[time-nuts] 31bit ADC, 1000 samples per second

David davidwhess at gmail.com
Tue Dec 11 00:54:37 UTC 2012


On Mon, 10 Dec 2012 23:01:49 +0000, "Poul-Henning Kamp"
<phk at phk.freebsd.dk> wrote:

>In message <tilcc892a6ntfk64t2nljm92idr92df078 at 4ax.com>, David writes:
>
>>The gain drift is specified at 2ppm per degree C.  There are
>>provisions for a calibration cycle but that requires external
>>multiplexing.
>
>That's the really smart thing about this particular chip:  You can
>drive the geophone calibration signal in through the second input
>and out onto the first, which eliminates the external calibration
>mux and the errors that usually cause.

I see now from the data sheet that the 3000 ppm typical gain mismatch
applies between the PGA gain settings and not between the multiplexor
input channels so you can use the internal multiplexor for external
calibration.  The other instrumentation delta-sigma ADCs I am familiar
with do a gain and offset calibration for every conversion which
knocks the drift down by an order of magnitude unless you disable that
feature.

>And yes, in seismology you do care about frequencies from 0.1Hz and up
>(0.01Hz if you're involved with the CTBT) and most voltage references
>will be stable enough for that.

Voltage reference 1/f noise will be a problem but only because the
ADS1282 input stage is already chopped allowing that kind of
sensitivity at low frequencies.  Competing instrumentation
delta-signal converters have the same problem if you consider that a
problem. 

>For longer term you calibrate your entire system, starting with the
>geophone, so the voltage reference is caught that way.

What they need is a ratiometric geophone . . . :)



More information about the time-nuts mailing list