SPI-FI
Let’s
turn our attention to the other side of the chip
for a moment. Here you find the familiar SPI that
delivers the digital goods up the chain of command.
And here you also find a problem—namely, so many
bits, so little time, and so few pins.
I
just finished saying that the ADC can run at up
to 75 MHz, delivering 16 bits of data each clock.
Excuse me, but last time I looked, there weren’t
many chips out there with a 1.2-GHz SPI! As a practical
matter, the feasible sample rate is dictated by
the chip’s maximum SPI clock rate, which is 40 MHz.
You can get up to 2.5 Msps in Single Channel mode,
which outputs 16-bit words. But, as you can see
in Table 2, adding channels
cuts the maximum sample rate, more so because 24
bits are output for each sample in Multi-Channel
mode (see Figure 2).
|

(Click
here to enlarge)
|
Figure
2—In Single Channel mode, the 16-bit sample
data is output directly. Multi-Channel mode
requires the addition of an 8-bit header including
a channel designator, parity bit, and a data
valid flag. |
There’s
another seemingly odd thing about the interface.
The good news is that the over-sampling ratio can
be set independently to optimize the speed/resolution
(ENOB) trade-off for each channel. The data rate
out the SPI port is dictated by the channel configured
for the highest real sample rate. But data is output
from every active channel at the high-speed channel’s
rate, even data from channels running at a lower
speed (i.e. higher over-sampling ratio). That means
data from lower-speed channels is output at an artificially
high rate with dummy outputs (previous data with
new data flag not set) possibly outnumbering real
ones (new data and new data flag set).