Wednesday, October 15, 2008

Why does the Juno-60 sound different from the Juno-106?

As you already know if you've been following this blog, I have a Juno-106 that I bought new in 1984. I've always been a fan of its sound, and I keep mine going despite the reliability issues with the 80017A IC's. I've never owned its prececessor, the Juno-60. But many people who have owned both say that they don't sound quite the same -- the words usually used is that the 60 is possibly a bit more mellow (that dreaded means-anything-anyone-wants-it-to-mean word, "fatter"), while the 106 is possibly brighter, "cold" (another one of those non-words), and some people even say it sounds "digital".

So let's start by getting a couple of things on the board out front. The VCO, VCF, and VCA circuits are nearly identical. "How can you say that", you might ask, "when the 106 uses those 80017A's that keep failing, and the 60 doesn't?" Consider: The 60 uses a VCF filter circuit based on the IR3109 quad OTA. It uses a BA662 VCA to control the resonance. A second BA662 serves as the voice's VCA proper. As for the 106? The notorious 80017A is really just an encapsulation of three ICs and some resistors. The IC's? An IR3109 and a pair of BA662s! It's the same circuit, just in a smaller package. A lot of people don't realize that when you look at an ordinary IC, most of what you see is packaging; the actual integrated circuit is a tiny bit of silica embedded in the plastic. Roland bought a bunch of unpackaged 3109s and 662s and had someone encapsulate them, and voila, the 80017A was born. Similarly, the Juno-60's DCO circuit: the counter IC that times the DCO, the reset transistor, and the wave shaping circuitry are encapsulated into the much-less-infamous (because it seldom fails) MC5534 in the Juno-106. The voice circuits are, for all practical purposes, the same.

So assuming that there is a difference in sound (and I've heard enough reliable witnesses say there is), where could it be coming from? Let's take a look at the rest of the audio processing: the portion that follows the summing amp, which combines the six individual voice signals into a single mono signal. A lot of Juno players don't realize that there is not a highpass filter per voice; Roland cleverly placed the HPF switch on the panel to suggest that the HPF precedes the VCF, but it isn't so. There is only one HPF circuit for the whole synth, and it works on the summed mono output of the voice circuits. Following the HPF is a seventh VCA, which is tied to the VCA level control on the panel (that bit looks like a kluge; maybe I'll write about it later), and then the chorus circuit, which takes the mono input and produces a stereo output. There's a bit more stuff for the master volume control and the various outputs, but all of that is bog-standard IC amp and buffer circuits.

The chorus circuit on the two synths looks the same; there may have been minor improvements that have eluded my quick scan of the circuits, but they both use the same bucket brigade ICs, the same control circuits, and the same gain make-up circuits (there is no noise reduction, which anyone who has heard either synth on headphones has already realized). However, I found some differences in the HPF circuits. First of all, on both synths, the HPF is not a voltage-controlled filter. It's basically a set of four passive RC filters. The panel or recalled setting controls an 1-to-4 analog demux which routes the signal through one of the four.

The one on the Juno-60 is pretty straightforward. Here's the portion of the schematic:

The four outputs of the analog demux are on the right (the two inputs from the control CPU are on the bottom, and the mono signal enters at the top left at pin 3). The pins are labeled as to the corresponding position of the HPF switch on the panel. As you can see, going up from output 1 to output 3, the signal gets routed through progressively smaller-valued capacitors; the smaller the cap, the higher the cutoff frequency. The output for position 0 has no cap; it's just a straight wire, so position 0 of the HPF is actually no filter at all. It's straight through.

Now here's the corresponding circuit from the Juno-106:

As you can see, it's more complex. First things: I am pretty sure that the CPU is sending the two input signals, A and B, inverted. Therefore, the pin labeled "Y0" corresponds to position 3 of the HPF switch, "Y1" is position 2, etc. The opposite interpretation doesn't make sense when you look at the circuit.

Now note the first difference: The straight-wire output corresopnds not to position 0, but to position 1. The circuits for positions 2 and 3 look pretty similar to the ones on the Juno-60. But what's all that business connected to the position 0 output, around IC4b? Well, it sort of looks like a Sallen-Key filter, as used on the Yamaha GX1. What's it doing? Note C8, the 0.01 uF cap shunted to ground. That's a lowpass filter! This part of the circuit is acting like a bass boost. (C6, I think, is just there to keep IC4b from self-oscillating.)

Second difference: Note IC4a. In both the 60 and the 106, the "seventh" VCA that I mentioned earlier immediately follows this HPF circuit. Although the VCA is an oddball part ("uPC1252"; the only data sheet I've found is in Japanese, but some Googling reveals that it was manufacturered specifically for dbx), it doubtless is based on an OTA circuit, and like all OTA circuits, it loads the input some, particularly as the gain is decreased. The Juno106 uses IC4A to buffer the input to that VCA. The Juno-60 doesn't have that buffer; it couples almost directly, only separated by a DC-blocking capacitor. That means that the 1252 VCA is loading down the outputs of the passive filters on the 60, which introduces high-frequency rolloff. So the HPF actually acts a bit more like a fairly wide bandpass filter, particularly as the VCA level control on the panel is turned down and the input impedence of the 1252 decreases. That doesn't happen on the 106 because the buffer amp provides a constant high input impedence for the HPF output.

Third difference: The analog mux used is a different part. The Juno-60 uses a 14051; the 106 uses a 4052. They work basically the same way, but possibly the properties of the analog portions of the two circuits are different. I need to look into that some more.

To me, the biggest difference is in the configuration of the filters, with the 106 providing one "high pass" position which is actually low pass. And, the loading of the filter circuit on the 60 is probably significant; circuits with some rolloff above 8KHz or so are often perceived as "warmer" by listeners. Maybe I'll have to get a 60 so I can compare them myself.


kroffe said...

Very nice and comprehensive runthru the Juno-60 and 106. I already knew about the 80017A and all that, but the exact function of the HPF had slipped my mind..:)

However, your quote:
"To me, the biggest difference is in the configuration of the filters, with the 106 providing one "high pass" position which is actually low pass."

I percieved the circuit for position 0 on the 106 as both straight signal (R25) AND bassbosst (IC4b), but I guess that's what you meant?

Thanks for the effort of writing it down anyway! ;)


Dave Cornutt said...

I actually wasn't totally sure which caps were doing what in that part of the circuit. But yeah, I think the combination of C8 and IC4b are providing bass boost, with the rest of the signal going through unaltered -- there isn't any actual high pass in this part of the circuit.

rob said...

I don't have a 106, but was surprised at how close the 60 sounds to the jp4.

Apart from the the HPF and the envelopes, oh ok and the chorus... but the basic tone is ridiculously similar.

Playing chords immediately differentiates them, as the JPs oscs each tune differently.

After reading your other article on Juno DCOs the idea of multiple master clocks (one for each osc) seems too delicious to ignore.

Do you think it would be possible?

Dave Cornutt said...

I absolutely think it would be possible, if someone was designing a DCO polysynth from scratch. I'm not sure about modding an existing Juno-60 or 106, but I won't say it isn't doable. However, I have another approach I might try someday. I've always wanted to pull the ROM from the module board, extract the code, disassemble it, and see about doing mods to the firmware. If I do, one thing I plan to add is an offset table that will add a small, randomly-chosen offset to each voice. That would eliminate the "perfect" intervals between voices, and unison mode in particular would sound a heck of a lot better.

Camille said...

great post guys!
VCA Examen

robert said...

Nice post. I agree with kroffe and think the details of your interpretation of the 106 position 0 are little off. During position 0 IC4a takes the sum of two things, (1) the unfiltered signal from R25 and (2) the crazy filter signal from R24.

Here's what's happening in the crazy filter: Stage 1 consists of C8, C9, and R22. You are right that C8 is connected to ground and will filter out high frequencies, but high frequencies will also pass through C9 without much impedance. If you work out the math Stage 1 has one pole and one zero. High frequencies begin to roll off because of the pole but then the response flattens out at higher frequencies because of the zero. So Stage 1 is a bass boost. Stage 2 consists of R18, R19, R20, C6, IC4b. It is really just an active version of Stage 1 with the same type of frequency response. See "adding bass boost circuit" here.

The combination of the two stages has two poles and two zeros. The two poles cause the roll off to be steeper and the two zeros flatten the response out at higher frequencies.

Adding the unfiltered signal to the crazy filter signal still gives a two pole bass boost but with shifted frequencies. I guess this gave better design flexibility or let them use smaller caps. I'm not sure.

I agree that it is a bass boost circuit and it sort of looks like a Sallen Key filter, but I don't think it technically is.

rob said...

Here's the page that must have planted the multiple master clock idea in my head:

orpheus said...

Hey Dave,

You probably know this by now, but the problem with the 80017A's isn't really that they fail, it's the epoxy coating. I didn't believe this myself, but voices 1 & 3 failed on my 106 and I couldn't get a replacement at that time so I tried it.

I removed the 2 failed voices with solder wick very carefully. I've read a better way is to douse the whole track until it's molten solder then simply slide out. I then soaked the 2 voices in acetone for about 3 days (after 1 day I cut away around the sides) and was able to clean off all the epoxy. They now work again as new!

The only thing that didn't work out is instead of soldering them back into the board I soldered in some sockets so I can plug the voices. This was a bad idea and they do work their way lose when I move the synth around. Anyway, hope this helps.


Unknown said...

i think that's right;
hpf 0 is boost
hpf 1 is hpf+original signal
hpf 2+3 (just made arbitrary settings for what
sounded most effective)

looked at all this quite closely for a synthedit
project. it sort of worked out ok.(but i used
very sharp 'naive' waveforms for the DCO, as the
original shape is quite sharp) -you can get quite
close to the sound by poking in the right direction
and studying the schematics to get an idea of the
signal path.
what i did wrong was not to have 6x filters+VCAs