Hi Jud:

Glad you got your DAC up and running. I know you are shy just yet about who did the DAC board for you, but can you tell me which USB>I2S board you have that is taking 352.8/384kHz? Is it an XMOS-based interface, because I have been trying to track down some XCORE firmware code to license for 352.8/384.

So tell us about your power supply. How about some pictures!

Lastly, what value of output coupling caps are you using? Is it a tube output stage?

I am the owner of the MusiCap brand of film-and-foil (not metalized film) polypropylene capacitors. I have them made to my specs with specialized films, end spray, solder, silver-plated wire, dimensions, curing, etc. I have high-end loudspeaker and electronics OEM clients around the world, some of whom I have been supplying since 1986 (while at my former co-founded firm Hovland Company which we closed in 2009). I'd be happy to send you a sample pair to try in your DAC since I think we have a similar ear for sound.

Just let me know what value you would want (3.0uF/200V or 4.0uF/400V would be typical and are in stock). Ignore most of the old cap shoot-out reviews you may see (most used parts from a couple generations ago, and some did not follow orientation directions and/or used values/dielectric thickness inappropriate for the application)--you need to hear and judge for yourself!

Best,

Alex Crespi

UpTone Audio LLC

Hi Jud:

Glad you got your DAC up and running. I know you are shy just yet about who did the DAC board for you, but can you tell me which USB>I2S board you have that is taking 352.8/384kHz? Is it an XMOS-based interface, because I have been trying to track down some XCORE firmware code to license for 352.8/384.

So tell us about your power supply. How about some pictures!

Lastly, what value of output coupling caps are you using? Is it a tube output stage?

I am the owner of the MusiCap brand of film-and-foil (not metalized film) polypropylene capacitors. I have them made to my specs with specialized films, end spray, solder, silver-plated wire, dimensions, curing, etc. I have high-end loudspeaker and electronics OEM clients around the world, some of whom I have been supplying since 1986 (while at my former co-founded firm Hovland Company which we closed in 2009). I'd be happy to send you a sample pair to try in your DAC since I think we have a similar ear for sound.

Just let me know what value you would want (3.0uF/200V or 4.0uF/400V would be typical and are in stock). Ignore most of the old cap shoot-out reviews you may see (most used parts from a couple generations ago, and some did not follow orientation directions and/or used values/dielectric thickness inappropriate for the application)--you need to hear and judge for yourself!

Best,

Alex Crespi

UpTone Audio LLC

Great Jud!

I'm as happy as you are with your project now reaping the sweet fruits: Delivering nice music with the SQ of your dreams...

Welcome to DSD native playback, I knew you would not be disappointed.

Regarding PCM playback to the maximum DAC sample rate by A+ iZotope up sampling I agree is the wise way and where I achieve the best SQ.

When you have the time please show us some pictures.

Kind regards,

Roch

Great Jud!

I'm as happy as you are with your project now reaping the sweet fruits: Delivering nice music with the SQ of your dreams...

Welcome to DSD native playback, I knew you would not be disappointed.

Regarding PCM playback to the maximum DAC sample rate by A+ iZotope up sampling I agree is the wise way and where I achieve the best SQ.

When you have the time please show us some pictures.

Kind regards,

Roch

Folks, excuse the delay in response - schedule's been pretty full recently, both work and Real Life.

Yes, I'll be posting pictures. Right now it looks like just what it is, a few PCBs and a tranny sitting on a small piece of black plywood. The Omega Mikro ribbons and maple chassis are off in the (hopefully not terribly distant) future.

What I can say about the PCBs at this point is that the output stage is not tube, and there are 4 output caps, 4.7uF value. The problem with saying more right now is twofold: (1) Mostly, I don't know! and (2) The same person is supplying all the PCBs, including the PSU. But I will try to get you the info re USB>I2S interface.

Folks, excuse the delay in response - schedule's been pretty full recently, both work and Real Life.

Yes, I'll be posting pictures. Right now it looks like just what it is, a few PCBs and a tranny sitting on a small piece of black plywood. The Omega Mikro ribbons and maple chassis are off in the (hopefully not terribly distant) future.

What I can say about the PCBs at this point is that the output stage is not tube, and there are 4 output caps, 4.7uF value. The problem with saying more right now is twofold: (1) Mostly, I don't know! and (2) The same person is supplying all the PCBs, including the PSU. But I will try to get you the info re USB>I2S interface.

Very cool project Jud! Looking forward to seeing pics of the final product, as well as your further thoughts once the final version is up and running.

Good choice going with Mapleshade/Omega Mikro. I am very happy with my Mapleshade 4" platforms and associated Mapleshade brass footers and isoblocks. Pierre is a great guy. I have had a number of conversations with him over the phone. Pierre is scary-smart. I love his out-of-the-box thinking.

Very cool project Jud! Looking forward to seeing pics of the final product, as well as your further thoughts once the final version is up and running.

Good choice going with Mapleshade/Omega Mikro. I am very happy with my Mapleshade 4" platforms and associated Mapleshade brass footers and isoblocks. Pierre is a great guy. I have had a number of conversations with him over the phone. Pierre is scary-smart. I love his out-of-the-box thinking.

Pierre is scary-smart. I love his out-of-the-box thinking.

Pierre, an aeronautical engineer, was partly responsible for perhaps the most successful military aircraft procurement project in the past several decades, the one that produced the A-10 "Warthog." Pierre was part of a team that stubbornly resisted the usual military wish for additional capabilities not in line with the original mission, and the desire for "sexy" capabilities such as speed rather than those actually needed for the mission, such as pilot protection. Decades later the Warthog is still being used for new missions for which its capability set suits it far better than newer, more exotic and expensive planes the Pentagon and military contractors would dearly love to see assigned to those missions. See A-10 vs. F-35: The Air Force's Latest Budget Bungle | Mother Jones .

You can see some of the same traits in his audio designs. He's not a crank who feels the accepted wisdom is wrong automatically; rather he himself doesn't accept the "accepted wisdom" until he's seen it pass scientific and engineering tests, as well as real-world listening tests. Thus unlike a lot of high-dollar cables, Pierre and his partner's designs don't look like garden hose. But after having faced down senators, Pentagon brass and engineers from military contractors, I don't think that bothers him.

Ah I can say Jud after reading all of these blogs on this custom hand built DAC is, cool

Jud: this sounds like a cool project, but I am confused by something. You claim that the DAC converts at 352.8 (and 384), and that you therefore are using Audirvana to oversample/filter, and then let the DAC solely convert to analog. That all makes sense, but then you also claim the DAC also converts DSD natively? This is where I get lost, DSD has a ~2.8 mHz sampling rate, and would not be compatible with the supposed 352.8 native conversion rate of the DAC? What is really going on here?

Also, a suggestion for the future. Since this DAC uses coupling caps, I would like to suggest that Clarity MR capacitors are excellent for this. I have tried a number of different highly rated caps, and the MRs have proved to be the most neutral and transparent I have found. I have compared them to no cap at all, and I am not even sure I can hear them at all. I have not tried the super expensive caps out there, like Duelund, but these clarity MRs are really, really good.

Hi barrows: Please allow me to jump in for Jud for a moment. I think his point is that the sigma-delta DAC chip in his unit (an ESS or a Crystal I forget), like virtually all of its ilk, would normally oversample everything in a couple of steps to its native rate of 352.8/384kHz (some automatically go to 768 internally). So by upsampling in s/w--and with iZotope giving full control of the filter parameters--he is able to bypass the DAC's internal filters.

As for DSD, if you look at the spec sheets for the typical modern DSD-capable chip, you will see that when they sense a DSD stream they route the signal around the usual SD conversion and filters.

Where they fail, however, to allow realization of the potential of the DSD format, is in the way most of those chips create their own digital/analog DSD low-pass filters at the last stage (my engineer friend explain the two different methods used--both seem like elegant hacks that defeat the spirit of the format). But that's for another discussion...

I've been playing around with a few commercial and DIY DACs lately, and it is quite startling just how much they all are held back (musically) by their filters. Reducing they influence of the internal filters by using quality, customizable upsampling in the computer beforehand, can allow a very common sounding DAC to punch well above its price class.

As you probably know from my other posts, I use a filterless NOS 1704K, so I don't have to fight any internal filters. But of course I have to upsample or endure the nasties from aliasing artifacts--at least with Redbook material.

Sorry Jud, din't mean to sidetrack your blog!

--Alex

Hi barrows and Alex. My DAC uses one chip that routes PCM less than 8x rates first into an interpolation filter, then into a sigma-delta modulator. Upsampling in software avoids the interpolation filter but not the SDM; DSD avoids both.

Re the caps, I really appreciate the suggestion, barrows. Maybe at a future date when the rewiring and chassis expense is behind me. In the range I need, the Clarity caps appear to be about $70-$80 apiece, and the unit takes 4 of them. So right now I'm looking in a lower cost range.

S.dad: The ESS processes audio data at ~2.8 MHz (or higher, I forget), so external oversampling to just 4x rates does not come anywhere close to feeding the DAC at its native rate. But the ESS is an exception to most DACs, most current SDM DACs operate at 8x rates, 768 kHz and 705.6 kHz. But, while oversampling in software to 4x rates does not feed the DAC chip at its native rate, it is likely that any audible effects of the oversampling/filtering process will be dominated by the first 4x rate as done in SW. This is because the subsequent oversampling is at such a high rate, that all artifacts it causes will be so far above the audioband as to be entirely inaudible.

Exactly how the ESS handles DSD is not revealed publically, but the block diagrams do not indicate a separate data path at all for DSD, it appears that DSD is subject to the same SDM modulator path as PCM (this makes sense because PCM is oversampled to DSD rates). Interestingly, the digital volume control of the ESS 9018 works perfectly with native DSD, draw whatever conclusion you might like from this…

Of course, what you are doing with a 1704 is a totally different thing, as that chip has no onboard oversampling/filtering, and is designed specifically to have the oversampling/filtering done externally, of course,being an RTR converter it has no DSD compatibility… So we can be sure Juds DAC is not using 1704s.

BTW, just to be clear, I'm not using an ESS chip.

Barrows, thanks for the reply. Yes, I know Jud is definitely not using PCM1704. That's just what I use and like! And yeah, no DSD for me except when Audirvana converts it for me on the fly. But the results of that have not been impressing me.

The ESS chips have always seemed quite unusual to me, and I notice in the other forums that people need to run it with clocks a lot faster than 22.5792/24.576 just to get it to cooperate. I know that they are very popular though.

Why do you refer to the 8x rates of present day DAC chips as 768kHz? 48kHz x 8 = 384K. Can you tell me what other mainstream SD DAC chips are doing 16X to 768? Seems to me that the top Wolfson and T.I. chips are still 8x. You know I am not an engineer, so I am likely missing something here.

As for DSD processing, have you looked into all the weird ways that the DAC chip makers have accommodated DSD? I have been told that a lot of them create short FIR filters at the output using internal RC networks. Not necessarily the purest approach.

Thanks,

ALEX

S-dad, so sorry it is a bit of my mistake here. Most conventional (TI 1792, Wolfson, etc) SDM chips can indeed run with an 8x oversampling filter as you note. I was getting a little confused between the standard OSF and what OSF can be implemented externally. For example, Ayre runs their own external OSF in the QB-9 DAC (I am not talking about the new version with ESS chip here, but the original version which uses a TI 179x chip) at 16x rates, indeed this is possible because the actual modulator in the TI 179x chips runs at 64x rates. You can bypass the OSF in a typical TI 179x and provide an external OSF at 8x, as you note. So, the TI SDM chips can run with an external OSF at 8x or 16x (as Ayre does), but the actual conversion takes place at 64x rates. Sorry for the confusion.

As I mentioned in my previous post though, even if the DAC chip applies additional oversampling beyond the 8x level, it is entirely likely that the only audible effects one can hear will be those produced by the first 8x oversampling filter, as the oversampling to even higher rates should only produce artifacts so high in frequency as to be completely inaudible (and those frequencies are likely filtered out in most DACs by an analog filter in the output stage).

As to DSD filtering, DSD DACs are generally going to apply low pass RC filter somewhere, either in the chip as you note, or in the output stage. Some kind of additional low pass filter, with a fairly low corner frequency, is considered a necessity for DSD, do to the high levels of noise just outside the audio band. I took a quick look at the data sheet for the fairly common TI 1792 DSD DAC chip, and indeed it offers a choice of four different FIR filters in its analog output section (on the chip) for DSD playback.

What goes on inside the ESS 9018 is harder to fathom, as the data sheet is not readily available to the public. I do know that it runs at a very high rate, at either 5, 6, or 7 bits (user selectable) and it also has a very high degree of user programmability/flexibility. It also features an onboard asynchronous sample rate converter and DPLL, in addition to the oversampling filter. My experience with it suggests that it sounds better when the ASRC and DPLL are rendered moot, by synchronously clocking it, few manufacturers use it this way though: the only ones I know who do are Ayre in the new version of the QB-9, and Wavelength Audio in their ESS equipped DACs. Of course Ayre also uses their proprietary OSF with the ESS, probably sending data to the chip at 16x rates.

....standard OSF and what OSF can be implemented externally. For example' date=' Ayre runs their own external OSF in the QB-9 DAC (I am not talking about the new version with ESS chip here, but the original version which uses a TI 179x chip) at 16x rates, indeed this is possible because the actual modulator in the TI 179x chips runs at 64x rates. You can bypass the OSF in a typical TI 179x and provide an external OSF at 8x, as you note. So, the TI SDM chips can run with an external OSF at 8x or 16x (as Ayre does), but the actual conversion takes place at 64x rates. Sorry for the confusion.

[/quote']

 

Since I feel strongly (from experience over the years with various DACs' filters, and with making simple DACs sound remarkable with custom iZotope upsampling in A+), we are mostly on the same page: Bypassing the internal filter choices of a modern SD DAC chip by using an external filter--whether hardware or software--is always a good thing. And indeed, regardless of the chip's ultimate super-high internal conversion rate, getting up to 352/384 pretty much pushes up any residual effects of it remaining filters. It sounds great!

 

As I mentioned in my previous post though' date=' even if the DAC chip applies additional oversampling beyond the 8x level, it is entirely likely that the only audible effects one can hear will be those produced by the first 8x oversampling filter, as the oversampling to even higher rates should only produce artifacts so high in frequency as to be completely inaudible (and those frequencies are likely filtered out in most DACs by an analog filter in the output stage).

[/quote']

 

Yep, full agreement. BTW, I am curious as to the ESS-based Buffalo which you run (per your current profile listing): Are you using one of the Sabre's built -in OSFs or are you upsampling externally with s/w?

 

As to DSD filtering' date=' DSD DACs are generally going to apply low pass RC filter somewhere, either in the chip as you note, or in the output stage. Some kind of additional low pass filter, with a fairly low corner frequency, is considered a necessity for DSD, do to the high levels of noise just outside the audio band. I took a quick look at the data sheet for the fairly common TI 1792 DSD DAC chip, and indeed it offers a choice of four different FIR filters in its analog output section (on the chip) for DSD playback.

[/quote']

 

See, but once again they are hampering the (already hamstrung, noisy 1-bit) DSD format by applying a lousy FIR filter to it that has similarly lumpy response as you see in the graphs for the PCM filters, and of course the chip designer's choice of impulse response. I know I don't have all my facts correct on this, but my friend and associate John Swenson, explained it to me in detail.

There are definitely vastly better ways to filter DSD--but it means keeping the signal entirely away from any DAC chips! When the dedicated DSD board for Sonore's Rendu ships, you will see a truly great way to decode DSD.

 

 

What goes on inside the ESS 9018 is harder to fathom' date=' as the data sheet is not readily available to the public. I do know that it runs at a very high rate, at either 5, 6, or 7 bits (user selectable) and it also has a very high degree of user programmability/flexibility. It also features an onboard asynchronous sample rate converter and DPLL, in addition to the oversampling filter. My experience with it suggests that it sounds better when the ASRC and DPLL are rendered moot, by synchronously clocking it, few manufacturers use it this way though: the only ones I know who do are Ayre in the new version of the QB-9, and Wavelength Audio in their ESS equipped DACs. Of course Ayre also uses their proprietary OSF with the ESS, probably sending data to the chip at 16x rates.[/quote']

 

All that work to make a 1-bit/high-fs format work in chips that are better suited to PCM. (And of course in the studios that have to do 4-bit wide DSD just to be able to do anything with what was meant as an archive format.) It is not as if preamps are short on inputs or that dedicated DSD DACs have to be really expensive. We don't expect our DACs to handle our FM radio or vinyl replay. Why force them to take DSD as if it was just another digital sample rate?

End of rant. Cheers!

S-dad: first, sorry to jud for the derail here, but I will take this chance to respond to s-dad's inquiries. Yes, right now I am using the onboard OSF in the ESS, and I agree that this is likely a limitation in my DAC, still it sounds great, but I suspect that it could be better. I am currently investigatiiong options to that end, but it will be awhile... I use the slow filter option in the ESS, which is still a pretty strong filter.

barrows, I don't consider it a derail at all. My project is aimed at two things, within the constraints of my budget, skill, knowledge, and so forth: (1) Trying to minimize the number of format or rate conversions, filtering steps, etc., in the recording/playback chain; and (2) Applying some of the Mapleshade/Omega Mikro philosophies and equipment regarding vibration control and materials (particularly wiring) to see whether the sound is affected (hopefully for the better). The discussion you and Superdad are engaging in is right in line with the first of these aims.

Where I do have to do conversions, such as upsampling for RedBook, having this DAC has helped me to dial in my software upsampling settings, with no in-DAC interpolation to take into account. Where I don't, such as with DSD, I've really enjoyed the results.

What's left with regard to my own experimentation on the first goal above is to download some of the 2L material in DXD (8x rates) to hear what PCM sounds like without either software or hardware interpolation filters applied, just SDM modulation.

Hey Jud, Cool! I have not tried Audirvana recently, do you know what kind of filter options they are providing from Izotope? For instance, do they have minimum phase (no pre ring) type filters, and options for minimum phase/slow roll off filters (no pre ring and very little post ring)? Or are the filter on offer just well implemented symmetrical impulse response filters?

Even though I use my server in my main system most of the time, occasionally I am testing gear with my Mac, in which case I use Pure Music, but lately I have been disappointed with the lack of continued development of Pure Music (especially for DSD), as it does not seem that Pure Music even supports gapless playback of DSD. Maybe it is time for me to pick up Audirvana again…

I look forward to reading more about your DAC project, at some point you might also try rolling the opamps… there are even some discrete opamps available to try (if a little spendy).

Barrows:

Man, oh man, are you in for a treat! Audirvana Plus (Jud and I highly recommend beta 1.4.9.7-- http://audirvana.com/delivery/AudirvanaPlus_1.4.9.7.dmg) supports all available features of iZotope in a great way. You have full control over:

Steepness, Filter max length, Cut-off freq., Anti-aliasing (=final max. attenuation about), and Pre-ringing (1.0 is linear--equal pre/post; 0=all post).

You can also control what you upsample to (Max, Power of 2, or just 2X).

There is a long thread here on CA about iZotope settings. I have turned average DACs (with good clocking, output, PS, etc.) into great sounding ones by careful settings balance. I prefer power of 2 upsampling.

I recommend Integer Mode 1, and non-iTunes integration.

Have fun and let us know what you think!

ALEX

Thanks Alex, that is good to know. Interesting that it appears impossible to find any mention of the filter options at the Audirvana site. I will read up some more on the iZotope page. I knew iZotope itself has lots of options in their SRC SW, I just did not know how many of these were available through Audirvana, last time I tried it these options did not exist.

BTW, Jud, maybe you would share your settings here on the blog?

Not that he needs me to speak for him, but as of last week, Jud's settings were (from an e-mail he sent me):

Steepness, 3db;

Filter max length 2 million samples;

Cutoff freq., 1.02;

Anti-aliasing 200;

Pre-ringing 1.00

Mine are quite different:

Steepness: 7

Filter max length: 1,150,000 (this one I may bump up a bit based soon, though I am against the full 2 million)

Cutoff freq.: 1.04

Anti-aliasing: 200

Pre-ringing: 0.65

BTW, you can model things with the free trial of iZotope RX (the advanced version is the required one with the resampler IIRC). The settings scales match what Audirvana license from them, though it seems oriented towards downsampling. It works for graphically modeling even after the trial period expires. Check out the iZotope SRC thread here on CA for examples of the graphs people were making to model the filters. But your ears will tell you what is right!

ALEX