World's first NOS 24/384 filterless DAC

[PeterSt]

Edit, May 2011 :

I just changed the title for its "24/192" part into "24/384", because this is of course what the Phasure NOS1 DAC turned out to be.
I also removed the question mark about 24/192 being the first NOS filterless DAC, because by now we really know it didn't exist. As it still doesn't if we count out the NOS1 (which of course is filterless at all sample rates).

Also, the topic has been moved from the Chatter board to the Phasure NOS1 DAC board.

For those who are new, well, you can try to read it all through, because it really shows the history of how someone like me started to produce D/A converters, which after over two years ended up in the very best DAC existing at this time. Not because I say so, but because by now it replaced all the great names people owned, them people usually gaining a fair net amount of money after putting their old DACs to eBay.
You may ask for your DAC to be potentially replaced whether it has (you could do it in this topic), and or I'll wait a few days for the now NOS1 owner to respond himself but will answer myself if nothing pops up, or I will respond right away if I know it's a "new" one.

Anway, of course all ended up in a DAC which uses completely uncommon technology (interface, less is more approach with hardly anything in the signal path, no switches, NON-oversampling which measures as good as OS but showing the *real* figures, in-software filtering (e.g. by means of XXHighEnd), really zero in-DAC filtering, special differential setup of the 8x PCM1704U-K chips with better figures than even the chips themselves, super low noise and jitter levels) ... all from an external DAC.

But maybe it is more efficient to start reading somewhere at the back of this topic.

Lastly for here in this edited first post, maybe it's nice so see some new posts in this topic, because it has been a great time throughout creating it with many of us, still having the nice memories about it. Or maybe I will add some newbies to the topic myself later ...

Regards,
Peter


Guys,

I must be honest with you. I have been working on other things than software lately ...
See below, which as far as I know is the first 24 bit 192KHz DAC explicitly assembled for non oversampling and filterless mode.
36 hours ago the champagne was opened because it produced sound for the first time.

Some technical insight for those interested :

- Balanced passive I/V conversion by means of two balanced DACs (PCM1704) per channel as the base;
- Balanced or RCA out;
- Output 1.5 VRMS (2 VRMS possible);
- Possibility for oversampling to 211KHz with 1.2ps RMS total jitter;
- Possibility for filtering (2 pole Bessel 100KHz);
- Drives long interlinks sufficiently to avoid a pre-amp or other means of buffer;
- S/PDIF connected, and due : directly Firewire connected without soundcard into I2S;
- Terminals for an external clock;
- The most stable shunt regulated power supply imagineable, both channels completely separated throuhgout;
- Everything operating in pure Class A;
- Housing : WxDxH : 43 x 26 x 8.5 cm = 17.2 x 10.4 x 3.4 " worth of 15Kgs weight;
- The various elements not designed by me, but chosen as *the* combination for the best DAC ever (so 1 % credit for me really );
- Name I dedicated to this combination : NOS1.

You may ask yourself : Can I build this too ?
If you are like me, forget it. I drove various people to complete madness and received some burnouts myself.
Anyway, it took me over a year thinking about it and "designing" various concepts, 7 weeks throughput of ordering modules and parts, three weeks of understanding, building and getting more parts, and occupuying the dinner table. The (programming of the) Firewire connection still to do.

Background

I didn't start this project well over a year ago to perceive better sound, and I was very happy with my current NOS 18 bit 96KHz DAC (the TwinDAC+). However, since the hires material is on the horizon closer and closer, including the *fact* that a pre-amp should be avoided encouraging for digital volume control - that by itself taking out bits, I got obessed of wanting more. Technically more.
Being a convict of non oversampling filterless DACs, I ran into a problem : 24 bit 192KHz DACs do not exist.
Many months of thinking and designing brought me the concept of a 32 bit 384KHz DAC completely software realtime controlled (yep, the firmware part), of which the hardware arrived May 2008. But after I squeezed out the first sines from it, I coulnd't find enough time for it.

While I started this off thinking 32 bits would be needed for a proper digital volume control, underway I created the digital volume for XXHighEnd which very sufficiently operates at 18 bits or even 16 when the output voltage of the DAC isn't too high and the gain of the amplifier isn't too much. This brought me to again searching the internet and it came to me that the PCM1704 should theoretically be able to do the job, but nobody just made something out of it. Or ?
For over a month I had bookmarked a company which expressed specs that looked promising but kind of vague in the mean time. I kept on coming back to this because there wasn't anything else and in the end I started to send an email to that company ...

Ok Peter, why don't you mention the name of that company ?

This is a strange story, and possibly I can't even justify it 100%, but I have the feeling that right now this company shouldn't receive dozens of emails from you guys asking away because they just can't bear it. Maybe later, but not right now. This is all my fault and prices would double instantly ...
People who know me, will know that the occasions I asked someone for help in my professional life can be counted on one hand. Until one month ago that was. Right now, some email boxes have been overflown, just because I started some DIY project with modules provided as kits and a withgoing FAQ that says all is for skilled people only. Hmm ...
Add to that this isn't about a "DAC kit" as such, but a couple of modules which makes it a DAC. Add to *that* again that I asked for modifications which were politely met, but that these functional modifications ended up in technical stuff beyond my knowledge, and no existing manual could explain to me how to deal with things. Just being honest here ...

So even if you know what company this is about, please don't go there right now, and if you indeed know, you also know this is related to respect. I'm sure you will understand in a later stage ...

First listening impressions (with 0.9w-3 and which is theoratically not the best (thing to do))

Since rather many combinations of settings exist, all with their own merits, right now it is rather hard to determine what is the best setting. But no matter what setting I tried ... it is one big step ahead in audio playback. To mention a few things I definitely recognize (just 16/44K1) :

• Enourmeously increased resolution.
  
  This is unrelated to the higher samplerate (because I'm just not using that right now) and also this is unrelated to the nos/filterless principle;
  It is related to the enormeous stability of the PSU, which, mind you, must deal with such small voltages (.0001 etc. mV). And, knowing that creating 1V of output here easily degrades the .0001mV output there, you will know what this is actually about, hence what I was after.
  Also, of course I know how accurately things were trimmed by me, but more importantly : how accurately they *can* be trimmed by design.
  
  Cymbals now receive an additional dimension, which I generally express as : now you can see the size of the cymbals much better. Also, compared to reality (live cymbals) it comes creapily close now.
  
  
• A bass response of which I didn't think it was possible.
  
  For the first time in my life I "heard" a powerful bass response as how we all perceive the large heavy amplifier being able to produce good bass response. We all know that the latter is non-sense, since the class D and other 100g chipamp stuff can do it just the same. What I did not know is how a very good PSU for a DAC incurs for the same but in exponential form. But it can well be that it is the other way around from what I sad right above : the small voltage spikes needed for the higher frequencies won't tear down the more current eating bass waves (never mind this all occurs at micro levels). That the I/V (current to voltage) conversion occurs in a high current domain is another thing that for sure contributes.
  
  As everyone I use a bunch of test tracks and throughout the subwoofers speak where they did not before. This is the most occurring, because - apart from XXHighEnd anomalies as some may recall - usually this comes together with a more profound "just low bass" region (say 40-200 Hz) while right now there is no spur of this. Oh, the bass is more profound allright, but it doesn't incur for any colouration or the tempt to change the cross over. It is just all good good good.
  
  
• Compared to hires it is now an apples with apples comparison.
  
  Since this DAC allows for NOS/Filterless at the higher resolutions, I now can compare 16/44K1 with 24/96 and 24/192 without having to hop over to a oversampling sigma delta DAC.
  I sure did not recognize any different nature at the higher resolutions (tried 192 as well), but I also did not perceive it as better. This latter is not meant to be related to the DAC but merely is a confirmation of what I already though before but couldn't prove (apples and oranges thing) : redbook (hence 16/44K1) seems to suffice and any higher samplerate plus bitdepth doesn't bring more to our brains. Something for another discussion !
  
  
• There's an unary "experience" listening to my speakers now, and I don't know how to put it into words yet.
  
  It is not "black" as we often hear or experience, but in that area. An "emptyness" maybe. A being swallowed by the music, you being in that black hole. It's not a negative, but a strange one. Again things are happening in mid air (I talk about this more often), but now they more happen in 3D while not being able to point at it.
  It sprung to my mind that bass is more directive as ever (I use subwoofers in stereo setup) and besides that the veils from drums suck on you. The latter is again not a negative, but at trying to express what I feel, this has to be in the equation somehow. Hmm ... it could well be that absolute phase is wrong (just change it in XXHighEnd of course), which btw *is* very well possible because of the various taps which can be used, and each "stage" inverts the absolute phase.
  Anyway, it looks that the "boost" of music is more profound now, where boost becomes inverted boost (suck) when indeed the absolute phase is wrong.
  
  
• Very good micro detail at the larger instruments.
  
  This may sound strange and also looks similar to the beforementioned higher resolution, but I am fairly sure this is about the inheritent speed of the DAC (about slew rates and such, possibly even feedback ).
  Under "increased resolution" I talked about the other dimension in cymbals, and here I talk about yet another dimension : the attack of cymbals. Ok, maybe you must have the experience of holding a drum stick a couple of times, but when a drum stick hits a cymbal this doesn't need to be one surface touch. I mean : when a ride cymbal is "ridden" this *is* one touch only, but when a smash cymbal is hit, before the stick goes away the cymbal hits the stick again. It is this which just is audible now.
  
  Brushes, even more delicate and far worse to express properly as I found lately, can't avoid the "feedback" of a cymbal and snare (or tom). That too is audible now, though not exactly in the leage of the "larger instruments". I mean this opposite to the more commonly known detail like breath in a flute or sax. So it is not about this.
  Similar IMO is the wood of a bass becoming audible. I hear you say "man, I heard this long time". But no, I don't think you did. I know my abouts, and this is in another leage. This is about jitter (in good combination with the other virtues of this DAC);
  
  I don't know whether it is a world record, but the overall jitter of 1.2ps seems to incur for what I just described. Why do I think that ?
  Because this low jitter only can be achieved in oversampling mode of the DAC and it is there where this is happening. This is *not* because of the oversampling itself hence impeeding higher "resolution", because at least I claim to know that oversampling only degrades. And for sure it won't "create" the beauties I hear here. If that would be so, XXHighEnd could do it just the same, and it can't.
  
  All 'n all this subject is not really honest, because it uses oversampling mode which net I don't like. It colours the music, and all starts to sound similar because of it. Nothing different from what I was used to.
  But for the more or less solution, the direct Firewire connection comes into play, because that will incur for less jitter because of the I2S connection beind it. And to get the real merits of this : at non oversampling mode the DAC depends on the clock of the incoming data. At oversampling mode this is just unrelated ...
  
  
  

Might you want to know : none of the modules I used, emerged from more or less known DIY projects as known from e.g. DIYAudio. I can tell you that each single part was designed from highly qualified engineers and that the designs themselves testify high quality engineering. Not that "DIY engineering" wouldn't be any good or the engineers being out there are not trustworthy for their profession, but what I have here contains all the consistent knowledge of the most qualified people imagineable, for whom I make a deep bow.

This is not a matter of simple is less, and looking at the picture below I can tell you that the DAC (ok, SMD) comprises 1/25th of the total PCB surface. 24/25 comprises of PSU stuff and getting the output current from the DAC into the output terminals.
Oh, I didn't count all the switches needed for the various input and output settings.

[glynnw]

I think in the commercial version, the cardboard should be black

[pedal]

Very interesting!

My previous DAC was a NOS filterless construction. (DDDAC 16/44 only). It had the most dynamic sound I have ever heard from CD. Very "immediate" sounding.

BUT: With no filter, there will be a lot of HF noise at the sampling frequency, and above. This noise can be quite stressful for wideband amplifiers. Also it will create IM distortion at audible frequencies. Do you have the possibilty to measure this noise?

[SeVeReD]

You need to send this to me for, ah hmm, evaluation.
(ya, that's the ticket)

[PeterSt]

Very interesting!
[...]
BUT: With no filter, there will be a lot of HF noise at the sampling frequency, and above. This noise can be quite stressful for wideband amplifiers. Also it will create IM distortion at audible frequencies. Do you have the possibilty to measure this noise?

Yes, but it won't be different from what you are used to (at 44.1). But still it is so that with the filter it sounds less good/natural.
Just take a look at what manishandher produced by means of (RME's) Digicheck which is just about that. As commonly known, nos DACs should not be "measured".
Not an excuse of any kind, but just the net result of filtering which debets to the sound. As I say it : not filtering makes all "snappier" (or more dynamically if you want), but also more natural.

Btw, don't confuse the "HF noise" you mention with HF noise as the left over of downsampling from a higher sample rate. This sure is *not* anything you'd like at all, and this happened for XX at downsampling from 352800 to 176400 at first, which really needs a filter (the AA filter as is there since). This is important, because nothing in there (the DAC) is downsampling again. Not at oversampling (to 210 KHz) and not at nos.

[PeterSt]

You need to send this to me for, ah hmm, evaluation.
(ya, that's the ticket)

Not necessary. You'll accept it blindly.
But honestly, yeah, I'd dare to invest the time of building, send it yo you, pay for the shipping charges, and be sure it won't come back.
A kind of Internet means btw, because how else to judge products from far away countries without a local distributor and dealers ?
What goes with it, would be a "maybe some other user tried this very product before you" and each scratch would allow for some discount.

[PeterSt]

You may know I am a provocer of "no such thing as a real bad recording exists until the real proof of it is there".
Well, one of those recordings prone to that is Musicology from Prince, expecially the 2nd track "Illusion, Coma, Pimp & Circumstance";
I recall being at someone else's place, pointing out the extraordinary excursion of the woofer being 2cm or so in one direction, which at normal circumstances should not be more than a few mm's only.

Today I thought about "control" of the DAC in that area, and if one track would prove that, it would be this one.

As you already guessed, it worked out beautifully (to the sense of how "beautiful" can workout for Prince ).

This is a track with rather low synth fundamentals (a bit Madonna like) which in this case before worked out the most rough as possible. It just made vomating (a bit of a dutch expression) the woofer before. Just completely out of control as how it came to me, although without really knowing, and virtually blaming it on the "bad recording" (like completely overstreered). But today ?
Ha ! not so today. Today it appeared to be an indeed low synth bass line, but with a bunch of higher frequency harmonics, just "good" disco like.

Might you have the album, try it. But uhhm, output should be at an average of 90dB @ 5m distance. Run this at a lower output, and it's just some fumbling around.

Good control of a DAC ... who would have guessed that ...

[xp9433]

Peter,

This NOSDAC project is interesting - but only if we could eventually get the opportunity to buy one or construct one ourselves. What is your thinking about how that might happen?

Cheers
Frank

[SeVeReD]

[/li][li]Compared to hires it is now an apples with apples comparison.

Since this DAC allows for NOS/Filterless at the higher resolutions, I now can compare 16/44K1 with 24/96 and 24/192 without having to hop over to a oversampling sigma delta DAC.
I sure did not recognize any different nature at the higher resolutions (tried 192 as well), but I also did not perceive it as better. This latter is not meant to be related to the DAC but merely is a confirmation of what I already though before but couldn't prove (apples and oranges thing) : redbook (hence 16/44K1) seems to suffice and any higher samplerate plus bitdepth doesn't bring more to our brains. Something for another discussion !

Since you haven't built the FW input yet, you must be still using the FireFace?  so to listen to hirez you still had to increase the the sample rate right?  Is that something you'll have to do always? or when you feed it directly with Firewire... how low can you go? and will it have to increase like with the FireFace?

I was off 4 days and only got to listen once this whole time :/  holiday stuff and building a computer for my daughter at her mom's place... and, out of all the computers I've put together, this is the first time something wacky happened and it didn't fire up right away... had to take it apart piece by piece to see what was up ... 2 days wasted to find out I think I either didn't seat the cpu correctly or some "arctic silver" paste spilled over when trying to place the heatsink fan ... :/ bleh  anyway, just got it going tonight yea, but wasted time... Casual listening I've been lazy and haven't gone back to W3, but you're still using it over V7... I must bring myself to give it another try with different settings.

[PeterSt]

You are completely right of course, and I didn't mention that on kind of purpose;
Firstly, it is relatively less important than having to use a complete other DAC as how it was before (with me).
Secondly, I didn't want to confuse things unnecessary, because I first have to see whether the larger buffer still makes a difference. And in other words too : whether XX can still make a difference by means of the various settings.

The latter is very complicated because not is all known (not by me anyway), although this project / DAC might be just the test pilot to find out. I mean, whether all is PSU impeeded (which whould be unable to influence in this case) or the jitter (which is completely detached from the incoming stream in this case *if* oversampling mode is used (which I won't hehe)) ... it shouldn't make a difference anymore. Shouldn't, but it will take some ages to test everything, while I'm not even through the base of it all.

I didn't use 0.9w-3 because I think it is better or whatever; instead it was just lazyness with the combination of nothing being shielded yet, using unshielded cables all over ... well ... look at the picture. It is just impossible to be the best right now, hence it is useless to try to squeeze out the best of it at this moment. Nevertheless it is the best I ever heard.
Also, it can hardly have been broken in right now.

A complete other matter is that indeed I'm trying to get to the buffer of this firewire connection, and at this moment I can't judge yet what can be done because I don't have the drivers running yet (the hardware should all be set already). If I want I can program the complete firmware and driver (I mean, right from the beginning) so I could make it the best for "our" purpose. But it is not to underestimate, because that would need just another development board (besides the one I have now), and the additional $1500 it takes is not funny, not even knowing what can be achieved with it.
I have to keep up that I am going to produce XXX DACs already , otherwise you don't get access to all the stuff anyway.

[pedal]

I have to keep up that I am going to produce XXX DACs already , otherwise you don't get access to all the stuff anyway.

Wow! Are you expanding your enterprise into hardware products? That's cool.

When it comes to designing and building a DAC there are several critical points. PSU is of course one of them. Personally I would like my future "ultimate" DAC to feature a BNC clock input. I belive an external clock with huge PSU can be good. I am not an engineer, so I wonder if this option is easy to implement in your new DAC prototype?

Another thing is that most DAC's are working from x48kHz samplingfrequency. (48, 96, 192, etc). But the CD is 44,1kHz. So inside the DAC, there has to be a samplerate convertion (SRC) chip in front of or integrated inside the DAC chip. I believe this SRC converters are "not good". My ideal DAC, then shoud be runninig on 88,2 then. Eventually the SRC process could be done in a "better" way. Maybe the SRC can be done in a software, implemented as a plug-in in the digital loop available in your FF800 software? (I know there are software solutions in the marked).

What do you think?

[PeterSt]

Hahaha, wait a minute pedal ...

What I expressed for features and all were not much commercial expressions, but merely what I had in mind for myself. I little show-off if you like ...

Personally I would like my future "ultimate" DAC to feature a BNC clock input. I belive an external clock with huge PSU can be good. I am not an engineer, so I wonder if this option is easy to implement in your new DAC prototype?

As the original features show, a terminal is there for an external clock. What they don't show, is that a BNC input for a word clock is there too. However, for me, myself and I it would not be connected. But it can be ...
The configuration as I have it in mind  (and which is just there, although not everything is connected) is as follows (besides the beforementioned in the first post) :

- As said, BNC Word clock input;
- Direct TOSLink S/PDIF input;
- Firewire input 2x (useless by itself, by see below);
- AES input (over Firewire);
- S/PDIF input (over Firewire);
- ADAT input (over Firewire);
- TDIF input (over Firewire);
- Direct I2S receiver;
- I2S input 3x (over Firewire);
- Direct S/PDIF receiver 2x;
- I2C interface;
- Each of the above also available as output (not Firewire);
- Some more Pro stuff;
- Last but not least : DSD input.

Huh ?

The last mentioned "feauture" needs some additional explanation :

During the process the ESS Sabre DAC came available (I think March this year); Besides it has the best specs ever, its internal working is very much similar to my own design about the 32/384 DAC for jitter specs and more. My "DAC" will contain the ESS Sabre as well, and although it is a (heavy) oversampling DAC, as said, the specs are special. Otoh, the jitter specs are not better than the DAC I have running right now ... in oversampling mode.

Here all are apples and oranges again, currently knowing that the oversampling mode of my "NOS1" just doesn't touch it. So far I tried each night, and it really doesn't last for one track ...

Besides that, when I really implement the ESS Sabre (due here for a couple of weeks) it won't fit in the same cabinet. But then I anticipated on that with having two cabinets with sharing connections where needed (like the inputs, outputs, some PSU parts and routing switches).

So do I overdo it ? probably yes, and out of all available options and combinations only one will be used. However, there is a significant difference with how I had it before :
Before, in my case, always the Fireface was the intermediate. It was the "routing" device, but it always routed over S/PDIF. That now can be avoided, because the DAC can be its own router. For example : any CDPlayer, DVD(A)player, DATDevice, SACDPlayer, SATReceiver and the like (??) can use the DAC as a DAC, and it is not PC-dedicated. In the mean time, there's nothing that makes it dependent on the PC (the direct Firewire connection causing that, and - upside down - a soundcard is not involved hence does not disturb).

Did I mention USB input ? No. But optionally it is there too, but now completely without reason *and* currently it doesn't support 24/192 (and not DSD of course ... afaik).

What can I say ? not any element was designed or created by me. But with some soldering and driver programming it just allows for all.
This "all" is just about backups. Right now I again enjoyed for several hours the SUPERB bass coming from this single - far from optimal - connection. All still unshielded as the picture in the first post shows.

Peter

[pedal]

I agree with you, that the new Sabre DAC seems to be the new champ. Today I did some research and learned that actually I have friend waiting for delivery of same (or similar) DAC-kit as you. So, perhaps I can lay my ears to it within the end of the year. Can't wait!


Q: Regarding the I/V converter following the DAC-chip: Will you use a SS (opamp), transformer or tube circuit?

[LydMekk]

Pedal: LMC ?

[PeterSt]

Q: Regarding the I/V converter following the DAC-chip: Will you use a SS (opamp), transformer or tube circuit?

Neither. But it is passive anyway.

If it interests you : any transformer just kills transients, "sharpness" and sprankling following that, never mind it's passive.
OpAmps ... well ... YMMV but if they're not noisy they colour the sound (ah, I am very much generalizing here).
Tubes ? hmm ...

Tubes can colour, but IMO this is unwanted. I think they can just as well be neutral if picked properly for the job.
People say tubes are slow, which I personally don't believe. What I do believe is that tubes don't last forever and I can't stand that I'd never know when they are worn out. But that's personal.

All'n all passive is the more sure way to go for. If it can be achieved of course. Anyway it was one of my sure objectives and requirements, friendly met after politely asking.
But who knows ... when the I/V was setup in an active design (as it originally was so), what would have brought *that* ?

If I may say so ... my TwinDAC+ also takes this explicitly into account ... but it just hasn't got the drive for longer interlinks. But hey, weren't we suppose to use pre-amps ?