The question is how close is the sound output between the SC-8820 (with SC-55 map) and the SC-55 among a few DOS games. Are the above recordings a rare example where the sound output is somewhat different or is this common? It seems unknown whether they sound near identical 99% or merely 50% of the time.

Also, do the other software synthesizers perform as well as this VST plugin?

x86++ wrote:

The question is how close is the sound output between the SC-8820 (with SC-55 map) and the SC-55 among a few DOS games.

One could have asked that same question 18 years ago when the SC-8820 came in 1998.
I think many people were happy for many years without nit picking, but yes, a third party (community) emulated version of SC-55 would be ideal.
For the time being, nothing tops SC-VA besides owning the hardware.

I tried a few games and once you stop doing side by side comparisons, it sound, well, like Roland 🤣

Or if you think of it as a SoundFont from Roland I think it's very nice. But the price is quite steep.

x86++ wrote:

Are the above recordings a rare example where the sound output is somewhat different or is this common?

String Ensemble 1 is the basis for most orchestral-sounding music, which is a rather common application of General MIDI. That's why I consider the example that I posted to be non-trivial.

I don't quite get why Roland would even choose to focus on emulating the 8820. There are hundreds; if not thousands of compositions specifically for the SC-55, and far fewer for the later models.

And their target market is probably not DOS gamers.

kode54 wrote:

And their target market is probably not DOS gamers.

I agree... I think this software was mainly developped for the japanese market. From what I've seen online, it seems they have a bigger community interested in legacy midi equipment than anywhere else.

Here is a very detailed review of the Sound Canvas VST by a japanese website. Meanwhile the release of this VST has been largely ignored in western music-making websites.

http://info.shimamura.co.jp/digital/review/2015/12/73045

It seems they sometimes have midi shows where they showcase older synths and modules. Google "Think Midi 2015". How cool is that? You would probably never see this level of interest in Europe or in the U.S.

MMaximus wrote:

kode54 wrote:

And their target market is probably not DOS gamers.

I agree... I think this software was mainly developped for the japanese market. From what I've seen online, it seems they have a bigger community interested in legacy midi equipment than anywhere else.

Yep. There was a lot of music composed for the SC-88 and up in Japan, even for games on systems like the PC-98 and x68000. That's why I've got three Sound Canvas units: an SC-155 to cover old western dos stuff, and a M-GS64 and SC-8820 to cover later Japanese composed tracks, which are my primary interest to be honest. The M-GS64 is pretty much superfluous, but it comes in handy occasionally.

@NRS
The above insights are an interesting finding. It seems that Roland was appealing to a non-SC-55 market. In either case, thank you for providing details on a way to benchmark GM sound.

Particularly for GM in DOS games, an interesting question is whether the software synthesizers are reasonably competitive with this VST plugin option. This could be judged subjectively along with cost and cross-platform capability.

While the emulation seems decent, it's useful to know that Roland clearly mentions the following on their website:

"The Sound Canvas VA is not designed to play Standard MIDI Files. Depending on the data in your music files and how it’s processed with the host application, the mix balance, tone, and timing may vary slightly from the sound of the original hardware."

James-F wrote:

Midi IS Sound Canvas.

No. General MIDI is Sound Canvas. MIDI as in "pre-1992 MIDI" is MT-32 (as far as home computer music is concerned, obviously).

gdjacobs wrote:

Okay. Can you please tell me WHY it's better? What do you like about it compared to Bassmidi with a good sound font?

usually when Roland put effort into something they will do some small tweaks ..
it is definately the most new.. so anyone who thinks that newer = better would think it would be better.. 😀

James-F wrote:

Midi IS Sound Canvas.

No. General MIDI is Sound Canvas. MIDI as in "pre-1992 MIDI" is MT-32 (as far as home computer music is concerned, obviously).

IMHO General MIDI is the standard all of the cheapest and most basic sound cards had to comply, including the horrendous Roland-supplied Microsoft GS Wavetable Synth and Apple's QuickTime. Sound Canvas is actually the Roland GS, in which most MIDI supporting games were composed. And Yamaha had it's competing standard, the Yamaha XG. Remember these modules were not made for gamers. They were made for composers (including game composers) that could afford a 600 dollar peripheral in addition to their 2000 USD MMPC and their 500 USD MIDI controller. Today's Master Race is centered around multiple GPU's. Back in the day what separated the peasants from the elite was the sound card.

Okay, i found the file i was referring to before.

This is how the Doom intro should sound on a SC-55, notice the whispy, sweepy effect on the halo pads is very distinct and defined, and becomes a deliberate part of the tune.
https://www.youtube.com/watch?v=tykAn_7aRP4

Here is the doom intro recorded on a real SC 8850 (from this site - though the "play midi files for you" function doesn't appear to work any more for those interested):

I can't remember which is the sc-55 map and which is the 8850 map but they both sound very similar to me - i'm guessing the 8850 (and 8820) use the same halo pad patch in both maps?

And here is is on the SC VA with the SC 55 map:

And here are only the halo pads:

To me, these seem reasonably similar and the halo pad whispy, sweepy noise is way less pronounced and distinct than on a SC-55

Not sure how the 8850 compares to the 8820 that the SC VA vsti is trying to emulate.

According to wikipedia:

Roland ED SC-8850 1999 GM GS GM2 64 128 1640 63 24-bit @ 32kHz Desktop module that can connect via USB. Contains a new native map as well as SC-55, SC-88, and SC-88Pro maps for backwards compatibility, however compatibility with MIDI files made for these modules is poor

Roland ED SC-8820 1999 GM GS GM2 32 64 1608 63 24-bit @ 32kHz Mobile MIDI device with no display, successor of the SC-88ST Pro. USB-powered, but does not act as a full USB audio device. GS Advanced Editor or another program is highly recommended when using this synth. Despite similarities, patches and effects occasionally play differently from the 8850 - some samples on the 8850 are stereo while the 8820 versions are mono and vice versa.

Anyway, that's just a small example, see what you think.

Hmm, been playing around with my SC-155 and SC-8820. The string sounds in the SC-55 map were definitely changed somewhere after the 55 models, so the VSTi is probably quite accurate to the 8820 (from the samples, I suspect not entirely accurate, but I don't have it to test against). As was demonstrated, they're "cleaner" and not as deep sounding. I wonder why Roland did that; it must have been a conscious decision.

lavadrop wrote:

Sound Canvas is actually the Roland GS, in which most MIDI supporting games were composed. And Yamaha had it's competing standard, the Yamaha XG. Remember these modules were not made for gamers. They were made for composers (including game composers) that could afford a 600 dollar peripheral in addition to their 2000 USD MMPC and their 500 USD MIDI controller. Today's Master Race is centered around multiple GPU's. Back in the day what separated the peasants from the elite was the sound card.

Well, what DOS games actually uses GS or XG instead of the stock standard 128 GM instruments? Can't be many.

PhilsComputerLab wrote:

lavadrop wrote:

Sound Canvas is actually the Roland GS, in which most MIDI supporting games were composed. And Yamaha had it's competing standard, the Yamaha XG. Remember these modules were not made for gamers. They were made for composers (including game composers) that could afford a 600 dollar peripheral in addition to their 2000 USD MMPC and their 500 USD MIDI controller. Today's Master Race is centered around multiple GPU's. Back in the day what separated the peasants from the elite was the sound card.

Well, what DOS games actually uses GS or XG instead of the stock standard 128 GM instruments? Can't be many.

This is one of those things DOSBox could have helped out with greatly, still could, via logging, like how firefox gives the option help bug fix with reporting.

GS is not just additional instruments that are selectable via Controller 0. GS also provides for different drum kits (GM only defines the Standard drum kit), in particular, the "Orchestral" drum kit used by many games. GS furthermore defines (although it may not have been the first one to do so) controller 91 as reverb and controller 93 as chorus, both of which are used by many games. GM on the other hand defines no controllers other than modulation, volume, pan, expression, sustain pedal, RPN, all notes off and reset controllers.

The GM standard was kind of vague in concept. The Sound Canvas was the pioneering GM device and other hardware that followed was actually designed to "emulate" it rather than the standard. Games used and I know at least Yamaha supported extended GS features.

General MIDI standard is not vague as described here: https://en.wikipedia.org/wiki/General_MIDI.

Also, Sound Canvas is not a standard and they produced many models with and without proprietary extensions, such as GS: https://en.wikipedia.org/wiki/Roland_Sound_Canvas.

How rigorous was the testing to label something as GM compliant? You can write all the paper in the world, but if there's no method of evaluation (even self evaluation), it's useless.

gdjacobs wrote:

How rigorous was the testing to label something as GM compliant? You can write all the paper in the world, but if there's no method of evaluation (even self evaluation), it's useless.

I don't know how rigorously it was tested, but the spec wasn't terribly hard to meet. The problem, in terms of compatibility, was mainly that the specification gave no detail whatsoever about the characteristics of each GM patch. Attack/release times, relative volumes, etc, did not have to follow any rules. This meant that a GM track could sound drastically different if it was played on a device other than the one it was composed on, even though both devices could be 100% GM compliant.