WK240: How is the "GM Trumpet" voice different from the "Trumpet" voice?

[Little-Acorn]

Or more generally, how are the GM voices different from their non-GM counterparts? My Casio WK-240 keyboard has both.

 

Does GM stand for General MIDI? Do the GM voices do something that the non-GM voices don't? Like maybe send out MIDI commands from the USB port in some special way?

 

I've looked thru the User's Manual that came with the keyboard, couldn't find any mention of GM voices in particular, yet. I'll keep digging. But a few hints would be handy. 

 

Thanx all!

[- T -]

To begin with, keyboard voices do not "send" anything.  Rather, they receive signals, either from the keyboard's keybed, or from the MIDI-IN port (if equipped) or from the receive side of the USB-MIDI port.  The difference between GM and non-GM voices is mainly one of quality and realism.  Many keyboards these days reserve voice bank 0 for the GM voices, and default to this for incoming MIDI signals from external sequencers and MIDI file players, while the keybed defaults to the non-GM voice banks, which are normally referred to as the keyboard's "panel" or "native" voice banks.  The arrangement of voices (one after the other) in the GM voice bank must follow the General MIDI standard so that any GM song file will play the correct voices.  Here is a link to a Wikipedia article on the GM spec and the arrangement of GM voices.  Non-GM voice banks will not necessarily follow the GM spec for arrangement of voices so a GM song file sent to them will not necessarily sound right.  Where the GM file expects to find, say a "bright piano", it may find a trumpet instead.  If a MIDI song file does need the non-GM voice banks, it will include the codes to switch to the "native" voice banks, but obviously will only play correctly on keyboard models that have those same banks.  Typically, keyboards that include a GM voice set will provide some means of accessing that voice set from the local keybed.

 

http://en.wikipedia.org/wiki/General_MIDI

 

A keyboard's non-GM voice banks will typically use larger sound samples than those in the GM voice banks.  The larger the samples, the better the sound quality, but larger samples mean larger memory to store them and consequently, higher production costs.  Also, GM voices use single samples for each note, while "native" voices may use two, three, four or more samples per note.  Some high end keyboards use 8 samples per note - particularly for the piano voices.  This is mainly for realism articulations depending on how hard a note is struck.  A real piano note sounds differently when it is struck harder - particularly the "attack" (the initial "thump" of the hammer striking the string) and the "sustain" (the ongoing "twang" of the string).  The "native" voice banks can be programmed to select the correct sample(s) to emulate this, depending on how hard (velocity) a key is struck.  This same technology can be applied to the guitar voices, the strings voices, and even the "breath" articulations of the brass and woodwind voices.  The marquee piano voice in the Casio XW-P1 is a dual sample voice, while the marquee piano voice in the CTK/WK-6XXX/7XXX is a triple sample voice for more dynamic realism.

 

Some years ago, some keyboard manufacturers adopted the GM2 voice standard which provided some improvements, mainly in the number of voices that can be active at the same time (increase from 24 to 32) and the total number of different voices in the bank (increase from 128 to 256).

 

http://en.wikipedia.org/wiki/General_MIDI_Level_2

 

With today's new keyboards, you will find a mix of either or none.  Some manufacturers have stayed with the GM2 standard, while others have stayed with or gone back to the GM1 standard.  Still others have abandoned GM voice banks all together (or never offered them to begin with) and offer only "native" or "panel" voices.  Obviously, these keyboards can not play MIDI files which use GM voice banks, unless the file is edited and "revoiced" to match that particular keyboard.

 

For a while, Roland offered further enhancements over and above the GM2 voice set with its GS voice spec, while Yamaha did the same with its XG and XG LITE voice sets.  The Roland GS spec is now defunct, but Yamaha still includes XG voice sets in its higher priced units and XG LITE voice sets in its lower priced units.

 

Up to this point, GM, GM2, GS, XG, XG LITE voice banks have been discussed as being in addition to (or included with) a keyboard's higher quality "native' or "panel" voice banks, but it is possible to find low priced entry level keyboards that offer only a GM or GM2 voice bank.  In that case, the GM or GM2 bank IS the "native" or "panel" voice set.  I believe at one time Roland offered some lower priced units that had only a GS voice set while Yamaha did the same with XG only units.

 

So, to sum up, a keyboard's "native" or "panel" voices are there for better selection, quality, and realism over that offered by the GM voice set, while the GM voice set is there to make the keyboard compatible with the many thousands of MIDI files which use the GM voice sets.

 

Hope this helps !

[Little-Acorn]

GREAT explanation! Thank you, Ted!

[- T -]

You are most welcome, sir !

 

Hope you are enjoying that WK-240 !

[BrettM]

Thanks for this explanation Ted, the problem is that the "panel" sounds on the Casio boards I have listened to sound exactly like the GM sounds with only a slight variation, e.g., an octave lower, a DSP added or more chorus or reverb. Or is it just me? Why would they bother making lower quality sounds? Wouldn't they want standard midi files to play back as well as possible?

[- T -]

Brett,

 

There are many who would dismiss your criticism as being purely (or mostly) subjective, but I believe there is a sizable element of truth (fact) to your observation, attributable to several circumstances.  To begin with, if you are comparing panel voices to GM voices through the keyboard's internal sound system and speakers, it is doubtful that you would be able to hear the subtle differences.  Even good quality headphones have difficulty with this.  For a reasonable comparison, you need a high quality stereo amplifier, with a good damping factor, feeding a three-way speaker system with at least a 6 inch (8 inch minimum is much better) woofer.  Damping factor is the amplifier's ability to maintain exact control over speaker voice coil excursions - keeping "overshoots" to an absolute minimum.  The larger the speaker, the more important the damping factor.

 

BUT . . . this is just the beginning.  I don't think it is so much a matter of panel voice quality not outpacing GM voice quality, but rather, a case of technological advances improving the quality of GM voices to be near that of panel voices. For many years now, I have been able to avoid the dreaded Microsoft Wavetable Synth on my PC by dedicating my old Roland SC88 to that service.  At one time, that was Roland's flagship GS Voice unit, but for all its glory, it can not hold a candle to the GM voices on the CTK-7000/WK-7500, except in a very few cases.  Then, there is something else going on in the background.  Turning aside from the GM voices for the moment, this very subject has been a common topic, of late, on Casio, Yamaha, and Roland forums.  Having never been a Korg owner/user, I can not speak for that market.  Listen to a Casio sponsored professional demo of a CTK/WK marquee piano voice and compare it to that on your own local unit.  There really will not be any comparison.  Yours will just not match that of the demo.  Currently, the difference seems to be most pronounced in the top of the line Yamaha PSR (S950) and their flagship arranger (Tyros 5), but it is also quite obvious with the Casio CTK/WK units, and is attributable to the fact that the demo voices on the demo units are very highly tweaked prior to the demo - mostly one end or the other of the envelope (depending on the particular voice), the release time (for piano voices) and the compression (depending on the rest of the sound system).  Recent critics seem to be asking the question "So why don't the manufacturers provide us with pre-tweaked voices?", so let's consider that.  You can not tweak something that does not exist, so you must first do all the design and manufacturing that we find in the currently marketed units before tweaking can begin.  Professional tweaking would normally require several persons for several hours per voice, but let's assume we have two highly motivated individuals that can get the job done in one hour per voice.  Even our lowly CTK-7000/WK-7500 comes with 800 voice presets.  That's 1600 design-engineering man-hours for a complete tweaking job, so let's admit that we can not afford that, and go for just 25%.  That's still an additional 400 man-hours over the current market offering, plus we must now scrap the original un-tweaked ROM and replace it with a tweaked ROM, and our $600/$700 keyboard now comes off the shelf for around $800 to $1000 or more, and those pre-tweaks still have no way of predicting what type of sound system you are going to connect it to, so all of that expensive professionally designed and engineered pre-tweaking may come to exactly naught in your particular sound system.  I think we should look at the GM voices as a fixed set of elementary school water colors that we use pretty much as is, but in the panel voices, the manufacturers have provided us with a nearly limitless sound palette where we can mix, blend, and create to our hearts' desires.

 

The current stock answer seems to be that, if you are not satisfied with your preset panel voices, you probably need to spend some time doing some tweaking, and this would seem to stress the importance of the higher priced units with more substantial user storage space and the need for user voice file clearing houses for file sharing.

[BrettM]

So Ted, you are basically saying that the GM tone pallet is the starting point and the panel sounds are the GM tones that have been tweaked to some extent. They may be very similar to the original GM starting point, depending on the sound engineer, and the two are likely to converge as technology improves. However, the manufacturers leave the original GM tones in there as the raw original waveforms. 

 

Does this mean that the GM tones are a better starting point for our tone editing? Should we be able to reproduce the panel sounds from the GM sounds? What about the panel sounds that don't have a GM counterpart?

[- T -]

Brett

 

No !  The GM voices and the panel voices are totally seperate.  Panel voices are definitely NOT just "tweaked" GM voices.  They are designed and engineered completely separate.  Each set has its own bank of wave samples (even though they may be stored in the same ROM chip), and as I said earlier in this thread, many of the panel voices have two or three (or more in the higher end units) sets or layers of samples that can be switched on or off (usually by velocity or note-on time) for added realism, and panel voice samples are normally considerably larger than GM voice samples for higher quality sound.   Even though both panel voices and GM voices share the same wave shaping parameter templates within a given keyboard or sound module, changing the values of certain parameters for the GM voices may have little or no effect as they are just not set up to respond to those parameters.  What I have said above, is that if you are not satisfied with your panel voices, you need to tweak THEM - not the GM voices.  Tweaking GM voices usually does not improve them much, mainly because of their small sample sizes and cost saving design.

 

GM voices are provided strictly to ensure device to device to device tone compatibility for Standard MIDI files.  Keyboards and tone modules call up the various voices by tone NUMBER not by tone NAME, but all GM voice modules use the same (or nearly the same) voice name/numbering schemes, so that a single SMF song will sound the same (or nearly the same), no matter which GM unit it is played on.  Panel voices do not necessarily follow the same name/numbering scheme from one unit to the next. Voice 2 on a Roland unit may be a bright piano, while voice 2 on a Yamaha unit may be an EP and so on, so a MIDI Song file made for the Roland Panel voices would not sound the same when played on a Yamaha unit.  Using GM voice banks for SMF songs avoids these voice name/number discrepancies, but trades off some sound quality in the process.

 

Going back to your original comment, I think your problem may be that in the Casio units, being toward the lower end of the marketing scale, there really is not a lot of noticeable difference between the GM and the panel voices, mainly for the reasons I stated in my previous reply, but when you get into the $1500 USD and above units, the differences become quite noticeable.  Even so, If you want to improve the sound of your Casio, work with the panel voices, as that is where the room for improvement resides.

[BrettM]

Thanks Ted, I get it now.