daX6fantom wrote:
Just wondering, has there been any update to these external midi delay sync problems??
I just noticed these problems on my GT, and after much mucking around, have discovered that despite what some people have said, even using standard midi with clock sync, it still freaks out and delays sound horrible.
Anyone know a workaround, or have access to the super-secret beta GT operating system that fixes these problems that has been rumoured?
An analysis of the V-Synth GT architecture (including the SH-4A and SH-2A cores) in the context of Roland synthesizers reveals that the "MIDI jitter" or desynchronization issue in later models (V-Synth GT) compared to earlier models (Original 2003, XT) stems from changes in interrupt processing priorities in the TAU (Timer Array Unit) and the operating system.
Q.: Why did the bug appear in the GT (2007) when it did not appear in the Original/XT (2003-2005)?
A.1.:
Complexity of the interrupt hierarchy (Interrupt Latency):
Earlier V-Synths (Original/XT) used previous-generation chips with a lighter bus load. In the V-Synth GT, the SSI Serial Sound Interface (R8A02021) & WSP chips simultaneously serves two sound engines (AP Synthesis and Vocal Designer). When the TAU timers generate an event to send/receive a MIDI packet, the main processor may be busy with high-priority tasks like maintaining the audio buffer or calculating sound modeling parameters. This creates micro-delays (jitter).
A.2.:
32-bit timer cascading:
Accurate MIDI synchronization requires a stable "tick." In the SH-2A architecture, using timers in cascaded 32-bit mode for long intervals sometimes conflicts with the DMA (Direct Memory Access) mechanism. If DMA blocks the system bus to transfer samples from P2ROM memory, the TAU timer must wait for the bus to become available to update registers or trigger an interrupt, which delays the MIDI event by several milliseconds.
A.3.:
Task sharing between cores:
In the original V-Synth from 2003, the MIDI stack ran on a dedicated auxiliary controller or had hard-coded time slices. The GT version based on couple R8A77800N & R8A02021A saw a significant integration of functions into a single SoC. All control functions (interface, MIDI, USB, engine control) were relegated to a single core SH-4A (R8A77800A). This resulted in a fluctuating MIDI data stream during intensive use of controllers (such as the D-Beam or Time Trip Pad).
A.4.:
Differences in the implementation of timer blocks:
-
V-Synth Original/XT: Used a simpler, yet predictable, MIDI clocking scheme. The software code was optimized for a smaller workload, ensuring an immediate response to timer interrupts.
-
V-Synth GT : Increased computational complexity resulted in MIDI event processing becoming dependent on the current CPU load. The bug is not actually a hardware issue with the TAU timer, but rather a problem with the scheduler, which prevents the MIDI timer from prioritizing synthesis calculations.
Bottom Line: The desync is the result of a bottleneck in the SH-4A & SH-2A chip's attempt to simultaneously manage complex graphics, two synthesis engines, and a MIDI stream through a single interrupt system.
...
[attachment: Roland WSP][attachment: IMG_0455.JPG][attachment: IMG_0451.JPG]
vsynthesis wrote:
yeah its not happening on all machines but seriously - why dont you set the gts tempo internally to the same as the sequencer - i am not saying you should have to do that and its f...lame that roland still cant sort out such pathetically simple things - when was midi clock invented ? didnt they test this - the jp8000 i have does this.........so yeah i agree roland are a bunch of w.......rs when it comes to design testing and using buyers as testers but seriously when all is said and done if you have a simple option to get it working ' ok ' then go for that - set the tempo internal to same tempo as the song and turn off all midi clock sync . . .
To minimize MIDI jitter on the SH-4A chip (as well as the SSC, which is equipped with the complementary Renesas R8A02021ABG SH-2A core) on the V-Synth GT, You should be utilizes methods aimed at reducing system bus load and modifying the TAU timer interrupt handling logic.
1. Software Optimizations (Scheduler & Priorities)
The main problem is Interrupt Latency. When the TAU timer is ready to emit a MIDI event, the processor is busy calculating the vocal engine.
- Interrupt Prioritization (IPR): In the SH-2A architecture, the Interrupt Priority Registers (IPRs) allow you to set the priority for each TAU channel. This optimization involves setting the maximum priority (level 15) for exsample:
UART/MIDI-related interrupts, higher than that of the DMA controller handling the screen graphics or enothe one processing of interupts.
- Disabling Background Tasks: In the GT system settings, it is recommended to disable USB-MIDI when using standard DIN connectors. This frees the processor from handling the USB stack, which has high overhead and often interrupts TAU timer interrupts.
- MIDI Clock Filter: If the GT is not the Master device, disabling External Clock reception (unless critical) significantly reduces the load on the timers, as the processor does not need to constantly adjust the phase to the incoming stream.
2. Hardware Hacks and Modifications
Hardware issues with the V-Synth GT are often related to the power supply and thermal conditions of the R8A02021 chip, which indirectly affects clock stability.
- Core Voltage Stabilization:
The GT's processor power lines are heavily loaded. Installing additional low-ESR tantalum capacitors in the BGA361 chip helps smooth out voltage drops during sudden changes in computational load (for example, during deep patch modulation). This reduces the likelihood of synchronization errors due to noise in the clock circuit.
- Cooling (Thermal Mod):
The R8A02021 processor in a 27x27 mm package gets noticeably hot when running two synthesis engines. Overheating leads to timing degradation within the crystal. Enthusiasts often install a copper heatsink on the chip (via a thermal pad, like analogue CPU heatsinks), which stabilizes the system bus.
- MIDI Circuit Isolation:
Some tinkerers
replace standard optocoupler at the MIDI input with faster ones (such as the 6N137 series) to minimize the delay of signal edges before they reach the controller.
3. Lite Mode for Jitter Correction
There's an unspoken "hack" for using patches: if jitter becomes unbearable, it's recommended to switch the Vocal Designer engine to bypass mode (unless it's used in a patch). This frees up significant resources in the R8A02021 chip, effectively returning MIDI stability to the level of the old V-Synth/XT.
Summary: The main "cure" is to reduce the CPU load and protect the chip's power supply.
Attachments
V_SYNT_GT.zip
(4.8 MB, 339 downloads)
· file not available interactive circuit diagram (moded)
Снимок экрана 2026-02-23 в 10.53.44.png
(833 KB, 10216 downloads)
· file not available Roland WSP
IMG_0451.JPG
(3.9 MB, 10211 downloads)
· file not available main board (party)
IMG_0455.JPG
(1.4 MB, 10203 downloads)
· file not available Serial Sound Controller (SSC)
Last edited by Vladistone on 24 Feb 2026, 09:08, edited 34 times in total.
