FMsynth

[tolkien]

Fantastic work adrdesign. Eres una máquina tío!
I will play with them this weekend.
Thanks so much mate.

[commodorejohn]

Originally Posted by emufan

@commodorejohn: i still cannot find them, but i remember, there were some 4143 (+-) byte files.
but maybe 3 out of 100 i found. so we can skip those, i think.

Yeah, I've run into those. Usually if you trim the fat with a hex editor (look for the SysEx start string - F0 43 - and trim everything before that; extra data on the end isn't so important, but the ending byte is always F7) you can get them working again, though sometimes they'll have removed the SysEx header/footer and you'd need to copy them from another bank file.

[daxb]

Originally Posted by adrdesign

The patches are in .voice format, which generates 8363hz, 16384 bytes long samples. Note I´ve selected 16384 by default because it is long-short enough to have an idea how it will sound and render at a reasonable time. You can change manually the length if you need.

I`ve no knowledge or experiences with synthesizer and such banks or how the dx7 sounds (I only know the mc-303). So, excuse me for my question. If I render a voice in FMsynth does it sound equal to the sound of the dx7 and how much it differs? Can we have similar results like the dx7? When it comes to rendered samples from the voices. I guess it is only interesting for programs the make use of the 5 octaves sample. For trackers (e.g. Protracker) only the 1 octave sample is interesting. The other octaves seems to be just doubled or halved (not real octaves) what can be done in the tracker. So what comes usefull when we want 8svx samples and do batch processing. Problem seems to be which values to choose for length, frequency, ...? Manually, you can do what you want of course but then there exists more powerful softsynth tools. You may explain a bit.

[Amiga1992]

I am sure it will not sound like a DX7, because it's an 8-bit sample at low sample rate. There will be a lot of aliasing and stuff.

[adrdesign]

Originally Posted by daxb

I`ve no knowledge or experiences with synthesizer and such banks or how the dx7 sounds (I only know the mc-303). So, excuse me for my question. If I render a voice in FMsynth does it sound equal to the sound of the dx7 and how much it differs? Can we have similar results like the dx7? When it comes to rendered samples from the voices. I guess it is only interesting for programs the make use of the 5 octaves sample. For trackers (e.g. Protracker) only the 1 octave sample is interesting. The other octaves seems to be just doubled or halved (not real octaves) what can be done in the tracker. So what comes usefull when we want 8svx samples and do batch processing. Problem seems to be which values to choose for length, frequency, ...? Manually, you can do what you want of course but then there exists more powerful softsynth tools. You may explain a bit.

I agree with you with no need to use multisamples. The proper length usually takes more than 32768 bytes, here the 16384 value is intended for preview purposes, based on a plain a600 speed.

I think you have to balance between quality/needed pitch/length. So, I choose 8363 based on Fairlight CMI library (also has kinda dx7 softsamples in the DX disks), because it retains enough quality and allows to use higher octaves. Moreover, I think crispy render rate adds some "analog" character the FM synthesis doesn´t have.

so no standard length

Synthesizers usually not only have the synth engine, after generating the sound, the DSP adds some chorus, reverb and so, thats why the Sound Blaster Fm Synth didnt sound like the DX7. Amiga doesn´t have any DSP, so it WONT sound like the DX7 (at least with this software).

FMConvert needs two values to render a sound: one for the length, the other for VELOCITY.

When creating a whatever-synth patch, Velocity is usually linked not only to VOLUME but also things like CUTOFF, RESONANCE, LFO (or all at once!)

So, when you play the same .syx in a real DX7 (i had some in my hands) you experience richer sound. For example, an Electric Piano (whitney houston type) should play dark/bright when pressing the key soften/harder (thats velocity).

The .voices I´ve prepared renders a note are like they were played at VELOCITY=100. If you need different velocities/lengths you could try batch convert

FMConvert <bank.syx> -l(length) -v(velocity 0-127)

Sorry for my english and hope u understand

[adrdesign]

@akira: I would convert these to wav, but the script Daxb did for rexx only converts one folder each time (there are more than 250 folders).

BTW: if you want it for testing, from cli:

>rexxmast
>rx MsynthVoiceTo8svx.rexx

it will ask for the folder where you .voices are. Samples are rendered in RAM: by default

[daxb]

@adrdesign:
Thanks for explanation.

I didn`t know that you have to process so many drawers. But makes it really sense to batch convert all of them to 8svx? If the voices are already different to dx7. For individuell use yes, but as a standard sample bank for the public? However, you can temporary copy the files to convert to one drawer. Or if you really need I can change the script to your needs.

[adrdesign]

Originally Posted by daxb

@adrdesign:
Thanks for explanation.

I didn`t know that you have to process so many drawers. But makes it really sense to batch convert all of them to 8svx? If the voices are already different to dx7. For individuell use yes, but as a standard sample bank for the public? However, you can temporary copy the files to convert to one drawer. Or if you really need I can change the script to your needs.

I agree again with you. I wont release the library in 8svx because it would be a hard work to prepare each patch with its proper length and loop it. People would forget the chord option wich is handful in 4ch tracking.

BUT: There are some VST compatible with DX7 .syx (such FM7/8). There are LOTS of .syx and there isnt already a way to preview the patches they have in a confortable way. THERE ISNT.

So, If I knew how to program, my ultimate tool would:

1) Scan for all syx in a folder. Detect if they are 4096-long. if not, trim them for good.

2) Ask for length (<65534) and velocity (0-127), defaults would be 16384, 100.

2) Convert each .syx with "FMConvert <file.syx> -l<length> -v<velocity>" and put the .voices in a folder with .syx name.

3) Finally convert each .voice to .8svx (1 octave) saving in the same folder the .voice is.

I imagine a copy of such bach conversion (in .wav format) in ableton for previewing all the syx we have, even in 8bit, they would be enough to have an idea how the patch sounds in FMx.

Really, I think musician community would appreciate it A LOT.

So, we´ll have enough if I we could add recursive scanning to the tool you did.

[daxb]

Originally Posted by adrdesign

1) Scan for all syx in a folder. Detect if they are 4096-long. if not, trim them for good.

2) Ask for length (<65534) and velocity (0-127), defaults would be 16384, 100.

2) Convert each .syx with "FMConvert <file.syx> -l<length> -v<velocity>" and put the .voices in a folder with .syx name.

3) Finally convert each .voice to .8svx (1 octave) saving in the same folder the .voice is.

That is doable and more or less easy but first, I would ask if that can be done on a system (pc?) more efficient and better (quality)? At least if the target is a sample format (8svx, wav, ...). Btw. you have to decide what should be the target of the voices and/or samples. Just a conversion as accurate as possible ("no" filesize or quality limit. 254000 Bytes if FMsynth is used). Should the samples Protracker compatible (64k sample size limit). Something between that or another target?

1)
I fear we need file information (SysEx header) of .syx files to process them proper. If I understand it right the filesize isn`t always the same. Further, will truncate 6 bytes from leading and 2 from trailing always work?!

FMconvert docs say:
"Acceptable are 32 and 33 voice banks without SysEx header and checksum (size 4096 or 4224 Bytes)."

2)
Easy. Do you want a requester asking that or just in command line? Process on all sys files or ask each sys for new values?

What do you mean with "put the .voices in a folder with .syx name."? As default FMconvert saves the .voices in current directory. Why a drawer with ".syx" as name? Or do you mean that the bank name should be used as drawer name? E.g. bank "rom1a.syx" ---> drawer "rom1a".

3)
I guess you mean the lowest octave, right?

[adrdesign]

Thanks for your interest!

I think we can stay in AmigaOS, because we must use FMSynth for the rendering. I didn´t know if you can call a rexx command inside an amigados script inside WinUAE from a Windows batch (!)

Selecting octave (1-5) output would be the same to selecting sample rate. If you add that option you can somehow get different qualitys.

1)
Me too. I havent found any >4104 .syx. If someone puts one, or better two here, I could play a little with.

If not, just ignore them.

2)
For the length: button requester with some values (like 16384, 32768, 65534, 131072, 254000, custom). Custom would open a text field asking for an integer <254000.
A text field requester to enter velocity (1-127)

It would process all .syx files at once, please! Entering values for each .voice file would be a pain in the ass! BTW: a different script with the option to take a single .voice and render all its velocities with a common custom length would be great for Electric Pianos.

-About the drawer´s name (sorry for my english):
processing "rom1a.syx" will create drawer "rom1a/" and will move all the .voices in (by default fmconvert puts all the .voices in the same drawer the .syx was, not in the current path).

3)
If you dont add the 1-5 octave requester, the lower octave then.

Sorry if Im abusing asking too much. I warn you It would be enough if the script you did could read recursive folders. Anyway, thanks in advance !

EDIT: NOTE if we release this in any form, ALL PEOPLE INVOLVED WILL BE CREDITED, but,I think this library would be named YamahaDX7_4_Amiga_by_DAXB_and_friends for the better

[commodorejohn]

Originally Posted by adrdesign

Synthesizers usually not only have the synth engine, after generating the sound, the DSP adds some chorus, reverb and so, thats why the Sound Blaster Fm Synth didnt sound like the DX7. Amiga doesn´t have any DSP, so it WONT sound like the DX7 (at least with this software).

The original DX7 doesn't have any effects processing; they didn't add that until the DX7-II, which had a different sound anyway due to using a cleaner 16-bit DAC (compared to the original's 14-bit DAC, which is noisier but has a pleasantly organic sound for a digital synthesizer.) The Amiga is still going to sound different than a DX7, but not because of a lack of DSP effects.

[adrdesign]

Originally Posted by commodorejohn

The original DX7 doesn't have any effects processing; they didn't add that until the DX7-II, which had a different sound anyway due to using a cleaner 16-bit DAC (compared to the original's 14-bit DAC, which is noisier but has a pleasantly organic sound for a digital synthesizer.) The Amiga is still going to sound different than a DX7, but not because of a lack of DSP effects.

Thanks for your explanation. I thought it had a DSP from 1st version. So, I´ve heard many FM synthesizers, could you tell how the DX7 synthesis sounded alike?

-Sega Megadrive/Genesis (warm)
-SoundBlaster 16 (digital)
-MAME (Yamaha coin-ops: OutRun, Enduro Racer alike, with darker and rustier sound)
...

[commodorejohn]

The basic sound of one operator modulating another at a given pitch ratio and modulation level is pretty much indistinguishable across the whole Yamaha FM line.

The main thing that separates the DX7 from the single-chip FM sound generators used in consoles, arcade systems, and lesser Yamaha synths is that it has six operators per voice (versus four in most of the single-chip solutions, and two in the OPL line) with more routing "algorithms" to combine them, much finer control of operator frequencies (which makes it easier to create unusual timbres,) and much better modulation options (a completely programmable five-stage envelope versus a simpler ADSR or ADSDR design, and LFO control of individual operator levels.)

(On the other hand, some of the other FM chips, in particular the OPL line used in the Sound Blaster and clones, had multiple waveforms for the operators, while the DX7 has only sine, so they can create some complex and unique timbres of their own despite being much, much simpler in overall design.)

So they're very similar in overall character, but very different in the details of their capabilities. Which is why everyone should buy several and love them all very much

[commodorejohn]

P.S. if you want some demos of the DX7's sound specifically...

DX7:
[ Show youtube player ]
[ Show youtube player ]

DX7-II:
[ Show youtube player ]
[ Show youtube player ]
[ Show youtube player ]

DX7s (DX7-II sound generator minus the DSP and split/layer capability):
[ Show youtube player ]
[ Show youtube player ]

[commodorejohn]

Addendum: well heck, I was wrong! Even the DX7-II didn't have DSP effects! (I thought it did, but it only has some basic performance/pan effects.) Which means that it wasn't until the SY77 (an expanded six-operator architecture with sampled sounds) and the V50 (a double TX81Z in a keyboard) that Yamaha added anything more complex than a basic chorus effect to their FM synthesizers.

[daxb]

The raw stuff is ready and working. Needs some testing and maybe some finetuning. Let me know. On top of the script there is a configuration section. You need to change the path of FMsynth and FMconvert that fits on your system.

Following external common requirements are needed:

RequestString V1.2 by Thomas Rapp (included in archive but can get from thomas web page too. Need to be in Sys:C directory or change path in script))
FMsynth V3.7
FMconvert V1.2
rexxsupport.library (comes with OS))
rexxarplib.library V3.8 (get it from Aminet if you don`t have it)
reqtools.library (get it from Aminet if you don`t have it

All files with ".syx" suffix and valid "F04300" header will overwritten and converted to voice files. Then all voice files will converted to 8svx. If started a second time voice files are deleted first and recreated. The 8svx files will overwritten if they exists alreday.

ps: I guess the 4224 size syx is broken or just unknown format.

Originally Posted by adrdesign

EDIT: NOTE if we release this in any form, ALL PEOPLE INVOLVED WILL BE CREDITED, but,I think this library would be named YamahaDX7_4_Amiga_by_DAXB_and_friends for the better

Please don`t add my nick name if you make the library.

BTW: a different script with the option to take a single .voice and render all its
velocities with a common custom length would be great for Electric Pianos.

Sorry, I don`t understand. A voice file has a velocity (or one for each instrument?) value that is used by FMconvert. Do you want a script that use FMconvert with all velocities or a script that change the velocity value in a voice file?

[adrdesign]

Originally Posted by daxb

The raw stuff is ready and working. Needs some testing and maybe some finetuning. Let me know. On top of the script there is a configuration section. You need to change the path of FMsynth and FMconvert that fits on your system.

Following external common requirements are needed:

RequestString V1.2 by Thomas Rapp (included in archive but can get from thomas web page too. Need to be in Sys:C directory or change path in script))
FMsynth V3.7
FMconvert V1.2
rexxsupport.library (comes with OS))
rexxarplib.library V3.8 (get it from Aminet if you don`t have it)
reqtools.library (get it from Aminet if you don`t have it

All files with ".syx" suffix and valid "F04300" header will overwritten and converted to voice files. Then all voice files will converted to 8svx. If started a second time voice files are deleted first and recreated. The 8svx files will overwritten if they exists alreday.

ps: I guess the 4224 size syx is broken or just unknown format.


Please don`t add my nick name if you make the library.


Sorry, I don`t understand. A voice file has a velocity (or one for each instrument?) value that is used by FMconvert. Do you want a script that use FMconvert with all velocities or a script that change the velocity value in a voice file?

Im testing it ATM. I´ve notice you also will need WBRun (and parm.library included), RequestChoice and RequestFile.

About multiple velocity, just ignore it. It would be a mess converting 127 times the same syx just for a single patch. Im happy with your work. SO PRO!

I´ll tell later.

EDIT: Tried a folder with 3 syx. It stop after opening FMSynth to convert the voices. And gives me an End

[daxb]

Oh, I forgot about WBRun. If you don`t want to use it then just start FMsynth manually before the script. I could use OS command Run but then the tooltypes are ignored. RequestChoice and RequestFile are part of the OS, so I didn`t mentioned it.

About your error. Example output looks like this: Do you get the same or different output? Are the voice files created with the new drawer? E.g.: "Ram:4104/ROM1A/BASS_1.VOICE". Maybe the header/footer of .syx file isn`t correct removed?

Ram:4104/ROM1A.SYX ---> Header and footer removed!

ROM1A.SYX ---> Converted to voice(s)!

Now calculate and save 8svx:
Process on: Ram:4104/ROM1A/BRASS_1.8svx
Process on: Ram:4104/ROM1A/BRASS_2.8svx
Process on: Ram:4104/ROM1A/BRASS_3.8svx
Process on: Ram:4104/ROM1A/STRINGS_1.8svx
Process on: Ram:4104/ROM1A/STRINGS_2.8svx
Process on: Ram:4104/ROM1A/STRINGS_3.8svx
Process on: Ram:4104/ROM1A/ORCHESTRA.8svx
Process on: Ram:4104/ROM1A/PIANO_1.8svx
Process on: Ram:4104/ROM1A/PIANO_2.8svx
Process on: Ram:4104/ROM1A/PIANO_3.8svx
Process on: Ram:4104/ROM1A/E.PIANO_1.8svx
Process on: Ram:4104/ROM1A/GUITAR_1.8svx
Process on: Ram:4104/ROM1A/GUITAR_2.8svx
Process on: Ram:4104/ROM1A/SYN-LEAD_1.8svx
Process on: Ram:4104/ROM1A/BASS_1.8svx
Process on: Ram:4104/ROM1A/BASS_2.8svx
Process on: Ram:4104/ROM1A/E.ORGAN_1.8svx
Process on: Ram:4104/ROM1A/PIPES_1.8svx
Process on: Ram:4104/ROM1A/HARPSICH_1.8svx
Process on: Ram:4104/ROM1A/CLAV_1.8svx
Process on: Ram:4104/ROM1A/VIBE_1.8svx
Process on: Ram:4104/ROM1A/MARIMBA.8svx
Process on: Ram:4104/ROM1A/KOTO.8svx
Process on: Ram:4104/ROM1A/FLUTE_1.8svx
Process on: Ram:4104/ROM1A/ORCH-CHIME.8svx
Process on: Ram:4104/ROM1A/TUB_BELLS.8svx
Process on: Ram:4104/ROM1A/STEEL_DRUM.8svx
Process on: Ram:4104/ROM1A/TIMPANI.8svx
Process on: Ram:4104/ROM1A/REFS_WHISL.8svx
Process on: Ram:4104/ROM1A/VOICE_1.8svx
Process on: Ram:4104/ROM1A/TRAIN.8svx
Process on: Ram:4104/ROM1A/TAKE_OFF.8svx
This is the END!

[adrdesign]

Originally Posted by daxb

Oh, I forgot about WBRun. If you don`t want to use it then just start FMsynth manually before the script. I could use OS command Run but then the tooltypes are ignored. RequestChoice and RequestFile are part of the OS, so I didn`t mentioned it.

About your error. Example output looks like this: Do you get the same or different output? Are the voice files created with the new drawer? E.g.: "Ram:4104/ROM1A/BASS_1.VOICE". Maybe the header/footer of .syx file isn`t correct removed?

Ram:4104/ROM1A.SYX ---> Header and footer removed!

ROM1A.SYX ---> Converted to voice(s)!

Now calculate and save 8svx:
Process on: Ram:4104/ROM1A/BRASS_1.8svx
Process on: Ram:4104/ROM1A/BRASS_2.8svx
Process on: Ram:4104/ROM1A/BRASS_3.8svx
Process on: Ram:4104/ROM1A/STRINGS_1.8svx
Process on: Ram:4104/ROM1A/STRINGS_2.8svx
Process on: Ram:4104/ROM1A/STRINGS_3.8svx
Process on: Ram:4104/ROM1A/ORCHESTRA.8svx
Process on: Ram:4104/ROM1A/PIANO_1.8svx
Process on: Ram:4104/ROM1A/PIANO_2.8svx
Process on: Ram:4104/ROM1A/PIANO_3.8svx
Process on: Ram:4104/ROM1A/E.PIANO_1.8svx
Process on: Ram:4104/ROM1A/GUITAR_1.8svx
Process on: Ram:4104/ROM1A/GUITAR_2.8svx
Process on: Ram:4104/ROM1A/SYN-LEAD_1.8svx
Process on: Ram:4104/ROM1A/BASS_1.8svx
Process on: Ram:4104/ROM1A/BASS_2.8svx
Process on: Ram:4104/ROM1A/E.ORGAN_1.8svx
Process on: Ram:4104/ROM1A/PIPES_1.8svx
Process on: Ram:4104/ROM1A/HARPSICH_1.8svx
Process on: Ram:4104/ROM1A/CLAV_1.8svx
Process on: Ram:4104/ROM1A/VIBE_1.8svx
Process on: Ram:4104/ROM1A/MARIMBA.8svx
Process on: Ram:4104/ROM1A/KOTO.8svx
Process on: Ram:4104/ROM1A/FLUTE_1.8svx
Process on: Ram:4104/ROM1A/ORCH-CHIME.8svx
Process on: Ram:4104/ROM1A/TUB_BELLS.8svx
Process on: Ram:4104/ROM1A/STEEL_DRUM.8svx
Process on: Ram:4104/ROM1A/TIMPANI.8svx
Process on: Ram:4104/ROM1A/REFS_WHISL.8svx
Process on: Ram:4104/ROM1A/VOICE_1.8svx
Process on: Ram:4104/ROM1A/TRAIN.8svx
Process on: Ram:4104/ROM1A/TAKE_OFF.8svx
This is the END!

As you stated some files were included with the OS, I supposed you were using some ClassicWB preinstalled environment. Shame on me, I was pretending to use a basic 2.1 68k setup wich seems incompatible, even adding those commands and libraries...

So I´ve tested it on ClassicWB Ful v28 following your indications and finally IT (edit: will) WORK LIKE A CHARM.

Thank you DAXB! I will play a little with this, then decide what to do. You will be not credited never ever.

EDIT2: When I try converting to 1 octave 8svx it stops. When I try to convert to multioctave 8svx, it seems to be working but after stop I don´t know where the samples are.

EDIT: I have the same problem with some SYX. I think, could it be due to .voices filenames, since they contain " / : wich can commit error on a DOS filename?

[daxb]

Please add "trace i" at line 17 in the script and save it. Then launch it from Shell like this:

rx Convert_Syx.rexx >ram:log

LHA the log file and upload it to this thread.

About filenames, yes some voice names have bad characters. I`ll see if they can renamed before converting.

8svx files should be in the same path where the voices are.