A high-end digital amplifier for everyone

[kiwone]

Thanks for your answer and thanks to inform me that you speak also in french but in this forum I think it would be rude ton in Moliere's language but, maybe, if I can explain some wishes I'll write you an email in french.

do you thin if it's possible to built this nearly famous Multichannel DPA as good as a Bel Canto ( if you know this one) and if it possible to built this amplifier just with monaural module amplifier like in very High End analog amplifier? I don't know if it's better but it just like this we could built an amplifier with 1,2,....,8 channel and choose the power of the different module ( different power for the subwoofer channel.)

[kiwone]

Hello everybody

I put a little "up" to this thread because it is very important for me and I want to know more about this project and built this High End Digital Power Amp.

So please do not let die this thread, I need all of you!!!

Thanks

[magnus]

Hi,

I'm following this thread as I have been waiting for something like this.

I would think a lot of people would be interested in a 5.1 setup. Now, the thought that pops up in my head is you could do the following:

1.
Get a 5.1 decoder that will output digital signals (S/PDIF or I2S I suppose) for all channels. Quite likely this is a "hacked" receiver based on the fact that it is hard/impossible to get a license from Dolby for home use. I know for a fact that this has already been acomplished at avsforum but I can't find the thread just now. In any case, it can be done. Another alternative is a completely homebuilt decoder which is certianly possible but highly doubtful IMO.

2.
It would then be feasible to distribute these signals to the speakers via, say, Cat5 cable. The reason for using Cat5 would be so that you could send volume information as well, embedded in the stream data. I'm no I2S expert so tell me if I'm totally wrong here. The reason for doing this would be to have an "analogue" volume control. You would send a 16 bit word that would set the rail voltage and thus the volume of the amp. The reason for doing this would be to avoid a digital volume control.

3.
Also on the wishlist is to have a stereo amp with digital x-over filtering done in the FPGA. This way you could do away with the lumped x-over filters (save for a cap on the tweeter for safety perhaps). Great idea, no?

Comments? Feasible?

I hope so.

/Magnus

[kevyjo]

Forgive me, I'm way out of my league here, but I find this whole discussion very intriguing, and I'd like to contribute my envisonment of "the ultimate audio/home theater system", but that takes pages and pages, I thought it might be innapropriate, or unworkable to type such a long thread response. I've thought for years that the adversarial relationship between the disperate camps of audiophilia/home theater enthusiasm is stilly, and the ultimate evolution of audio would bring everything together in a no compromizes package. I am not tech savy enough to understand the tech details about this subject yet, but it seems that DAX ultimate focus is universalizm, which I find very intriguing.

[DAXgroup]

Hi everyone,

We have updated the DAX website with the documentation for the platform and the current output stage. The documentation being a little raw, we decided to include our senior design project report, describing the fundamentals of the project. Keep in mind that the platform undertook multiple changes since then, so the report is not necessarly reflecting the current hardware and software implementation.

Regards,
DAX Group

[kevyjo]

I cant seem to open the new links, get an error. One thing I'd like to know, you have talked about the flexiblity/programmability of this platform; could it also incorporate speaker dynamics, and incorporate a digital crossover? I know this is a big leap, but I figured this development stage would be a place to throw out such possiblilities. Like i said, my understading of this subject is very limited, im sort of looking "outside in", but think of the possibilities: If an audio/HT system had an interactive "brain" that incorporated speaker dynamics, you could take advantage of the dramatic improvement in fidelity, imaging and dynamics that active crossovers/bi/triamplification can get you. If this system could indeed be smart, it would also open up the possiblity of digital processing automatically compensating for specific speaker driver shortcommings. Say you have a modest set of bookshelf speakers. The 'brain' sends a customized signal to the woofer, and another to the tweeter, optimized for the performance of the driver. The brain would be programmed with the drivers characteristics, and would know what its ultimate limitations were, so they would never overload and never distort. At low/moderate listening levels, a bookshelf speaker may be capable of 20hz bass extension if the drivers were driven to their limits, but not beyond. as listening levels increased, the brain would automatically cut the signal at the lowest bass levels as to not overdrive the woofers. putting all of this together, you could have a system of modest size that could have pretty much shocking dynamics and fidelity. What do ya think?

[nowater]

I agree with kevyo -- a "brain" allowing DSP and digital crossover is a great idea, and brings many more people's interest to the project. Maybe based on several TAS3103 chips, allowing numerous channels or active driver crossovers to the output.

And an option for the output module to be power DACs or ordinary DACs.

Grant

[Jaka Racman]

Hi,

before hijacking this thread into " A high-end digital multichannel amplifier with digital crossover and speaker- room equalisation and triamplification and everything else anyone comes up with including kitchensink ", would anybody please at least read article which has been up for at least 24h and understand what still has to be done before this becomes basic "A high-end digital AMPLIFIER".

In line with that my proposal would be to go with multiphase output stage. This is the quotation from classd.org :

As explained in the Sharp section, driving a power stage with DSD is about the least intelligent way to make a high power audio signal. Philips work their way around the problem by operating 8 (or any power of two) small power stages which are paralleled using common-mode coils. Each of the 8 power stages is assigned 1/8th of the DSD data volume which means that each bit is represented for 8 clock cycles. One could also say that the DSD signal is passed through a rectangular FIR filter 8 taps long outputting 9-level data. The result is that each output stage switches at a fraction of the DSD rate, resulting in lower switching losses and lower distortion. As the output stages are small, switching speed is also much higher than in an equivalent single power stage, resulting in even lower distortion.

As such this topology is a very original and effective implementation of the multiphase power stage first proposed by B&O.
The modulation index problem is addressed by doubling the length of the FIR filter. When it has 16 taps, a 17-level output signal is produced of which only the middle 9 actually occur due to the 50% modulation limit of DSD. These 9 levels map onto the 9 levels of the power stage, allowing 100% modulation.

I have some idea how to implement this, but being an analog guy I will have to simulate this before in spice which will take me some (lot of) time.

Best regards,

Jaka Racman

BTW, DAX, I am really impressed by the originality and quality of the design.

[phase_accurate]

I find the CLASSD.ORG homepage quite interesting and I have also been discussing things with the owner some time ago.

But I don't know where he takes his conclusion from that you could drive a DS output stage only to 50 % of its output voltage capability. When I simulated a DS amplifier it was perfectly possible to drive it up to almost 100%.
I am of course aware that an ordinary DS modulator would deliver an idle-output signal that is consisting more or less of half the clock frequency.
As this can be rather high, switching losses might be high as well (apart from other RF related problems). OTOH conduction losses (while idling) in the output transistors are smaller at a higher switching frequency, for a given output filter. So there is always a tradeoff.
With DSP preprocessing it should be possible to generate a DS signal that doesn't switch with too high a frequency while retaining good resolution (the bit-flipping technique is one of them).

According to a book about DS -modulators and -converters, any asymmerty in the output stage's behaviour (of a one-bit converter !) would result in ONLY an offset voltage. And this definitely NOT bad news.

Regards

Charles

[tiroth]

I'm out of my depth here, but is there an issue with power supply pumping in a power DS output stage, if such an offset exists?