World's first NOS 24/384 filterless DAC

[manisandher]

Hmm ... Did you actually ever try to open the box and look for chips ?

I haven't, but Romy has! Well his Model One at least:

http://goodsoundclub.com/Forums/ShowPost.aspx?postID=6566#6566

Not sure if you can make the chips out though...

Mani.

[manisandher]

I'm not sure if this is relevant, but Ritter talked about the 'upcoming' Model Two back in 1999. At that point, only the Model One was actually available. However, my particular Model Two was one of the first to be built... and that was in 2003!

Perhaps they were waiting for commercial 24/192 ADC chips to become available for the Model Two? However, Ritter talks about 'custom, discrete, full-ladder converters'...

Unfortunately, I'm not willing to take my unit apart to have a peek inside those PMADC-1 modules and see what chips they're using. Sorry!

Mani.

[Per]

Peter,

Won't a lot of the things you describe as problems with OS dacs and advantages of NOS dacs be masked by the following electronics / speakers anyway?

Per

[PeterSt]

Hi Per,

Yes, of course. But I guess this is about those zillion things being wrong. I mean, the data coming from a DAC (and also the software as, we know now) can be so totally different (over setups/designs) that it easily overrules. I also could say : you have no idea what can be squeezed out of your loudspeakers. So, also the lousy ones.

If you take for the example the relatively very tiny influence software can have (remember, bit perfect), but eveybody hears it through no matter what speakers and further system, well, then you know the truth of it.
Fact remains that that too (speakers and everything) matter a LOT.

I don't want to tease you all, but I think I have entered the stage which might be a kind of end stage. Btw, important to know is that everything I actually (or really) did was about speed speed speed. Nothing else. Thus for example, not a high quality well known whatever amplifier, but a very fast one. Now :

I don't know if someone has the OST version of The Wall, but yesterday I played it for the first time, and (thus) I have no relative judgement of it. Man, this was totally creapy. This album is so much ful of "real life" sounds that in the end I got tired from it. This starts somewhere (IIRC) at the second track with someone whisteling. I looked in the room, looked again, looked at the speakers, and I just refused to believe this was not someone in the room whisteling.
Well, we all have this experience sometimes (the crying baby is a good example), but now it never stopped. Really, it tires (but with a happy face).

Later throughout the album my wife walked into the room, and I asked her "am I crazy, or does this all sound so realistic ?". She misinterpreted the question and thought I was talking about the performance of the system instead of the album. Her response : "it sure does. You know, lately, when I'm sitting outside, I'm constantly thinking someone inside is talking to you or anything" (outside is with the doors closed btw).

At a certain stage I thought "this must be the best recording I ever heard", which is total nonsense of course (unless plainly true, which theoretically can be so), but then the conversation I had some 2 years back with Peter from 2L (recording engineer) slipped through my mind; he told me something like "you don't want to know, but at 352.8 things suddenly work". So yes, I was using 352.8 indeed, though upsampled ...

All 'n all I can't be sure what will be happening to your systems, but if I look at all the improvements I could create for myself (with a. XXHighEnd and b. the Phasure NOS1 DAC) it should be so that you will be totally shocked by now. Maybe a slower speaker or a slower amp will debet on it, and maybe it *requires* the fast "back end". We'll have to see that. But I tell you : this just starts to be unbelieveable (indeed says me myself and I ).

Peter

[Per]

Thanks a lot for your reply, Peter. VERY exciting indeed...

I must say that judging from what you have done with the software / computer side of digital playback with XXHighEnd I think we all might be in for some real (pleasant it seems to) surprises with your dac - IF your ideas and electronic artistry works out in real life with different setups judged by different users, that is.

Can't wait to hear the feedback from the first customers of the PHASURE NOS1 DAC. Might turn everything upside down in regards to building a highend setup. Pondering about some things / questions but that is for the ideal amp / speaker setup thread http://www.phasure.com/index.php?topic=1182.0

All the best,

Per

[bgjohan]

Peter

In relation to measuring the noise in your DAC project you stated that noise would be (at least 3 dB) better if measured with balanced lines.
I take that to mean same signal from the DAC's digital section, but processed through the balanced section of the analog stage in your DAC, rather than processed through the unbalanced section and measuerd at the unbalanced analog output (as in your posted measurement pictures).

In other words, you are not speaking about the difference in the DAC's performance when fed a balanced source (digital IN) signal versus an unbalanced source (digital IN) signal. 

Would be interersting to hear your thoughts on feeding unbalanced versus balanced digital source signals to an external DAC - in general as well as related to your DAC project (listening and mesurement findings).

Best regards

Bjorn

[PeterSt]

Hi Bjorn,

Ok, after taking some deep breaths on how to answer this, here is my well meant attempt :

First of all (don't get scared) : There is no such thing as balanced or unbalanced digital-In in the Phasure NOS1. It is direct i2s (you will learn how, later) and no such thing as balanced/single ended exists there.

But in general the answer could be : it may matter in jitter because a balanced connection will eliminate noise, if there in the first place. Notice though that any normal means of us consumers connecting DACs, such noise won't be there. Merely think about pro-use, and hundreds of meters digital connection.

The real merits of the base of your question is different though, although you will not have noticed that yourself;

When I told about this noise, and better with balanced etc., this is was not about the elimination of (possible !) noise in the outgoing interlinks (or otherwise I don't know what text you are referring to), but merely about this :

The DAC chips themselves (and there are 8 of them) for the half of them are lined up in balanced (differential) operation. This means that what's fed as positive voltage to the one chip, is fed as negative to its counterpart (just like balanced line). This eliminates a type of distortion which may occur at one side of the 0 voltage line (think like negative voltage creating another amount of (inherent) distortion than positive voltage). It "balances out". Thus, whatever the text is you refer to, this will eliminate distortion to begin with.

While this is one half of the line up of the DACs, the other half is setup parallelled. Notice this was created later, and wasn't so at the beginning of this topic. This causes the output to be higher in dB while the "PCB-inherent" noise (not distortion this time !) keeps the same. Thus, relatively the noise drops.

Now, the first setup of the DAC had a choice for Balanced and Single Ended (unbalanced), and it was a 4 chip version. As per my explanation above, both imply something for distortion and noise respectively. It was *or* balanced *or* single ended;

The current version is somewhat smarter :
It still has the choice for balanced and single ended BUT with the balanced setup the parallel configuration is also at work (this is because only 2 chips (per channel) are useful at creating the differential (balanced) setup; 4 don't add anything more). So, now there's less distortion *and* less noise at the same time.
When you choose for Single Ended, all chips are parallelled and there's eve less noise (but not as few distortion as possible).

Well, at trying to be as vague as possible (as it seems to me), you must also think about how you use the configuration options. I mean, while there are actually two options to choose from (balanced/SE), in either case you can use half of the output lines vs "added". Notice that "added" is wrongly put for Balanced (you'd just be using a negative and a positive line), but looked at it the other way around : you can use half of the balanced output for a SE connection, and you can also use half of the SE output for a SE connection. Although this just works perfectly, both these options are actually stupid, because they produce half of the possible output, either with unnessecary distortion or unnessecary noise (if you could ever follow my above story haha). BUT :

It produces less net output towards the main amps just the same, so, if indeed working without preamp, it may just give you the attenuation needed, not causing to loose one bit of resolution (6dB) otherwise. Otoh ... I don't think there will be a switch for this option (other than internally) because that switch would be in the signal path, and I don't like anything in the signal path which is not really necessary.
So, the story about it was merely to support the way it works and to hopefully make it more understandable, than that it will be used in practice.


Lastly, because this is as important I think :
Notice (from above story) that the balanced output is nothing which is created in the analogue domain (hence no transformers or anything). Instead, it just starts in the DAC chips and from there goes (via the analogue stage) direct out to the XLR pins.


Peter

[bgjohan]

Peter

Thanks for your explanation - provides a good outline of technical side of the operation of (earlier and latest version) NOS1.

Let me just state my understanding of these operating principals and how you intend to implement those in the NOS1, and you can correct me if I have misunderstood.

1. Switching btw Balanced and SE operation.
The switching of internal operating mode of NOS1 btw Balanced and SE. In other words, not a switching btw Balanced vs SE digital input.
In effect, the DAC does not even need a Balanced input as as a balanced +/- digital signal can be generated in the DAC from a digital SE(+) IN signal.

2. Balanced vs SE output
Balanced vs SE output is linked to the selection of operating mode; i.e. Balanced (XLR) output only active when Balanced operating mode selected and SE output only active when SE operating mode selected.

Bjorn

[PeterSt]

Bjorn,

In effect, the DAC does not even need a Balanced input as as a balanced +/- digital signal can be generated in the DAC from a digital SE(+) IN signal.

It is hard to explain (further), because the phenomena just are not applicable ! But, the operation of the DAC makes you confused, I'm sure.

Think about a pre-amp. Does it need balanced inputs of some kind in order to have balanced outputs ? no, because it is totally unrelated. But :
*If* a preamp would have balanced inputs, I'd put that through all over (doubling really everything for parts etc.) and you could even have separate power supplies for the plus and min part (and notice that you'd need plus and min power rails anyway, but they usually come from the same PSU).

With more DAC chips you could do anything, and you could even let one chip deal with the plus voltage (from the music data this time !) and the other chip with the minus voltage, while at the end they come together again as a normal signal (with +/- to drive your speakers). If this were necessary it would have been done, but the PCM1704 does that internally already ! (eliminating cross-over distortion).
So, the useage of more chips just allows the trick for balanced operation to be applied, and it is totally unrelated to the input. Also notice that if the input were balanced, this is just an analogue thing (no matter it is carrying functionally digital data). Within the DAC it is no analogue thing at all, BUT since there are two outputs (separated in plus and min) you just as well can let that continue in the I/V conversion to the outputs (where it has become pure analogue). Thus, a trick which avoids the transformers normally needed to make a SE analogue signal balanced.
Less = More --> Less = Speed.

Balanced vs SE output is linked to the selection of operating mode; i.e. Balanced (XLR) output only active when Balanced operating mode selected and SE output only active when SE operating mode selected.

True. But by the grace of me not applying something strange, like using half of the balanced output only, making that SE again.


I'm afraid this didn't help much, but keep in mind it is all about technical decrease of distortion and noise. With technical I mean : not subject to coincidental picking up of noise which would be eliminated by balanced cabling. Instead you can just calculate on it with this as the best example :
When the inherent noise of the DAC board (think PCB) is -120dB while the output harmonic distortion of the DAC chips (and all) is -130dB THD+N(oise) will be -120dB. When I'm able to increase the output of the DAC chips (and all) with 3dB because more chips contribute to the output (which is current at first), the THD+N will be -123dB. Why ? because the signal got higher but the (PCB) noise remained the same.
Nothing much different from the good old analogue stuff, and louder recording implied less noise (S/N ratio !). -> The recorded signal is higher but the equipment and tape noise remain the same.

The above was about the parallelled part of it, which is there anyway (4 chips). So as said, when all is parallelled (SE output) S/N ratio gets better again, but THD gets worse.

To finish this off, when the noise floor is above the HD floor, HD doesn't matter anymore, because it disappears in the noise. But when you improve the noise (lower the noise floor) HD get apparent. All 'n all, one of the tricks is to have both at the same level, which is "efficient" so to speak (hence, why improve further on HD while you can't hear it anyway because of the noise).

Oh well
Peter

[Telstar]

To finish this off, when the noise floor is above the HD floor, HD doesn't matter anymore, because it disappears in the noise. But when you improve the noise (lower the noise floor) HD get apparent. All 'n all, one of the tricks is to have both at the same level, which is "efficient" so to speak (hence, why improve further on HD while you can't hear it anyway because of the noise).

a-a! This is a false belief.
Our ears are able to listen 10 to 20db below the noise floor.
To measure HD under the noise floor the notch filter trick can (and should) be used.

So, in first instance the noise floor should be very low (and <120dB IS very good for a low-level signal), then the harmonics should be another 10db under it.

[PeterSt]

Haha, yes I agree. But it is difficult enough already to explain it all (from the angle we talk about here).

But it is also about practice. I think I told this before :

Take a 1 bit recording (RME has (or had) such a file for download). This is -96dB. Now try to set your equipment so loud you can hear the music play. The very first you'll hear (in a random system) will be a lot of noise, but you will be able to hear the music playing as well (unless you really have too much noise). This proves it.

Sorry for the wrong suggestion !
Peter

[manisandher]

Hmm ... Did you actually ever try to open the [PM Model Two] and look for chips ?

FWIW, it looks like the DAC section uses UltraAnalog chips, bearing a striking resemblance as they do to the older 20bit chips used in the Spectral SDR-2000 (compare the attachment to pics on Romy's site; http://goodsoundclub.com/Forums/ShowPost.aspx?postID=6566#6566). This isn't surprising seeing as Keith Johnson (one of the designers of the Model Two) happens to be the Director of Engineering at Spectral. I'm not sure about the ADC chips though...

Looking at the specs for the SDR-2000, it seems that the people at Spectral share your obsession speed...

Mani.

[PeterSt]

I must admit that those boxes look veeery similar (while it should be a cover for the real chips only) although they seem to be not equal in size.

UltraAnalog ... hmm ...

[bhobba]

Hi Peter and All

Finished reading the tread.  Wow what a lot of info.  Will need to read it again and maybe again again to fully glean all the stuff it stands.  But right now for me I still don't understand how you get the 384/32 into the DAC?

Thanks
Bill

[PeterSt]

Funny Bill ... but you REALLY are the first one to ask. I guess it is known that I don't tell everything which is regarded a secret (propriatary) in the first place ? Or everybody thinks this is no problem ?

Your number is on the low side though.
Peter


PS: But 24 bits !