to the centre tap of the output transformer. This doesn't alter the operation of the ring modulator in any way except the carrier voltage is now applied to the centre tap of the input transformer and has to be inverted. This now make it much easier to under stand.

If you now replace each diodes with perfect zero volt drop diode in series with and a 1 volt battery. This is not quite a real diode but it is close to emulating a diode with 1 volt forward voltage. Now the signals have to be max +/- 1 volt peak.

Now we can stat to see what is happening.

The centre tap of the input transformer is now the carrier voltage and the top tap of the input transformer is the sum of the input voltage and the carrier. The bottom tap of the input transformer is the carrier voltage minus the input voltage. (This part is true even with real diodes). In this circuit only one diode/battery will ever conduct at one time and that one diode will not start to conduct until the sum of the carrier and the input voltage is greater than 1 volt.

You can see that there are 4 quadrants and you can consider the circuit as being in four different states.

Input plus and Carrier plus..... Output can only be Plus
Input plus and Carrier minus.... Output can only be Minus
Input minus and Carrier plus.... Output can only be Minus
Input minus and Carrier minus... Output can only be Plus

Each of the above states use only one of the 4 battery/diodes and as other diodes are backed off.
In each of the above states you only need consider the one input winding and the one output winding that is attached to the relevant diode.

You can see that the output equals the absolute value of the input voltage plus the absolute value of the carrier minus 1 (clipped at zero) but with the polarity of the output is adjusted depending on which of the 4 states the system is in.

This is what my Kyma sound emulates.

Now if we go back and replace the diode/battery components with real diodes we find that although you don't have to wait for the sum of the carrier and the input to be greater than 1 volt before you start to get an output and although more than on diode can now conduct at the same time, the differences are not that different.

It can be seen that this circuit adds the input and the carrier first and, then gets distorted (smoothed) a bit by the log law of the diodes, and then switches the polarity of the output as an exclusive OR, but it does not perform multiplication in any way. For multiplication the input and carrier would have to be logged first, and only then added together and then anti logged. This circuit does nothing like that and that's the main reason why it sound so different from a true multiplier. I believe the minor imperfections of the transformers and small capacitances have very little bearing on the over all sound.

I hope this makes sense.

Pete