A SAD day for Mr. Roland (with bonus: Bitscope or Bitnope?)

Once upon a time, back in the early 90s, I bought something from the old Smithys shop in Camperdown, and they had four dinky little Roland RS09 keyboards with clunky old fashioned-looking grey rocker tabs on their panels sitting unloved and unwanted with $100 price tags, right in the middle of the workstation keyboard era. I didn't give them much thought.

Shortly after that I went into another music store (might have been Venue) and found a slightly more modern-looking example of the same model with coloured buttons, and played it through headphones. It sounded magnificent, with its simple controls belying its ability to control textures by balancing an organ and string section, and by selecting whether either, neither, or both of the two voicings went through the rich-sounding ensemble circuit. I subsequently got distracted by other things, and by the time at the end of the decade when I started hankering for an RS09, the prices being asked in music stores had seriously rocketed skyward, to $400 and beyond, so I thought that was that.

In 2009 I saw Stereolab at the Metro theatre supported by Crayon Fields. Crayon Fields used an RS09, which sounded great through a PA! Stereolab had had keyboard troubles that night-Laetitia banged her Little Phatty several times against its stand to keep it working, and near the end of the set Joe Watson's Wurlitzer piano stopped working altogether. The support act obligingly set up their RS09 for him, and it stormed through French Disko, prompting Laetitia to exclaim "The Roland Rocks!" By this point RS09 prices were climbing towards $A1000.

Then one fateful day in 2017 I walked into a shop and found a grey rocker-tab RS09 sitting there with a $240 price tag. A couple of sliders appeared a bit bent, but all knobs were present and it powered on. This one was not going to get away from me, so I bought it immediately and left the shop.

All the keys and controls worked, and it was still a great sounding keyboard, if maybe a bit more trebly through headphones than I remembered, and there also seemed to be more chorus animation through one of the channels. Maybe it could be adjusted? Still well worth the price. In line with what I had heard at the Stereolab gig, the RS-09's "thin" sound was well-balanced for a keyboard combo amp, and sounded top-notch in mono. Plus-full polyphony-all 44 keys at once!!!

One day I tried isolating the output channels into my combo amp, and discovered to my consternation that while one channel sounded rich, the other only gave a slow pitch variation effect. Something was wrong, and it was time to investigate...

The original 1979 RS09 (which I own) has physical rocker-tab and slider-toggle switches for organ cancel, organ filter, string footages, and ensemble in/out, plus its ensemble circuit uses four SAD-512D BBD chips, two for each output channel, with each channel's delay lines modulated by slow anti-phase triangle LFO pairs, running at different speeds per channel. One delay line per channel can also switch in a fast sine wave LFO (again, slightly different speeds per channel) in addition to the slow LFO-much richer than the Juno or even Dimension D circuits, which only ever use a single LFO and one delay line per output.

In 1980 the RS09 was updated with Jupiter-8/TR808 style colourful keyswitches to electronically control registrations and ensemble switching, as well as changing some legending from light grey to a garish orange (gotta move with the times), plus changing the nice solid-looking "waterfall" piano keys to so-called "diving board" types, which are more vulnerable to accidents if they are ripped upwards. The electronic switches are not as reliable as physical switches due to dust contamination, however as a positive they moved to Panasonic BBD chips, which are somewhat easier to find than the Reticon SAD BBD chips used in the old version.  

Luckily, both the official owners' manual (from Roland) and the unofficial scan of the service manual (from elsewhere) both refer to my earlier version, so time to dive in.

After unscrewing the top and propping it up gently, I found a very well-laid out circuit board (typical of Roland in the analogue era), with the four modulation circuits converging to a central dual op-amp that mixed the final stereo signal. As I keep the keyboard at home, and my "real" scope is in my workshop, I thought I would check it out with the aid of a Bitscope Micro USB oscilloscope I bought in an Element14 runout sale two years ago.

Bitscope monitoring the A440 test on Mod1 and Mod2. Note the 44 decay generator capacitors on the voicing board at bottom left. The joys of full analogue polyphony!

The BitScope is a cute little device, Australian-developed and made with a good software suite thrown in. Its dependence on an FTDI driver makes it finicky, however-I eventually worked out that my Mojave-running MacBook needed to be started up with the scope attached to its working USB port for it to work.

I quickly found that its performance noticeably suffered in two channel mode, and it was very noisy with low-level (ie small-signal audio) signals. The 500ns pulses that controlled the BBD delay time also looked very slewed on the display, and almost seemed too much for its (nominally) 40MSPS performance, but on the other hand its limited storage memory made it struggle with very slow signals. Nonetheless, it helped solve my puzzle.

I had been hoping that the dead part was not one of the rare and expensive SAD512D BBD chips-maybe the LFO circuit to one delay pair had packed up? Roland conveniently provided LFO test points and all four LFO outputs appeared to be working perfectly. OK.

I then checked the BBD clock test points. All BBD clocks appeared to be working OK, and modulating in response to their LFOs. Hmmm, maybe it's the output circuitry.

At the final mixer opamp, all the delayed audio signals were present except one-Modulator 4. Maybe it's the BBD output noise filter? Fingers crossed.

First I checked the bases of Mods 1 and 2 initial single-transistor output noise filters. Nice signals moving in opposite directions to each other, showing clock noise ready to be cleaned.

Checked the bases of Mod 3 and Mod 4's initial single-transistor output noise filters. No signal on Mod 4-not even clock output. Starting to worry.

Finally checked the signal into Mod 4's BBD. It was fine. Dammit! New BBD needed! Or I could cut my losses and continue to enjoy it as-is. Hmmm, let's see our options...

(to be continued)