The 707/727 must be the worst machine in this regard! It honestly must be easier to burn
new sounds for practically *anything* else.

Going over the improved schematics which were posted a while back, it is finally possible
to see the pin numbers on the ICs. This means that the "scrambling" is basically explained.
With some kind and helpful explanation from Colin Fraser, I have learned how to verify
that these are as they appear, which I will do this weekend. If this was the only issue to
consider in making new sounds, it'd be really easy.

The sounds used are of varying quality. For example, the crash and ride cymbals each fill a
whole SRAM chip. They are each 32kB, which is decent. Since they are each on their own
chip, with their own clocks and addres counters - making new chips for these is fairly
simple! One could replace these two cymbals with their own samples. Just look at the
schematic and see IC19 (crash), and IC22 (ride) - note the pin numbers listed for A0-15,
and D7-D2. Once I socket my chips, I will try this. (after I get like three other big projects
off of my bench, which I've had little time for this summer)

What I would really rather do is replace other sounds, such as the kicks, snares, hats, etc.
The toms and hats are all 8kB samples. All of the remaining sounds are only 4kB. This is
pretty lo-fi. As each sound is decoded, a special envelope signal is matched up with it.
Also, lots of the remaining sounds are paired together: bass1+bass2, snare1+snare2,
rim+cow, clap+tambourine. So each pair is 8k, with the even addresses being one sound,
and the odd the other. And if this wasn't all difficult enough, new samples in these places
would still get the same pre-programmed (analog) envelope contour applied to them,
which may or
not sound any good. I bought better quality scans of the service manual from Mark
Glinsky, so I can read the timing chart from "707svc-07.tif". Now that I am beginning to
understand all of this, I am less inclined to want to drop new ROMs in. I will probably try it
eventually anyway. Meanwhile, I certainly will try new sounds in place of the cymbal ROMs,
since they are relatively easy.

What I am more inclined to do is do away with the Gate Array and DAC entirely. I love the
707 interface, but this is a lot of work to get lo-fi samples. Roland did this because it was
cost-effective for them, but so what. Another consideration is that some of us have
already put tunable sample clocks in these machines - yet another thing we no longer
need the gate array for. Easier to just get a more EPROM-friendly machine like a Linn,
Drumtraks, DMX, etc, but of course this loses us the best X0X interface in the world! I'd
rather "maximize the potential" of the 707. Since I already fit my 707 for a board of clock
circuits, I am going to try to make a better board with improved clocks, along with voice
EPROMs and DACs. A full 8-bit sample for each sound, put through the 707 audio board.
The tricks in interfacing to the original 707 voicing circuits are sending gates to sound the
samples at the correct time, and applying the articulation. The former I think is easy, but
the latter I haven't tried yet. In the 707, the envelopes for the sounds (except cymbals) are
timed to change the DACs +reference. It should be possible to use them from IC38 and
IC39 without multiplexing them first.

OTOH, maybe new SRAM samples are enough The grunge isn't horrible, but new samples
would need to be specially compressed and treated for the analog envelopes. The
envelopes themselves could be redone, but this means either replacing parts each time the
samples are swapped out, or adding yet more controls to my already knob and switch
covered 707. I think that reproducing the address circuitry around the ride and crash
cymbals is a little easier and will probably sound better. Once I socket my machine and rip
the ROMs though, I guess I'll try to make new sample sets.

CJ - trying to decide!