AIC3254 MiniDSP Instruction Set/Programming

D. Hartl said:

The instruction set of the miniDSP is not published. TI recommends using Purepath Studio to build processing blocks for the miniDSP.

You should at least correct the reference in the datasheet so users won't go wandering around looking for a non-existent table :) Are there any plans on releasing the instruction set (why it is not published btw)?

Yes, i am also agree. Why they are not published yet. It could be very interesting to see this stuff.

Is there at least a description of the mini-DSP architecture.  I recall 1152 instructions per sample, but is there anything more?  Plus how complex could this instruction set be?

Hi, Guys,

It is a strategic decision by the management (at a paygrade WAY higher than mine!) of TI to not disclose this information... Sorry...

-d2

Dan,

1. In the short term is there an NDA or some other process I can go through to get relevant information?

2. Longer term, please ask for a re-consider.  I'm trying to make design decisions and it makes a difference on power, probably cost, definitely space, if I need to add a separate DSP vs using the capabilities of the codec.  Is there some competitive disclosure issue?

3. Maybe for this and other things you should have the equivalent of Apple's MFI program, where a company and/or individual has to register and sign serious NDA agreements, etc. but then has access to important design information.

Thanks,

David

I have to agree with David here, having a NDA/MFI style program is a good idea. Restricting access to miniDSP instruction set won't stop your competitors from creating a similar product, and there ARE alternatives, the bad thing is that their management thinks similary. 

Not publishing miniDSP programming information is an upper management decision...  WHY????  I hate relying on a company's development software for unimaginative, out-of -the-box generic algorithms, and having no other options!!!  I want it to double as a math co-processor for my microcontroller, among other things!  I pride myself on being a very creative algorithm designer and assembly language programmer, capable of amazing engineering with a minimum of resources, BUT ONLY IF ALLOWED TO DO SO!!!  TI, how can people shine if you don't allow it??  Are all of your customers deemed dumber than your own engineers, so why bother???  Given what your development environment offers, I'd rethink that one!!!  Are your products much less special than you advertise, so don't dare let your competitors see details??  Management decisions like this HOLD US BACK!!!  Let us be amazing with your products by giving us the needed resources!!  Don't force your long-time customers to go looking at THE VERY NICE KALIMBA DSP from CSR (Cambridge Silicon Radio), or better yet, beg Freescale to give us ultra-small, ultra low-power mobile DSP56300 derivatives!!!  Actually, that IS the best option for all of us given the longevity, code-base, huge wealth of information, and lovely parallel assembly language!!!  Come on, everyone, off to Freescale to beg for the DSP56300 MicroDSP Family!!!!!!!!  Long Live DSP 56K!!!!!!!!!!!!  Thanks for pissing me off yet again, TI!!!!  It's all for the best!!!

- Adam Roman

Adam,

I wish we had a <Like> button...

-d2

I like your response, d2!!  I like TI just fine, how else would I know to be so disappointed when it doesn't play nice??  TI engineering, like.  TI management, no like.  YOU control the <Like> button, don't you know that??  So, ummm... where's our instruction set info???  I don't see it yet.  Don't make me come out there!!!  ;~P

-Adam Roman

P.S. Freescale doesn't care about my comments either.  No new DSP56300 roadmap.  You win, TI.  For now...

Long Live TMS9900!!!  Oh, wait...  Nevermind.

Stop nagging them, they're in rocks already. Owing to their wise "upper management," perhaps.

I'm critically disappointed. The company which is famous for its openness for developers takes curious advances in the area that has never been their latest tribute. DSP software must be open, or else you end up with certain companies that have no chance to survive whatever the DSP technology feats you're achieving.

Sorry.

I'll nag all I want, it's been over a year, and this issue is still the same stinking pile of crap.

The idiot-oriented PurePath Studio environment does me NO GOOD whatsoever, as I have an application that is actually CREATIVE and DIFFERENT.  It is entirely reasonable and possible, but must be accomplished on the assembly level with a custom OS, which is one of my specialties, IF I HAVE THE INFO I NEED!!

And now there are even more CODECs with miniDSPs we don't have full access to like the AIC3263.  TI management, I still hate you for this nonsense.  Why not make this info NDA-based or open source already???  To protect your precious algorithms?  Guess what, they're not that special.

I encourage people to hack the miniDSP and publish the findings so people can get on with some different applications of these otherwise fine products.

And TI should thank whoever does this for taking initiative to correct their management mistake.

Adam, I run in your same difficulty: not been able to assy program the AIC3254 at once.

Have you thought of a similar chain: ADC FFT filter chain FFT-1 DAC ?

Any good DSP should do.

Andrea Modiano