Jump to content
IGNORED

New Berkeley DAC? This article implies as much but ... ?


Recommended Posts

This is very interesting. They are basically saying, we don't do true 1-bit DSD, never have and never will, and 99% of manufacturers claiming they do, don't really do native DSD because they convert to multi-bit, which sounds better anyway.

 

I wonder if they understand difference between multi-bit PCM and multi-bit SDM and if they understand how multi-bit SDM converter chips work inside. So is their DAC R2R ladder or is it delta-sigma converter, and why do they think their running oversampling and delta-sigma modulation in hardware/realtime inside DAC is not a problem, but doing inverse process is?

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
I'm interested in the correct percentage. I'm willing to bet it's closer to 0.05% of DACs have a true 1 bit architecture. Maybe a thread discussing true 1 bit DACs is in order?

 

OK, can someone now give an official technical definition of "true 1-bit architecture", something that is specific enough?

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
I'm not speaking for Berkeley Audio Design in any way, but I assure you they understand this stuff very well. Brilliant is an appropriate description for Michael “Pflash” Pflaumer.

 

I'm still somehow getting picture from lot of marketing material, that companies don't really understand how the chips they use really work inside...

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
I was only commenting about offline processing; otherwise if you do on-the-fly aren't you possibly risking the same unnecessary noise and cpu resources (varies with server horsepower, of course) too near the signal path, almost the exact reason Berkeley wants it removed from their own signal path.

 

What is the issue if there is good isolation from the computer?

 

I have for example split processing and playback to separate computers. Processing computer runs in my office while playback computer runs in the listening room.

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
Why do we need any bits at all for DSD?

 

Just filter the stream in the analog domain and pure magic comes out.

 

Sometimes the simplest way is by far the most elegant.

 

Because some of the analog filtering ways involve multiple bits. Doesn't mean that you would change any of the DSD data at all though. ;)

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
Dunno, ask Berkeley. All i was saying is that if you have a small cpu (less than Zuma, Alix, etc) then on the fly conversion can make your computer work too hard and could cause hiccups, etc (documented reality; has happened to me in small servers). And how is that better than having the DAC do the work? If heavy lifting is the issue (Berkely's issue, not mine) then why not do it offline.

 

That's why I decided to split the heavy lifting to a server suitable for that purpose while playback is being handled with such a small computer (~10% CPU load on 200 MHz ARM9).

 

Reason why it is better than having DAC do the work is that I have not yet seen DACs with something like quad-core i7 inside. So in that sense I completely agree with Berkeley that it is better done with computer. And if it would be inside, the noise source would be much closer to the sensitive analog stages than if it's in a different room altogether...

 

My issue with offline conversion is that it takes a lot of diskspace. And what if I want to do conversion different way tomorrow? And since I have various different DACs in different rooms, I would need different kinds of conversions for each. Managing and updating set of conversions is quite an effort and time consuming.

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
I can let you in on a cyber Monday deal for a 4TB disc drive for peanuts.

 

And then I would just have to remember to make conversions of every existing and new album for every DAC, and preferably with each of the 56 different DSD-to-PCM conversion setting combinations....

 

Just something like 448 sets of the same content!

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
The Alpha 2 uses their own code on a SHARC FPGA.

 

SHARC is a DSP processor. FPGA is another thing and is programmable logic. So they even have a small computer inside...

 

But I was trying to ask what kind of D/A converter it has...

 

(Current/old Alpha seems to have a single AD1955 which is delta-sigma converter chip... (my integrated amp has two of those in dual-differential configuration))

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
Just an observation, but if I spent $15k on a DAC and another $2K on an USB2S/PDIF converter to drive that DAC, I would convert all of my music library to the optimum form for that DAC without a second thought. :)

 

For such expensive system I would buy a purpose built computer and not worry about such annoyances as converting files. Investing another $15k on a computer would buy pretty nice one.

 

And in any case for that money you should get a system already that is not really sensitive to computer...

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
  • 2 weeks later...

Problem with word clocks is that it needs frequency multiplier to produce master clocks for modern delta-sigma DACs. And that multiplier tends to also multiply any jitter by the same amount...

 

So better is external master clock from something like M2Tech Evo Clock (super clock & master clock outputs). The provided 22.5792 and 24.576 MHz clocks are native master clocks for many DACs.

 

However, DACs that can use external master clock are really rare compared to pro-gear that can use word clock which is sort of standard. For pro-gear the reason for using word clock is just to sample-synchronize more channels than single ADC or DAC can provide and thus provide a mean to have more channels. 32-128 channels is pretty ordinary studio...

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
Multi-bit delta-sigma audio very definitely is PCM, and represents coarsely quantized whole output sample values being sent to a linear PCM DAC.

 

OK, so they just demonstrated that they don't understand how multi-bit delta-sigma DACs or multi-element 1-bit delta-sigma DACs are constructed... :D

 

I'm becoming to conclusion that I don't buy DACs anymore from people who use DAC chips, unless the price is very low (then I can accept it for price reasons).

 

IIRC, Charles said once that people who use op-amps don't understand how to design an analog circuit. Now I feel that people who use DAC chips don't understand how to design a converter...

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
Miska - Your bold statements are starting to make me think you don't understand what you're talking about. Suggesting Berkeley Audio Design doesn't under stand this is preposterous. Please explain what lead you to this conclusion.

 

I have explained this so many times that I'm just getting tired and may soon give up and just stop caring...

 

Because multi-bit delta-sigma DAC output stages have equally weighted bits controlled by unary coded sample values, compared to PCM DACs that use binary coded (two's complement) encoding of 2^x weighted bits. (Wolson uses 14 non-equally weighed 1-bit delta-sigma modulated elements and differs in this respect and thus ends up having total of 78 elements)

 

So you can think multi-bit delta-sigma DAC as array of 1-bit DACs. For example Sabre has 64 pieces of 1-bit DACs run in parallel. Multi-level output is constructed from MASH-like structure of cascaded low-order modulators. Then the output is scrambled through the unary thermometer-coding to dynamic element matching where same bit position is every time allocated to different equally weighted output element.

 

Big difference already between SDM and PCM is that SDM typically has odd number of output levels which may be a non-power-of-two number like 25 (dCS). While PCM has always even number of power-of-two output levels.

 

I have also earlier shown in other threads how to use such array of 1-bit DACs to form a linear phase analog filter (that doesn't have any phase distortion) for use with DSD decoding and won't repeat it here.

 

P.S. I'm getting extremely disappointed by the quality of responses I see, especially compared to all the books, scientific papers and lectures over several decades on delta-sigma converter designs. (including bunch of AES papers on the topic)

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
Further evidence that multi-bit delta-sigma data is PCM is the fact that recovering an audio signal after D/A conversion simply requires a low-pass filter with a flat frequency response in the pass band. If it were delta modulation, as has been claimed, an integrator would be required, which has a 6 dB/octave attenuation slope in the pass band.

 

Wrong, I have implemented five different any-bit delta-sigma modulators in software. It is not delta modulation, but none of the output samples encode absolute output values.

 

Reconstruction filter, be it for SDM or PCM converter is the integrator. It is as much required for both PCM and SDM. On PCM it is required to cut out images that repeat at multiples of sampling frequency and on SDM it is required to cut out the uncorrelated noise.

 

HQPlayer has various integrators for DSD-to-PCM converter that all have flat pass-band. One of the latest additions called "wec" there is actually kind of software simulation of multi-element DSD DAC. I am now working on hardware D/A implementation of the same.

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
Based on Berkeley's statement, Michal doesn't know what he is talking about. Plus, I know other engineers who refuse to get involved in a discussion with Miska because they think he is arrogant and plain incorrect.

 

Avoiding discussion is not good for anybody. And just stating that someone is "wrong" without being ready to defend one's position in open debate is not very constructive.

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
What are the current ADCs if not PCM?

 

Delta-sigma converters with digital conversion to PCM (notice that decimation filter block = integrator). Two of the top-class most used ones:

 

CS5381 (Lynx Hilo, RME, etc):

5381blkdiag_mag.gif

 

AK5394 (lot of pro-audio gear):

AK5394AVS.gif

 

Also notice especially pictured out in CS5381 the delta-feedback loop from modulator through inverting DAC back to ADC. There you have the difference (delta) + integrator (sigma).

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
The guys that actually build and sell very good sounding dacs on a commercially successful basis wins the credibility prize any day of the week.

 

Software doesn't count? Well, it just reinforces that I need to publish discrete DAC schematics and Gerbers this winter. No way I'm interested in producing hardware, too much uninteresting hassle related to selling hardware to consumers. And I just don't have time to think about pretty looking CNC aluminum case work that costs €€€ (I would need to choose between improving software and having pretty box).

 

Personally I think the whole disagreement resembles a semantical masturbation exercise, somewhat akin to agreeing or disagreeing with Clinton's assertion he did not have sex with that woman. I say he did. Thenagain, come to think of it - he really did not. Tough call.

 

As long as we talk about multi-bit SDM and PCM that's not gonna go away. Sonic already demonstrated that by assuming any number other than 0 meaning PCM, even if it's value of pi or golden ratio. :) (be warned, your bank account has been defined to be PCM, so beware of any jitter there)

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
or, as I thought m post already suggested, dCS likes selling their customers very expensive additional boxes with "better" oscillators inside, or perhaps even more accurately, likes their customers' money.

Why did dCS not put the better oscillator in the DAC itself from the start?

 

What I like about in dCS is that they actually design their own DACs. There are not that many such companies. Meitner, Playback Designs and MSB comes to mind. Others?

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
I have searched the web quite a bit and nothing I have ever found con confirm that the Da Vinci DACs are a discrete resistor design--nor does the company ever claim that.

 

Same here. HiFi-News only mentioned that it's an array of four DACs with two as complementary pair and two time-delayed (staggered). I couldn't find any detailed enough photos either.

 

I am betting that that they are actually an NOS PCM1704, or maybe one of the other T.I. chips that allows for turning off of its built-in filters (though that would then not be an R2R design). If anyone can find a photo or quote to prove otherwise, I would be interested.

 

I originally thought that it would have been PCM1704, but after checking out the HiFi-News review and looking at the measurements I'm not sure. Seems to be linear within +-0.5 dB down to around -93 dB which is spot-on PCM1704 (-90 dB in datasheet) in staggered configuration (+3 dB). Also the complementary pair is just like PCM1700/PCM63P/PCM1704 are constructed.

 

Maybe Larry could enlighten us on this topic?

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...