eZdsp with TMS320F2812 - A/D converters

Hello,
	for my project I've bought the Spectrum Digital board with the 
TMS320F2812 DSP. I have chosen it because it has A/D converters and PWM 
units on board. Yesterday I met a man from a company in Germany and he 
told me that the converters of the processor are very bad. He would 
measure the DC signal, and what he got was something that from 12-bit 
signal only 7 bits were stable. He said that although the signal was 
taken from the voltage generator and was really stable DC signal, the 
output from A/D converters was "changing", the 5 bits were constantly 
changing. He said, that he had to use external A/D converters and now he 
doesn't have this problem.

Have you seen this before? Do you possibly know where it could come 
from? Or it is like this with this processor?

Greetings,
Pawel

Pawel Szczupak wrote:
> Hello,
> 	for my project I've bought the Spectrum Digital board with the
> TMS320F2812 DSP. I have chosen it because it has A/D converters and
PWM
> units on board. Yesterday I met a man from a company in Germany and
he
> told me that the converters of the processor are very bad. He would
> measure the DC signal, and what he got was something that from 12-bit

> signal only 7 bits were stable. He said that although the signal was
> taken from the voltage generator and was really stable DC signal, the

> output from A/D converters was "changing", the 5 bits were constantly

> changing. He said, that he had to use external A/D converters and now
he
> doesn't have this problem.
>
> Have you seen this before? Do you possibly know where it could come
> from? Or it is like this with this processor?
>
> Greetings,
> Pawel

TI states the typical effective number of bits for the 2812 A/D is 10.1
bits (S/N 62dB), thats typical, min could be much worse, its not stated
in the datasheet I've read. Add to that noise of external reference (if
used) plus various other external noise sources and I can see getting
only 7 stable bits.

The performance of the A/D's that go into DSPs depends not only on the
A/D itself, but where it placed relative to the core. The same A/D used
in different DSP's can have a wide range of performance. They also tend
to have "sweet spots", that is clock  frequency multiples of the core
clock where performance is the highest, some manufacturers will state
these clock frequencies in the datasheets.

Also the best material to make a DSP core is not the best material to
make a high performance A/D, and vice versa,  so putting both on one
die is a compromise.

So yes an external A/D is going to have much higher performance as
compared to an internal A/D for the above reasons. The TI A/D is pretty
fast so you could oversample/average to increase your effective number
of bits.

On Sat, 05 Mar 2005 10:48:57 +0100, Pawel Szczupak
<pawl6@OHNESPAM.op.pl> wrote in comp.dsp:

> Hello,
> 	for my project I've bought the Spectrum Digital board with the 
> TMS320F2812 DSP. I have chosen it because it has A/D converters and PWM 
> units on board. Yesterday I met a man from a company in Germany and he 
> told me that the converters of the processor are very bad. He would 
> measure the DC signal, and what he got was something that from 12-bit 
> signal only 7 bits were stable. He said that although the signal was 
> taken from the voltage generator and was really stable DC signal, the 
> output from A/D converters was "changing", the 5 bits were constantly 
> changing. He said, that he had to use external A/D converters and now he 
> doesn't have this problem.
> 
> Have you seen this before? Do you possibly know where it could come 
> from? Or it is like this with this processor?
> 
> Greetings,
> Pawel

What you have been told is pretty much true.  Consult the latest
errata documents for the 2812 from TI's web site.  There are some
possibilities to improve the operation.

One involves using two channels for references, inputting 0.5 volts to
one and 2.5 volts to the other, and performing a software calculation
to use in adjusting or correcting the readings of the other 14
channels.  We are using this one.

The other is to provide your own external reference voltage to the
ADC.  The newest versions of the chip allow you to do so on two pins.
We have not tried this one.

Consult the TI errata documentation and contact TI for app notes.

-- 
Jack Klein
Home: http://JK-Technology.Com
FAQs for
comp.lang.c http://www.eskimo.com/~scs/C-faq/top.html
comp.lang.c++ http://www.parashift.com/c++-faq-lite/
alt.comp.lang.learn.c-c++
http://www.contrib.andrew.cmu.edu/~ajo/docs/FAQ-acllc.html

> One involves using two channels for references, inputting 0.5 volts to
> one and 2.5 volts to the other, and performing a software calculation
> to use in adjusting or correcting the readings of the other 14
> channels.  We are using this one.
> 
> The other is to provide your own external reference voltage to the
> ADC.  The newest versions of the chip allow you to do so on two pins.
> We have not tried this one.
> 
Hello,
	Thank you both for answers. I will check the methods.
  As you told, there is a pin for external reference voltage. I am going 
to try it :)


Greetings,
Pawel