Beginner's First DSP Study Board

[technix]

For a beginner in DSP designing a board for their first experiments with a proper DSP, what should be the target specs?

Or in my case:

1. ADSP-BF531 or TMS320VC5509A? I have samples of both chips.
2. Maxed out SDRAM? A healthy amount of Flash (4MB parallel NOR or 32MB SPI NOR?)
3. How to get the signal in and out of the chip? High speed ADC/DAC plus BNC's? I2S?
4. Will a Rigol DS1104Z be adequate at seeing what came out of it, and a home Hi-Fi system adequate at hearing it?
5. What are the other points I need to mind?

[diyaudio]

3. How to get the signal in and out of the chip? High speed ADC/DAC plus BNC's? I2S?

Simplest yet the toughest setup is a talk thru as its the full system chain without processing.     

[technix]

3. How to get the signal in and out of the chip? High speed ADC/DAC plus BNC's? I2S?

Simplest yet the toughest setup is a talk thru as its the full system chain without processing.     

That is the problem: I/O. How to implement that?

[diyaudio]

3. How to get the signal in and out of the chip? High speed ADC/DAC plus BNC's? I2S?

Simplest yet the toughest setup is a talk thru as its the full system chain without processing.     

That is the problem: I/O. How to implement that?

For simple stuff assuming audio, I cannot comment on high speed ADC applications.
Use a Audio Codec Chip. example...WM8731
https://d3uzseaevmutz1.cloudfront.net/pubs/proDatasheet/WM8731_v4.9.pdf

https://www.mikroe.com/audio-codec-proto-board

[SiliconWizard]

I would use I2S and interface with any ADC/DAC board that is fit for the purpose. You can find many third-party ADC/ DAC boards with I2C interfacing. That's more flexible. You may also add SPDIF and/or optical I/Os.

I don't know the TI DSPs very well, and the ADSP line is rather good, so I'd tend to favor those, but I'm biased. One thing to consider before choosing is the development tools. How expensive are they? Etc. As far as I remember, the ADSP dev tools were expensive. That may have changed.

As for the Rigol oscilloscope, it's more than adequate to debug your board and see what's going on. Don't expect to use it to precisely measure distortion figures, though.

[technix]

I would use I2S and interface with any ADC/DAC board that is fit for the purpose. You can find many third-party ADC/ DAC boards with I2C interfacing. That's more flexible. You may also add SPDIF and/or optical I/Os.

I don't know the TI DSPs very well, and the ADSP line is rather good, so I'd tend to favor those, but I'm biased. One thing to consider before choosing is the development tools. How expensive are they? Etc. As far as I remember, the ADSP dev tools were expensive. That may have changed.

As for the Rigol oscilloscope, it's more than adequate to debug your board and see what's going on. Don't expect to use it to precisely measure distortion figures, though.

I don't have ADC/DAC boards, but I have the chips.

Will this arrangement for ADSP-BF531 work, using Raspberry Pi as the optional host controller and one of the I2S targets, and a codec on the other I2S:

* Form factor: Raspberry Pi HAT
* 64MB SDRAM (MT48LC32M16)
* Signal switches for selecting SPI master/slave boot modes (so it can work with or without the Pi)
* CS4272 or WM8978 on one of two I2S ports, controlled using SPI
* Raspberry Pi I2S on the other I2S port
* Boot ROM device: W25Q128FVSIG (16MB SPI)
* Should I just shut the PPI in, or what should I bring it out to?

If I get the Pi involved it can be used to inject code (using the slave SPI port on the DSP) and test data. Actually on the Pi there is shairport-sync software that implements an AirPlay target, allowing me to stream music into it from my iPhone. With a DSP in the middle it would be interesting to meddle with, for example, those Taylor Swift songs.

Also what is the TMS320VC5509 equivalent?

[rsjsouza]

technix, I can't comment on the BF, but there are a few resources for the VC5509A:

I would check what is included in the development kits that used to be sold: the DSK5509A and the EVM5509A Plus - I used the EVM kit which has more connected peripherals (SD card, display, USB slave, etc.) but some of the parts there may be a bit outdated - the codec TLV320AIC23, for example, has average specs for today's standards.

Also, this is a guide for some pre-defined EDA symbols and some design rules that may be interesting.
http://www.ti.com/lit/an/spraa30/spraa30.pdf

[technix]

technix, I can't comment on the BF, but there are a few resources for the VC5509A:

I would check what is included in the development kits that used to be sold: the DSK5509A and the EVM5509A Plus - I used the EVM kit which has more connected peripherals (SD card, display, USB slave, etc.) but some of the parts there may be a bit outdated - the codec TLV320AIC23, for example, has average specs for today's standards.

Also, this is a guide for some pre-defined EDA symbols and some design rules that may be interesting.
http://www.ti.com/lit/an/spraa30/spraa30.pdf

The kits for the 5509A is surprisingly expensive here on Taobao... It might be cheaper if I cook my own.

Do keep in mind that this is actually my second DSP project, since there will be a SigmaDSP-based Bluetooth headphone receiver project before that, and that DSP really handles just some equalizing and mixing, so those big chips are not warranted.

[coppice]

For a beginner in DSP designing a board for their first experiments with a proper DSP, what should be the target specs?

Or in my case:

1. ADSP-BF531 or TMS320VC5509A? I have samples of both chips.

Why not choose something more relevant to 2017?

[rsjsouza]

technix, I can't comment on the BF, but there are a few resources for the VC5509A:

I would check what is included in the development kits that used to be sold: the DSK5509A and the EVM5509A Plus - I used the EVM kit which has more connected peripherals (SD card, display, USB slave, etc.) but some of the parts there may be a bit outdated - the codec TLV320AIC23, for example, has average specs for today's standards.

Also, this is a guide for some pre-defined EDA symbols and some design rules that may be interesting.
http://www.ti.com/lit/an/spraa30/spraa30.pdf

The kits for the 5509A is surprisingly expensive here on Taobao... It might be cheaper if I cook my own.

Yes, that is what I had in mind - I think only the DSK is still soldthese development kits are older (thus more expensive) but at least they still have all design information publicly available for you to create your own.

Do keep in mind that this is actually my second DSP project, since there will be a SigmaDSP-based Bluetooth headphone receiver project before that, and that DSP really handles just some equalizing and mixing, so those big chips are not warranted.

I see. The SigmaDSP is pre-programmed device, right? In this case, if that does what you need, then there is really no need to dive into a fully programmable processor unless you are interested in the journey.

[coppice]

The SigmaDSP is pre-programmed device, right? In this case, if that does what you need, then there is really no need to dive into a fully programmable processor unless you are interested in the journey.

The SigmaDSP isn't preprogrammed, but it is only capable of a narrow range of tasks - basically audio filtering.

[technix]

The SigmaDSP is pre-programmed device, right? In this case, if that does what you need, then there is really no need to dive into a fully programmable processor unless you are interested in the journey.

The SigmaDSP isn't preprogrammed, but it is only capable of a narrow range of tasks - basically audio filtering.

All I need on that Bluetooth adapter is audio filtering (to null out the nonlinearity of the hopefully not too shabby AFE and the likely quite mediocre earbuds,) adjust volumes and route signals (it have 4 inputs and 4 outputs.) That is why to me the SigmaDSP like ADAU1761 is adequate for me. The Bluetooth receiver chip, CSR64215, actually comes with its own DSP, but that thing is truly preprogrammed - mask ROMed really, and for my purposes that isn't flexible enough and have to be replaced.

This is about those fully programmable clunkers, not that little portable thing that only have to respond to three buttons and a constant on audio stream.

[Marco]

All I need on that Bluetooth adapter is audio filtering (to null out the nonlinearity

Correcting for nonlinearity is not all that simple and I don't see how you could do it with SigmaDSP ... of course it's only audio, so I wouldn't worry about it.

[technix]

All I need on that Bluetooth adapter is audio filtering (to null out the nonlinearity

Correcting for nonlinearity is not all that simple and I don't see how you could do it with SigmaDSP ... of course it's only audio, so I wouldn't worry about it.

I wonder how much computing power do I need for correcting that nonlinearity? The source signal is capped at 2-channel 96ksps 24-bit so no point think beyond that. I sure hope the ADAU1761 SigmaDSP being adequate.

As of the full blown DSP experiment board, is it a good idea to, as I planned, to build it as a Raspberry Pi add-on board? I wish that the Pi can serve as one of the two sets of signal source and sink, configuration source and communication partner. It would be interesting to build an audio filter in that DSP, and run an Apple Music song through it.

[Marco]

I'm not sure how it's done, or if anyone even bothers ... it's only audio after all. I know that non-linearity with memory effects are often captured with Volterra series ... and because of momentum speakers almost certainly will have a memory effect. So google using that.