Re: Texas TMS1232 Delta Sigma converters

On Tue, 01 Nov 2005 17:11:13 +1000, Adrian Jansen <adrian@xxxxxxxxx>
wrote:

>John Larkin wrote:
>> On Tue, 01 Nov 2005 10:00:49 +1000, Adrian Jansen <adrian@xxxxxxxxx>
>> wrote:
>>
>>
>
>>
>> What's a TMS1232? I can't find it.
>>
>> Make sure you're not picking up 50/60 Hz line noise, and that your
>> voltage reference isn't noisy. 19 bits it thin air.
>>
>> If you can't reduce the sample rate, all you could do is add some
>> analog filtering ahead of the ADC, but that equivalently slows it
>> down, so you can't filter too hard.
>>
>> Is this a weighing application? If so, vibration will be a serious
>> problem, and the only fix is lower sampling rates or more averaging,
>> which amounts to pretty much the same thing.
>>
>> John
>>
>Sorry, I meant ADS1232, too used to typing T for Texas parts.
>
>They did some neat tricks in the internals of this, using a 10 Hz
>sampling frequency automatically rejects a lot of 50/60 Hz signal.
>
>One of the objects in using this widget is that we can cut the external
>part count down from around 60 precision components on an existing
>design to less than 10 non precision ones, so I prefer not to add analog
>filtering unless we *really* have to.
>
>My bench power supply is quiet, 5.00 volts with less than 0.1mv residual
>noise, and using fully isolated double transformer with separate
>inter-winding shields. Still get some hum pickup on the bench of
>course, but its about as low as I can get till I do a proper board layout.
>
>The app isnt for weighing, and I can jigger the test transducer so its
>inputs are essentially true zero for test purposes, its just noise
>picked up between the transducer and the ADC I'm trying to eliminate.
>As well as the ref noise and stuff of course.
>
>Thanks for the tips.


Past about 18 bits, the reference problem is serious. Most references
have a heap of 1/f noise, which will be a problem around 10 Hz...
bandgaps are bad. We're using an Analog Devices Xfet ref, ADR421,
which is pretty good. The Intersil things are interesting too, basicly
a voltage trapped on a capacitor forever.

One could build a 0.xx Hz active lowpass filter (a non-trivial
problem) or maybe average a bunch of references, four maybe, but the
math is discouraging here, improvement as to the square root of the
number of refs.

John

.