Texas Instrument Digital Signal Processors

[flatpack]

may mga users ba dito at programmers ng TI DSP?

[zer0w1ng]

I do DSP programming but not on TI chips.
TI DSPs are expensive.

My first DSP experience was on a TI trainer on a seminar years ago. I even forgot now its assembly syntax.

Other DSPs such as DSPIC are similar. You only need to know how their MAC (multiply-accumulate) instruction works.
Most DSP algo's will be using this instruction.

Floating point DSP also is easier to program than fixed point or integer core.  But also is more expensive.

['yus]

pasingit lang po:
anong pinagkaiba ng mga DSPs sa microcontrollers? 

[zer0w1ng]

pasingit lang po:
anong pinagkaiba ng mga DSPs sa microcontrollers? 

DSPs are specialized microcontroller/processor for processing real-time signals.
Most of them have fast A/D and D/A for input and output or a CODEC.
It has a single cycle MAC instruction (multiply and accumulate), reverse addressing (for FFT) and circular buffers (FIFO) built-in in hardware. It also has zero cycle loops.
These optimizations are needed to make it fast as possible to be able to process data and output it real-time without delays.
They are mostly used in audio or video applications.  Digital equalization, MP3, JPEG compression and MPEG4 are its examples.

The most important feature of a DSP processor is a fast MAC instruction.
Algorithms of DSP uses this, like matrix dot multiplication in FIR filters.

[jaja]

patanong din po kung may single cycle MAC instruction (multiply and accumulate), meron bang division and accumulate na sigle instruction? kung wala, paano ito ginagawa sa DSP? tanong lang po. salamat.

[zer0w1ng]

No division on chips that I know. Divisions also costs more cycles.

And division is rarely needed in DSP algos.
Also you could use multiply with numbers less than one to achieve this.
For example: X/2 = X * 0.5

[jaja]

Thanks zer0w1ng. that a good style, multiplying by fraction or decimal.

But there are times where you really need to divide. For example we have a PSU with rated power of 50W and a variable output voltage from 6V to 12V. P=V*I, I=P/V. In this case, what would be the best approach. tanong lang lang po, i'm a hardware engr. thanks again. 

[zer0w1ng]

Thanks zer0w1ng. that a good style, multiplying by fraction or decimal.

But there are times where you really need to divide. For example we have a PSU with rated power of 50W and a variable output voltage from 6V to 12V. P=V*I, I=P/V. In this case, what would be the best approach. tanong lang lang po, i'm a hardware engr. thanks again. 

If the processor has no divide instruction, you could use binary division or successive subtraction.
Binary division is like long division by hand by shifting and subtracting.
It is difficult to explain in detail without pictures.
This is a common algorithm just search the net on how to do it.

The easiest is successive subtraction but this is not the fastest.
Just subtract the the dividend by the divisor until the result is less than the divisor.
The number of subtractions is the answer. And the last result is the remainder.

[razalas]

patanong din po kung may single cycle MAC instruction (multiply and accumulate), meron bang division and accumulate na sigle instruction? kung wala, paano ito ginagawa sa DSP? tanong lang po. salamat.

Wala po. Tama sila successive subrtraction lang, like what we can do sa uC. Ang gasgas na FIR filter, which is a common type of digital filter using DSP uses MAC, so why bother to add operation like division?

[jaja]

salamat sa dagdag na karunungan.