Anyone work with SHARC processors ?

[diyaudio]

Hi

I'm looking at doing DSP development later this year, vie been looking into the SHARC processor platform however, the development boards and hardware tools are really really expensive 600 USD (I cannot afford it, it cost as much as a new Rigol scope  if not more) can anyone recommend or advise  an alternative development board or advise an alternative platform.

Cheaper platform mini SHARC running a ADSP21369. 185 USD (without shipping)


Alternative platform but limited the DAC is limited to 4 outputs. 195 USD (without shipping)


ADAU1442, ADAU1445, and ADAU1446 SigmaDSP Products.  600 US (without shipping)

[Hideki]

SigmaDSP is not SHARC. It's set up with a (slightly buggy) graphical drag and drop environment and you can't write programs in C for it.

While $185 may be a decent price for the devboard, don't forget that the compiler costs $3500 and the cheapest JTAG emulator is $1200.

[diyaudio]

@Hideki

I actually meant the sigma as an alternative, I understand the SigmaDSP is not SHARC, Note there is SIGMA for SHARC as well.
http://www.analog.com/en/dsp-software/sh_sigst_00/sw.html

If I skip the cost of Visual DSP++ and use the Sigma for sharc, I can then get by using the USBi cable used it to upload the firmware to the processor using this
http://www.analog.com/en/evaluation/bfext-usbi2ezb/eb.html

I understand the critical point is this,  I can only then use the custom made blocks within the sigma studio unable to design and develop custom DSP algorithms of my choice.

[Hideki]

Note that SIGMA for SHARC still needs the $1200 emulator.

I guess it depends on how much you want it to do. Maybe the ADAU chips are enough for you.

At the low end, an ARM Cortex M4F can probably do more processing than a SigmaDSP - GCC is free, ARM JTAG is usually cheap (or already on the devboards), but finding suitable devboards for audio may be hard. STM32F4DISCOVERY at ~$17 plus an external I2S audio codec board perhaps.  A10-OLinuXino-LIME is only 30 euros for an 1 GHz Cortex A8, etc.

[diyaudio]

Note that SIGMA for SHARC still needs the $1200 emulator.

I guess it depends on how much you want it to do. Maybe the ADAU chips are enough for you.

At the low end, an ARM Cortex M4F can probably do more processing than a SigmaDSP - GCC is free, ARM JTAG is usually cheap (or already on the devboards), but finding suitable devboards for audio may be hard. STM32F4DISCOVERY at ~$17 plus an external I2S audio codec board perhaps.  A10-OLinuXino-LIME is only 30 euros for an 1 GHz Cortex A8, etc.

Never worked with ARM boards I will look into the STM32F4DISCOVERY and a cheap JTAG.

[Hideki]

You don't need it for that board since it has the ST-LINK/V2 built in.

[diyaudio]

You don't need it for that board since it has the ST-LINK/V2 built in.

wow thanks they are very cheap, 20 USD, Im no expert at DSP processors I wonder how well the ARM M4 processor performs against the SHARC ADSP21369,  would like to hear what people as to say about that.
 

[dannyf]

The CM4's dsp is no match for the sharc.

[diyaudio]

The CM4's dsp is no match for the sharc.

I had a gut feeling that is the case I found this seems like alternative hardware.
http://www.openadsp.com/proclass1.asp?class1id=7
 

[Hideki]

The M4F is a designed as a microcontroller, not a DSP, so it doesn't have all the neat (but sometimes hard-to-use) functionality that can make DSPs do things faster.
It's clocked at half the rate or slower (NXP has 204 MHz parts, ST 168 or 180 MHz versus 400 MHz SHARC). While the FPU is reasonably fast it can't do anything in parallel or load from multiple memory buses. The fixed point instructions are however not that bad. It can beat SigmaDSP but not SHARC.

For a hobbyist, the total price including compiler and other development hardware usually means using a DSP is out of the question, even if they have more performance.

This PDF compares SHARC and Cortex M4F, A8 (and A9).
http://www.dspconcepts.com/sites/default/files/white-papers/2011%20AES%20-%20DSP%20vs%20Micro%20rev%202.pdf

Your "alternative" hardware looks like chinese clones of Analog Devices and I'm sure it comes with equally cloned (well, direcly copied) software as well.

[diyaudio]

The M4F is a designed as a microcontroller, not a DSP, so it doesn't have all the neat (but sometimes hard-to-use) functionality that can make DSPs do things faster.
It's clocked at half the rate or slower (NXP has 204 MHz parts, ST 168 or 180 MHz versus 400 MHz SHARC). While the FPU is reasonably fast it can't do anything in parallel or load from multiple memory buses. The fixed point instructions are however not that bad. It can beat SigmaDSP but not SHARC.

For a hobbyist, the total price including compiler and other development hardware usually means using a DSP is out of the question, even if they have more performance.

This PDF compares SHARC and Cortex M4F, A8 (and A9).
http://www.dspconcepts.com/sites/default/files/white-papers/2011%20AES%20-%20DSP%20vs%20Micro%20rev%202.pdf

Your "alternative" hardware looks like chinese clones of Analog Devices and I'm sure it comes with equally cloned (well, direcly copied) software as well.

@Hideki

Many thanks for the PDF will go through this later ! Not many chinese clones of Analog Devices available on ebay search yields rather shy results.
 

[Hideki]

I just noticed this new product that they say can "execute up to three times more audio processing algorithms than its predecessors".
http://www.analog.com/en/audiovideo-products/audio-signal-processors/adau1452/products/product.html

Maybe an alternative if a cheap devboard appears.

[legacy]

blackfin is too tiny for you ?

[diyaudio]

I just noticed this new product that they say can "execute up to three times more audio processing algorithms than its predecessors".
http://www.analog.com/en/audiovideo-products/audio-signal-processors/adau1452/products/product.html

Maybe an alternative if a cheap devboard appears.

uhmm not bad. 225 USD for the dev board.
http://www.digikey.com/product-detail/en/EVAL-ADAU1452MINIZ/EVAL-ADAU1452MINIZ-ND/4571726

http://www.bdti.com/InsideDSP/2014/02/05/ADI

[diyaudio]

blackfin is too tiny for you ?

wayyyyy to expensive for me.

[legacy]

it may cost less than $200, i payed $70 USD for a tiny BF532, brand new!
the problem is … you can't use VisualDSP++ with it, cause the cheapest jtag is $200 USD
i mean the cheapest jtag which VisualDSP++ v5 will accept and use
you could buy an open source jtag, e.g. ICE bear, which costs $50, but .. you have to forget to be able to use it with VisualDSP++
so you have to use opensource toolchain, that's not fun at all

anyway, if you want a very cheaper BF532 board i can suggest you a canadian seller

[diyaudio]

it may cost less than $200, i payed $70 USD for a tiny BF532, brand new!
the problem is … you can't use VisualDSP++ with it, cause the cheapest jtag is $200 USD
i mean the cheapest jtag which VisualDSP++ v5 will accept and use
you could buy an open source jtag, e.g. ICE bear, which costs $50, but .. you have to forget to be able to use it with VisualDSP++
so you have to use opensource toolchain, that's not fun at all

anyway, if you want a very cheaper BF532 board i can suggest you a canadian seller

Sure why not throw me the seller,im almost certain the  EVAL-ADAU1452MINIZ should be okay.. 

Why does the analog guys make things so hobby unfriendly .

[legacy]

Sure why not throw me the seller

this board is made by soc-robotics and it is sometimes sold on ebay at cheaper price (i bought 2 of units at $70 each), unfortunately it has no jtag included and it is using an AVR644 to load the blacken's binary code through the serial port into ram in order to burn the spi flash which is shared by blackfin and AVR644: it's done sector by sector, and it's a bit slow, but cheap.

[JoeN]

Why does the analog guys make things so hobby unfriendly .

With regard to microcontrollers, only, right?  Everything else seems to have liberal sample policies and the ICs themselves are as easy to work with as anyone else's.  They have a lot of interesting parts.  But yeah, they make entry into their microcontrollers nearly impossible for a hobbyist, which is 100% the opposite of Atmel, Microchip, TI, ST, Silicon Labs, Zilog, Freescale, and Xilinx and Altera if you are counting FPGAs.  Everyone else has a cheap programmer and free (sometimes size limited or optimization limited) software at this point.

[diyaudio]

Curiosity..
 
Can anyone provide further input, how will an FPGA  match against the SHARC ?

I came across the video.

[rsjsouza]

Not sure if you are still looking for alternatives, but did you look at the floating point C6748 DSP from TI? It has much cheaper boards and free SW (here and here). Also, the IDE (Code Composer Studio) is free depending on the emulator you choose.

[Hideki]

An FPGA makes sense when you want to duplicate the same type of processing N times, like you do in a mixer, but writing HDL instead of C code is very very different.

RME uses FPGA in their products.

[paf]

An FPGA makes sense when you want to duplicate the same type of processing N times, like you do in a mixer, but writing HDL instead of C code is very very different.

RME uses FPGA in their products.

No, No, No.

On a an FPGA, you can do what type of hardware blocks you need, and you can to different types of blocks with different sizes if you want.  If  you need  32 counters with sizes from 32 to 64 bits (all different) you can.

You need  20 FIFOs all with different widths, and with different sizes, you can do it.

[Hideki]

Yes, yes, yes. I'm talking about doing DSP-like processing - not counters and FIFOs. My guess would be that the FPGA does indeed handle lots of channels in parallel and large mixing matrixes instead of implementing completely custom circuitry per channel. That appears to be the RME way as well.

DiGiCo has three TigerSHARC DSPs on the board they display in their video, so there is absolutely some separation between what does what.

It's easier to run different code on the DSPs than to reconfigure the FPGA every time you want a different effect.

[SArepairman]

before yoou make your decision check the internet for cracked programs. I got pretty upset at the numbers that came up in this thread.

if you are a hobbyist there is nothing wrong with cracking a 10,000$ program aimed at a company that has employees and shit being paid ten times that per year.

its senseless to use the "free" tools which are often buggy crap (like I switched from pirate Eagle cad to pirate altium  ). If I continued to use eaglecad I would probably be dead by now.

its not like you are mugging people at the bus stop  and the company you are "stealing" it from does not give 0.00001% of a shit about you.

spend your money on nice test equipment and components instead. stuff that would actually hurt people to steal.

this specifically made me upset:
"
For a hobbyist, the total price including compiler and other development hardware usually means using a DSP is out of the question, even if they have more performance.
"

the fuck should I have to use some ghetto technology because of a software license 

people need to get on cracking this shit ASAP

if you sell the chip for 50$ and you expect me to pay 5g for the software you are smoking crack

[nctnico]

This old thread on Dsprelated pretty much sums it up:

http://www.dsprelated.com/showmessage/143215/1.php

It seems AD isn't interested in hobbyists messing with their high end DSPs.

[dannyf]

-you are a hobbyist there is nothing wrong with cracking a 10,000$ program aimed at a company that has employees and shit being paid ten times that per year.-

wow.

[legacy]

what do you think about the SOC robotics's Blackfin boards ?

[diyaudio]

This old thread on Dsprelated pretty much sums it up:

http://www.dsprelated.com/showmessage/143215/1.php

It seems AD isn't interested in hobbyists messing with their high end DSPs.

@nctnico

Thanks for the link. Actually I got the feeling that Analog Devices are making hobbyist  feel like an "outcast pesent" during any attempt to gain accessing to their Sharc/Blackfin processor line. I must admit, I did came across some cracks too visual DSP++ and some hardware work around (strange they all were from china), its just to much effort and time effort to follow through.
 

[diyaudio]

before yoou make your decision check the internet for cracked programs. I got pretty upset at the numbers that came up in this thread.

if you are a hobbyist there is nothing wrong with cracking a 10,000$ program aimed at a company that has employees and shit being paid ten times that per year.

its senseless to use the "free" tools which are often buggy crap (like I switched from pirate Eagle cad to pirate altium  ). If I continued to use eaglecad I would probably be dead by now.

its not like you are mugging people at the bus stop  and the company you are "stealing" it from does not give 0.00001% of a shit about you.

spend your money on nice test equipment and components instead. stuff that would actually hurt people to steal.

this specifically made me upset:
"
For a hobbyist, the total price including compiler and other development hardware usually means using a DSP is out of the question, even if they have more performance.
"

the fuck should I have to use some ghetto technology because of a software license 

people need to get on cracking this shit ASAP

if you sell the chip for 50$ and you expect me to pay 5g for the software you are smoking crack

There are cracks floating around for visual dsp++ (all from china).

[Marco]

Yes, yes, yes. I'm talking about doing DSP-like processing - not counters and FIFOs. My guess would be that the FPGA does indeed handle lots of channels in parallel and large mixing matrixes instead of implementing completely custom circuitry per channel. That appears to be the RME way as well.

FPGAs are horrible at floating point, a FPGA with lots of multiplier blocks might be okay at fixed point ... but why go through all that pain? It's performance/Watt is still going to be poor ... and if you don't care about performance/Watt just get one of the many ARM boards, most of the ARM processors have DSP extensions.

[Hideki]

Well, someone obviously found out that FPGAs are faster than DSPs for doing certain things. We just don't know the finer details of what those things are.

And we don't even know what diyaudio wants to use the DSP-like processing for

[nctnico]

I agree. If he gets a board like this he has 4 ARM cores running at 800MHz and audio in/out:
http://uk.farnell.com/freescale-semiconductor/mcimx6q-sl/imx6q-sabre-lite-dev-board/dp/2124398

[Marco]

Could just use any board with USB audio, ODROID-U3 is probably best for FLOPS/$.

[legacy]

There are cracks floating around for visual dsp++ (all from china).

i bought a blackfin DSP board on ebay and i got also a CD inside the parcel with a cracked copy of VisualDSP++ V5
i was shocked: the seller said that they are used to provide it this way in order to use the boards they make X_X

fortunately i got a regular licensed (old v4.5) version of the AnalogDevices suite

btw, the problem is jtag: B1000 series is the only one cheap i could find compatible with VisualDSP++, but … it requires v5 + an update
so … what the china-guy has sold me ... is the only way to "test" the jtag cable i bought, 'till i will buy an update for the whole suite.

[nctnico]

There are cracks floating around for visual dsp++ (all from china).

i bought a blackfin DSP board on ebay and i got also a CD inside the parcel with a cracked copy of VisualDSP++ V5
i was shocked: the seller said that they are used to provide it this way in order to use the boards they make X_X

fortunately i got a regular licensed (old v4.5) version of the AnalogDevices suite

btw, the problem is jtag: B1000 series is the only one cheap i could find compatible with VisualDSP++, but … it requires v5 + an update
so … what the china-guy has sold me ... is the only way to "test" the jtag cable i bought, 'till i will buy an update for the whole suite.

Why would you buy the whole suite for hobby purposes? Anyway selling cracked copies is 'how they roll' in most of Asia. There are many shops where you can buy every CAD and software development software package a few dollars.

[diyaudio]

Well, someone obviously found out that FPGAs are faster than DSPs for doing certain things. We just don't know the finer details of what those things are.

And we don't even know what diyaudio wants to use the DSP-like processing for

For now, its purely experimental audio DSP related. (I do admire the more powerful boards as educational suggestion, FPGA (although the SHARC would have been a great platform to stick to)
Counter productive aspects are the development of the DSP algorithms, IMHO I guess AD made up with their unfair policy  by introducing a budget sigma ADAU1452 development board, this tiny processor is mean (295-MHz) and a powerful audio routing matrix, compared to its predecessors it seems like the best in its category. (The guys at minidsp are probably rubbing their chin with ideas)

They even demonstrated this at CES


AD still seems to be sneaky in this area, sneaky as the communication uses a commercial proprietary USBi interface to upload the code to the processor, so there is no other way to change the
program firmware in production (from what I see) and a micro-controller interface to manipulate the parameter registers.

I would like to save up for a low level DSP board maybe the C6000 or one of the already suggested models depending to my budget, FPGA is out of the question for me its just to to of a complicated specialized area for me. (its interesting to see what others has been saying in this area though)


   
 

[nctnico]

TI has similar software to create chains of DSP algorithms. I have tried it to see if it was useful for a project. I'm not convinced it saves any time compared to writing software in C using a DSP library. The TI software had several serious limitations on how the blocks could be used. Also some algorithms like the echo canceller worked very poor. In the end most people will want to be able to write their own algorithms.

[diyaudio]

TI has similar software to create chains of DSP algorithms. I have tried it to see if it was useful for a project. I'm not convinced it saves any time compared to writing software in C using a DSP library. The TI software had several serious limitations on how the blocks could be used. Also some algorithms like the echo canceller worked very poor. In the end most people will want to be able to write their own algorithms.

@nctnico

Cannot comment on the TI processor designer, but the AD sigma series has been used in a few high grade audio products to name a few.
 
Hypex DSP2W DSP


B&W Zeppelin Air


I think its safe to assume its safe for reliable "audio" solutions. However you got a point, designing your own custom DSP algorithms is still preferred, although I find dsp math beyond me as its fairly complex and aggressively mathematically no matter what processor one end up using, this area is very deep and complex and requires loads of time and patience to truly master (like anything worth doing in) bah!
 

[legacy]

Why would you buy the whole suite for hobby purposes?

i bought the full suite in order to run my desk, i am an electronic consultant, so i am required to buy the tools i need for that.
I bought a cheap board for hobbies porpoise and i got an hacked copy of the suite inside

[rsjsouza]

In the end most people will want to be able to write their own algorithms.

Yes, I used that for a while but I agree that I preferred instead to design the filters in a math pack such as Matlab/Octave and export the data to my own algorithm on the DSP...

Cannot comment on the TI processor designer, but the AD sigma series has been used in a few high grade audio products to name a few.

Bose has some TI content on it; this article says it uses a previous family (C6713) and the clock specs on this page are too similar to the C6748's sister device OMAPL138 - not sure if this is the only device with 456MHz, though.

[Berni]

If you really want a DSP the C6000 family from TI is not too bad. There are boards with them, sometimes they might be ARM+DSP but you can use the DSP on its own. C compiler is free for it and I bought a JTAG to USB clone for it for about 15 bucks from Blackhawks or someone. But keep in mind that these large DSPs are not the easiest thing to program.

But yeah a FPGA is a very good way of doing DSP too, for very parallel algorithms they can fly. But programming and debugging them is annoying as there is no single steeping trough code and looking at internal memory can be difficult.

[diyaudio]

I found what seems like a viable alternative as it does ADI Blackfin & SHARC processors, seems like the only decent clone available, from china of course, I cannot find it on Ebay though. I've been asking some of my sellers on ebay to stock it. Cost $350 not to bad compared to $1,350 for the original.

* 5MHz JTAG TCLK frequency
* Support ONLY ADI Blackfin & SHARC processors
* USB 2.0 interface enabling download speeds of up to 255 KB/sec
* 1.8V, 2.5V, and 3.3V compliant and tolerant
* Multiprocessor support
* IEEE 1149.1 14-pin JTAG connector
* PlugFest Certified USB Bus-powered design
* CE-certified
* ROHS Compliant
* Solder points for attaching a wire to reset the target board
* ADZS-HPUSB-ICE's JTAG design
* ADZS-HPUSB-ICE's Interface protection circuit design




The video exists but its not in English, what I understand they demonstrating how it works on the EZKIT and various other Blackfin, SHARC demo boards. 
http://www.openadsp.com/english/product_detail.asp?proid=6

Let me know your thought around this one.