Hi,

I read an article at:
http://www.eetimes.com/General/PrintView/4210546

When the author talked about MIMO, he said:
...........
MIMO factor

A case in point is MIMO multiple antenna applications. With floating point in each DSP core,
developers can “leave complex numeric algorithms such as MIMO applications to floating point,”
explained Bier. For something like MIMO application, which is “super touchy” and requires “high
accuracy,” Bier said that not having to do matrix inversion is a huge win for developers. 
............

I don't understand why he said floating point can avoid matrix inversion. What do you think
about his opinion?

Thanks.

I think what is meant is that traditionally you do the algorithm model in
floating point (probably double precision in Matlab or the like). Then
convert the algorithm to fixed point for the product implementation. 
If the target platform supports native floating point, you can skip the
latter step and reduce time-to-market.

MIMO comes in as an example that is sensitive to numerical precision and
difficult to analyze (compared, to, say, a simple FIR filter, which
actually isn't -that- simple anymore, if optimized in fixed point).

fl <r...@gmail.com> writes:

> Hi,
>
> I read an article at:
> http://www.eetimes.com/General/PrintView/4210546
>
> When the author talked about MIMO, he said:
> ...........
> MIMO factor
>
> A case in point is MIMO multiple antenna applications. With floating
> point in each DSP core, developers can “leave complex numeric
> algorithms such as MIMO applications to floating point,” explained
> Bier. For something like MIMO application, which is “super touchy” and
> requires “high accuracy,” Bier said that not having to do matrix
> inversion is a huge win for developers.
> ............
>
> I don't understand why he said floating point can avoid matrix
> inversion. What do you think about his opinion?

It sounds to me he is saying that you can avoid implementing a matrix
inversion in fixed-point, not avoid it completely, but I agree that the
way it is stated seems to say completely. I wouldn't have written it
that way.
-- 
Randy Yates
Digital Signal Labs
http://www.digitalsignallabs.com

On Sun, 26 Aug 2012 07:07:56 -0400, Randy Yates wrote:

> fl <r...@gmail.com> writes:
> 
>> Hi,
>>
>> I read an article at:
>> http://www.eetimes.com/General/PrintView/4210546
>>
>> When the author talked about MIMO, he said: ...........
>> MIMO factor
>>
>> A case in point is MIMO multiple antenna applications. With floating
>> point in each DSP core, developers can “leave complex numeric
>> algorithms such as MIMO applications to floating point,” explained
>> Bier. For something like MIMO application, which is “super touchy” and
>> requires “high accuracy,” Bier said that not having to do matrix
>> inversion is a huge win for developers. ............
>>
>> I don't understand why he said floating point can avoid matrix
>> inversion. What do you think about his opinion?
> 
> It sounds to me he is saying that you can avoid implementing a matrix
> inversion in fixed-point, not avoid it completely, but I agree that the
> way it is stated seems to say completely. I wouldn't have written it
> that way.

And, depending on the matrix you're inverting, just doing it in floating 
point doesn't guarantee that you'll magically get good results.

-- 
My liberal friends think I'm a conservative kook.
My conservative friends think I'm a liberal kook.
Why am I not happy that they have found common ground?

Tim Wescott, Communications, Control, Circuits & Software
http://www.wescottdesign.com