A high-end digital amplifier for everyone

[nowater]

Originally posted by DAXgroup

Without being rude, we put to much efforts in that project to accept negative comments stating this amplifier cannot compare to existing products. What are your motivations for describing our DAX as a "poor-performing power DAC"? Please.

…. We also think the sonic quality of our amplifier should not be attacked based on a few specifications and a large amount of misconceptions. …. How many amps out there are showing 1% THD but people are ready to sell their house to buy one. …. It's a matter of defining if you're building an amp to show numbers to your friends or if you are ready to make compromising and realizing the REAL benefits of a given system. ….. New technologies are there to replace the previous ones. The only way we can achieve this is by WORKING on these new solutions.

…. This project can bring a lot to people who are willing to learn about digital audio, and people who knows about digital audio can bring a lot to the project.

Firstly, my apologies for a misunderstanding. I did not mean to describe your DAX as a poor performing power DAC. I meant to describe power DAC’s as poor-performing. I have gotten the impression that improving a power DAC's performance pretty quickly becomes a matter of improving the power supply. See TACT.

Secondly, I am not technically strong in digital, so I am here to learn about digital audio.

So how can I justify jumping to conclusions about power DAC’s? Because I can read and research.
Take a look here:
http://www.classd.org/oem_products/p...ti_dig_amp.htm
http://www.classd.org/qna/qnafs.htm
and here:
http://audioholics.com/techtips/audi...mplifiers.html

Of course, not all that is written is true, but sometimes the writer is authoritative. Putzeys impresses me. He declined to join this thread.

So, perhaps you guys at DAX are inadvertently asking the technically strong people on this thread to accompany you in hurtling towards the brick wall at the end of a dead end street, at least regarding the power DAC. There’s nothing wrong with that. If you break through, you’re all heroes.

I accept your comment that high distortion amps can sound good. Is it the goal of this project to design such an amp?

Finally, can you explain how exactly the project will be inferior if it uses an analog class D power stage? What do you stand to lose, apart from distortion and a purist principle? If the answer is to see how good a digital input Class D amp can be, then see TACT.

I hope, instead of being seen as negative, my posts are seen as “testing” the strength of the project and its goals.

Grant

[phase_accurate]

Whether some people want it or not, audio is an evolving field. We cannot stick at building tubes or class-A amplifiers forever and hope it will still be the state of the art. New technologies are there to replace the previous ones.

I am personally a fan of switching amplifiers but I would also like to mention that it isn't worth USING new technologies just for the sake of new technologies if they don't have other advantages than being new ..........

I am convinced that it is easier to make a high-precision DAC at low levels and high precision amplification in analog.

But it is still worth trying of course !

Regards

Charles

[kevyjo]

Ok, I couldn't resist another post. As far as the discussion about "state of the art", I think the jaded attitude many audiophiles have towards new technology is understandable, though I personally feel its excessive. Decades ago, a hifi system was just that, a system that strove for high fidelity. In the last 30 years or so, we've seen so many "new technologies" that offered wiz bang effects and gee wiz features that really had nothing to do with fidelity, it is no wonder why hard core audiophiles stick to tubes/vinyl/class A ect. I personally have been exited about digital amplification, be it based on class D or direct amplification, and have been frustrated by peoples casual dismissal of it as just another gimmick. How this technology may be implemented may seem like a trivial question compared to the issues you are dealing with right now, but implementation is where I see many new technologies fail to really catch on. We could have better audio systems NOW if we just brought together all the off the shelf technologies that already exist, let alone implementing new ones. If this device is perfected, maybe it could serv as a leapfrog into a new type of audio system.

[Jaka Racman]

Hi,

nowater, why am I under impression that you want to persuade what you consider a group of technically savvy but otherwise misguided people who are willing to share their knowledge (DAX) into making you a digital playground?

I am well aware of Bruno Putzeys opinion regarding all digital power amplifier. In fact I posted the same link before in another thread about Philips digital amplifier.

Now about the merits and problems of various approaches:

Analog modulated switching amplifiers can have feedback taken either from switching stage or even from output filter. Thus they are capable of power supply and output stage distortion rejection. But modulator stage is always a comparator comparing two slow signals: one is usually triangle shaped (either carrier or output from integrator in self oscillating designs and another is almost dc relative to the switching frequency (either output from error amplifier or reference value for hysteresis modulators). This works well as long you have only one half bridge switching leg working. Comparator timing decision is made ahead of switching leg transient so there is no interference. Things radically change when you have a multichannel amplifier. Every switching leg transient introduces several millivolts of noise on slow changing comparator inputs. Result is several nanoseconds of jitter from the output stage. Countermeasures include separate power supply stages for error amplifier, comparator and output stage and elaborate grounding scheme, Bruno Putzeys own US patent 6046636 is a proof of that.

Digitally modulated switching amplifier usually has no feedback, so it has no power supply and output stage distortion rejection. So theoretically it is inferior. But it has very well defined modulator timing with almost no jitter and no crosstalk from other channels in multichannel amplifier. Since modulator can be made almost ideal, only power stage remains the problem. One solution is to use Pulse Edge Delay Error Correction (PEDEC). You can read more about it in the following links: http://bogo-united.oeb.tdk.net/graph...ier_system.pdf and http://pearlx.snu.ac.kr/Publication/PESC0203.pdf One might argue that we have analog modulation scheme again, but in case of PEDEC, slopes of signals at comparator inputs are much steeper resulting in less jitter susceptibility.

Another advantage of analog modulation scheme is that you can include output filter in the feedback loop. This is hard to do in the digital domain, but there are remedies. One is multiphase approach. Basically in most switching amplifiers you set corner frequency of output LC filter somewhere between 40 and 80kHz, resulting in max 40dB attenuation at 400Kkz switching frequency. Another trade off in filter design is that you need a relatively small inductance (typically around 20uH) to have correct damping factor with speaker impedance around 6Ohm. With this values there is a few A of ripple current in the inductor, which is consequently removed by output filter capacitor. Now if there is a way of reducing inductor ripple current to a few tens of mA, you might eventually get away with output capacitor. Ripple can be reduced by multiphase ripple cancellation technique, first introduced in Crown BCA amplifier http://www.crownaudio.com/amp_htm/amp_info/bca_op.htm. Another article worth reading of multiphase technique would be: http://bogo-united.oeb.tdk.net/graph...tions/4917.pdf.

It is obvious that multiphase modulation has been conceived with analog modulators, but advantages are lost by noise coupling to the comparator inputs. Easiest way of ripple cancellation is achieved by phase shifted fixed frequency PWM modulated signals which are then summed by inductors
of output stages.

My idea on how to do it on raw sigma delta data stream (let's say this is 2.8MHz DSD) is as follows:you take 8 consecutive data samples from N to N+7 and count number of ones in it. At N+8 you output another 8 cycles long data stream which now has all the counted ones at the beginning and the remaining zeros at the end. At the same time you count ones in another sequence of input data stream from N+8 to N+15 and then output counted ones first. So basically input DSD data stream is transformed into 2.8MHz/8 PWM modulated data stream with 8 level resolution. Now if you take the same algorithm, but process input data from N+1 to N+8, you get another PWM modulated data stream, but this time phase shifted 12.5%. If you again repeat this with another input data set from N+2 to N+9 and so on you eventually have 8 phase shifted PWM modulated data streams, each with 350kHz switching frequency. But the resulting switching frequency after summation is again 2.8MHz not counting ripple current cancellation. This results in almost negligible current ripple at the output. One can also use one transistor and diode (as in Crown BCA) instead of half bridge for each PWM output stage thus reducing complexity. I really don't know how the noise spectra of such manipulation would be, but I hope it would not be worse than before. I wanted to simulate this first before posting the idea, but since you asked for it, here it is.

Best regards,

Jaka Racman

[pburke]

Originally posted by DAXgroup
[B]
40W is maybe not much to drive a classical speaker, but for a lot of people, it is more than sufficient.

let me chime in as a potential user of a DIY amp - 40 watts, even if for each driver, isn't going to be of any interest for me. I have magnepan speakers, and I am seraching for amps that can control these inefficient speakers.

200watts each, and I'm game. I don't think I am alone with this - many owners of inefficient speakers are starting to look at digital as an option.

Peter

[kevyjo]

Do you know what the ultimate power potential would be? Or has that not yet been determined? The above post speaks to an issue, yes there are many people out there with high end 40watt systems, but that demographic may not be very open to digital amplification (alot of them have tubes).

[phase_accurate]

Jaka Racman wrote:

Things radically change when you have a multichannel amplifier. Every switching leg transient introduces several millivolts of noise on slow changing comparator inputs. Result is several nanoseconds of jitter from the output stage.

I have a distinct feeling that this is even a problem with a single amplifier channel. Since a (analogue) sigma-delta modulator samples at discrete time-intervals it might be less susceptible to this effect than a PWM amp (when the next decision is taken, the overshoots of the last transition etc are already over).
That is just what my imagination tells me, I don't have any mathematical proof for it.

But it makes me enough curious to try it out once.

Regards

Charles

[amt]

Ive been following this thread and thought I would thrown in my observations. I do not have an engineering backround and am just an avid diy enthusiest. I think that digital amps are most definetely the future but as for the basic music buff/DIYer in the here and now, I echo pburkes' opinion. I too have Maggies and the promise of high efficiency and little heat etc. is most attractive for high watt amplification.

Now a 40watt amp for active speakers is fine but I see this as really a commercial product rather than for diy guys. Why would we chose a new, sonically unproven design, when so many wonderful 20 - 50 watt designs are already available. Other than class A, heat really isnt a concern. Ive got several Gainclones that run cool, sound great and are dirt cheap. A 200 watt monoblock that doesnt heat my house, break my back to move, cost a small fortune, or pop the fuses in my house during turn-on (Thresholds monos I borrowed) and you will truely have something. Use for subs alone would be huge I think.

I dont think your average, non-EE type diyer is going to thrown his Aleph away quite yet. Im not sure I would replace any of my 6 amps with one either. Now for packaging in a COMMERCIAL active speaker, multichannel HT in a box, that 40watts is fine. Are we helping design a commercial product or one that can be flexible enough for the wide range of needs we here have?

Im not beening negative, just voicing an opinion that might be held by many of the less vocal and non technical.

amt

[tiroth]

I think the response is again "you have to walk before you can run." At the moment we have exactly zero end-to-end DIY digital solutions. This project is the first step in realizing that, and it is great that it is being designed from the ground up as flexible enough to accomadate various topologies. I think the emphasis is on gaining the knowledge to move from 1st generation to 2nd generation, not on replacing all existing analog amplifiers.

[Jaka Racman]

Hi,

to those that a concerned with 40 watt power limitation of current design, this is only a technology demonstrator. Digitally modulated switching amplifiers can scale far better than analogly modulated, since power stage and modulator stage a totally separated. Here is an estimate of currently possible technology:

-supply voltage around 400V due to 450V electrolytic capacitor rating.
-switching frequency up to 350kHz with use of Coolmos FETs and SiC schottky diodes.
-current capability around 10A per switching leg with such frequencies. Paralleling of switching legs is required, but with summation on output inductors, not by direct FET paralleling. Ideal for multiphase design.

So with modulating index at 0.8 and full bridge design, one can output 720V peak to peak sine wave or 8100W into 8ohms. Enough to drive Maggies (I also own them). It may sound preposterous, but it is technically entirely possible.

Best regards,

Jaka Racman