Chrisoft

This article is about my hands on of the EDIROL STUDIO Canvas SD-80. Even though I'm a newcomer to the STUDIO Canvas owners community, veterans may still find something interesting here.

STUDIO Canvas acquired

Ever since I learned that ZUN uses a SD-90 as one of his major synthesizers here, I've always been dreaming to get my hands on one of those. But these things are not particularly easy to find in China. There are plenty of deals on ebay though. However, after a failed purchase attempt made in 2016 which also led to the ridiculous suspension of my ebay account, I left the idea alone for multiple years. Things changed earlier this month, when owner203 helped me acquire my very own SD-80 and kindly delivered it from Japan.

I did not go for a SD-90 since they are all ridiculously priced in 2019. The deal I tried to make back in 2016 would cost me around $300 (shipping included). Now they are all way beyond $400. This forced me to resort to lower-end models, namely the SD-80 and SD-20. There's also the SD-50 which I almost bought back in 2014, before I found out it is essentially a stripped-down version of Roland SonicCell without the SRX expandability. Therefore although the SD-50 retains the compatibility with the synth engine found in SD-90 (both are from models supporting SRX, namely the XV-5080 and SonicCell), it comes with a complete new set of waveforms and patches which makes it sounds totally different. So the SD-50 is definitely a no go.

The only choices left are SD-80 and SD-20. SD-20 can be found very cheap online (from $59). However SD-20 does have half of the polyphony and most editability of other STUDIO Canvas' axed. Also gone are 3 entire sound banks (and presumably some waveforms). On the other hand, SD-80 costs a lot more (starting from ~$200), but is capable of generating everything the SD-90 can (the only missing part in SD-80 compared to SD-90 is the audio interface). After deciding no more compromise should be made, I went straight for the SD-80.

The module arrived two weeks after the order was placed. It certainly wasn't the happiest SD-80 in the world: the chassis bears quite a lot of scratches all around. The value knob takes a lot of force to turn and has a good amount of friction on it, to which I found a quick solution by simply pushing the knob downwards a little bit while turning it. Aside from the looks, the module itself is in perfect working order. It came with a power cord and its rack mount ears, but without any of its original paperwork or the CD-ROM. As most of the documentation and drivers can be found online, the only things I've lost are the demo midi files, which is still a shame.

Top cover	Label with S/N removed. Visible on top left corner is a foot pad that seems to be a slice of cork.
Left half of the front panel. Buttons yellowed out, showing the module's age. Scratches and dents are visible.	Wow! Such Optical! Much MIDI!

The configuration

Connection to the computer

Without the audio interface portion, it's impossible to use the SD-80 with only a USB connection: you need something for it to connect to in order to make a recording.

To begin with, I wanted to get something that accepts S/PDIF input for the audio interface. But those kind of things are scarce today. The cloest model I could find in Chinese market was from an unknown vendor and comes with a very sketchy chip, whose official driver only supports up to Windows 7 32-bit. As a crappy digital interface may sound worse than a more decent analog one, I went for one of the cheapest audio interfaces availble: a Focusrite Scarlett Solo (which is more than 5 times the price of the weird digital interface and probably still better than the one comes with SD-90 as it supports up to 192 KHz sampling rate).

For a audio equipment novice, connecting things up wasn't that straightforward: I tried connecting the phone output on the SD-80 to the line in of the audio interface and it kind of works. There is just one tiny problem: the left and right channels got mixed up into one, and is only on the right channel in a stereo recording. At first I thought it was my cable. But after a couple minutes of researching, I found out that every input on an audio interface is actually mono. WHY? Guess I'll never know.

So I got another TRS to XLR connector. After connecting the left channel of output 1 on the SD-80 to the microphone input on the audio interface and the right channel to line level input, it produces stereo recordings as expected. The resulting audio is very unbalanced though, due to the microphone input being much more sensitive than the line level input. In order to balance them I played a sine wave on the SD-80 and adjust the input level of each channel so that while turning down the volume knob on the SD-80, the level indicators on all channels stop blinking at the same time.

The 'direct monitor' switch on Scarlett Solo enables live monitoring of the input without operating system intervention and therefore, without latency. However the direct monitor feature on Scarlett Solo is limited to mono only, so in order to get stereo monitoring, a round-trip through the computer is required.

Settings things up in Linux

If you are absolutely insane and using Gentoo Linux just like me, make sure you've already got working audio and have the following kernel configuration items enabled:

Device Drivers --->
    <*> Sound card support
        <*> Advanced Linux Sound Architecture --->
            [*] USB sound devices
                <*> USB Audio/MIDI driver

Both SD-80 and Scarlett Solo should then just work out of box (if you are using pulseaudio [1]).

At first I thought only the 'generic' USB mode of SD-80 is supported by this kernel module, however it turns out the 'vendor' USB mode works just fine as well. It also suffers a lot less stuttering.

The USB vendor ID and product ID is 0582:0029. Changing the USB mode to generic makes it 0582:002a.

In order to monitor the input, load the loopback module first:

pactl load-module module-loopback latency_msec=1

(latency_msec means exactly what the name suggests)

then move it to the audio input of the audio interface:

pacmd move-source-output 7 alsa_input.usb-Focusrite_Scarlett_Solo_USB-00.analog-stereo

where 7 is the index of the loopback followed by the name of the input of your audio interface. Usually you can use tab completion, however if you are still unsure, you may also use a GUI utility such as pavucontrol.

I haven't tried JACK yet. But using QjackCtl should make everything a breeze.

The editor for SD-80 works perfectly using wine.

Setting things up in Windows

Setting things up in Windows is somehow more tedious than it should be. Roland didn't bother releasing a driver for Windows 10, but a web search indicates the driver for Windows 8/8.1 works on Windows 10 with a minor tweak. As the device works with a generic driver on Linux even when the USB mode is set to 'vendor', it will probably work on Windows without the Roland driver too. Although I'd rather install the official driver for Windows 8/8.1 because there is one.

The tweak is simple. Open up RDIF1023.INF in a text editor, copy everything under section [Roland.NTamd64.6.2] to section [Roland.NTamd64.7] and you are good to go. Driver signature enforcement has to be disabled as stated in the quick start manual.

Settings up monitoring is as simple as clicking on a checkbox in the control panel or a button in your DAW. You just don't see one use the command prompt for this purpose on Windows.

What does it sound like?

Well, it sounds super 'ZUNish', if you've ever listened to Touhou music. It's definitely the best sounding module back in 2002 (barring those extendable modules such as XV-5080).

All instruments are from the 'solo' set. The trumpet is 'Tp.Dark vib', which is the famous 'Romantic Tp' with a significant lower filter cutoff. Everything else is the basic variant.

In general the STUDIO Canvas sounds more expressive than most other tone generators, virtual or not. This is probably related to its default non-zero modulation level (10) on every part. [2] Technically this makes it GM2-incompatible as section 3.3.2 of the GM2 specification says the default value for modultion depth should be 0.

Also the STUDIO Canvas has a bunch of waveforms of instruments played with vibrato, which is absent on some expandable modules such as the XV-5080 [3], making vibrato sounds much more natural than simulating with LFO modulating the pitch and other parameters.

Relation to EDIROL HyperCanvas and the Cakewalk TTS-1

There's a rumor saying HyperCanvas and TTS-1 is essentially the 'classical' set from the STUDIO Canvas. And I believed the rumor has been around for quite some time. Finally I can check it out myself.

The first impression is they do sound quite alike. But as soon as the drums kick in, you'll find out they actually sound different.

It's quite obvious that the drums set of TTS-1 is actually taken from SC-88Pro (except Elec. Snare, which is taken from Electric Set). The piano sounds almost identical[4]. Sax, marimba and sweep pad also sounds identical. Bass seems to heve different velocity curves and guitar in TTS-1 sounds more bright.

Further more, if you have Super Quartet, which is also made by Edirol, you may find out they also share many similar sounds, namely 'Rockabilly', 'Jazz Bass', 'Rock Bass' and 'Acoustic Set', which correspond to 'Rockabilly', 'Jazz Bass', 'Fingered Bs2' and 'St.Standard'[5].

These old Edirol software synthesizers might have a very different engine compared to the XV engine found in the STUDIO Canvasses considering the DSP power of PCs of that time period. So it's normal that the TTS-1 produces subtly different sound even when the same samples are used.

SOUND Canvas VA doesn't simulate anything it claims to perfectly, but it's the best result someone who doesn't physically own a SC-88Pro can achieve I guess.

All audio files above were recorded using Cakewalk by Bandlab. The Cubase Pro purchase was just for more instruments to be added to my HALion library.

Here are more bonus clips recorded using arecord on Linux.

And finally SD-80 in a mix, featuring two signature instruments ZUN used a lot in recent games ('Romantic Tp' and 'Reed Romance') and also 'St.Sm Choir'.

Working with the SD-80

Fighting against latency

Even though Scarlett Solo can achieve a very low latency, there's no way I can squeeze out enough processing power out of my ULV dual core i7 to handle it. Therefore I was forced to live with ~11 ms latency, which is not that high, but still a quite significant amount to deal with.

Cakewalk's glitchy latency compensation makes things even worse: it works in some projects but completely out of order for the others. In order to listen to a full mix, I have to make a short recording, which is really annoying. As this is much more expensive time-wise, I started using more guesswork before trying a full mix and it's probably not good.

Recording

Recording is simple and works as expected. However I seem to suffer from quite a high SNR value, which hovers around ~90 dB. Perhaps it's my crappy cables making me pay.

Editing the sounds

The best thing about SD-80 is its editability. Neither SD-90 nor SD-20 came with the same level of editability upon launch. [6]

As claimed by the Sound On Sound review, the editor for SD-80 is very similar to the one for XV-2020. The editor exposes a generic sample-based synthesis engine quite comparable to the one found in HALion (with more restrictions, of course). It lets you take full control of the synthesis engine and even create new patches not found anywhere else.

The synthesis engine consists of four layers (with up to two samples [7] for each layer, plus frequency modulation), four filters (one for each layer), eight envelope generators (amplitude EG and filter EG for each layer), two LFOs [8] and a 4*4 modulation matrix. There are also 90 post-processing effects (some of which are combinations of multiple simple effects, hence the name 'MFX') to choose from for each channel.

Restrictions aforementioned include inability to use external samples so you are essentially locked in to 589 built-in waveforms (which is probably fine because it's not what the STUDIO Canvas line is aimed at), limited number of envelope nodes, limited modulation routing, restricted number of different MFX's that can be used at the same time (3).

The editor has quite terrible user experience. Guess it's common for such decades-old software. It comes with a twenty-page pdf manual (which contains a lot of excerpts from the XV-5080 owner's manual), which focus on the internals of the synthesizer, and a html manual, which focus on the interface of the editor.

I've also tried editor software for other STUDIO Canvas modules using the same generation of XV engine. The SD-20 editor works seamlessly with SD-80, and it only offers basic parameters editing just like TTS-1. The SD-90 editor, despite being extremely similar to the SD-20 editor, doesn't work at all, which is not surprising as it's essentially the SD-20 editor plus an editor for the audio mixer and audio effects processor found in the SD-90, which the SD-80 lacks.

By the way, if you have hi-dpi display and use Windows 10, be sure to use 'System (Enhanced)' scaling behavior for the SD-80 Editor. It scales perfectly that way.

I'm not going to scatter the post with even more pictures, so please follow the links for selected screenshots of the editor:

main screen / 'part survey' / layers / amp EG / LFO / layers mapping / modulation matrix / patch options / 'part all' / MFX editing / Rhythm layers editor / Rhythm amp EG

Impact on the workflow

First of all I've to record and thus, deal with latency, noise and clipping. It's pretty annoying as already mentioned above.

Also the SD-80 doesn't integrate very well with any DAW. There's no way your settings of this hardware synth get saved automatically with your project like a virtual synth. Also forget about automation and all sort of things. You have to use MIDI events and system exclusive messages (especially tuning some obsecure parameters) for this purpose, and DAW software doesn't have great support for system exclusive messages in general.

Quirks

The SD-80 is GM2 compatible. However many midi files you randomly find may not play on it even when the device is in GM2 mode. If the midi file being played has any sort of bank selection event not recognized by the SD-80, the corresponding part will produce no sound at all.

If your midi file has NRPN messages, things will get even worse. These things are essentially undocumented for the SD-80. All I could say is good luck keeping your eardrum intact. [9]

The manual says 'This set ("classical" set) is also used when GM2 data compatibility is important'. However upon receiving the GM2 on message, the module automatically selects 'contemporary' set for every part. It seems impossible to change this behavior.

The XG Lite mode has way more instruments available than what the manual has listed. Bank 18 patch 1 gives you 'Piano 1d' (persumably 'Mellow Grand Piano' in XG specification), which is not listed in the manual. It actually has 489 normal patches, 49 sfx patches, 2 sfx kits and 9 drum kits [10]. It's quite irony to find out that they have more patches for their competitor's standard than their own legacy standard. However these sounds are relatively lame when compared with authentic YAMAHA XG synthesizers of the same period. Particularly some sounds in different variant slot of the same patch number sounds almost identical. Roland obvious put less effort in these sounds in terms of sound designing.

MIDI timing messes up if events flood in within a demisemiquaver worth of time. This quirk resembles the problem QMidiPlayer for Windows suffered from in its early days. But I'm pretty sure the cause is completely different. [11] Also this could be a common issue of old MIDI devices, as my old YAMAHA keyboards does the exact same thing.

The sound generated by the SD-80/90 is phasey as could be heard in ZUN's early works. If unsure, take a look at the spectrogram.

The module, as all electric appliances do, generates heat. The manual says "A small amount of heat will radiate from the unit during normal operation", and the chassis do get quite warm even when it is completely idle. This didn't become an issue until later.

Finally there're a couple of quirks in Roland's former partner Cakewalk. Whenever the input signal level exceeds 0dB, instead of clipping the audio, it produces a loud cracking or popping sound in the record. Cubase and arecord have no such problem.

Another quirk with Cakewalk is its metronome. Metronome settings in Cakewalk is saved on a per-project basis and there's no way to change the default value (unless creating a template, which is useless for existing projects). By default it sends the metronome to the default MIDI device, which result in records with metronome clicks in them.

Beneath the chassis

Although I cherish the module quite a lot and there are warning text on top of it (which I can't read because it's in Japanese and ... well, English), I disassembled the unit.

CAUTION!! 注意!! ATTENTION!! (/a.tɑ̃.sjɔ̃/)

As most other vendors does, Roland build their audio equipment 'like a tank'. The entire chassis is made out of aluminium and steel, and has an absolutely crazy amount of screws on it. The front panel came off after taking out 5 screws on top and bottom, after which I took out the value knob and the rubber button sheet for a cleanup.

The top cover was freed after taking out 11 screws. Under the top cover lies the guts of SD-80. All boards except the power supply board uses SMD components extensively. The main board is made exclusively using SMD components, which is probably quite impressive back in the early 2000s.

The power supply module is surrounded with thick plastic sheet persumably for insulation. Rated voltage for the module seems to be changable by using different pin layouts on the input side. The power supply module is made of two separate boards. The two boards are conncted with two wide connectors that only have a few pins on them. I would guess the pin layout on the connectos sets the rated voltage of the module but I'm not sure. I would definitely test that out someday.

I took note of the engravment of every chip on the main board, which could be found below.

What the manual doesn't tell

A very sensible thing to do after the disassembly is to search the engravment on the chips online. Searching 'RA08-503' found on the largest Roland-labled chips [12] brings up a whole new world to me: service manuals for Roland synths. Sadly there are no manuals for any STUDIO Canvas models, nevertheless, they are very useful for learning more about the insides of a Roland synthesizer.

There are a couple of models bearing the 'RA08-503' chip found in the SD-80. Examples include XV-5080 and MC-909. This chip is referred to as 'XV' in their service manuals (also in the SD-80, as you will find out later). It's safe to guess they handle the most important work of a sound module -- tone generation. Also I observed that all modules supporting 32 parts and 128 polyphony have two of these chips in them, while the ones with only one chip only supports up to 16 parts and 64 polyphony. Maybe that's the limitation of the single chip. Also one can tell the chip also handles some weird job such as LED indicators from the circuit board schematic.

A very lonely chip on the right side has '6417706' on it. Turns out this is a microprocessor implementing the SuperH architecture. It's the same CPU as found in MC-909 (which makes sense as they are produced around the same time). Linux kernel has support for this processor[13]. The processor has a maximum clock speed of 133 MHz and is underclocked to 128MHz in the MC-909. It's probably underclocked even more in the SD-80. There's an unpopulated D-sub connector persumably for debugging near the CPU.

Testing mode

According to the service manuals available for other models, they all have a hidden test mode. Some also have the ability to update system firmware. Entering testing mode often involves turning the power on with a combination of keys pressed. Every model have its own way to enter testing mode and there's no obvious pattern. At that point, I was desprate to dive into it. So I simply tried all button combinations consisting of one to three buttons. The result didn't let me down. I found three combinations that make the SD-80 boot into special modes.

• INST + SHIFT + PART▶ = Test Mode
• INST + PAGE◀ + PREVIEW = Program Updater
• SYSTEM + PAGE◀ + PREVIEW = Program Updater

Below is everything I found about these modes.

00 Version Check
1.03 0022 2002/07/29

01 Device Check
      ALL OK!!

02 MIDI Check
MID1:x MID2:x THRU:x

03 LCD Check
PAGE:Sel / ENC: Contr

04 SW/LED Check
    ooooo ooooo

05 Encoder Check
  Value(0-23) = xx

06 Sound Check
Push button to check
   Left channel
   Center channel
   Right channel

07 Effect Check
Push button to check
     MFX(Delay)
    System Delay
    System Reverb

08 Factory Reset
   Push [PREVIEW]

Program Updater
Version: 1.03

Program Updater
Prog:12D3[12D3]1.03

Program Updater
Updt:66FB[66FB]1.00

Program Updater
Boot:5170[5170]1.01

Update by MIDI
[ENTER][SHIFT(Exit)]

Update by USB
[ENTER][SHIFT(Exit)]

Messing around sans the chassis

As the two XV chips are arranged in a master-slave manner, I tried figuring out which one is acting as the master. I threw a bunch of midi files at it while measuring the temprature of the XV chips... with my fingers. Both chips turn quite toasty but IC19 is always warmer than IC27, sometimes it's even a little difficult to keep my finger stay on that chip.

When the action gets more intense, IC27 starts to warm up. Before I was just going to conclude that IC27 is acting as the master, I realized that I will never come into a meaningful conclusion without further reverse engineering: I have absolutely no idea how the load is distributed between the two XV's!

Frustrated, I entered testing mode to find out whether there's anything interesting if it's run with chassis removed. To my surprise, the device check failed with NG:XV2 and a bunch of other errors in the following tests. I thought I was doomed, but a reboot of the module solved the problem completely and it never show any trace of abnormality that day.

The other day, however, the problem returned. After messing with the SD-80 editor and creating random complex patches for an hour, I randomly decided to check out the testing mode again, where I was greeted by the NG:XV2 failure the second time. Again, a reboot solved the problem for the day. No amount of hardwork will put the SD-80 into a buggy state if it boot straight into normal mode. At this point I thought it could be a bug in the test or the device check could be quite sensitive to temprature, which is probably not a very good thing as there're ninno vent holes for airflow on the module whatsoever.

A few more days, I found my SD-80 frozen after keeping it on doing almost nothing for a day. In yet another case, it just randomly froze during playback after half day of usage. I'm unsure these unstability are specific to my machine.

Chart A

Chart B

Things to do besides imitating ZUN

The SD-80 does not like QMidiPlayer very much. So the first thing to do is quite clear.

Giving it a total makeup is the second thing on my bucket list. I'll probably ditch the original top cover and front panel altogether and make some custom acrylic glass parts for it.

Porting Linux to it might be a very fun (also atrocious) thing to do. The internals of the SD-80 is capable of doing much more than what it does as a STUDIO Canvas. It's got the same main processor as the MC-909, just think about the possibility out there (this is also the reason why I call it a 'trapped beast' in the title of this post). The only thing against this is that I am shy of any experience with this level of hardware hacking.

Setting up a web service where people upload their midi files and have them rendered with the SD-80 also sounds pretty cool. But I'm afraid I'll receive something from Roland by then and it wouldn't be fine for me. Is it really illegal to use a instrument on a time-sharing basis?

Seriously though if I could pull it off, I'll probably add a donation button and buy more classic sound modules for the site. Eventually it will turn into an online museum for sound modules... screw it I'm talking utter nonsense again.

Also somehow extracting the waveforms and creating a instrument bank for HALion or Kontact is tempting. But the odds of being sued by Roland is even higher even though they did not explicitly disallow sampling their early products. To be honest I found it disturbing about the sampling restriction on these 'unconventional' instruments.

Newer Roland sound modules?

Newer Roland sound modules such as SonicCell and the latest INTEGRA-7 seems to maintain the compatibility with their original 'XP' synthesis engine to some extent -- as they all support SRX expansion in a certain way. The synthesis engines are obviously improving over the years as Roland says the SRX expansion sounds built into INTEGRA-7 will sound a little different compared to earlier modules.

However, disappointment strikes as soon as I saw the 'READ' button on screenshots of the editor software for the INTEGRA-7. I felt Roland really need to make more use of the USB bandwidth: the presence of that button implies the editor software still can't reflect the realtime status of the synthesizer. [15] This is a huge drawback of using a hardware synthesizer that Roland still fails to fix to this day.

Also, the INTEGRA-7 is super expensive for an amateur and Roland ended their budget [16] SOUND Canvas / STUDIO Canvas product line years ago. So no more Roland sound modules for me I guess.

If you want to learn more ...

• Gigadenza, owns multiple sound modules, including the latest INTEGRA-7.
• Romantique Tp, a Touhou music addict that I came across on Steinberg user forum.