Just got a Yggdrasil!

[PeterSt]

9 hours ago, crenca said:

Can anyone speak (or perhaps have a link) about what how low level but "dirac" like spikes in the signal from the DAC effect the output signal/wave of typical amps?  It occurs to me that this sort of distortion could have outsized effects on an amplifier...which would lead to the inaudible being audible

 

 

My idea about this is that it will be harmless. The level of it is too low. The question is a good one though because in my opinion many of the perceived effects are "generated" outside of the DAC itself, but caused by it anyway. The bandwidth of amplifiers was given already (maybe it was not in this topic ?) and while people tend to think 1MHz bandwidth is a good thing, it may be not at all because what with the poor speakers. Or, with 200KHz bandwidth of the amp, how does the amp itself respond to that (speakers are fine now). It is all not so easy.

 

Not addressed to crenca, but generally :

 

What I don't understand is how people keep on thinking that such a thing will be inaudible. I mean, when I don't see any changes in whatever measurements it can still easily be a matter of life and death, so to speak. USB isolation is such an example (and don't dig the Intona hoopla because that uses a (commercially) biased base). So when all is right, no Intona or our own Phisolator for that matter shows a hoot of a difference. Still the audible difference is day and night.

IOW, what you should digest is that once differences *are* visible (with some reference as the base) then you can completely forget a good result. If I would not behave like that, then I would be selling stuff which is only liked by half of the people, because it would be subjective to my ears. So yes, we humans can really like distortions, but never for infinitely long because it implies flavors.

 

The sad thing of course is that all should be measurable but *or* we don't know what to measure, *or* we don't have the equipment for it. Or, like with the USB isolation example, we don't know to apply the measurement without influencing/molesting the signal itself. OK, I don't know how to, which does not mean nobody does. But those who may know don't have the DUT (device under test) which in our case may come down to a complete audio system. Example of that : telecom companies will have the super expensive gear but they don't live in my listening room.

 

Anyway, if something does not measure right, it will BE wrong, audibly. If you can't accept that then all Mani-contributions are without sense to you. Mine the same of course. This is a pity, because when we'd all take each other's opinions with some respect and value, audio will progress again (I always explicitly say that when I wouldn't have criticizing customers, no good product would have been born in the first place as I don't have the pretension that my ears know all).

 

[Ralf11]

" if something does not measure right, it will BE wrong, audibly."

 

what does this mean?

 

do you claim that any measurable distortion can be heard?

[semente]

6 hours ago, Ralf11 said:

" if something does not measure right, it will BE wrong, audibly."

 

what does this mean?

 

do you claim that any measurable distortion can be heard?

 

Some distortions will be difficult to hear, some people will have more trouble hearing some of them than others, some distortions are universally recognised as being pleasant to certain individuals.

 

See the AP presentation on youtube that @Jud attended and the piece on Analogue Warmth at Sound on sound.

 

R

[gmgraves]

6 hours ago, Ralf11 said:

" if something does not measure right, it will BE wrong, audibly."

 

what does this mean?

 

do you claim that any measurable distortion can be heard?

 

Let's look at this another way. Take two 100 Watt amplifiers. Amp A has 0.0001%  THD. Amp B has 0.00011%THD both at 100 Watts. That means that amp B has .00001% more distortion than does amp A. Mani would say that this extra 1/100,000th of a percent disqualifies amp B from consideration because it measures as being higher in THD than amp A. In the real world, however, that tiny amount of of added distortion is completely inaudible to anybody and any sonic differences between these two competing amps will not by caused by that minuscule amount of added distortion. Inaudible to the human ear is inaudible. And that 45 microvolt anomaly in the DAC's waveform is one of those things that are totally inaudible. To put it in perspective, 45 microvolts is about as much voltage as is fed to the antenna terminals from a simple folded dipole on a good suburban FM  station that gives full quieting.

[semente]

4 minutes ago, gmgraves said:

 

Let's look at this another way. Take two 100 Watt amplifiers. Amp A has 0.0001%  THD. Amp B has 0.00011%THD both at 100 Watts. That means that amp B has .00001% more distortion than does amp A. Mani would say that this extra 1/100,000th of a percent disqualifies amp B from consideration because it measures as being higher in THD than amp A. In the real world, however, that tiny amount of of added distortion is completely inaudible to anybody and any sonic differences between these two competing amps will not by caused by that minuscule amount of added distortion. Inaudible to the human ear is inaudible. And that 45 microvolt anomaly in the DAC's waveform is one of those things that are totally inaudible. To put it in perspective, 45 microvolts is about as much voltage as is fed to the antenna terminals from a simple folded dipole on a good suburban FM  station that gives full quieting.

 

Which would mean that the differences Mani could hear and recorded have a different cause? 

 

R

[Jud]

38 minutes ago, semente said:

 

Some distortions will be difficult to hear, some people will have more trouble hearing some of them than others, some distortions are universally recognised as being pleasant to certain individuals.

 

See the AP presentation on youtube that @Jud attended and the piece on Analogue Warmth at Sound on sound.

 

R

 

I remember the presentation well, particularly the demo of slew rate limiting that no one else seemed to mind much and I thought was really damn irritating.

[Jud]

8 hours ago, PeterSt said:

So yes, we humans can really like distortions, but never for infinitely long because it implies flavors.

 

8 hours ago, PeterSt said:

Or, like with the USB isolation example, we don't know to apply the measurement without influencing/molesting the signal itself. OK, I don't know how to, which does not mean nobody does. But those who may know don't have the DUT (device under test) which in our case may come down to a complete audio system. Example of that : telecom companies will have the super expensive gear but they don't live in my listening room.

 

8 hours ago, PeterSt said:

I always explicitly say that when I wouldn't have criticizing customers, no good product would have been born in the first place as I don't have the pretension that my ears know all

 

Perhaps you are getting better at expressing yourself in English or I am getting better at understanding you; either way, nicely said.

[manisandher]

57 minutes ago, gmgraves said:

 

Let's look at this another way. Take two 100 Watt amplifiers. Amp A has 0.0001%  THD. Amp B has 0.00011%THD both at 100 Watts. That means that amp B has .00001% more distortion than does amp A.

 

 

Haha... It also means that both amp A and amp B have an order of magnitude lower THD than the Yggy, due to the Yggy's glitching . It wasn't for nothing that the designer of the AD5791 chip said:

 

"... 

I always assumed the code-change-dependent glitch energy would make THD unacceptable to audio buffs."

 

Mani.

[Jud]

3 minutes ago, manisandher said:

 

 

Haha... It also means that both amp A and amp B have an order of magnitude lower THD than the Yggy, due to the Yggy's glitching .

 

 

Mani.

 

An order of magnitude higher would be 0.001%, right?

[Jud]

I'd like to be fair here and quote the entire Head-Fi post from the chip designer:

 

Quote

This is exciting!  I work for ADI and designed the AD5791 & AD5781 DAC IC's in the Yggdrasil and Ragnarok.  But that was about 2009.  I'm delighted a new market has been found for these products.  As the Yggy blurb explains, I never considered audio when I designed these DACs.  In fact the main motivation at the time was medical imaging - the nice folks who make MRI scanners are awfully fussy about performance, and for good reasons!
 
To be honest, I'm genuinely surprised this makes a good audio DAC.  It's terribly expensive and I always assumed the code-change-dependent glitch energy would make THD unacceptable to audio buffs.  (There are some ingenious board-level tricks around this but they are difficult and expensive to achieve - I'd love to know if the Yggy designer is using them ... feel free to get in touch!)  On the other hand, the AD5791 architecture is unbeatable for noise.  I really don't know what you audio perfectionists are hearing when you listen - is it THD ... or noise?  Some combination of both, I imagine ...
 
Anyway, after years of designing products for specialist scientific, medical and industrial equipment I really like the idea of seeing my design in something I could use myself ... must get one of these!  Given the prices, and my lack of sophistication when it comes to appreciating audio fidelity, I think it'll be the Ragnarok ...
 
Roddy

 

[manisandher]

14 minutes ago, Jud said:

 

An order of magnitude higher would be 0.001%, right?

 

Yep. Actually around 0.0025%.

 

Mani.

[manisandher]

11 minutes ago, Jud said:

I'd like to be fair here and quote the entire Head-Fi post from the chip designer:

 

When I first quoted him in this thread, I did say:

 

On 4/18/2017 at 7:14 AM, manisandher said:

[Though in fairness, he does go on to say that he's glad that someone is at least trying his chip for audio purposes.]

 

I didn't feel it needed to be said again.

 

Mani.

[Jud]

7 minutes ago, manisandher said:

 

When I first quoted him in this thread, I did say:

 

 

I didn't feel it needed to be said again.

 

Mani.

 

The bits I thought were intriguing were where he says (1) there are board-level "tricks" to get around the glitching energy, and (2) that the chip is "unbeatable for noise."

 

Edit: I don't know whether the "tricks" he refers to were used in Yggy, and whether the measurements with some level of glitching show either that said tricks weren't used or that this is the level of glitching one would expect to measure after the tricks *were* used.

 

As usual, it's all perfectly clear to me.

[manisandher]

On the other hand, the AD5791 architecture is unbeatable for noise.  I really don't know what you audio perfectionists are hearing when you listen - is it THD ... or noise?  Some combination of both, I imagine

 

It seems strange to me to use an ultra-low noise DAC chip, if the chip's THD remains high due to glitching. Of course, it could be that the reason the AD5791 was chosen had nothing to do with its ultra-low noise, but for other reasons. Perhaps the (lack of) availability of audio R2R chips?

 

Mani.

[Jud]

11 minutes ago, manisandher said:

 

 

 

It seems strange to me to use an ultra-low noise DAC chip, if the chip's THD remains high due to glitching. Of course, it could be that the reason the AD5791 was chosen had nothing to do with its ultra-low noise, but for other reasons. Perhaps the (lack of) availability of audio R2R chips?

 

Mani.

 

If I'm remembering right, Jason and Mike have pretty well said as much (lack of availability of favored R2R chips) at Head-Fi, or maybe that's just me remembering my own "reading between the lines."

[jabbr]

9 hours ago, PeterSt said:

 

The sad thing of course is that all should be measurable but *or* we don't know what to measure, *or* we don't have the equipment for it. Or, like with the USB isolation example, we don't know to apply the measurement without influencing/molesting the signal itself. OK, I don't know how to, which does not mean nobody does. But those who may know don't have the DUT (device under test) which in our case may come down to a complete audio system. Example of that : telecom companies will have the super expensive gear but they don't live in my listening room.

 

My wife has given up trying to ask me what is in the series of 70lb boxes that have been arriving, and immediately brought down to my basement workshop/lab ... oh dear I am running out of room for cases of wine

 

I will say this, I am spending about 25% of the cost on shipping so that gives you an idea of what a formerly $100k piece of test equipment goes for  

I think this is being driven by a new round of technology, essentially digital (google "SDR") so basically hook up a 4-6 Gsps 12-14 bit ADC/DAC** to an FPGA and Voila! Now this is all on a chip and the prices of this technology are decreasing but there are already products (think digital oscilloscope).

 

** http://www.analog.com/media/en/technical-documentation/technical-articles/Maximizing-the-Dynamic-Range-of-Software-Defined-Radio-MS-2735.pdf

http://www.analog.com/media/en/technical-documentation/data-sheets/AD9625.pdf

 

Now why would this bandwidth be necessary? Well you want to measure phase right?

[Jud]

If anyone wonders what all this glitching is about and what are some techniques for helping to make it go away, here are a couple of blog posts from TI's web site:

 

https://e2e.ti.com/blogs_/b/analogwire/archive/2013/06/14/what-s-with-all-this-glitch-ing

 

https://e2e.ti.com/blogs_/b/analogwire/archive/2013/07/01/dac-essentials-glitch-be-gone

 

[jabbr]

6 minutes ago, Jud said:

 

If I'm remembering right, Jason and Mike have pretty well said as much (lack of availability of favored R2R chips) at Head-Fi, or maybe that's just me remembering my own "reading between the lines."

Yep. Achieving better resistor trimming than the PCM1704 is not cost effective. New DAC chips have different specs: http://www.analog.com/en/products/digital-to-analog-converters/high-speed-da-converters/ad9164.html#product-overview and just think: the other day they might have sold 59 or whatever of those in like one hour  

[manisandher]

On 4/18/2017 at 10:41 AM, PeterSt said:

... and re-engage those files 

 

I'm going to do things slightly differently...

 

I'd really like to give anyone interested an idea of what I heard when I had the Yggy here, and why I started exploring the Yggy's measurements, to perhaps find a correlation between these and the sonic signature (or flavour) I was hearing from the Yggy.

 

If you have the inclination, download the following files and take a listen. In particular, listen out for anything resembling "bold incisiveness" or "artificial detail" in the Yggy file, that doesn't exist in the original file. (At some later point, I will link to the 24/48 capture of my regular DAC playing the same original file through the same playback/recording chain, to show that it wasn't the playback/recording chain causing the difference between the Yggy file and the original file.)

 

Original 16/44.1 file: https://drive.google.com/open?id=0B0PU5LO5jVjfZFI4ZnR1SHhiSmc

 

Yggy 24/48 capture: https://drive.google.com/open?id=0B0PU5LO5jVjfd3FfY1c4WW9FQVk

 

Note that the Yggy file is 6dB down in level to the original file. You'll have to account for this of course. (And I hope the change in format between these files doesn't prove too troubling.)

 

Mani.

[Jud]

4 minutes ago, jabbr said:

Yep. Achieving better resistor trimming than the PCM1704 is not cost effective. New DAC chips have different specs: http://www.analog.com/en/products/digital-to-analog-converters/high-speed-da-converters/ad9164.html#product-overview and just think: the other day they might have sold 59 or whatever of those in like one hour  

 

Mike Moffat has said he prefers the PCM63 to the PCM1704.  (Actually he said the PCM1704 "sounds like ass.")

[manisandher]

5 minutes ago, Jud said:

 

Mike Moffat has said he prefers the PCM63 to the PCM1704.  (Actually he said the PCM1704 "sounds like ass.")

 

Yeah, he also thought the PCM1704 was only capable of accepting a 96kHz input... until I pointed out to him that it could actually accept rates up to 768kHz. His come-back was, "it said so in the datasheet". In any event, I reckon he was thinking of the PCM1702, which many people claimed was worse-sounding than the PCM63 when the former first came out.

 

Mani.

[jabbr]

15 minutes ago, Jud said:

 

Mike Moffat has said he prefers the PCM63 to the PCM1704.  (Actually he said the PCM1704 "sounds like ass.")

 

That's amusing if you think about it. Even the designer of the AD chip they selected (and got to work after considerable work) admitted the was surprised it could be made to work for audio. So let's just imaging that someone could say: "the AD5791 sound's like ass" and he would retort: "In your puny hands it does but given my skills it sounds great", right? So ...  maybe what he really doesn't like is the DF1704 DSP chip that is what has the 96k limitation.

 

Some people favor the PCM63 but I favor the PCM1704 because it *doesn't* do upsampling (you don't need to use the DF1704 in front of it)

 

But from a commercial POV, if you don't have a large stash of them you are SOL. I happen to have a very small stash (8 or 16) sitting on the shelf for a very very rainy day  Maybe @PeterSt will trade me an NOS1a for my stash :rotfl:

[audiventory]

45 minutes ago, Jud said:

If anyone wonders what all this glitching is about and what are some techniques for helping to make it go away, here are a couple of blog posts from TI's web site:

 

https://e2e.ti.com/blogs_/b/analogwire/archive/2013/06/14/what-s-with-all-this-glitch-ing

 

https://e2e.ti.com/blogs_/b/analogwire/archive/2013/07/01/dac-essentials-glitch-be-gone

 

If each DAC work on own half of amplitude (positive or negative) such overload 7fff > 8000 and 0000 > 7fff are probable as glitch reason.

[Jud]

1 minute ago, audiventory said:

If each DAC work on own half of amplitude (positive or negative) such overload 7fff > 8000 and 0000 > 7fff are very probable as glitch reason.

 

Right, overshoot due to transitions in M(ore)SB's caused by changes in L(ess)SBs.

[Jud]

7 minutes ago, jabbr said:

maybe what he really doesn't like is the DF1704 DSP chip that is what has the 96k limitation

 

Ah.