Can I open the descrambled file in an audio editor?

[weavermedia]

Hey, very nice code to descramble the BIN file! 👏

After descrambling I tried opening the file in ocenaudio to check out the audio contents. It asked for the file data format for decoding. I noticed the SR-JV80 boards seem to be 8-bit and I guessed a sample rate of 32kHz.

In the opened file in ocenaudio I can mostly hear the audio contents but they're quite distorted. In the House board descramble, for instance, I can hear vocals, drums, keyboard loops, etc. but everything is heavily distorted.

Do you have any recommendation for opening the descrambled files to examine the audio contents?

[pekd]

The data is compressed with a DPCM codec, but there is no public documentation of it for now. Essentially there is a block of "exponents" somewhere at the beginning and then there is 8bit data that has to be shifted with the "exponents". If you directly listen to the 8bit data, you're missing the necessary shifting and accumulation operation and everything will sound heavily distorted and high pass filtered.

If you want to play around with it and be the 4th person to re-do what 3 others did in the past already (although without publicly documenting it), this decodes the data once you found the exponents (called coefs in the code) and the compressed sample data (called deltas in the code):

public static int[] decode(byte[] coefs, byte[] deltas) {
	int[] result = new int[coefs.length * 2 * 16];

	int value = 0;

	for(int frame = 0; frame < 2 * coefs.length; frame++) {
		int coef = coefs[frame / 2];

		int exp;
		if((frame & 1) == 0) {
			exp = coef & 0x0F;
		} else {
			exp = (coef >> 4) & 0x0F;
		}

		for(int i = 0; i < 16; i++) {
			int offset = frame * 16 + i;
			int sample = deltas[offset];

			int delta = sample << exp;
			value += delta;

			result[offset] = value << 14; // padded to 32bit
		}
	}

	return result;
}

You will end up with a series of concatenated samples, which will result in DC offset if a sample didn't end at exactly 0. You would then have to decode the sample metadata to avoid this problem by decoding all samples separately.

Keep in mind that I do have much more information about this, including knowledge about the sample metadata as well as a tool which can encode samples to this codec most of the time, but I simply didn't have time to make a polished implementation and upload it to GitHub so far. I might finish this eventually though.

[weavermedia]

@pekd Thanks for the info! So I assume there are no public wave ROM dumps available, similar to the Yamaha ROM dumps on the Sector 101 Soundcloud page?

[pekd]

Not that I'm aware of. The full set of ROM dumps (SR-JV80, SRX, as well as the extended SRX-like "SuperNATURAL" ROMs in the INTEGRA-7) exists of course, but it's not public for now.

If you have an INTEGRA-7: you can extract the SRX ROMs from the internal SD card, but this is a bit of work. IIRC they are descrambled already.

If you have the RolandCloud SRX plugins: you can extract mangled and already descrambled versions of the SRX ROMs from the plugins, but they are different than the ones you would get from the physical boards and you can't write them to a custom replacement board and use them in a real SRX compatible module either. But if you're only interested in extracting the samples, then they work just fine for that specific purpose. Similarly, you can patch in your own edited samples or ones extracted from other ROMs to make them available in these VST plugins. This was done in the past already, specifically to "port" missing SR-JV80 sounds to the plugins.

[FlamemusiK2K]

Hey there, I find all of this quite fascinating, and I've spent quite a lot of time on this as well, more specifically on the ability to insert custom samples. I don't know if anyone is still on this topic, but for the last 2-3 months or so I've been trying to import custom samples into the SRX VSTs without success (using code). Is there anyone still passionate about this that could explain how the structure works/how to import samples into a ROM and make it work on the VST? Or show how to achieve this with a full, complete script? I do get/understand the picture, but not fully. I know about the nature of the DPCM encoding, what it looks and sounds like, when you import the raw .bin (ROM) file into Audacity, the right sample rate needed (32 000 Hz), the right bit depth, and the endianness required. Also about the ROM internal structure, how there are different types of ROMs (8 MB, 32 MB, 64 MB, etc.), how the samples are placed, the samples' metadata, all the other essential infos and stuff, etc... but to for anything to work you need a solid script. also, I am just trying to import custom samples for sound designing purposes. Can anyone help out ?

[SynthfulDwarphus]

The new-material SRX expansions (I believe SRX-01/02/03/04/05/10/11/12) should be all, or primarily, 44.1kHz, while the SR-JV compilation SRX expansions (SRX-06/07/08/09) should be all, or primarily, 32kHz. (There's an outside chance it's actually 31.25 kHz). Also, all SRX expansions are 32MB. All XV-series (including Fantom FA76 keyboard) and Integra7 PCMS engine use identical 32MB internal PCM-ROM. That 32MB should be approximately 8MB 32kHz (or 31.25kHz) of SuperJV/JV80 internal PCM in its entirety, approximately 3MB 44.1kHz of JD-800 PCM (with possibly a bit of JD-990 mixed in), and the remainder presumably new-material at 44.1kHz. All SR-JV expansions are 8MB. All SuperJV-series (including XP keyboards) have identical 8MB 32kHz (or 31.25kHz) internal PCM-ROM. There are no 64MB (or 16MB) packages (other than Roland marketing lies).

[FlamemusiK2K]

hi, yeah, I do know the difference between the SRX expansions and SR-JV expansions. The SRX expansions are much larger than the SR-JVs. The SR-JV expansions are much more smaller because they were made for the JV/XP series (JV-1080, XP-80 etc..) and because they are older. So, they are 8MB ROMs, you could stack up to 4 expansions on the JV 1080. But, on the other hand the SRX expansions are 32 MB and were made for the XV series/ up and when Roland is talking about 64 MB ROMs and stuff, they are talking about stacking up the ROMs. You could see that Roland augmented the full capacity of their ROMplers overtime.

But, I'm more interested about the technical aspect and code. I know the SRX VST are based on the XV series ROMs. Much more specifically on the XV 3080 model. If you dive deep, you can see that in the header and even the patches .So logically no other types of ROMs would work on the VST unless they are at base from an XV model ROM or follow a similar structure as an XV series ROM. Which size is at ~32 MB. These ROMs have a specific structure to them if you don't follow it exactly nothing will work (at least for the VSTs).

there is a lot of layers to this which is very interesting.

[FlamemusiK2K]

check these images out. There is an Internal ROM expansion (coming from one of the SRX VSTs0 , an SRX expansion and an SR-JV expansion.

[SynthfulDwarphus]

The VST is newer software and something else altogether and likely has many differences internally from the 25-/30-year-old original XV & SuperJV hardware synths.

I understand the virtual "XV5080" and ""JV1080" emulation VST replicates the synth engine from Fantom-S and/or -X (some extra features & parameters beyond the XV), and the effects engine is close to or same as Fantom-G (1 insert fx max tied to each Patch and 16 available simultaneously in Performance/Studio Set/whatever it's called now, with no serial chaining available). So it's really a "Frankenstein" emulation, and the XV-3080 hardware is no different than rest of XV-series, using the identical 32MB internal PCM-ROM (1.083 waveforms), although different XV models (& Fantom FA76) just use different integrated tone generator/effects chips, with different amounts of effects RAM connected for different effects capability (more blocks and algorithms available at once), and then have either one or two of the chips for more or less polyphony.

This Github project is to descramble the original SRJV-/SRX-series PCM-ROM ICs so they can be read out properly. Once read, the data still has Roland proprietary FCE-DPCM 2:1 data compression to contend with (which some individuals have apparently already deciphered).

You can tinker with the VST PCM files, and do resarch online to see about decoding them, since they're likely in a different format and possibly using different encoding and compression (if any).