A high-end digital amplifier for everyone

[hifiZen]

This looks very appealing. I have been waiting a very long time to build a DIY DSP engine... if you want to hear the story, read on, if not skip to my comments at the end.

I first concieved of a DIY DSP project way back in 1998, still a 2nd year EE student. I did the first rough board design using Analog Devices ADSP-21060 DSP. It was of course, not a very good design, and too expensive with low-latency SRAMs and heaps of discrete logic. It featured the CS8412 DIR with an AD1890 for SRC functions, CS4390 DACs, and a PIC micro to handle front panel display, reset, user input, etc. I never built it thankfully, it would have been hell to debug, and I doubt it even would have worked properly. Worse, I couldn't get dev tools for the SHARC! But a second opportunity to make it a reality came up a few years later...

My graduation project for April 2001 was an audio DSP board centered around the much friendlier ADSP-21065L SHARC chip. This processor has I2S in/outs. I changed input/SRC to AD1892, and DACs to PCM1702 with a big analog output circuit. Although the 4 layer board design was complete, the availability of some key parts and free or cheap development tools for the SHARC processor prevented me from ever building it and getting it working... and that was the end of that. At least I got credit for it.

Since 2001, I've been working full time, but I've kept the project alive in my head, waiting for the break I needed to make it work. The board architecture has evolved a lot... first removing the DIRs, DACs and analog output stages so they could be put on separate modules and interchanged. I upgraded from AD1892s (integrated DIR/SRCs) to AD1896 SRC chips when they came on the market. Having all I/O modules off-board makes possible many new options like LVDS receiver boards for I2S directly from hacked consumer equipment, or a separate SPDIF/IEC958 board, or a separate module with a PMD-100 HDCD decoder from an old CD player, or, or... Of course all these interchangable modules required flexibility that discreet logic couldn't handle, which led me to add an FPGA as a configurable I2S mux and data reformatter. I replaced the PIC micro with a more useful Atmel AVR. Only recently have affordable, powerful FPGAs and free dev tools arrived on the scene with enough gates to do some real DSP with. Since free dev tools for the SHARC are still not available, it seems only natural to eliminate the troublesome SHARC at last (I still hold a grudge against AD for not releasing free dev tools, or at least a worthwhile educational package).

So I've been anxiously waiting 5+ years to make this a reality, and I've just been laid off from my company here in the Valley... hmm, what to do???

After I finish moving back to Canada, I'll have lots of time on my hands unless I find a new job right away, and I am looking to brush up my Verilog skills. I'd really love to get my hands into this project, doing a board spin and/or code development. Since a digital power amp isn't really a goal for me, my primary interest would be adding other DSP functionality, such as DSD->PCM conversion, and audiophile-grade IIR and FIR filters for tone control / speaker crossovers etc. that users could apply in standard DAC designs and so forth.

So a few questions for the DAX group:

1. What language are you using? VHDL or Verilog?
2. What input word width and sample rate are you using?
3. How soon can you get this puppy ready for us to sink our teeth in!?

[magnus]

Actually, like the previous poster, I had this idea some time ago, 1995 to be exact. Needless to say I quickly found it impossible to realize at the moment but ever since then it's been lurking in the back of my mind as something that is indeed possibe but just too much work/too costly for a single person.

And now it's here!

I can only second the opinion of the previous poster and say: Let's get on with it.

Regards,
Magnus

[jwb]

It's true you can't get the dev tools for SHARC except 90-day evaluations and lots of money. But is the situation actually different for FPGA dev tools?

[magnus]

It's free: Look

Originally posted by jwb
It's true you can't get the dev tools for SHARC except 90-day evaluations and lots of money. But is the situation actually different for FPGA dev tools?

[Gridstop]

Yeah, Xilinx's free tools are pretty great.

Unfortunately unless you're up for designing a DSP/processor core from the ground up as state machines/logic it won't get you very far

Opencores HAS made some good progress with basic FIR and IIR filters and a few embedded microcontrollers, maybe some progress could be made?

All of opencore's stuff is GPL'd or LGPL'd, which is really cool. But if you use them and distribute your design in any way be sure to read up on them so you are aware of your responsibilities under the license.

[hifiZen]

Awful quiet on this thread lately... must be final exam time.

Gridstop: thanks for the pointer to opencores... looks like it'll come in real handy!

[DAXgroup]

So a few questions for the DAX group:

1. What language are you using? VHDL or Verilog?
2. What input word width and sample rate are you using?
3. How soon can you get this puppy ready for us to sink our teeth in!?

We coding in VHDL. The input word width is only dictaded by the input stage (CS8416 and SCR4193), this means it supports 16, 20 and 24-bit.

At this moment we are putting all our time in producing the user manual which will answer 98% of potential questions and (you guessed right) on our final exams and projects. Unfortunately, we cannot predict production time as it will be function of the number of people willing to invest in the project. Judging by the feedback we received so far, we are hoping to start producing something as early as possible in 2004.

Gabriel,
DAX Group

[hifiZen]

Excellent!

Well good luck to all of you on your final exams. We'll be waiting patiently to see your user manual.

[kiwone]

very good project! CONGRATULATION

I'm very interested by this project, I've been searching, for 8 months, on all french forum to find a schema or an other guy who has already built this type of amplifier. But nothing.

And now I find your project, it's in english but I try to understand and to write. But my project was builting a Multichannel digital power amplifier (for my home cinema) as good as or better than the Bel Canto Evo 6. With your project it's not possible because your module has just a digital input and the pre-out from my pre- amplifier or all analog. This is the problem, if Ihave to put 6 DAC or ADC between my pre-amplifier and my "Multichannel digital power amplier" it will be very expensive!

That why a few times,seeing if other people or interested by this Multichannel D.P.A. and seeing if some very experiment people (Brian Brown or you) could suggest some answer and schema.

Thanks a lot if you can understand my english and if you can find a solution for the HI END Multichannel D.P.A.

[DAXgroup]

But my project was builting a Multichannel digital power amplifier (for my home cinema) as good as or better than the Bel Canto Evo 6. With your project it's not possible because your module has just a digital input and the pre-out from my pre- amplifier or all analog.

As you said, our system has one digital input (coax or optic). However, as we said, the hardware is very flexible and adding one or multiple analog inputs is pretty straightforward. Most ADCs output I2S data at 192khz (the same type of data already being fed to the fpga), so one could simply duplicate the existing logic in the fpga to have more audio channels. In the same fashion, more output stages could be added.

In fact, adding an analog input is one of our next priority, since it makes the amplifier much more versatile and we still have a lot of analog recordings


And by the way, if it can encourage you, nous parlons tous francais...(french is our first language).

Rgds,
DAXgroup