Re: AD/DA pairs

[fox@xxxxxxxxxxxxxxxxxxx (Jeff Fox)]

>Hi!
> 
>> I'm not sure... certainly the cpu has the power, the problem would
>> probably be the analog parts. The D/A is only 8(?) bits,
>
>Bidirectional 6 bits. According to Jeff's specs the analog coprocessor
>reads a 20 bit word, sends the lower 6 bits to a DA-output converter, bit
>nr. 10 indicates a cpu interrupt and the highest 6 bits are rewritten to
>the memory word from an AD-input converter.

F21a,b and c were 6 bit.  F21d does have 8 bit A/D D/A pair but
otherwise what Soeren said is correct.  It does an instruction/data
read with optional write-after-read cycle.  The accuracy and linearity
of the A/D D/A pair has not yet been tested except in OKAD.

It was also mentioned that with overmodulation a D/A can produce
higher precision by using the integration properties of the output
device.  So there are a few more bits of resolution available for
slow signals like audio.  This same thing can be done with the
digital outputs.  Modulation at high frequencies produces precise
analog output when matched to the integration properties of the
output device.  The addition of an external comparator and the use
of two digital pins can be used for arbitrary precision A/D as well.  

It is also possible to use a differential amplifier after the D/A
to get higher bit precision with relatively less oversampling.  So
extending from 8 bit to say 12 bit in audio frequencies will cost
a little for exteneral parts, a few cents in quantity, a fraction of
a dollar in singles.

Audio is also only a very narrow bit of the spectrum covered by the device.
The free running analog coprocessor makes very high sample rates
possible and minimizes programming overhead to play and record samples.

Jeff Fox