Yes indeed Senso. Have looked into TI and Analog Devices nice offerings. At modest resolutions 'systems in a chip' are easier to find, which greatly limits the needed board work. Working at Precision Analog (PA) near the noise level limits is needed for much lab work, which pushes toward separated chip functions between DSP, conversions and signal conditioning. That has my current attention.
Since signal conditioning has lots of factors that are influenced by particular applications, coupled with alternative tech solutions, maybe that is best left as a daughter board that is easier to tailor to different lab and instrument applications.
That leaves DSP with DMA for shared couplings, and A to D conversion as a very high traffic region that has pushed this area into buses of 16 bits in width, or more. Much of that bandwidth stems from the handling of multiplexed analog sensor sources into the A to D. Byte buses to a separate microcontroller is probably adequate if the DMA can handle it, because processing brings data compression into play. A tall order. But PA is a big, wide, wonderful world. Waiting.
If we had a good open systems offering that was attractive to academic projects, maybe that could bring easier entry to the needed support software by companies wanting to forge 'relationships with a future' thru open systems. Suits me.
This Newby is perplexed in hunting for a landing spot for these highly related subjects, from among your terrific collections. Maybe the ARM topic was the closest. The ARM Wiki states that as of 2009, an amazing 90 percent of portable devices are employing ARM and most include DSP. But when I deploy SEARCH here for DSP or AUDIO I get zero hits.
I need a $10 or $20 dedicated DSP chip, DMA coupled to a host controller and Muxed ADC with 16 or 18 bit resolution. Also some hi resolution DAC. All good basis for lab or instrument applications. Availability of support software is crucial item. Any joy?
Righton sqkybeaver. Currently having difficulty navigating this wonderful wormhole.
Had started with 'web platform' board and bus connector to support a separate I/O board working much deeper in analog. Definately at a cost premium but not wasted. This will take time and could move up to at91 for interface and run time support.
Sharc has expensive compiler unless it can be negotiated for open systems use. Causes much uncertainty.
Feeling need for my board having good DMA connection to microcontroller via Parallel Master Port PMP. Not sure if Sharc has DMA compatablity with at91.
Only landed here for now, until any shared interest possibly shows up.
Not ready to navigate a separate forum page but would like company.
Catching flex during early stage definition of the 'Premium Digalog I/O' adapter.
Having typical -- Good news/Bad news - - progress.
Have found a terrific 18 bit A to D with both serial and parallel addressing control. With the addressing mode being selectable at run time. It is from Analog Devices AD7679 series. A $20 chip, central to my application. Working to match it up with a 4:1 external mux chip.Targeted to run at full tilt data bandwidth but have some room to back-off. Could plan to proto with serial but use 8 bit parallel data bus for eventual real time acquisition and processing.
My earlier feeling of DSP as being needed is confirmed in Spades. Compresses data in real time and does some pattern recognition based on FFT. This is viewed as a co-processor providing relief for an upstream microcontroller. Will probably retain analog signal conditioning at front end.
This is an eight or ten dollar chip from Analog Devices; catch their advertizing as - -
SHARC® Processors, dominate the floating-point Digital Signal Processing market, delivering exceptional core and memory performance complemented by outstanding I/O throughput. Starting at 319 MFLOPS per dollar...
How many of the DMA channels remain to be assigned?
Defining an instrument interface, now centered on use of Analog Devices AD7679 A to D converter with 18 bits of resolution. Keen part has both serial and parallel interfaces.
Ultimate I/O bandwidth of over 400 K bytes per second implies use of DMA in the PMP mode but SPI is especially handy for prototyping stage, and its use for some purposes will likely remain.
Per 38.1 of the PIC33 ref this would need 4 of the 8 DMA channels.
