[sdiy] Yamaha DXy DCO's

[Don Tillman]

   > From: "Theo" <t.hogers at home.nl>
   > Date: Wed, 18 Feb 2004 15:28:24 +0100
   >
   > The whole point of FM was NO FILTERS, a variable modulation index
   > takes over that role.  

Yes, exactly.

   > Good sounding and affordable digital filters where not doable at
   > the time, so pp looked for workarounds.

Yep.

A single digital filter, depending on implmentation, requires roughly
5 "real-time multiplies".  By "real-time multiply" I mean at least a
12-bit by 12-bit binary multiply operation and an at least 12-bit
accumulate operation that can be peformed fast enough for each sample
cycle, and for each voice.

In 1983 technology a fast multiply was an expensive operation.  My
memory tells me that the only chip that could do this at the time was
made by Hughes and performed a 16-by-16 multiply and accumulate in 
200 ns and cost about $300.00 (!!!). There's some nontrivial support 
circuitry involved too.

Then, do the math.  At a 20 KHz sample rate, you've got 50 us per
cycle, 10 us per filter, 600 nS per voice, so that might work with a
simple enough filters, but it wouldn't be appropriate for consumer
pricing.

The cool thing about the DX7 is that it didn't use filters.  And it
was implemented with no real-time multiplies, just simple adds and
table lookups.  The DX7 also deserves a lot of credit for turning a
technical limitation into a whole new expressive tool.

  -- Don

-- 
Don Tillman
Palo Alto, California
don at till.com
http://www.till.com