Selecting a DSP for audio effects (Delay, Echo, Chrous, Flanger etc)

[iampoor]

Hi

Recently Ive been interested in learning more about dsp's and audio processing. I amspecifically interested in creating delays, echos, chorus, flangers etc. for guitar and pro audio applications. I think it would be a cool way to make music, and gain more knowledge of electronics. 

Now this leaves me in quite an awkward position. I honestly am not sure where to start as far as what development board to start out on. There are so many options! Most designed for audio seem to optimized for decoding audio, surround sound etc. TI released a paper in 2005 about coding audio effects, but it appears to only work with a few specific dsp's that are pricey and the development board is 2000$ (Worth more then my car! haha). I have read some articles and understand the basics, Im just notsure how to get started. ANy good places, or lines to look at?

I know this is a a very vague question, but honestly my google searches have just led me to a bewlildering array of options, and it hasnt been very productive. I would like to find a book or some great articles and a reasonably priced but reasonably powered development board to start, unless anyone has a better idea.

[Someone]

Most usable solutions will cost hundreds or thousands of dollars, as you'll want an active community around it to help you. One option might be to get a Tiptop Z-Dsp and the numberz programmer/development header:
http://www.tiptopaudio.com/numberz.php?goto=features

The spin DSP is a bit old but you get a ready to go platform and a customer base to sell designs into.

[Marco]

32 kHz sampling and a tiny underpowered DSP ... why still do something like this in this bloody decade. It's just like SigmaDSP.

Hardware which in this day and age only belongs in devices running on button cells.

[Richard Crowley]

http://spinsemi.com/products.html

[iampoor]

Most usable solutions will cost hundreds or thousands of dollars, as you'll want an active community around it to help you. One option might be to get a Tiptop Z-Dsp and the numberz programmer/development header:
http://www.tiptopaudio.com/numberz.php?goto=features

The spin DSP is a bit old but you get a ready to go platform and a customer base to sell designs into.

Thats a very cool system, but it seems limiting, in the fact that I couldnt make designs on it and carry them over into products. I okay with spending some money on developent tools, I just dont want to spend a fortune, Ill pay what I have to pay, the few development boards I have found just seem to be packed with features for a different market. All I would really want in the dsp, support circuitry, and maybe a dac and adc on board.

32 kHz sampling and a tiny underpowered DSP ... why still do something like this in this bloody decade. It's just like SigmaDSP.

Hardware which in this day and age only belongs in devices running on button cells.

The spindsp?

http://spinsemi.com/products.html

Wow, this looks super cool! Only downside is r. Barr seems to have passed away 4 years ago. I wonder how much longer the dsp will be supported and manufactured....

[old gregg]

I've always been interested in Spin products, without going futher though.

sadly Yes Keith Barr died few years . He had quite an interesting past in the audio side of electronics.

There's also the SHARP dsp used in few products (the one I remember is "Strymon" which uses it for an all line of product). But I think the dev board is really expensive.

There's another one that I don't remember the one is seem simpler to use. I'll do more research on it.

[Marco]

Last time I did audio DSP was in school on a low cost floating point DSP ... this was 2 decades ago. I implemented overlap add FFT filtering for efficient filtering with very long filters (not zero latency though, one because it was too hard for me and also it was still patented then ... patent has since lapsed obviously, I think there is some source code out there too, not that it can run on the archaic sigma/spin DSPs).

That the most popular audio DSP platforms aren't much cheaper than the starter kit I used then AND aren't even capable of doing what it did at the time boggles my mind.

Personally I'd look at LV2 plugin programming on Linux on your PC. If you want to run them in a standalone application you can use an ARM SBC like the 50$ A20 boards.

[Hideki]

At the low end (well, it is £249) there's the OWL http://hoxtonowl.com/

If that is still too expensive you will have to roll your own. Marco's suggestion of getting a cheap Cortex A board of some sort is a good one. They usually come with terrible audio, so you might need to add a separate I2S codec.

Until you have decided on the hardware... it's easier to start prototyping the software on a PC, using whatever is the appropriate plugin standard for your OS. VST for Windows, AU for Mac, LV2 or others for Linux.

[iampoor]

I've always been interested in Spin products, without going futher though.

sadly Yes Keith Barr died few years . He had quite an interesting past in the audio side of electronics.

There's also the SHARP dsp used in few products (the one I remember is "Strymon" which uses it for an all line of product). But I think the dev board is really expensive.

There's another one that I don't remember the one is seem simpler to use. I'll do more research on it.

Have you read through the whole website? Lots of good info!

Yes, definitely! He ust have made quite a pretty penny being a founder and owner of two world class companies...

What DSP do they use? Id be interested to know, just for comparisons sake. Plus I want to know how powerful the Strymon effects are! haha

Cool, let me know what you find.

Last time I did audio DSP was in school on a low cost floating point DSP ... this was 2 decades ago. I implemented overlap add FFT filtering for efficient filtering with very long filters (not zero latency though, one because it was too hard for me and also it was still patented then ... patent has since lapsed obviously, I think there is some source code out there too, not that it can run on the archaic sigma/spin DSPs).

That the most popular audio DSP platforms aren't much cheaper than the starter kit I used then AND aren't even capable of doing what it did at the time boggles my mind.

Personally I'd look at LV2 plugin programming on Linux on your PC. If you want to run them in a standalone application you can use an ARM SBC like the 50$ A20 boards.

Arnt most "audio DSP" fairly crippled? Most audio DSP's for running reverb/echo algorithms seem to be pretty standard, but with some added instructions and a fast memory buffer. I have had a hard time wading through "standard" Dsp's that would also be good for this purpose and not hobbled down with extra features...

Ill look into that! Any  recommendations on where to start? The A20 looks extremely fast and powerful!  O0

At the low end (well, it is £249) there's the OWL http://hoxtonowl.com/

If that is still too expensive you will have to roll your own. Marco's suggestion of getting a cheap Cortex A board of some sort is a good one. They usually come with terrible audio, so you might need to add a separate I2S codec.

Until you have decided on the hardware... it's easier to start prototyping the software on a PC, using whatever is the appropriate plugin standard for your OS. VST for Windows, AU for Mac, LV2 or others for Linux.

The OWL is awesome, but Im eventually planning on building my own hardware interface for it (Thats what I experienced in). Hence Ive been trying to find a dsp first, to base a hardware box around and then know what my limitations are.

What makes their standard audio performance terrible?

[Marco]

Ill look into that! Any  recommendations on where to start?

Not really, LV2 programming is what I would look at ... dunno if it would turn out to be a good idea though

If you need very low latency (ms range) for stuff like stomp boxes and studio mixing I don't know if the A20 would work well, you could probably get preempt_rt to run on it ... but I have no idea how low you can get the latency that way.

For very low latency you might want to use boards which are better documented and can more easily be programmed close to the metal like the ones with processors from TI/Freescale ... or the Raspberry Pi of course (someone already made a very rudimentary LV2 host straight to the metal for it).

[Yago]

Until you have decided on the hardware... it's easier to start prototyping the software on a PC, using whatever is the appropriate plugin standard for your OS. VST for Windows, AU for Mac, LV2 or others for Linux.

Could go even further down that route and forgo any hardware build:
http://www.ebay.com/itm/SM-PRO-V-MACHINE-VMACHINE-STANDALONE-VST-VSTi-PLAYER-WITH-V2-FIRMWARE-NEW-/221414762044?pt=AU_Musical_Instruments_Instruments&hash=item338d59263c
Or something of that ilk.

[kripton2035]

http://spinsemi.com/products.html

nice products. thanks for the link.

[Hideki]

What makes their standard audio performance terrible?

Little attention is paid to the ground and power for the codec, so there can be noise that changes a bit depending on what the processor is doing.
Some chips even have the adc/dac inside the processor, which is a bad place to be for the poor analog signals. Some only have a mic level input (or maybe no input at all).

Or you might get lucky and find something that's good enough

[iampoor]

Ill look into that! Any  recommendations on where to start?

Not really, LV2 programming is what I would look at ... dunno if it would turn out to be a good idea though

If you need very low latency (ms range) for stuff like stomp boxes and studio mixing I don't know if the A20 would work well, you could probably get preempt_rt to run on it ... but I have no idea how low you can get the latency that way.

For very low latency you might want to use boards which are better documented and can more easily be programmed close to the metal like the ones with processors from TI/Freescale ... or the Raspberry Pi of course (someone already made a very rudimentary LV2 host straight to the metal for it).

Gotcha.

I realized I didnt pout this in the OP, but I a looking at building stompboxes. 

Im a pretty big noob at programming, so I just want to clarify a few things (please dont laugh  )
LV2 programming, would be programming in a language "above" assembly, correct? Programs would be written in say, C++ and then complied down into assembly right?

Im guessing the main advantage of this is that if process architecture changes, the program can simply (I using this term "lightly" be loaded into a different compiler and still be compatable as long as the new chip has the same/better specs.

The disadvantage is that if have too much latency, then your boned right? 

Until you have decided on the hardware... it's easier to start prototyping the software on a PC, using whatever is the appropriate plugin standard for your OS. VST for Windows, AU for Mac, LV2 or others for Linux.

Could go even further down that route and forgo any hardware build:
http://www.ebay.com/itm/SM-PRO-V-MACHINE-VMACHINE-STANDALONE-VST-VSTi-PLAYER-WITH-V2-FIRMWARE-NEW-/221414762044?pt=AU_Musical_Instruments_Instruments&hash=item338d59263c
Or something of that ilk.

Yikes, thats pretty cool! Definitly not what I personally looking for, but could come in handy.  O0

What makes their standard audio performance terrible?

Little attention is paid to the ground and power for the codec, so there can be noise that changes a bit depending on what the processor is doing.
Some chips even have the adc/dac inside the processor, which is a bad place to be for the poor analog signals. Some only have a mic level input (or maybe no input at all).

Or you might get lucky and find something that's good enough

Ah! Thanks for the heads up! How can I evaluate the performance of the audio codec fro a datasheet? Or is it more of a ... "hopefully you get lucky" type situation? I guessing that any reasonably designed chip would have separate power and grounds for the audio/digital side. If the manufacturer is honest in their specs, Im presuming the dynamic range would be crippled by any sort of digital noise.

Oh wait, so this is why their is an "audio DsP category, right? 

[Marco]

LV2 programming, would be programming in a language "above" assembly, correct?

You'll generally be programming in C++ (or C with C++ comments, pure ANSI is only for the anal).

LV2 is a plugin interface, when you incorporate that interface in your code you can load the resulting plugin into a host ... the host will then pass on data from midi keyboards or audio in etc to your plugin, and take the resulting audio output from the plug and pass it on (for instance to another plugin, or to the DAC). Generally the host will be some type of editing/compositing software, but it can also be a minimal program which just connects it to the JACK audio interface or even directly to the ADC/DAC/etc like the to the metal Raspberri Pi software I mentioned.

VST is another plugin interface, originally meant for Cubase but there are many hosts now including stuff like the SM PRO V (although it doesn't have stereo in and I see a lot of complaints about it's computational power) or if you have a lot of money on something like the Receptor.

LV2 is a little more linux, enthusiast oriented ... VST is a little more windows/Mac, professional oriented.

Ah! Thanks for the heads up! How can I evaluate the performance of the audio codec fro a datasheet? Or is it more of a ... "hopefully you get lucky" type situation?

The latter, although if you use a popular board you can just Google/ask other's opinions of course.

I guessing that any reasonably designed chip would have separate power and grounds for the audio/digital side. If the manufacturer is honest in their specs, Im presuming the dynamic range would be crippled by any sort of digital noise.

Optimist ... also they don't control the layout/decoupling on the SBC.

[diyaudio]

Hi

Recently Ive been interested in learning more about dsp's and audio processing. I amspecifically interested in creating delays, echos, chorus, flangers etc. for guitar and pro audio applications. I think it would be a cool way to make music, and gain more knowledge of electronics. 

Now this leaves me in quite an awkward position. I honestly am not sure where to start as far as what development board to start out on. There are so many options! Most designed for audio seem to optimized for decoding audio, surround sound etc. TI released a paper in 2005 about coding audio effects, but it appears to only work with a few specific dsp's that are pricey and the development board is 2000$ (Worth more then my car! haha). I have read some articles and understand the basics, Im just notsure how to get started. ANy good places, or lines to look at?

I know this is a a very vague question, but honestly my google searches have just led me to a bewlildering array of options, and it hasnt been very productive. I would like to find a book or some great articles and a reasonably priced but reasonably powered development board to start, unless anyone has a better idea.

The best audio DSP processor for audio applications (based on industry standards) is a SHARC DSP this is if you are serious... I bought the emulation kit  (ADZS-HPUSB-ICE) clone for $500 (the real deal cost $1,200), if you serious and want to develop your own algorithms you would use something like this.

Secondly, if you on a budget I would recommend the ADAU1452 its a 300MHZ, powered by sigma studio a drag drop widows IDE where you drag modules like delays, chorus, filters, compressors ect... and wire them up in a signal chain fashion, the parameters on the chip once programmed, can be controlled via a micro-controller, the only disadvantage is you cannot write custom audio algorithms for it...



 
   

[Legion]

The best audio DSP processor for audio applications (based on industry standards) is a SHARC DSP...

I don't know much about DSPs, but I play guitar and I own an Axe FX II which is arguably the best amp modeler and FX processor you can get right now. The Axe FX II uses two TigerSHARC DSPs, so that's the direction I'd go if the budget allows.

[Marco]

It's just a processor, there is no magic in it ... one floating point multiply add is as good as any other.

[diyaudio]

It's just a processor, there is no magic in it ... one floating point multiply add is as good as any other.

until you use the tool chain.

[lpc32]

I've no idea, but here are some boards for less than the price of a used car.
http://www.analog.com/en/processors-dsp/blackfin/content/blackfin_evaluation_kits/fca.html

[Marco]

until you use the tool chain.

Hard to beat x86 there.

[ampdoctor]

Accutronics/Belton makes a DSP "brick" that might not be a bad place to start if you're just getting your feet wet. I don't really know how much about it or how much flexibility it has but just putting it out there as a low cost starter alternative you might want to investigate for around 50 bucks or so.  Just google BTSE-99FX

[old gregg]

I've almost missed it but there's a french modular synth manufacturer that makes most of his products with simple AVR microcontrollers !I think this guy is really talented (as I understood he worked in software) he makes really great stuff sound out of what can be considered as "simple" 8bits computer. I Built one of his kit (shruthi) and he put a lot of effort and great though in such 'small" device, with a really simple user interface. all the source codes (in C++) and the hardware are open-source : http://mutable-instruments.net/

this may give you ideas of what can be possible without going into big money DSP.

 There's also one of the previous guest of "the amp hour", Todd Bailey who created a digital sampler with an 8bit avr micro-controller : http://narrat1ve.com/

(his video synth seems to be a killer one as well !)

However, for simpler effects as those described in the tittle, analog circuits remain valid. You can buy BBD chips or a Reverb tank for less money than a Dev board. Few op amps are you're done.

Otherwise, Spin semicondutor FV1 seem to be a cheaper solution for a digital solution. Some products are based on it (mostly from Musical instrument manufacturer).