Posted by & filed under GR-700, Jupiter 6, Jupiter 8, JX-10, JX-3P, JX-8P, MKS-30, MKS-70, MKS-80, Oscillators, Roland.

Probably not what Roland had in mind

What do Roland’s “Cross Mod” or “Metal Sync” really do? There are quite few people asking this question on the internet, but no-one has much of an answer. Here’s a few forum discussions I found about it:

https://www.harmonycentral.com/forum/forum/Keyboards/acapella-18/330134-

http://forums.rolandclan.com/viewtopic.php?f=59&t=54410

https://music.stackexchange.com/questions/57210/what-exactly-is-the-cross-modulation-on-roland-jupiter-and-jx-analog-synthesiz/71048

Ok, so Harmony Central was never going to provide a solid technical answer, but the Roland Clan forums or (even better) Stack Exchange might have turned up someone who actually knew what they where talking about. No luck this time.

So what’s to be done? We’ll just have to read all the schematics…

Contents

Roland synths with Cross Mod or Metal Sync

Firstly, which synths actually have these features? And do they all implement it the same way? The most likely answer to that second question is “No”, which might explain some of the confusion.

There are nineteen synths from Roland’s history between the Jupiter 4 in 1979 and the MKS-50 in 1986 (the last analog synth they released before the D-50 in 1987) which we might want to look at, but several are monosynths or single VCO/DCO instruments, and any sort of “cross mod” or “sync” feature needs more than one oscillator. Checking schematics and user manuals leaves us with the following eight instruments that have these features in some form:

Instrument Year Notes Cross Mod? Sync? Metal Sync?
Jupiter-8 1981 Dual VCO polysynth Yes (FM) Yes
Jupiter-6 1983 Dual VCO polysynth Yes (FM) Yes
JX-3P 1983 Dual DCO polysynth Yes Yes
MKS-30 1984 Rack version of JX-3P Yes Yes
MKS-80 1984 Rack version of Jupiter 6 Yes (FM) Yes
JX-8P 1984 Updated JX-3P Yes (AM) Yes
JX-10 1986 12-voice version of JX-8P Yes (AM) Yes
MKS-70 1986 Rack version of JX-10 Yes (AM) Yes

Additionally, there was the GR-700 guitar synth which is very closely related to the JX3P. I haven’t looked in detail at the other members of the GR guitar synth family, but I expect they are derived from one or another of the instruments listed above.

What do Roland mean by Cross Mod?

What Roland mean by “Cross Mod” depends on which instrument they’re talking about. There are three generations of “Cross Mod” in Roland history. Each generation uses the “Cross Mod” term in a completely different way (yeah, really..) and the time-periods of each generation overlap to some extent.

It’s no wonder there’s a lot of confusion about it, because for several years in the middle 1980’s, Roland were releasing instruments which used the same term in completely different ways. In fact, the MKS-30, MKS-80, and JX-8P were all released during 1984, and each instrument has a different definition of Cross Mod!!

The reason is probably marketing: Our poor gigging musician goes to the store to ask about the new Roland synth they’ve heard about, and they want to know if it has the special features of the mighty Jupiter 8 which came out a few years earlier and which they cannot afford. “Absolutely!”, says the sales assistant,”Cross mod! It’s right here, look!”. Only in reality, they’re likely being sold something else entirely.

So…I still don’t know what “Cross mod” means!

Ultimately, if you take “Cross Modulation” to mean only “One oscillator does something to the other”, you won’t go far wrong, but that doesn’t define things very much. Thanks Roland 😉

Generation One (1981-1984): Jupiter-8, Jupiter-6, and MKS-80

In the early VCO-based instruments, Roland used “Cross Mod” to mean frequency modulation (FM) of one oscillator by the other. The term is used this way in the Jupiter 8, Jupiter 6, and the MKS-80 module. These are the only synths that have this option.

Jupiter 8 Cross Mod and Sync

Jupiter 8 “Cross Mod” is FM from VCO2 modulating VCO-1. This is described in the manual as “ring modulation style effects” which is true in as far as they can both produce non-harmonically related pitches. Technically, they’re not similar at all.

Note that this FM is fed to the 1V/Oct CV input of the oscillator, so it is exponential FM, not linear FM.

The Jupiter 8 also offers classic hard sync from VCO-1 (master) to VCO-2 (slave) with its “sync” switch.

Jupiter 6 Cross Mod and Sync

Jupiter 6 “Cross Mod” is FM from VCO-2 to VCO-1, just like the Jupiter 8. There is a CEM3360 VCA which controls the amount of FM, and this can be set manually (the “Manual” slider in the Cross Mod section) or via Env-1 (the “Env-1” slider in the Cross Mod section). The Jupiter 6 is more sophisticated than the Jupiter 8 in this respect – it allows envelope-controlled FM depth, whereas FM depth is a fixed setting on the Jupiter-8.

The two Sync buttons allow hard sync in either direction. You can sync VCO-2 to VCO-1 or the other way around.

Generation Two (1983-1984): JX-3P, GR-700 and MKS-30

On the later DCO instruments they used “Cross Mod” to mean any type of interaction between the two oscillators. “Cross Mod” became merely a panel heading under which other options were placed – no frequency modulation “Cross Mod” in the earlier sense appears. It is impossible to do audio-rate FM on a Roland DCO synthesizer. The digital counters the DCOs are based on are simply not updated quickly enough. If you’re not clear how DCOs work, you could read the article about Roland’s DCOs which covers this in more depth.

These Generation Two DCO synths are the only ones which have the additional “Metal Sync” option, which we’ll look at in detail shortly.

JX-3P, GR-700, and MKS-30 Cross mod: Sync and Metal

The JX-3P in 1983 is the first of Roland’s DCO synths to offer two DCOs per voice and therefore have the possibility of some kind of interaction between them. The Juno-6 and Juno-60 had used DCOs in 1982, but only had one per voice.

The JX-3P offers three options in the “Cross mod” section: Off, Sync and Metal. The GR700 guitar synth and MKS-30 module have an identical architecture to the JX-3P, so we won’t treat them separately.

The “Sync” mode resets both the 8253 counter and the ramp integrator. This produces classic analogue “hard sync” sounds. Despite the underlying technology being a DCO not a VCO, the final output is identical.

The “Metal Sync” mode only resets the integrator, but leaves the counter running. This means that the slave oscillator always resets when it would have done, but also resets some extra times based on the rate of the master oscillator. This is somewhat like ring modulation and will produce similar metallic sounds. This is not an option that exists on a VCO-based synth, so Roland have taken their DCO technology beyond what their previous synths could achieve.

Note that neither of these “Cross Mod” options have anything to do with the earlier Generation One meaning of Cross Mod. Both of these are completely different effects, produced a completely different way.

Generation Three (1984-1986): JX-8P, JX-10, and MKS-70

The last generation DCO instruments are clearly based on the same voice architecture as the earlier Generation Two, but they do add one or two upgrades. One of these is another variation on “Cross Mod”. This time it’s a VCA before the source mixer, which performs audio-rate amplitude modulation (AM) of one DCO with the other. This is almost ring modulation, but unlike ring modulation, AM doesn’t completely remove the original signals, although it does add sum and difference frequencies. The result is similar but perhaps not as extreme, which might even be slightly more useful!

Note that these later instruments don’t include the Metal Sync, which is a shame as it’s a pretty unique feature.

JX-8P Cross mod: Sync1, Sync 2 and X-mod

The JX-8P is very much an upgraded JX-3P, so it’s not a surprise that the circuitry is very similar. For some reason Roland decided to do some of the switching in the voice using analog switch ICs rather than the logic gates in the earlier generation. I haven’t investigated what the advantage of that might be. They also added an extra VCA mentioned above.

There are four cross modulation options on the JX-8P: Off, Sync1, Sync2, and X-Mod.

The “Sync1” mode is the classic hard sync setting. DCO-1 acts as the master oscillator and DCO-2 is the slave.

The “X-Mod” mode is the audio rate AM setting. The output of DCO-1 controls the level of DCO-2.

The  “Sync2” mode is simply both the previous two modes switched on together. So you get hard sync and AM at once.

JX-10, and MKS-70 Cross mod: Sync1, Sync 2 and X-mod

The JX-10 adds more voices (twelve instead of the JX-8P’s six) but otherwise the voice architecture is identical. The MKS-70 is the rack version of the JX-10. In all cases, they offer the same Sync1, Sync2, and X-Mod modes as the JX-8P.

 

What is Roland’s “Metal Sync”? Let’s have a closer look…

We mentioned briefly in the Generation Two section above how the “Metal Sync” feature differs from a classic “Hard sync”, but let’s have a look at that in a bit more detail.

In a typical ramp core synth oscillator, the standard “Hard Sync” effect is generated by using the reset pulse from one “Master” oscillator to reset the ramp integrator of one or more “Slave” oscillators. This locks the frequency of the slaves to the master’s frequency, but the waveshape produced is determined by the ratio of the slave’s frequency to the master frequency. It looks like this:

Standard “Hard Sync” waveforms

In the diagram above, you can see the way the “Hard Sync” slave waveforms are reset to zero everytime the master resets. They never complete the full wave cycle they would have done (shown dotted for a couple of examples).

The “Metal Sync” effect which Roland invented for the JX-3P is similar but subtly different, and depends on the different technology used – it’s a DCO effect, not a VCO effect.

In a VCO, there is only a ramp integrator keeping track of the frequency, so when the master resets the slave’s integrator, the slave is completely reset to zero. In a DCO, there are two parts; a digital counter keeping track of the frequency and generating reset pulses, and a ramp integrator generating the waveform. To completely reset the DCO, you need to reset both the counter and the integrator. This is what the normal “Sync” mode does on the JX-3P and the subsequent instruments, and it produces waveforms exactly like those above.

But what would happen if you only reset the integrator? The counter would still carry on as if nothing had happened, and would cause the DCO’s ramp to reset at the time it would have done anyway, as if nothing had happened. The master frequency’s reset pulses no longer entirely determine the frequency of the output. Rather, they just provide “additional” reset pulses scattered across the waveform at the master frequency. The effect of this is complex, but it’s not entirely unlike ring modulation and can give similar non-harmonic sounds, hence the “Metal” name.

Roland “Metal Sync” waveforms

 

Let’s have a look at how that’s done in the circuit. I’ve chosen the MKS-30 schematic because it’s the clearest one I have, but the JX-3P and GR700 are virtually identical.

Roland MKS-30 DCOs

IC31 to 34 are the 8253 digital counter chips that act as the DCOs frequency reference. Each chip has three counters, so four chips provides enough counters for all twelve DCOs. IC31 and IC32 provide “DCO-1” for all six voices, and IC33 and IC34 provide “DCO-2” for all six voices.

We can see the DCO-1 counter’s square wave output on the left (in purple). It comes down via the black bus wiring to C12/R29 where it is converted into a short pulse. This pulse is used to reset the DCO-1 ramp integrator (in the red circle). It is also taken to DCO-2 to provide sync features. We can see it taken to NAND gates 5d and 5a (the red wires). These gates allow the different modes to be turned on and off. For normal “Hard Sync”, both gates are turned on and the pulse goes via the blue wire/black bus to reset the DCO-2 counter in IC33. It also goes via the green wire to the reset transistor of the DCO-2 ramp integrator (in the pink circle). The normal DCO-2 reset pulses come from IC33 via the pink wire/black bus.

For “Metal Sync” mode, only IC5a is turned on, allowing DCO-1 to reset the DCO-2 integrator but not the DCO-2 counter. This gives us the metal sync waveforms we’ve looked at above.

 

JX-8P, JX-10, and MKS-70 Cross Mod: Amplitude Modulation

 

The JX-3P and JX-8P use a simple single-transistor-plus-op-amp VCA with dual inputs to control the levels of the DCOs. The same circuit was used in the TR-909 drum machine – the example shown is from the TR-909’s rim shot.

 

The JX-8P circuit adds a second VCA to provide audio-rate amplitude modulation of DCO-2’s signal by DCO-1. This AM is labelled “X-Mod” by Roland, and appears as one of the cross mod options.

JX-8P DCO with mixer VCA (blue) and cross mod VCA (orange) highlighted

Let’s have a look what this section does in detail.

Each DCO has a pair of switches for selecting its waveforms, Ramp and/or Square. These are “c” and “d” (shaded) for DCO-1, and the output is via the red path. DCO-2 has switches “a” and “b”, and is output via the green path.

The two-input mixer VCA is based on IC60a (blue circle), with Q15 controlling the amount of DCO-2 (green path) and Q19 controlling the amount of DCO-1 (red path) going into the mixer.

The other VCA is based on IC60b and Q16 (orange circle) and this is the one that is used for “Cross Mod” amplitude modulation. The DCO-1 signal is fed to the base of Q16 via the 330K resistor and the vertical orange path. This controls the level of the DCO-2 signal on the horizontal orange path. The output from the VCA goes out via R104/10K to the mixer VCA.

The ~METAL signal below the red path switches the VCA on and the direct DCO-2 off, effectively replacing it with the AM signal from the VCA. It seems pretty simple!

To make sure we understand it correctly, we can simulate the circuit in LTSpice. This is the simulation schematic:

I’ve simulated just the two VCA circuits, and the effect of the ~METAL signal. Here’s an example waveform output:

For this demonstration, I set DCO-1 to a very slow triangle wave (not a possibility on the original synth). I experimented running the triangle wave from -15V to +15V, but you can see that only positive voltage really has an effect. This leads me to believe that the synth’s DCO’s must output unipolar positive waveforms, and they won’t be as loud as 15V (the op-amps would clip before then). A 5V ramp is most likely, since that’s the same level as the square wave that comes from the 8253 counter.

The second half of the image shows the situation with ~METAL high. The two DCO waveforms are simply mixed. Again, negative voltages don’t do anything much – the triangle only shows up when it goes positive.

I’m going to keep playing with the simulation and see what else I can discover.

Conclusion and Feedback

It’s no wonder there is such confusion about Roland’s Cross Mod and Metal Sync. It’s taken me quite some work just to untangle it all!

Any comments, corrections, or feedback is always appreciated. Please get in touch.

5 Responses to “Roland “Cross Mod” and “Metal Sync” – What do they actually do?”

  1. Sherm

    Very nice, article.
    It clears up a lot of the mystery surrounding cross mod. Have you seen this article by Eric Persing on the JX-8P? The first two pages has content on cross-mod, and specifically he mentions that cross-mod 3 is “a form of frequency modulation”. Is that a contradiction to what you’ve stated here (ie cross mod 3 is AM).

    One last suggestion. It would be great to see a graphic or 2 about cross mod 3 for the 8P like you’ve done with metal sync for the 3P.

    Thanks again!
    Keep up the good work. I’ll be browsing your other articles soon.

    Reply
    • Tom Wiltshire

      Yes, my article contradicts Eric Persing’s article. I’m standing by what I said, though. There’s several reasons why AM is a lot more likely than FM. One, audio rate FM is probably impossible to do on a Roland DCO (never seen it yet). Two, schematics don’t lie. Three, that’s a VCA doing the cross mod between those two DCOs.
      There’s potentially a question over whether the VCA does 2-quadrant (“AM”) or 4-quadant (AKA “ring mod”, RM) multiplication of the two DCO signals, and for this reason I simulated the mixer section in LTSpice and had a look at what it does. It seems to be a simple AM VCA.
      I hope this clarifies.

      PS: Thanks for the suggestion. That last section could probably use a graphic or two, couldn’t it? 😉

      Reply
      • Sherm

        Thanks so much for the response!
        It clarifies much and FWIW your article leaves little doubt !

        Reply
  2. Russell McClellan

    Thanks for the great article, Tom!

    I’ve done a similar investigation on the JX-3P for a polysynth project I’m currently working on, and came to the same conclusions regarding how metal sync works – although you’ve written it up much better than I would have :-).

    Curious challenge of how to implement a “true” cross mod on a JX-3P style DCO. Seems pretty hard to me – maybe you could add the modulation signal to the integration voltage from the DAC, but of course this would never change the overall periodicity of the waveform, only the shape of the saw inside the period. Might be cool though.

    One place where x-mod would be really easy to implement would be in the alpha-juno style pure digital oscillators (where of course, the frequency modulation could be done in the digital domain). In a way, I’m pretty disappointed they never did a “JX” version of the alpha juno with two oscillators. I think that would have been a pretty powerful synthesizer. I guess the novation peak is the even-higher-end version of this, and it does include oscillator FM.

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *