Roland S-750 Refurb

Overview I bought a used S-750 in May 2018 for £85, which is a bargain, especially as it has an expanded 18 MB of sample memory, and was from Air Studios. It was manufactured in February 1993 and is a fairly late model although it does have wire kludges to the main PCB. Whilst it is rather large and heavy I plan to upgrade the sampler and use it with 7 Roland sample CDROM’s permanently attached on a SD card in the casing.

New LCD

New LCD The back light of the blue and original LCD had faded. Rather than just replace the back light EL I decided to replace the complete LCD with a new grey and white one which I bought off eBay to ensure it worked.

It does work and look great, but its a few millimeter higher than the original so the perspex screen has to be fitted back in 2mm proud rather than smooth to the front panel. It is a very nice LED screen. It is not easy to remove the perspex bezel from the old LCD as the double sided tape had hardened over the years.

SCSI Drive I wanted to fit a SCSI2SD V6 drive inside the case rather than having it taking up space outside. The S-750 main board has the connection for the external SCSI 25-pin connector and an unused 50-pin connector. In fact its just the bare PCB holes with no socket. On investigation this proved not to be a 50 pin SCSI connector or even the 40 pin socket on the S-770 which is used to connect a hard drive internally. It is a IDC 2mm pitch 50 pin connector wired as follows:

50 pin SCSI

Pin Function
16 Data 0
18 Data 1
20 Data 2
22 Data 3
24 Data 4
26 Data 5
28 Data 6
30 Data 7
32 PARITY
35 ATN
36 BSY
38 ACK
39 RST
40 MSG
42 SEL
43 I/O
44 C/D
46 REQ

Its a completely non-standard proprietary Roland pin out, but it does have the SCSI bus on it with power. My plan is to put a 50 pin socket onto the Main PCB and cable it to a new PCB that maps this to the correct 50 pin SCSI pin out and attaches to a SCSI2SD drive with a 50 pin socket. This will provide 7 SCSI drives which I will load with sample CDROM’s. The only down side is that I cant swap SD cards without opening the case and getting the sampler out the rack., but that doesn’t worry me given I will have 7 drives to play with.

This new PCB has two 50 pin sockets and space for the SCSI2SD drive to be mounted to the PCB. The S-750 has a convenient set of 4 hols in the chassis to mount the 90 x 100 mm PCB.

Roland Quick Disk Emulation

Overview The Quick Disk drive in the early Roland samplers is prone to failure and long since obsolete. One of my background projects is to create an micro processor emulator, as I have 3 of these samplers. This post details progress to this goal.

How the data is stored Each side of a QD contains one bank of 64k bytes of data which contains sample data, wave parameters and the performance parameters and split points. When a QD disk is loaded it moves the sample data as 12-bits into DRAM wave memory and replaces the current 8-bit wave and performance parameter values in the SRAM chip with the new ones. However you can choose to just load the sample and wave data and not the 26 performance parameters, if you want to keep the original settings. You can also preview the name of the sample before loading the disk.

Variations The S-220 introduced the capability to store more parameters as it introduced new capabilities beyond the S-10 and MKS100. The S-220 OS recognizes three types of sample disk:

• Type I – S-220 Basic Information (F2 pushed during save)
• Type II – S-220 All Information
• Type III – S-10 or MKS-100

The different amount of parameter data does not affect how the QD works or is emulated.

Sample Memory The wave memory is built from six HM50464P 64kx4 bit DRAM chips, which provides 128k of 12-bit sample storage in total. Because this data is stored in 8 bit bytes on the QD the 12-bit sample data is stored as 2 complements in 2x 8 bit words onto the disk. A disk side therefore contains 48k bytes of sample data and up to 16k bytes of wave and performance data. The 29 items of wave data control how the sample is played back; loop points, envelope settings, auto bend etc.

The limitations of the QD storage size means Roland had to split the sample memory into 4 banks, with each bank requiring a single side of a QD. So a full memory load means 2 disks and 4 load operations.

The samples are replayed via a 16-bit DAC at 15 or 30 kHz, which provides 2.18 seconds or 1.09 seconds of sample time per bank. This is remarkable short compared with today’s software sampling but it is the same wave memory size the Emulator 1 used in 1981 at a cost of $10k.

The sample bank can have a name which is 9 alphabetical characters which is stored as two hex characters.

Use Cases The QD in the Roland S-10, MKS-100 and S-220 has only a few use cases;

• There is no formatting of the disk, as it has no sectors.
• Full Read – reads the complete sample bank of 64kb
• Monitor Read – reads just the sample bank name and structure
• Full Write – writes the complete sample bank of 64kb
• Verified Write – checks whether there is sample data or its an empty disk then writes the complete sample bank

Essentially the QD needs to do only 3 things; partial read, full read or full write. Partial read is controlled by the OS and is not a QD function.

QD Interface The interface to the QD consists of 10 lines:

1. /WRITE PROTECT – indicates whether the disk is write protected or not
2. /WRITE DATA – data is written here when WRITE GATE is high
3. WRITE GATE – enables recording
4. /MOTOR ON – Low sets the disk motor running
5. READ DATA – read data stream
6. /READY – reading or writing can take place
7. /MEDIA SW – indicates disk is inserted or not
8. /RESET set low for reset period after POWER ON RESET
9. +5V
10. GND

The read and write hand shaking is very basic and described in the Roland services notes. The QD uses standard MFM encoding (its a DD disk!) and has a clock rate of 101 kHz much lower than the DS/DD floppy drive clock of 250 kHz.

QD Read & Write Timing

Gotek Flash Floppy Solution Originally I was going to make my own ARM chip based emulation, but after seeing how the Gotek floppy emulator can be loaded with a completely new and flexible code set, I ditched my plans. There are three challenges:

• The clock rate is lower and not an integer, however the FF code can be configured to run at different clock speeds.
• The QD has no software sectors or notion of indexes and steps across tracks. Its just one track. This can be configured as simple 64kb FTE.
• The hand shaking and I/O pins are a bit different to a floppy. Once again this can be configurable in the code and mapped in cabling.

I/O Mapping Lets take a look of how the 10-pin QD connector maps to a standard Shugart floppy 34 pin connector.

A Solution Kief Fraser is the developer behind Flash Floppy and the wiki is here. In mid August a version of the code was announced that supports Quick Disk and I have tested this out on a Roland S-220. The Gotek drive casing is 2 mm wider than the QD drive and 2 mm smaller in height. I took 1 mm off each side of the drive using a craft knife and then designed a PCB that attaches to the base of the drive and to the QD metal mounting bracket, so the drive is at exactly the right place. The PCB is 150 x 120 mm and contain a JST 10 pin connector for the cable to the Roland main board, a 34-pin IDC connector to attach the drive to, along with 4-pin power.

The original 10 pin connector on the QD is an unusual JST like format but it is not the same and I have not found a modern equivalent. I have replaced this connector (at the end of the Roland cable) with a new 10-pin JST one, which mates with the JST socket on the PCB.

Roland S-50 Restoration

S-50 As Bought

Overview I bought a broken S-50 Sampler in July 2019 for £50, this was either going to be a bargain or a train wreck! The seller was local so I picked it up and traveled back by train over the Surrey Downs, reading the user and service manual along the way.

The external condition was good but rather dirty, and the seller had tried a new OS diskette but that had not solved the problem of a dead S-50, although he had re-flowed the FIP display solder joints. I was expecting lots of issues….

History The S-50 was announced at the Frankfurt Music show in 1986 and became available in early 1987 for £2195, and favorably reviewed in Sound on Sound. My example is an early model (serial number 790068) which dates the sampler as June 1987. At that time the S-50 was well outside my means, and I was buying cheap second hand analog synths like the ARP Odyssey and a Prophet 5 for £150. In 1989 I did manage to buy a 12-bit sampler with a loan, the Emax SE at £1200.

Initial Power On After checking the 15V and 5V voltage rails (which were accurate) I turned on the S-50 and got the usual “Please Insert System Disk” on the display, but although the diskette drive was being switched on it and rotating it would not load the very scrappy ancient OS diskette. I ordered a new OS diskette in case this one was faulty and tried some of the Roland library diskettes I have, but still no success in loading the operating system.

New Diskette Drive The next step was to replace the diskette drive, as the CPU was working and the CPU Board had power (Red LED lit at top right). I had a Sony MPF420-1 diskette drive that was bought for my Emax before I upgraded to a USB drive. The Sony drive works with Emu samplers as they have a very simple floppy interface design but the Roland S-50 needs a READY signal on pin 34 and a DISK CHANGE on Pin 2. On the Sony drive Pin 34 is DISK CHANGE and there is no jumper setting for assigning READY as PC’s didn’t need it.

After a few attempts, and lots of Internet browsing, I managed to kludge the PCB of the drive with two thin wires and PCB trace cuts. The drive has a 15 pin solder block where the READY and DISK CHANGE signals are located. Once back in the sampler and powered on, the OS diskette loaded up successfully and a sample bank was loaded into Preset 1 called BRASS. I eagerly played the keyboard, but no sound at all! I feared the digital board was dead…..

Analog Board & Relay

Audio Output Relay A common fault on all S-50’s is the relay across the audio outputs failing and ageing. This relay switches the audio outputs to ground at power on for a few seconds, so that audible clicks are avoided. Unfortunately the relay contacts wear out and cause audio distortion and with my S-50 very chopped up and distorted signals. After reloading the OS diskette and playing the keyboard for a few minutes, I eventually heard a very weak and distorted sound. The sampler works but definitely needs the relay replacing!

In the S550 Roland replaced the 6 pole relay with a simple transistor switch that does the same job during power on, making this a longer term solution. They also used this design on the S-750 and S-10 and its a pity they didn’t do this on the S-50! I was hoping to replicate the S-550 circuit back into the S-50 but unfortunately the relay is in a different position, after the output capacitor rather than before. So this idea was scratched.

Recapped Analog Board

I want a proven approach and buying a new relay is expensive. Fortunately Open Mirror in Australia have developed a replacement PCB that holds 3 cheap relays to replace the original, details are here.

I ordered the PCB from OSHPark and it arrived on 26 July, and the relays arrived on 30 July.  I bought 3 PCB’s at the same time, so I am building the spare two up and selling them to help other S-50 owners. The relays are easily obtainable and cheap, less than £2 each. By fitting DIL sockets they can be replaced in the future, and they are also quieter than the original at switch on

Front Panel Switches The switches on the sampler were extremely worn out, making it very hard to use the front panel, the buttons sank into the panel when pushed. The S-50 had clearly seen a lot of playing over the last 30 years. I bought a replacement set of 34 Alps switches with part number SKHHARA010 from Mouser. These were fitted along with replacing the electrolytic capacitors in the FIP drivers which live on the front panel board.

It looks like a previous owner had some of these capacitors  replaced, and not cleaned up the flux. So I replaced them all with new items and cleaned the board. I was hoping this recap might increase screen brightness, which was a little dimmer than a brand new display.

PSU Board

New Capacitors This Roland S-50 is a very early model from 1987, so the electrolytic capacitors have been installed for over 30 years and its time to replace them with new and improved versions. This is a big job as there are over 100 capacitors to replace.

I started by creating a spreadsheet of all the electrolytic capacitors and checking against my existing stock of high quality audio capacitors (Nichicon USW series) and power capacitors (Panasonic EEA-FC and Nichicon UFW), so it was a matter of ordering a few more different sizes when I ordered the new switches.

The large capacitors are secured to the board with clear jointing compound, a little messy but it does the job, it stops the capacitors moving when the S-50 is shipped or on tour. I am retaining the linear power supply and recapping, but I may design a Meanwell switched power supply to replace it. This would reduce the large amount of heat in the S-50 and provides a better long term solution.

PSU Board Recapped

Slider Refurb The S-50 has three sliders on the left hand side for Volume, Recording Level and Pitch Bend Range. They were all sticky and rusted up, so I cleaned then with Dexoit and then added fader lube, without stripping them down. This worked very well and they glide nice and smoothly.

Keyboard Testing the grimy keyboard was successful, except for one key not working. It needs stripping down and cleaning. The strip down is still to be done as at August 2019, as I am rather keen to play and sample with the S-50!

Final Power On  After completing the relay replacement, recapping the analog and PSU boards and replacing all the switches, I powered it on again on 03 August. It booted up and played perfectly, no audio distortion and a clear detailed sound thanks to the analog board recap. Even the VFD display was a bit brighter.

However the S-50 would only boot from a strange OS diskette that the owner had bought from Australia. It would not boot on diskettes I created or the OS diskette I had bought new. Turns out there are two jumper setting on the diskette drive that I needed to implement. From then on all the diskettes loaded fine.

Outcome The S-50 was then transferred from the workshop into the studio, placed on a keyboard stand and hooked up to a TV monitor via an HDMI converter. The sampler will be used to play some of the nicer factory diskettes but it will mainly used for sampling analog synths and getting the low-fi 12-bit sound. The S-50 cannot use a mouse, but does work with either the RC-100 or MT-100, neither of which I plan to add, so sampling is strictly via the S-50 control surface.

Roland S-10 OS Modification

Only half a display

Introduction I fitted a new OLED display to my Roland S-10 sampler in February 2019, but the new display only shows the first 8 characters, as the original display is configured as 2×8 lines when it actually shows 1×16 lines. The new display is not a drop in replacement. This problem also affects the Roland Alpha Juno 1 and 2 which uses the same display. The solution is to modify the operating system to change the assembler code that writes the characters onto the LCD.

Processor History Roland started off in the late 1970’s using the Intel 8048 processor with inbuilt mask ROM in the Jupiter 4, Pro-Mars, CSQ sequencers and the 184 keyboard. This had a tiny 1k byte of ROM space, so the programs had to be very small and tight! The Juno 6/60 used the 8049 which has double the ROM size, whilst the Jupiter 8 and MSQ-700 use the Z80 and external ROM chips. By the mid 80s’ Roland were using the Hitachi 630X processor family alongside the Intel 8051 family. The S-10/MKS-100/-S220 samplers all use the same Intel 8032 processor with 64k bytes of external ROM to hold the operating system. The UV EPROM chip used was the MBM27C512.

OS Change Process To correct the display readout I need to modify the code in the Roland S-10 Operating System which is written in 8051 assembler. Fortunately this has been done before to alter the MIDI notes off (in both the S-10 and Alpha Juno), so I am treading a proven path. The OS is extracted from the ROM into a binary file, then disassembled. I can then alter the display code to provide the corrected offset on the LCD, and assemble the code back into a binary. This can then be burnt into a new EPROM and put into the S-10 replacing the old OS chip. I am using a modern Micro Chip 27C512 chip with 70 ns read speed, and they cost £1.50 each.

Roland S10 OS Binary The latest OS version from Roland is 2.07 file, and this ROM has already been extracted and is available here. I used this version as it saved a step in the process and meant I could look at the disassembled code quicker, and see whether the project was possible.

OS 2.07 Features I suspect most S-10’s have the latest OS, but if you have an earlier version this is what you are missing:

• MIDI system exclusive capability.
• Fix to fast appregiations dropping notes.
• Fix for the GATE input silencing voices.

Disassembly I tried using a free Windows program from Spice Logic to disassemble the code, but this did not read the hex file successfully. I then used a Windows 10 cloud based disassembler and this gave me both a reference and disassembled file. Looking through the code I could see what looks like an LCD write subroutine at address 73C3 with two sets of mov instructions spaced 40h apart, which matches the spacing of the original LCD data structure. Here is the code:

Address_73C3:
mov a,#0x80
mov Register_68,a
mov P2,a
mov r0,#0x00
movx @r0,a
inc r0
clr c
lcall Address_3AE0
lcall Address_78ED
mov r0,#0x00
mov a,#0xC0
movx @r0,a
inc r0
lcall Address_3AE0
lcall Address_78ED

In theory all that needs to change is the second mov command to use an address of 0x88, so that all the 2nd line display commends have an offset of 8h in addressing rather than 40h. The code would be well written if all the LCD calls use this sub routine. Register 68 looks like the LCD display register. There are around 200  text messages in the S-10 operating system, including messages for the test routine, and an Easter Egg at address DB7E (SUZUKI GSX-R 400, HONDA INTEGRA GSI). This text is longer than 16 characters and was never displayed!

I have mapped the text messages in the code to the address lines, and then back to the LCD write sub routine at 73C3. I have not mapped them all back but I am sufficiently confident to change the code and burn a new EPROM along with a change in OS version to 2.09, as 2.08 is the unofficial modified version for correcting the all MIDI Notes Off.

Assemble and Burn The change is only two bytes in the whole code, so my initial approach is going to be to alter the binary code at the appropriate addresses rather than re-assembling the disassembled code, as this is more straight forward and the as51 assembler may not read the disassembled code correctly. The first stage is to buy some EPROM chips (Micro Chip 27C512) and burn the 2.08 code into one and check the S-10 works with no modifications by me.  I am using my trusty TOPS 2009 ROM burner, but there are more recent versions that do the same job.

The second stage is to modify the binary code in two places (OS version and LCD offset), then burn a new EPROM and see if it corrects the display writing on ALL 200 messages. If there are any exceptions I will go back to my message address map and see how the display is being called for that message that fails. I will post the results soon!