16
Messages
This section allows you to view all Messages made by this member. Note that you can only see Messages made in areas you currently have access to.
Messages - Bustar
17
General discussion / Re: Missing DSP, Audio, Analog.
My immediate lab case is relatively unbounded in several respects. As one consequence, you have likely noticed that my systems choices are not very steady. But many lab cases may harbor such uncertainties, when researching. My approach had been to reach for easily available increases in system performances to help cover evolving uncertainties of my immediate cases.
Reaching decisions and developing acceptance of a system synthesis has long made this type of effort interesting to me over a considerable variety of systems implementations. This frequently involves tech not previously available. With the general definition level lacking important details prior to implementation, there are usually uncertainties involved. Takes courage, but hedge when possible.
For my applications it may also develop that FFT rounding errors due to fixed point math may not be signifigant to my lab uses, due to uncertainties of the needed dynamic range. Other structural changes are also being found that may lower the expected need of processing speed, which can signifigantly simplify an initial implementation. Such an avenue could bring the audio and my lab instrument case needs into closer alignment. Work continues.
P.S.- - Just caught the timely news posting by sqkybeaver via Ian. Nice work within common audio limits. Very useful demo. Consistent with the previously comments, distinguishing the expected differences between common audio vs lab instrumentation cases. This refers to the link within - -Here’s an application using- - FFT on an PIC18F4550 8-bit microcontroller
18
General discussion / Re: Missing DSP, Audio, Analog.
Spent some time with your 5506 part. Still amazing what gets packed into a single chip. Is that in the $6.50 range?
Also spent some time exploring the TMS320C6712D, as an entry level chip with floating point in the 6000 set. Priced like about 2X your 5506?
Maybe the contrast between the 5506 versus the 6712 provides a contrast that distinguishes audio versus scientific lab or instrument use, thinking tenatively. Available chip footprints may have important differences, among others.
The reported rounding errors of fixed point may not be evident for audio uses. That may be a different story when working at 16 to 20 bits of resolution for instrument purposes, somtimes near or below the noise floor. Those processing FFT have important consequences from the need and application of filtering at pre or post processing stages.
The audio case seems relatively well defined and bounded in some respects. But recognizing and altering audio sounds may not be so clearly established and/or the tools may not be readily available, as in speech recognition for example. Ditto for finger print recognition, with these being of interest as a possible classes of post FFT usages. (cont.)
19
General discussion / Re: Missing DSP, Audio, Analog.
Each datastream is windowed so eight FFT phases are processing concurrently. So each 2exp 10 points/spectrogram line is offset by 2exp 7 samples and that requires 2exp 3 = 8 concurrent lines for each of the four datastreams.
There are mode options which involves choices of the bandwidths being processed. Each datastream has an upper bandwidth in the area of 10 kHz to 15 kHz. The low end has fine work to explore but D.C. coupling is likely to be a desirable option. Although much of this is in the audio frequency range, it is definately not involving audio nor audio tranducers. No obvious reasons that all of those fancy audio decoders could not be carried along for your uses and for the general interests of others of the open system persuasions.
The analog board for my case would include numerous options for signal conditioning, and ESD protection would apply to the inputs from remotely located sensing transducers. The signal conditioning technology may be solid state under program control or it can be differential op amps with discrete passive feedback elements or a mix thereof. Board size is not a known constraint. Layout and other design factors are critical to low noise and interference performance. This arrangement could be an open prototype base for a variety of lab or instrument applications. My uses could reasonably employ about two dozen system replications.
I mentioned 'real time' but there can be a TBD burst mode, because the content of the datastreams are not likely to be changing at a high rate. Such a burst mode could open up some processing time to handle the next paragraph.
FFT had been mentioned because it is a relatively familiar setting. There is another major DSP processing task which needs to be addressed and it is less familiar and less characterized. It applies to the 'handling' of the FFT spectrogram output. If the initial treatment finds this cannot be contained and adequately supported, it can justify adding a duplicate processing board. And of course, for the data cases selected to be archived, there is a need to have a waterfall display of the spectrogram frame data to aid retrospective evaluations.
At this stage, was not moving to exclude any options. Would just like the best-can-do. Cheers.
20
General discussion / Re: Missing DSP, Audio, Analog.
From the TI website I had I moved along to peek at the 3rd party offerings. My oversized jpg file shows a propriatary package that is 2.7 inches long and 1.5 inches high. The unit mounts a
TMS320C6711 DSP from TI, for logic a Xilinx XC3S400 FPGA, with added SDRAM and FLASH memory.
Later on in 2008, their press release shows a strongly enhanced version at link - -
www.eetimes.com/electronics-products/embedded-tool...
It features a C6455 DSP clocked at 1.2 GHz, and a Spartan 3 FPGA plus added I/O.
These are examples that suppout your favorable view of the TI sourcing.
Earlier, I had also found a similarly strong 3rd party offering using the Analog Sharc series.
These technically strong offerings did not have attractive costs and may have also involved unattractive support software factors.
In my 3rd party meanderings I also found that Seeed was offering a board for the ARM Cortex M3, And when I followed that upstream, found that ARM had announced the Cortex-M4 with optional Floating Point Unit (FPU) last fall. And ST Micro had announced they would be sampling that chip this year.
I had noted the activity here in DP involving the interest of those "Getting started with ARM". That combined with the previous record with ST in open systems seemed to offer some interesting prospects. So taking an avenue toward joining a community effort, I was expressing interest in the new ST offering of the Cortex M4 with FPU because it could possibly share a common board layout with a Plain-Jane ARM version. And perhaps it could provide a shared setting for software development. Not sure how much of that can hold up.
It is true the M4 FPU could offer performance on the weak side for me. And if a stronger entry can be made, the added chip cost is not an issue with me. (cont.)
21
General discussion / Re: Missing DSP, Audio, Analog.
I will repeat my apology to you Rsdio for causing confusion. I think you may find my explanations do have some logic for my meanderings.
22
General discussion / Re: Missing DSP, Audio, Analog.
I greatly respect your experience and the judgements which you have accumulated. I will look closer at the 320. It is not clear in my mind, but I have probably been there. But will keep an open mind, as I look.
Surely you can help me understand a couple of things this newbie is seeing, relative to my push into DSP floating point.
Consistent with part of your post, some are evidently experiencing difficulty in changing to an ARM base. But maybe it is running pretty strong, probably for a couple of reasons. As an origin point, ARM Cortex M3 is breaking in with an open sys mini-board offering at $50. ARM Cortex M4 with an embedded Floating Point Unit (FPU) option was announced last fall, and ST Microelectronics has recently announced they will sample that chip this year. A trend and history in that sequence??
Maybe that portens that an ARM M4 with FPU will surface soon as another open sys board offering. I am totally unsure how that may contrast with the status of the 320 offerings, or the contrasting access to open support software between the two. But I would like help in understanding that topic.
The A to D tech seems to have been making sizable advances. Relative to your discussion of bus structures, the AD7679 group has multiple series and modulo parallel interfaces in the same part number. I have not grown comfortable with the sigma-delta style but the SAR types have been coming on strong in deep resolution, and may strongly displace signa-delta eventually.
An open sys FP processor board would seem to need several bus I/O protocols. A variety of those are often supported out of DMA, as well as primative. That facilitates tailoring of the front-end to the application on the analog I/O board.
I am bushed and heading for the hay. Cheers.
23
General discussion / Re: what architecture do you use?
Not sure I have the option immediately, but shifting to double precision would seem to be a remedy. 'But Doc, I'm not sure I need that med.'
24
General discussion / Re: Missing DSP, Audio, Analog.
Back to your audio interests. This PA case probably carries audio along because so many of the audio decoders, etc. are built-in offerings with many DSP chips or their plug-in alternatives. But you are surely getting those at a lower cost point than I am using. This PA board could work at frequencies that would let you communicate with the bats. Ha.
No kidding, you could do some voice recognition with them. Expression might be more difficult. - - - Aha, the makings of a bat and mosquito fence.
Mentioning cost point - - This newbie is wondering about the reception that would develop for a one hundred dollar I/O board. Even it it is not any old board.
Wondering if the I/O board could be both a high end and low end analog/audio project. If the high end DSP is a pluggable daughter board we can unplug that for the low end case. And maybe a footprint for a cheap DSP can be worked into the I/O board. Yep, dreaming.
You mention Sharc. That low end unit at under $10 must be an advertizing loss leader. And they have structured it with a handful of family groups. Amazing field.
25
General discussion / Re: Missing DSP, Audio, Analog.
First crack out of the box, caught a neat third party packaged upscale DSP in multiplanar for the big BG pad arrays with on-board supports. Mature. With support software Bucks TBD.
And I haven't even dented the variety of thrid party offerings. Its inexhaustable.
Makes lots of sense that DSP should have some good std offerings, because so much can be done in programming, perhaps with less baggage than microcontrollers with all of the accumulated I/O and peripheral standards built-up along the way.
Perhaps this can permit added concentration on the I/O board where things tend to get somewhat more specialized in matching to the analog data sources. Would like to keep the 'open systems' approach to structuring for wider shared usage. This being directed into the deeper resolution more suited to instrument work or geo sensing cases. Definately looking to work near the noise floor and somtimes processing signals at some depth in the noise. But with a hugh dynamic range. This demands careful layout and power supply decoupling treatments.
For my own case, looking at four main signal sources multiplexed into the A to D (18 bits preferred resolution) with fast data delivery into the DSP perhaps thru DMA. Expecting to use 6 or 8 sets of signal conditioning of the 4 main sources because some may operate at more than one analog bandwidth and/or gain. The signal conditioning may use some program selected types or some analog with passive discrete parts, or a mixture thereof. With ESD protection of input sources a must.
Sampling rate of 10exp 5 to 10exp 6 SPS for the A to D can support a whole range of analog signal input frequencies depending upon how many input channels are in use.
26
General discussion / Re: New member
27
General discussion / Re: Missing DSP, Audio, Analog.
28
Bus Pirate Support / Re: PirateScope - yet another BP oscilloscope client
Today, just encountered 3D graphics as a function that can be inbedded with GSP + ARM by TI. As a common waterfall display of FFT results that could be a natural to show a spectrogram.
Then all of the interest in SCOPE aspects clicked in. The Precision Analog (AP) system of my application interest also is keeping general lab and instrument applications in mind. This could eventually play into SCOPE solutions in a couple of ways. The high resolution A to D function in the present definition can sample a single source at about a half million samples per second continuous. It you are observing a repetitive event or sampling at slower rates you could display the amplitude versus time AND its 3D spectrogram of frequency versus time versus amplitude (a waterfall display). That could be produced by some reprogramming of a common AP board configuration. Presently available hardware choices could permit that sample rate to be increased, if that were useful. That will take time to emerge.
29
General discussion / Re: Getting started in the ARM world
Does anyone have interests or experience in either of the cited aspects?
It would be very, very keen to combine these features to obtain a common support and source strucure for this combo. This relative to my interest of bringing these features together for a Precision Analog application and for lab purposes in general. As of yet, the fit of other parameters for AP remain to be evaluated.
The 3D graphics are a natural application to display FFT result repetitions as a spectrogram of frequency versus time versus amplitude. Sometimes this is called a waterfall display. (Color does not interest me.)
30
General discussion / Re: Missing DSP, Audio, Analog.
It was not clear when I refered to Precision Analog (PA) that I was referring to 16 or 18 bit resolution from the A to D converter as input to DSP. My sensor case is within the audio or sub-audio frequency range with several sources MUX'ed into the A to D, thus increasing the DSP context switching and somewhat raising the required processing speed demand. Windowing of the FFT processing also adds much to the required processing speed demand. Not a known limiting factor but bears watching.
Currently a starting point with ADSP-21261 as an $8 or $10 chip clocked at 150 Mhz is a base in a 144 lead LQFP. Aside from the crunching power, it has the handy feature of providing a byte wide port for off-loading FFT output data to a microcontroller. It also has SPI, etc.. Not sure if this can be imbedded but stuffing it with a 16 bit A to D could be most useful for openers. But keeping the converter separated does allow easier adjustments for general lab uses, and this could find the 20 bit sychronous parallel input port to be strong. This all provides an opportunity for processing down into the noise levels as a possible lab interest.