[ostewart]

   
I'll leave the detail to iFi Audio to answer, if they are willing to go into the detail that is, since these are more or less their trade secret for now and they probably don't want others to copy what they have done. Hence why they are being a bit cryptic on this discussion.
 
I asked them the same question awhile back, and the answer is yes, the DSD1793 can't do DSD128, DSD256, DXD, etc if you use the default configuration that TI tells you to use. But deep inside the silicone, the chip itself is actually capable of native decoding for those formats. The way iFi is able to explore the DSD1793's full potential is by employing a customized XMOS implementation that allows the DAC to do those native decoding function that are otherwise hidden by TI.

 
Great answer ClieOS, after having good contact with iFi for a while, it is true what you say. And they write all their own code for the XMOS chip and driver, so in there they can squeeze out more than the tech sheet states. They don't have the DAC in it's default configuration so they do decode natively.

 

[analogsurviver]

   
Great answer ClieOS, after having good contact with iFi for a while, it is true what you say. And they write all their own code for the XMOS chip and driver, so in there they can squeeze out more than the tech sheet states. They don't have the DAC in it's default configuration so they do decode natively.

Similar practice is used in lots of electronics - and while looking a bit deeper, one can find that basic unit and top of the line from the same manufacturer basically has the same chipset - only many features on lower priced models are simply disabled. Thus it is easy to "cover" the entire range with basically one single design.

 

[ostewart]

Great way of putting it

 

[earwaxxer]

I agree with the above - Its about the sound. - We tend to be paranoid about being 'ripped off', some of that is justified, and some is not. If you want to talk about snake oil in audio, I would look at the $5K+ kit thats out there. I would argue that most of that high priced audio jewelery is a 'rip off'.

 

[KmanChu]

  I agree with the above - Its about the sound. - We tend to be paranoid about being 'ripped off', some of that is justified, and some is not. If you want to talk about snake oil in audio, I would look at the $5K+ kit thats out there. I would argue that most of that high priced audio jewelery is a 'rip off'.

Absolutely, and at those much higher prices the front end electronics tend to be more different than better. Measurements go by the wayside for what becomes "involving" and the sound we like vs what is arguably technically better or worse (e.g. anything with vacuum tubes, which I love BTW.)

 

[michilumin]

I've had it for about a day. So there's going to be some time involved in critical listening and comparing it to say, the Schiit Loki (which we know is DSD64 only) and some ESS Sabre driven stuff.
 
I also fully agree on the $5k+ snake oil "noise".  Diminishing returns fade well before that level.  It'd be nice if the level of legitimate returns was $189.
 
As for an expertise pissing-contest, I'd rather not. But if you really want to, I will. Suffice it to say that no, this is not my first rodeo, and if you must have credentials, PM me, but in my recent correspondence with iFi, yeah, they have that. I work in DSD recording exclusively, meaning the ADC side of things, and I have an interest in these rates reaching their full potential. So I'm hoping, really, for the availability of DACs that can do DSD128 and DSD256 to take advantage of what we're recording.
 
I just want to know if this DAC is actually doing what it says, because otherwise, it's not a "complete" exhibition of what DSD128, DSD256 and DXD can do. And no, for $189, I wouldn't expect it to be, but hope springs eternal.
 
Yes I expected these sychophantic "Who cares if it can do what's advertised or not, how does it sound?" responses. But it does matter, again, as yes, there are supposed to be advantages in the higher rates (pushing the noise ramp further from the audible band, for example.)
 
Which is odd, when people tend to get up in arms if they find out their DSD DAC "converts to PCM internally".
 
So far what everyone is saying is "Well, magic."  - No. DAC chips are REALLY simple creatures.
They either take the data given to them, or they don't.
With some exceptions, but I have yet to see iFi's explanation on that. (If I'm missing it, or didn't see it, please, point it out.)
 
I agree that it's possible that the DSD1793 (which is actually the same silicon as the DSD1791 that was used in a lot of lower-end SACD players in the early to mid 2000s) may be able to be, ostensibly, overclocked. 200kHz is not THAT far off from 384kHz; it is possible.  32 bits, well, I'm not sure how they'd be doing that. I'd be interested to know how they are without truncation, and maybe the answer is "dither" in the XMOS chip. Fair enough.
 
So, while there's a lot of flag waving here about "I don't care what goes on in the black box" and some "Does it matter? Trust your ears." of the likes of Bedini Clarifiers and Auric Illuminator gels, and a little bit of "I want to believe" (believe me, I want to, too) - really the only thing to conclude here is that iFi has the XMOS chip downsampling or truncating. And yes, that means we may not be getting the full potential of DSD128 or DSD256, DXD, etc.
 
If everyone's OK with that, then why not just downsample in software, where arguably a CPU can do it with a lot more accuracy than the rather anemic XMOS chip.
 
True believers, yeah, not going to change your minds.  And I actually really like how the iFi Micro is shaping up, (though, same chip in question when their non-DSD dac uses the ESS Sabre chipset.)
 
But, "How does it sound?" - Well, I don't know, I can't tell what I'm *NOT* hearing, so, if this was *REALLY* doing DSD128 or DSD256, then maybe it would sound better than it does.
 
There is the idea of, well, it could indeed be running at DSD64 or PCM 24/192, and that's "good enough" for the data that's being resampled/dithered (hopefully) or truncated (hopefully not) at the XMOS chip.
 
In which case, we have a $189 really-good DSD64 / PCM 24/192 dac that can accept higher input rates. Cool. No sweat.
But I don't know why people get so defensive about just trying to find out.

 

[boxinghris]

Yes, it is 'odd' when judging a product by its sonic performance is viewed as sycophantic, very odd indeed.

We should all skip the listening tests in reviews and jump to the bench tests, where objective results will dictate how much we 'like' a product, and if that product is worth buying.

How the objective specifications relate to the subjective performance is irrelevant, although in the case of the iFi DAC it is not such a clear cut objectivist v subjectivist debate, as the objective performance is, I believe, as stated in the literature, and you may yet be able to 'like' your DAC and revel in both it's sonic and processing prowess...

 

[michilumin]

  Ifi has obviously tested the DSD1793, along probably pretty much any other chip, for what it can do on paper - and beyond. They must have found out some way to squeeze out the DSD256 out of DSD1793; they did say they used some features not mentioned by Burr Brown/Ti in their data sheet.
 
Now it would be interesting to see to how much of this will be made public - provided it really can do DSD256 in the first place. Foobar2000 with nano clearly recognizes DSD256 as such. JRiver 19 does not indicate what is it playing directly; only kbps is displayed. For DSD128 it is 11289 kbps, for DSD256 it is 22579 kbps - which is exactly the same kbps for the DXD352.8 kHz .

 
Gotcha. Yes. I understand that. The XMOS chip does take these data rates and, what it does after that, nobody's sure of.  The "very high end" (YMMV) PS Audio Direct Stream DAC uses the same XMOS chip, and actually converts to *extreme* rate PCM in the chain. So who knows. And if iFi is able to drive the chip at a 384khz (vs 200khz) PCM clockrate, great - kudos for sure.  Same with these high DSD samplerates. And I actually do believe that if dither is occurring at the XMOS level for 32->24, that's also fair, as dither is a legitimate method of getting more SNR out of a given bitdepth, at least perceptually, and perceptually is what matters.
 
I don't expect iFi to expose trade secrets, but, I really do wonder if it'd be "giving away the store" or not to say either "Yes, the DSD1793 is running at max *spec* rates" or "No, the DSD1793 is running beyond those specs."  
 
"Does it sound *good*?" - Yeah. Sure does.  Could it sound better? Dunno. That's what I'm trying to find out, and it's hard to navigate that field without stepping on toes and setting off land mines it seems. It's not like iFi/AMR epoxied over the DSD1793 in an attempt to hide it or anything. So yes, maybe they figured out some wonderful stuff.  But really, i'm not interested in "deconstructing" the iDSD (in a negative light), just curious on trying to figure out how it works.

 

[michilumin]

Yes, it is 'odd' when judging a product by its sonic performance is viewed as sycophantic, very odd indeed.

We should all skip the listening tests in reviews and jump to the bench tests, where objective results will dictate how much we 'like' a product, and if that product is worth buying.

How the objective specifications relate to the subjective performance is irrelevant, although in the case of the iFi DAC it is not such a clear cut objectivist v subjectivist debate, as the objective performance is, I believe, as stated in the literature, and you may yet be able to 'like' your DAC and revel in both it's sonic and processing prowess...

 
I'll state my point again: We may not be getting the full benefit of the higher rates if this isn't doing them. From all accounts, it appears to be a "friggin awesome" DSD64 DAC. Better than the Schiit Loki by quite a bit.  But, what if that's where it bumps its head?
 
While it may sound good indeed, it perhaps could sound even better.
I never said that bench tests are all that matters. They don't matter much at all. I honestly couldn't give two ****s about hooking it up to an o-scope. 
 
I'm merely interested in whether or not we're hearing the maximum advantage that could be exploited by a true DSD128 or DSD256 DAC.
 
And we may not be.
That doesn't mean "the iFi iDSD Nano sucks" , or sounds bad, it just means that if we want a true window into to these higher rate formats, this may not be it - at least not fully.
 
But then again, it may.  That's the point of asking.  I've bought the thing, and I'm not returning it, and I'm really following their dev on the iDSD Micro, which I'll probably get as soon as it's available.

 

[ClieOS]

Boys, for the benefit of the doubt, lets give iFi Audio the chance to explain before further discussion. I would think it is only fair as such.

 

[michilumin]

  Boys, for the benefit of the doubt, lets give iFi Audio the chance to explain before further discussion. I would think it is only fair as such.

 
Agreed, and will do so. Thanks.

 

[john57]

While we wait on the official IFI statement I wanted to mention a few things I found out today. On Computer Audiophile  a reviewer, John Bamford, stated that PS Audio Direct Stream DAC is the best he has heard but did not quite measured well as he was saying. My understanding that is that Direct Stream converts everything to high multi bit for DSP processing and then gets converted back to 2xDSD. The 2xDSD signal then goes directly to a passive analog filter then  to output transformers. I have not read this review but I have heard the Direct Stream and it is very good, much better than their previous product.  My current understanding of DSD is that you only need a low pass filter on the signal and you are done. Regardless of the DSD being 2x 4x or 6x you just need a low pass filter and it can be analog and you are done. I would think it would be silly to try to down sample  6xDSD without going thru a PCM stage and why would you do that?   IFI did stated that the filter used for DSD is analog while for PCM it is digital. IFI also stated that the volume control again is analog. IFI again also stated that no DSP is being used and the Xbass and 3D is done in the analog domain. I am making some assumptions and I will see if my educated guess will be close to the official IFI statement. I trust my ears, myself regardless of the outcome or by others.

 

[KmanChu]

 
I'll state my point again: We may not be getting the full benefit of the higher rates if this isn't doing them. From all accounts, it appears to be a "friggin awesome" DSD64 DAC. Better than the Schiit Loki by quite a bit.  But, what if that's where it bumps its head?
 
While it may sound good indeed, it perhaps could sound even better.
I never said that bench tests are all that matters. They don't matter much at all. I honestly couldn't give two ****s about hooking it up to an o-scope. 
 
I'm merely interested in whether or not we're hearing the maximum advantage that could be exploited by a true DSD128 or DSD256 DAC.
 
And we may not be.
That doesn't mean "the iFi iDSD Nano sucks" , or sounds bad, it just means that if we want a true window into to these higher rate formats, this may not be it - at least not fully.
 
But then again, it may.  That's the point of asking.  I've bought the thing, and I'm not returning it, and I'm really following their dev on the iDSD Micro, which I'll probably get as soon as it's available.

 
No matter what the iDSD sounds like, of course it could sound better!! It's only a ~$500 DAC. No matter how good it sounds there is probably an old school 16bit non-oversampling DAC that would cost $20k and sounds better. The iDSD is a tiny DAC, built to low cost (in absolute terms) with lots of features. Most multibit DACs that take 24 bit signals don't make use of the full 24 bit depth. Anyway, don't expect the iDSD to be any kind of indicator of what quad or 8x DSD can deliver. But it may be able to take a 8x DSD signal and make it sound better than a 4x DSD signal.
 
I find it interesting that you would question iFi's technical specs because they are a very forthright, open, and responsive company. This is basically AMR we are dealing with, not some fly-by-night operation. They have made very clear that they are philosophically committed to minimum digital signal manipulation and native conversion (as opposed to other companies who embrace DSP and processing in the digital domain.) Again, I understand you are not aggressively attacking the company but rather seeking clarification. Just understand that in their short time they have amassed a very loyal following for a reason: they have typically promised a lot and then over-delivered! 

 

[ostewart]

"The "very high end" (YMMV) PS Audio Direct Stream DAC uses the same XMOS chip, and actually converts to *extreme* rate PCM in the chain."
 
No trying to stir up any more. But @michilumin you do work for: http://www.superaudiocenter.com/index.htm and also have good contact and friendship with PS Audio and Mytek, and you also distribute Mytek (maybe you should be labled a member of the trade here on Head-Fi?) This is also your twitter, where you are clearly slating iFi/AMR: https://twitter.com/michilumin
 
iFi have already explained what they do in a simple way, I don't believe they need to tell you everything they do, because otherwise any other company can do the same, which is not their goal.
 
They have a loyal following due to their open attitude, and also explanation of what they do. They do care about audio quality and always say that measurements are important but not the last word in audio quality.
 
Read here about chip used: http://www.head-fi.org/t/711217/idsd-micro-crowd-design-the-headphones-used-at-ifi-we-reveal-all-well-a-fair-bit-anyway-page-76/690

 

[koolas]

Here are some snippets from DSD1793 white paper:

APPLICATION FOR DSD FORMAT (DSD MODE) INTERFACE

Feature
This mode is used for interfacing directly to a DSD decoder, which is found in Super Audio CD? (SACD) applications.
The DSD mode is accessed by programming the following bit in the corresponding control register:

• DSD = 1 (register 20)

The DSD mode provides a low-pass filtering function. The filtering is provided using an analog FIR filter structure.
Four FIR responses are available, and are selected via DMF[1:0] of control register 18.

Requirements for Bit Clock and System Clock
The bit clock (DBCK) for DSD mode is required at pin 4 of the DSD1793. The frequency of the bit clock can be N
times the sampling frequency. Generally, N is 64 in DSD applications.
The interface timing between the bit clock and DSDL, DSDR is required to meet the setup and hold time specifications
shown in Figure 39.

THEORY OF OPERATION

The DSD1793 uses TI’s advanced segment DAC architecture to achieve excellent dynamic performance and
improved tolerance to clock jitter. The DSD1793 provides balanced voltage outputs.
Digital input data via the digital filter is separated into 6 upper bits and 18 lower bits. The 6 upper bits are converted
to inverted complementary offset binary (ICOB) code. The lower 18 bits, in association with the MSB, are processed
by a five-level third-order delta-sigma modulator operated at 64 fS by default. The 1 level of the modulator is equivalent
to the 1 LSB of the ICOB code converter. The data groups processed in the ICOB converter and third-order
delta-sigma modulator are summed together to an up to 66-level digital code, and then processed by data-weighted
averaging (DWA) to reduce the noise produced by element mismatch. The data of up to 66 levels from the DWA is
converted to an analog output in the differential-current segment section.
This architecture has overcome the various drawbacks of conventional multibit processing and also achieves
excellent dynamic performance.

So, here are the conclusions to draw:

• Guys in TI tested this chip with 64 fS, and
• this is the clock they advise you to use, because they know it works
• they didn't test 128 fS or 256 fS, or tests results were not satisfactory
• they guarantee chip works at 64 fS, but they are not telling you not to use 128 fS or 256 fS
• they also tell you, whatever clock you decide to use, you must satisfy timing requirements shown on Figure 39
• they also tell you that SDM modulator for PCM mode runs at 64 fS

And here is what I think:

• iDSD plays any PCM worse than DSD128, because in PCM mode it uses 64 fS, and in DSD128 it has to use 128 fS
• Me and someone else also have noticed that iDSD plays PCM as good as DSD64, and the reason probably is 64 fS

So, if assuming all above is true, then can we have new firmware that will use 128 fS for PCM playback, or is it fixed to 64 fS in the chip and nothing can be done?