"Jon Harris" <jon99_harris7@hotmail.com> wrote in
news:K83Te.20215$cy4.1003@trnddc05:
> Is your DSP a 32-bit fixed point processor? If so, the way you
> justify controls whether you have 8 bits of "headroom" (also called
> guard bits) for the signal to grow during processing or 8 bits extra
> at the end to deal with round-off during processing. Based on what
> type of processing you are doing, it might make sense to choose one or
> the other. If you are planning on mixing a lot of signals together so
> that result may grow, the former may be a better choice.
>
> Some type of compromise might be the best bet. I haven't done much
> 32-bit fixed-point processing, but I would imagine you would leave a
> few guard bits at the top to avoid overflow and use the rest for
> additional precision at the low end.
>
> If you are using a floating-point DSP or aren't doing any significant
> processing, then it's a moot point and you can choose whichever is
> easiest or you like best. Some people light LSB-justifying so that
> the digital word put out by the ADC is exactly what ends up in the DSP
> (although I guess you still have to deal with sign-extending for
> negative values).
>
Most of the time, MSB first is used.
--
Al Clark
Danville Signal Processing, Inc.
--------------------------------------------------------------------
Purveyors of Fine DSP Hardware and other Cool Stuff
Available at http://www.danvillesignal.com
Reply by Jon Harris●September 5, 20052005-09-05
Is your DSP a 32-bit fixed point processor? If so, the way you justify controls
whether you have 8 bits of "headroom" (also called guard bits) for the signal to
grow during processing or 8 bits extra at the end to deal with round-off during
processing. Based on what type of processing you are doing, it might make sense
to choose one or the other. If you are planning on mixing a lot of signals
together so that result may grow, the former may be a better choice.
Some type of compromise might be the best bet. I haven't done much 32-bit
fixed-point processing, but I would imagine you would leave a few guard bits at
the top to avoid overflow and use the rest for additional precision at the low
end.
If you are using a floating-point DSP or aren't doing any significant
processing, then it's a moot point and you can choose whichever is easiest or
you like best. Some people light LSB-justifying so that the digital word put
out by the ADC is exactly what ends up in the DSP (although I guess you still
have to deal with sign-extending for negative values).
--
Jon Harris
SPAM blocker in place:
Remove 99 (but leave 7) to reply
"Markus" <Markus.Schweikhardt@web.de> wrote in message
news:1125948240.082138.139480@z14g2000cwz.googlegroups.com...
> Hi,
>
> I work with an 24bit evaluation-adc-board from AKM. The board sends the
> data in 32bits to the DSP. I can choose whether I would justify the
> 24bits at the MSB or the LSB of the 32bit word which is send to the
> DSP. Now my question: what`s the difference and what makes more sense.
> The DSP can both formats.
>
> Markus
>
Reply by Tim Wescott●September 5, 20052005-09-05
Markus wrote:
> Hi,
>
> I work with an 24bit evaluation-adc-board from AKM. The board sends the
> data in 32bits to the DSP. I can choose whether I would justify the
> 24bits at the MSB or the LSB of the 32bit word which is send to the
> DSP. Now my question: what`s the difference and what makes more sense.
> The DSP can both formats.
>
> Markus
>
The difference should be obvious and what makes sense depends on you.
Early in a design project I like having things left justified because I
can change the ADC resolution without changing the meaning of the word.
There are other reasons that left-justifying everything makes sense,
but it really boils down to personal preference.
--
Tim Wescott
Wescott Design Services
http://www.wescottdesign.com
Reply by Markus●September 5, 20052005-09-05
Hi,
I work with an 24bit evaluation-adc-board from AKM. The board sends the
data in 32bits to the DSP. I can choose whether I would justify the
24bits at the MSB or the LSB of the 32bit word which is send to the
DSP. Now my question: what`s the difference and what makes more sense.
The DSP can both formats.
Markus