CHORD ELECTRONICS DAVE

[VM72]

SMPS uses many very respectable audio gears producers: Mola Mola, Emm Labs, Meitner, Merging (NADAC), Benchmark, Purifi, Hypex…

 

[chesebert]

Nothing wrong with SMPS if done properly with audio application in mind.

 

[Mediahound]

The trolling that’s permitted in this thread is quite astonishing.

It's quite odd indeed. There are a couple of trolls here that insist on continually posting here how they can't understand how good Chord stuff is, despite all the evidence and experiences here.

I would imagine these posters may actually be working for jealous competitors or something like that because other than that explanation, there is no sane person who has a certain DAC they like, who literally spends all their time going to the threads of OTHER makers to badmouth them.

 

[FunkyBassMan]

I would imagine these posters may actually be working for jealous competitors or something like that because other than that explanation, there is no sane person who has a certain DAC they like, who literally spends all their time going to the threads of OTHER makers to badmouth them.

I suspect you're right.

Because their English and their arguments are always so bad, I imagine they're part of the ASR business to pump Chi-fi.

 

[griff500]

It's quite odd indeed. There are a couple of trolls here that insist on continually posting here how they can't understand how good Chord stuff is, despite all the evidence and experiences here.

I would imagine these posters may actually be working for jealous competitors or something like that because other than that explanation, there is no sane person who has a certain DAC they like, who literally spends all their time going to the threads of OTHER makers to badmouth them.

It’s a bizarre way to spend one’s time.

 

[GoldenSound]

These are all patch work solutions to solve multi bit delta sigma problems. The actual answer (IMO) is to go back to DSD and have a dynamic noise shaper - possibly in the future this can be controlled by some ML chip so it would apply different noise shapers based on actual signals. I think Emm Labs and PBD all have their own versions of the dynamic noise shaper, but I don't know more than that.

They're not patch work, they're extremely advanced maths that provides effective results. And it's not to solve multi-bit delta sigma problems, it's to solve delta sigma problems. Multi-bit is part of the solution not the cause of the problems.
Going to 1-bit delta sigma causes MORE challenges not less.

Delta-sigma causes issues because of the fact that when you have for example a sample that needs to be represented as '0.7', but your DAC is 1-bit, and therefore can only output '1.0' or '0.0', then the closest value you can pick is '1.0' and you have a quantization error of '0.3'

You can address this in a couple ways, firstly by having a higher native bit depth. If you have a multibit converter that can output say 6 values instead of 2, then you can now output '0.8' instead and your quantization error is reduced to '0.1'.

However you still have some error, and so you need to fix this. You do this by running the device at extremely high speeds so that the difference can (to massively oversimplify) be compensated for on the next sample, and the faster you run the modulator/DAC the better you can do this. Exactly how this is compensated for or predicted in advance depends on how the maths is set up and you can have pretty basic modulators/noise shapers and very advanced ones.
Chord uses an extremely effective one additionally helped by the fact that the DAVE for example runs at 104Mhz which is far faster than the vast majority of DACs on the market.

Using a 1-bit DAC means you have much higher quantization error and to get the same accuracy need to be running at vastly higher speeds than an equivalent multibit DAC, hence why hardly anyone uses 1-bit anymore. They're usually just included for DSD compatibility and not used otherwise.

R2R or native PCM converters have the challenge that to be accurate the physical components have to be within ridiculously tight tolerances and/or have quite clever compensation methods.

1-bit converters have the challenge that to be accurate and sound good they need to run EXTREMELY fast and are very reliant on the quality of the modulator/software.

The ideal solution arguably is to make something that is between the two, using a 5-7 bit converter, which can be done accurately with relative ease (far easier than making an accurate 16-24 bit R2R ladder anyway), and which wouldn't need to be run as fast as a pure 1-bit DAC, or even if you DO run it that fast, you've got more flexibility and options open for more effective noise shaping

 

[chesebert]

You answered your own question. FPGA and micro controllers are fast enough these days (they were fast enough a few years ago).

1-bit converters have the challenge that to be accurate and sound good they need to run EXTREMELY fast and are very reliant on the quality of the modulator/software.

 

[GoldenSound]

You answered your own question. FPGA and micro controllers are fast enough these days (they were fast enough a few years ago).

The FPGA speed is one small part of the whole equation.
High speed doesn't necessarily mean it can actually run particularly complex maths. Same as CPU frequency doesn't tell you how well it'll perform in games or various applications. Instructions per cycle is important here and can vary depending on what you're trying to do.

Secondly there's the issue of actually being able to design the FPGA code, you can hand someone the most advanced, fastest, most powerful FPGA on the planet, it's a bit useless if they don't know how to design code to do what's needed.

Thirdly there's the hardware itself. Not all DAC topologies are capable of running at the same speed.

And even so, there's also the question of WHY use a 1-bit design when you can use multi-bit of various types such as PWM to allow for additional performance with effectively no downsides.

 

[chesebert]

I agree, you do have to write your VHDL carefully and mind your gate propagation delays. I am sure there are EDA tools to help you map to different architectures and optimize for propagation delays and speed.

 

[GuiltyRocker]

The FPGA speed is one small part of the whole equation.
High speed doesn't necessarily mean it can actually run particularly complex maths. Same as CPU frequency doesn't tell you how well it'll perform in games or various applications. Instructions per cycle is important here and can vary depending on what you're trying to do.

Secondly there's the issue of actually being able to design the FPGA code, you can hand someone the most advanced, fastest, most powerful FPGA on the planet, it's a bit useless if they don't know how to design code to do what's needed.

Thirdly there's the hardware itself. Not all DAC topologies are capable of running at the same speed.

And even so, there's also the question of WHY use a 1-bit design when you can use multi-bit of various types such as PWM to allow for additional performance with effectively no downsides.

Perfectly said.

 

[GuiltyRocker]

PS Audio is mid-fi sound quality at hi-fi pricing. Most of the people who buy PS Audio are new to the hobby and don't know what they are doing. What did you expect?

That's your opinion, those are just labels you're throwing around, not accurate at all.

 

[GoldenSound]

I agree, you do have to write your VHDL carefully and mind your gate propagation delays.

It's not really that, even ignoring the FPGA specific challenges, the actual code itself just in terms of the maths is not simple. You'd need to find someone fairly specialised to be able to do it in MATLAB let alone implement it in an FPGA.
Then you add on the challenges of the analog side and how the two interact and it's quite clear why there aren't a hundred different vendors all doing the same thing

 

[chesebert]

Andreas Koch knows how to do it...I am sure there are others as well. Audio is a tough business to be in so the best people generally work for TI, AD, Apple, etc. and not trying to start their dac company.

 

[GuiltyRocker]

It's truly amazing that Rob Watts was able to do all the programming of the FPGA, come up with his own noise shaper and interpolation filter, as well as his own DAC topology. Incredible achievements that haters just can't stomach.

 

[GuiltyRocker]

Other DACs just sound dull and lifeless to me, overly soft and boring. I'm a Rob Watts man for sure and proudly, his DACs and Scaler has allowed me to enjoy music.