ANOTHER Example of Why I HATE DSD and Why Customers Who Bought Sony's Boloney Are So Annoying

[Miska]

First of all, on Windows, use a native ASIO driver and you don't need to configure or use DoP at all.

 

DoP is just an ugly hack to make DSD work on operating systems where proper ways are not available (Mac OS X) and there's no nice way to auto-detect DSD capability on a DoP DAC. On Windows there's ASIO that can handle native DSD recording and playback without any of the extra payload (33%) of DoP plus has auto-detection of such capabilities. Nowadays there's a native support for DSD also on Linux, so it can do same as ASIO on Windows.

 

And yes, HQPlayer supports all digital processing for both PCM and DSD, including volume control, convolution engine (for DRC) and speaker distance & level adjustments. There's also DSD rate conversion and I recommend upsampling DSD to higher rate if possible when using digital volume control to gain extra dynamic range for output. And of course more to come.

[Miska]

AFAIK, Korg DS-DAC-10 supports native DSD playback on Mac.

 

Great! I guess they have created their own driver and their application talk directly to it (something like ProTools is doing)? Or how do they expose DSD in non-DoP way to third party applications?

 

Technically there's no reason why ASIO couldn't be used on OS X too, it would just need drivers and thin layer under the ASIO API. So far, nobody (?) has bothered...

[Miska]

The infatuation with DSD is a marketing driven circus created to sell hardware no one really needs, and give audiophiles with too much time on their hands something to talk about on the forums.

 

Yeah, I spent only a year (and some €€€) designing and building DSD DAC that I'm not selling to anybody because I'm not interested on hardware business. Plus all the effort I've put into developing delta-sigma modulators to produce DSD from PCM source. All just for sake of marketing circus (with huge marketing budget of 0€)!

 

There is much more to be gained from relentlessy innovating hardware to the point that the SQ of a DAC now costing $20K can be had for $2K

 

That's precisely what DSD128 and higher allow, by running oversampling filters and delta-sigma modulators in software. You can strip almost all of the digital side of current DACs and they suddenly become much cheaper and simpler devices. Yes, I can see that certain $$$ hardware manufacturers are very, very afraid of such thing. And don't save effort on trying to play and hush it down.

[Miska]

I am one of the ones that fights on this forum against the new "trend" of infatuation over DSD.

I am glad that Charles Hansen took the chance to tell the truth about this DSD crap, instead of playing it into marketing schemes like many other manufactures.

 

Do you have other proposals on how to bypass crappy oversampling digital filters and delta-sigma modulators of DAC chips, other than DSD?

 

I already support also 8-bit delta-sigma output, but unfortunately only DAC that supports it is my own... I'm extremely happy if I see suitable multi-bit DACs appearing on the market. My own multi-bit can currently do 8-bit at 256x rate and doesn't have any hardware DSP at all, everything on software (in player application).

 

PCM DACs with hard-coded oversampling digital filters and delta-sigma modulators? No thanks, not going to buy.

[Miska]

Korg doesn't support DoP at all, so going the native DSD route was the only way for them. Sadly, their driver enabling 2.8224MHz, 5.6448Mhz DSD playback on Mac OS X is not available to third parties at the moment.

 

Do they support ASIO DSD or something like that on Windows for third party applications? Or is their DAC tied to their player also there?

[Miska]

I view this DSD fad as being analogous to the current 3D feature fad on televisions.

 

Something added that we really don't need, purely for the purpose of allowing manufacturers to sell more units.

 

I could say PCM is the thing I don't need at all. I'm playing everything as DSD, so I don't really need or want any PCM capabilities from a DAC.

 

IMO, making a pure DSD-only DAC is the perfect item for future. Extremely simple, inexpensive and high performing device. And performance can be software-upgraded every now and then.

[Miska]

Miska, the DSD DAC was tried and failed because the inherent DSD flaws. Why reinvent the wheel?

 

Could you be more specific? The ones I have perform very well, from both measured and sound perspective.

 

http://www.grimmaudio.com/whitepapers/discrete%20ad%20converter.pdf

 

How is that related to anything?

 

For fun, from the same guy:

DAC

 

See the line with "100 MHz 1-bit" arrow going to the DAC section... So this is practically a DSD DAC running at 100 MHz. DSD = PDM = PWM, PWM is PDM where adjacent 1's or 0's are joined together and don't cause a state transition.

 

I just run equivalent of what "SHARC1", "SHARC2" and "SHARC3" there do, in player software, but in better way (because of more available number crunching power), especially for DSD.

[Miska]

That is precisely the problem. If 99.99% of the content is in PCM, but you "don't need it", because there is a better mousetrap (DSD), you are obviously not in it for the music.

 

No, now 99.99% of the PCM DACs on the market run digital oversampling filters and delta-sigma modulator (bunch of DSP stuff) in hardware to produce data for the actual conversion section. What goes to the conversion section is similar to what DSD is. Same happens in inverse order in 99.99% of current PCM ADC when music is being recorded.

 

I don't want any of that processing in hardware, I want a DAC that just converts to analog and nothing more. I run all that processing in software because it is cheaper and the algorithms can be much more advanced because there's lot more processing power available at unbeatable price performance ratio.

 

So your PCM music content is going to be converted anyway to something else, so my choice is to do it in software. You want to stick with doing it in hardware. That's fine, but you should realize what is really going on, end-to-end. (and most of your PCM music content is end result of conversion too)

 

P.S. Most of my music content is RedBook PCM and when I'm listening it, it goes to the DAC as DSD...

[Miska]

Who cares?

 

I obviously do...

 

Availability of content, and price/performance of the hardware/software needed to convert that content into music in a listening room.

 

And I am trying to optimize the price-performance of that... And I'm always looking for new ways how to do that, primarily how to squeeze more out of what ever reasonably priced hardware is out there.

 

The way I see it is native DSD content is a classical music niche, and price/performance of PCM based playback systems is advancing rapidly. If conversion of that PCM content to DSD somewhere in the digital chain contributes to that advance - wonderful. This does not rationalize why manufacturers are falling over themselves to release DACs with native DSD playback capability, and somehow convinced the audiophile community we actually need this.

 

If you care about shortest possible path from microphone to your speakers, DSD is one vehicle for that. And there are number of companies offering DSD content so you can make your choice.

 

Rock and pop are the genres where there's not that much DSD content, but still if the only way to get Pink Floyd in hires is to get it on SACD I'm perfectly happy. I have "Dark side of the moon" and "Wish you were here" on SACD and these are better sounding than the equivalent CD versions, which I also have...

[Miska]

So my question is: how much do you have to pay for a pre- or integrated with a truly high quality volume control that will clearly best most digital volume controls? Does it have to be "high-end" as I've read in several places?

 

You can implement digital volume in different ways, and there can be quite big quality variations. But when digital volume is implemented decently, it is very hard to beat with analog implementation (objectively evaluated). Certainly a potentiometer won't be enough. And if there are any active components involved on the signal path as you pretty much need to have, unless it's an integrated amp, those can only add distortion and noise.

 

Best solution is to match gains properly, so that there's no need for huge attenuation values anywhere in the middle just to be amplified later in the chain. DACs can output at least 2V and some pump out 6V, even 2V should be enough to drive most power amps loud enough and at least what I've seen will drive quite many to clipping already.

[Miska]

If DSD is so darn great then where are all the NEW recordings by major artist ( not classical) ? If DSD was the "go to format" then why aren't the artist demanding their music be in DSD. I surely don't know and it's not on the DSD download sites....

 

If you mean by major artist some that make big bucks for large record companies like Madonna, then it is non-existent. It is mostly AAC in iTunes or MP3's in Amazon. No any other hires format either for that matter. Good exception being Mark Knopfler's Privateering (24/96, sounds very nice upsampled to DSD128). I'm not sure if new Daft Punk would sound any different in hires, because most synths use 44.1 or 48 rates.

 

OTOH, most of trance/house/jungle/etc is published on white label vinyl and sounds very nice as 45 rpm maxi's, I have quite a pile of these too...

[Miska]

Well, try true multibit DAC's with good separated digital filters. Or use your own software oversampling with NOS multibit... if that's your pleasure.

 

PCM1704 has pretty much hit the performance wall of multi-bit PCM DACs and there are no reasonable ways to improve from there any further. It has about 18 - 19 bits worth of precision. Sabre can do about 22 bits worth, but could perform better by removing all the built-in DSP and performing better DSP off-chip, and if they would focus on the analog and conversion side instead. Even better if the new combined TI/NS would do it, they have better analog designers.

 

Let the software guys do unconstrained DSP and hardware guys focus on the actual conversion.

[Miska]

the 64fs noise shaping envelope is too close to the audio band of interest to allow even one generation of layering before the noise would become apparent.

 

That's more problem of the implementation than any technical restriction. But 64fs is too low anyway, with 512fs you can do pretty crazy stuff already.

 

There's still no DSD editing platform within Pyramix, but that could/will change as demand increases.

 

This is the key problem. Editing DSD is not a technical problem, it is implementation problem. I have plenty of DSD processing algorithms, but I don't even consider writing a DAW, because I know too well the amount of features modern DAW's have from GUI perspective, and implementing such DAW GUI would take ages for a single guy.

 

I don't believe however you'd get the same difference in a studio pop recording, which has no basis in natural sound reality.

 

Yeah, there's no way to define "realism" of a programmable software synth. And if it's doing it's maths at 24/44.1 internally, there's no point in using anything else - you can capture the output straight without anything ever going into the real world.

 

This used to be one of my favorite software toys:

Reason - Complete music making, music production and recording studio software - Propellerhead

 

And even more this, now it's an App!

http://www.propellerheads.se/products/rebirth/

[Miska]

This is the standard answer from engineers. People that have been fortunate enough to insert a real good preamp in the chain, will nearly unanimously answer the preamp sounds better. The be fair, I only discovered this a few months ago myself, and had been squarely in the no preamp camp up to that point.

 

Some people like the sound of added distortion, especially from tube amps.

 

How many have actually even tried a good gain matched DAC (without a volume control) connected straight to a power amp and then compared different digital volume controls from a player, and then compare that to an analog volume control?

 

I cannot say anything for AR Ref 10, it is not available here, closest dealer is in Sweden and it costs around 30k€. Which is around 252x more than my software based digital volume control. I think software based digital volume has pretty unbeatable price/performance ratio.

 

(I wouldn't buy 30k€ audio equipment anyway, I rather design and build my own instead, saves quite a lot, because there's at least 60% markup on the price tag anyway)

[Miska]

I do not think your comments are fair at all if you have not listened to a world class preamp.

 

I've heard these (not Ref 10 though), but could you please explain technically, how adding an extra component between two could improve quality? Any extra component on the signal path can only degrade signal.

 

Properly designed power amp should have exactly same kind of input impedance as pre-amp inputs, so from a DAC's perspective there shouldn't be any difference.

 

The preamp also gives me flexibility to compare different DACs.

 

I use a plain source switch for that, no active components on a signal path at all, only a switch.

 

It seems that you give the impressions in here that your own design software and hardware are the greatest. The other alternatives are not worth to explore.

 

No, I buy equipment and software too. I only create my own when I think I can do something better for less or same money, or if something is not available at all on the market. I have not said anything about something being greatest, but why on earth would I create and use my own stuff if it wouldn't be the best for me?

 

But I have to also objectively evaluate things, in addition to listening. Yeah, I'm boring engineer in a way that if something sounds better, but measures worse, it is no-go for me. I require better sound and better measurement results, simultaneously.

[Miska]

For audio equipment, I look inside, and if a 20 k€ device has 500 € worth of components inside, I'm just not going to buy it.

[Miska]

If we accept that any format conversion produces losses in quality, and therefore should be avoided...´

(1) if most (99%??) or the digital sources are PCM, why do you propose an additional conversion to DSD?

- are you not able to do the same digital processes and calculations without getting out of the PCM "domain"?

 

Because it allows better performance for less money. For example the best multi-bit true PCM converter, PCM1704 costs around 60$ and has accuracy of around 18-19 PCM bits. Best delta-sigma converters have accuracy of 21-22 PCM bits and these converters traditionally do that kind of conversion DSP inside and cost around $10.

 

- are PCM dacs not abble to accept that stream without changing it?

 

Most of the "PCM" DACs and ADCs are delta-sigma converters so they have to employ bunch of DSP inside to do the conversion. DSD on the other hand is 1-bit delta-sigma data, intended to bypass this back-and-forth conversion DSP at both ends, in ADC and DAC.

 

So essentially what I'm doing, is moving the format conversion DSP from the DAC to the computer. This allows making the DAC much simpler because it doesn't need to do any DSP. And also if you compare processing capabilities of a quad-core Core i7 CPU on a modern computer to capabilities of that $10 DAC chip...

 

 

(2) I believe that, concerning "the price-performance" ratio, for me and for most consumers that are not technically gifted and tweekers (99% of them, probably)...besides "price" and "performance" a third variable is extremely important: "ergonomics".

Concerning ergonomics, nothing beats a simple solution where all-in-one box (the dac) performs everything.

If this one box performs admirably well (say a Chord QBD76 for example), it's natural than a consumer is not hard pressed to change completely the perspective and adopt your landscape.

 

Sure, no doubt about that. It becomes trade-off between ergonomics, capabilities and flexibility. And also price...

 

I go for maximum capabilities at lowest price.

 

I guess what I mean is that the musical end result of your solution must be considerably better (say, of a Chord QBD76) to be a "imperative" to the average user to feel the need to change...

 

My take is that since I anyway play content from my computer using a player software, I can better utilize capabilities of the computer to improve sound while lowering cost. In this kind of scenario, it is not even clearly visible where the DSP processing happens. You have the same two components, computer and DAC, just part of the process has been moved from DAC to the computer...

 

DRC, volume control and multi-channel speaker placement processing can be performed as side effect on the computer too.

 

True DSD-capability in a modern DAC kind of turns it into a "NOS" mode and allows bypass of the internal DSP (or part of it depending on chip). You still don't loose any of the traditional capabilities though. So you can even make comparisons!

[Miska]

By the way, 60% is about the operating margin on an iPhone - how far along are you with your own design?

 

How is that related to anything? I've spent seven+ years of my career in smartphone industry, so I believe I have some kind of idea how things work there.

 

I also know what kind of margins distributors and dealers expect to get from hifi equipment, so there are no big surprises there either. Of course, after all those margins, 24% VAT is added on top.

 

My own design? Software or hardware?

 

I don't sell any of the hardware I design and build, I spent a year designing my most recent DAC. And BOM is around $1000 (not including case). I'm not stupid enough to try to sell it to anyone, it would cost so much that nobody would buy it.

 

For software I know how many hours of work I have put there and I also know what kind sales volumes to expect.

[Miska]

I sussed a long time ago that I could build a decent playback system for not a lot of money, the knowledge made me feel that some things are rather expensive from the what is inside perspective. However in the cases where the unit has properly been designed then fair plays it can cost a lot to run an R&D section.

 

If I'd had to name one product where I see price, technology and originality in balance, I would name Devialet.

 

Look inside and you can see that you are actually going to get something and someone has actually spent some time and effort on designing it too.

[Miska]

The same happens to me with switching amplifiers intrinsic distortion (noise).

 

I don't use switching amplifiers, but noise is not distortion. Distortion is correlated with input signal, noise is not.

 

Non adding nothing to the sound, but with a better impedance matching that most of the DACs I know.

 

...and since properly designed power amp has same input impedance as properly designed pre-amp (10 - 100 kOhm), how does that change anything?

 

If DAC is badly designed then that's another story, but pre-amp won't fix it anyway.

 

"Pre-amp" should be called "pre-attenuator" these days, because only vinyl sources require amplification for the cartridge, DAC output tends to end up being attenuated and not amplified...

[Miska]

Since distortion is noise to my ears, is the same (out of technical and semantic definitions)

 

Well, if you compare white noise, harmonic distortion and intermodulation distortion, they sound completely different...

 

Ultrasonic noise is just inaudible hiss. Are you hearing a constant hiss from class-D amps? It is there even without input signal.

 

I don't know about the amplifiers you use, but saw your Devialet recommendation.

 

It was comment about cost/originality/substance ratio, not a recommendation.

 

I use traditional class-A/B amps. The main amp I use operates in class-A to 10W output power and in class A/B from there. And can spit out 90+ amps of current to challenging speaker loads.

 

A well designed pre-amp could do magic to some bad designed analogue output stage on some DACs, like impedance matching, even if the amplifier has the right impedance input.

 

There would be the same impedance mismatch between DAC and pre-amp too. So I still didn't get the point why there would be less impedance mismatch between DAC and pre-amp than between DAC and power-amp?

[Miska]

Yes, there are aspects that seem to need a few turns here and there This was a surprise for me, as I was inclined to think the shortest path was the best. I only know that Peter (PeterSt) recommends to skip the preamp with the NOS1, but I am too "chicken" to allow the risk of blowing the speakers if something odd occurs...

 

That's why I'm talking about gain matching. Match gain of the power amp such that it doesn't go too loud even if DAC outputs full level. Then you don't need much attenuation from that level in normal cases. Likely the adjustment range stays within 20 dB. (and in cases of accidents there is no damage)

 

This way it is possible to reduce amount of thermal noise introduced to the signal along the way.

[Miska]

Do you mean adjusting the software digital volume ? I am not too sure about the "match gain" definition... ?

 

No, I mean adjusting gain of the power amp, or if not possible placing a fixed attenuation on the power amp input.

[Miska]

Most modern amplifiers have far too much gain to be used directly with a DAC unless some kind of attenuation is used.

 

This is a problem, because such large attenuation and then again large gain back will just increase thermal noise levels and add distortion.

 

Also, many DACs don't like relatively long interconnects due to the higher capacitance. Many modern DACs use an opamp in the output with for example, a series 100 ohm output resistor to help reduce the problems of capacitive loading, but that isn't always quite enough. In such cases it may be better to use a preamp to drive the longer cable to the P.A.

 

That is a design problem of the DAC and only indicates poor analog design. Opamp itself doesn't make it any worse, there are suitable opamps and current buffers for the purpose and you can also design circuit properly around an opamp. You can also keep power amp closer to the pre and use longer speaker wires because that doesn't cause any problems.

 

My own discrete Class A preamp is capable of correctly driving a 75 ohm load, so longer cable differences are not a problem if the input of the P.A. is also terminated with 75 ohms.

 

My DAC output is designed to be specifically able to handle capacitive loads. I also use cabling (Supra) that has been properly designed to minimize capacitance on interconnects and inductance+resistance on speaker cables. Also shield is correctly connected only from the source end.

 

In addition, I have 30 cm between pre-DAC and power and then more length on the low-inductance speaker wires. This setup goes just fine to 1 MHz without notable attenuations.

 

75 ohm is just 27 mA at 2V output while 40 mA output capability is pretty normal for opamps. My DAC can output up to 400 mA so it can drive 5 ohm loads to 2V.

 

IIRC, JA of Stereophile measures output stages also to 600 ohm loads, which is less than 1/10th of what a proper power amp should have.

 

I agree with Roch regarding digital attenuation. With many DACs, more than 10dB attenuation leads to severe SQ degradation with 16/44.1. 24/96 does of course fare a little better.

 

Why would the input material matter? DAC's internal pipeline is fixed word length, so it is for example 32-bit regardless if you input 16-bit or 24-bit data.

 

But I'm mostly talking about software volume in a player application. I've been testing digital volume controls in various ways. For example I can first attenuate down to -120 dBFS level and then digitally amplify it back and it still sounds completely fine. If -120 dBFS attenuation still sounds fine, anything less will definitely have less degradation...

 

P.S. Now there's a new loudspeaker cable from Supra I want to try out Quadrax at 0.3 µH/m it is very low inductance cable. My current one is 0.4 µH/m and 4.3 Ohm/km. Interconnect is 52 pF/m.

[Miska]

How are you going to make a LARGE reduction in the gain of an amplifier without getting into a lot of major changes regarding VAS stage compensation etc. ?

 

It is not that major change. Alternatively you could place two resistor attenuation on the input.

 

With players such as the Oppo 981 etc. , 10dB fixed digital attenuation very noticeably degraded the SQ of 16/44.1 yet 24/96 was still quite acceptable.

 

I'm sure it is possible to find many badly implemented digital volume controls (not surprising at all for a Mediatek-based device), but it cannot be generalized.

 

It is of course possible to find a lot of badly implemented hardware, but for that reason measurements and looking inside are good.

 

Many sound better with more than 100 ohms resistors at the output. Even the capacitance of FETs such as IRF7905 used to replace the now hard to get special muting transistor 2SC2878, which is an NPN transistor with an unusually high reverse hFE of 150 can degrade the performance of many current op-amps. Perhaps it's time to go back to using relays for muting purposes ?

 

100 ohm output resistor again increases thermal noise and the sound "improvement" is probably mostly due to filter effect in combination with cable and input capacitance.

 

Muting transistors are bad, I use Omron relays. And IIRC, for example TEAC UD-501 also has relay muting. I think my Fostex HP-A8C also has muting relays. Mytek on the other hand doesn't have muting at all.