Any classical listeners use NOS DACs?

[semente]

6 hours ago, davide256 said:

 

Audio note is definitely not a modern DAC, one  that retains reviewer interest. The only thing I know of note for this company is their overpriced capacitors. Care to try again

with a NOS DAC that is current tech? Things have changed pretty fast since 2013 for all DAC's ...

 

 

 

Jud's made a sensible contribution to the thread you linked, explaining the problems Redbook when played through a NOS DAC and the advantages of using a NOS DAC with Upsampled Redbook vs. an OS DAC:

 

On 03/02/2013 at 9:48 PM, Jud said:

I think this thread may be setting some sort of record for most misinformation in lowest number of posts. Lots of marketing terms being used as if they were technical. So I think, though I apologize at the outset for the length of this, that we are going to have to start right at the beginning.

 

- So-and-so DAC has "no filtering." Unless you are listening to a waterfall of unpleasant static or white noise, that is not correct. The digital bitstream sounds nothing like music at all. If your DAC ever loses lock on the signal and you get the unfiltered bitstream, the resulting noise will be sufficiently unpleasant to make you bolt from your listening seat and turn it off. In order to turn that noise into music, low-pass filtering is a necessity. "No filtering" is a marketing term that actually means "we don't do our filtering using the same means everyone else does." Early Audio Note kits actually used their transformers to filter. These days I don't know what they use, but it has to be done, since Audio Note owners are presumably not happily sitting in their listening chairs listening to obnoxiously loud static.

 

- Not a "brick-wall" filter. Think of two filters, one with 24.1kHz to work with (44.1 minus 20kHz), one with 332.8kHz to work with (352.8 minus 20kHz). To achieve the same amount of filtering (loudness reduction of the frequencies above 20kHz), which one will have to have a steeper slope? Well, just picture the graph, or if you like, think of it as a hill. One hill begins at sea level and rises to ten feet within the next 24 yards. The other begins at sea level and rises to ten feet also, but takes 333 yards (about 1000 feet) to get there. Which hill is steeper? So a filter running on a CD-resolution (44.1kHz) input in a non-oversampling DAC must, mathematically, be far steeper than an equivalent filter running on an 8x-oversampled (352.8kHz) input in a typical DAC. The steeper filter causes artifacts, particularly aliasing (audible harmonics that were not present in the original signal), to a greater extent than the more gradual filter. That is why, within a very short time after CD players were introduced (a couple of years), the industry standardized on 8x oversampling. Modern NOS DACs try to avoid the very worst failings of the early steep ("brick-wall") filters by various means (including starting the filtering slightly below the very top of the audible range), but sheer math requires that a filter working within a more limited frequency range be fairly steep and thus cause a certain amount of audible artifacts.

 

- Upsampling vs. oversampling. These are marketing terms without well defined meanings. The proper technical term is "interpolation." Doing interpolation, just like doing decimation (converting the digital bitstream to music), requires filtering. Because of the math used to design virtually all filtering, called Fourier transforms, filters cannot be optimized for both time and frequency response. Thus every filtering step represents an imperfect compromise. Most DAC chips do the industry-standard 8x "oversampling" in three 2x steps, so the resulting musical signal has been compromised 4 times - three oversampling steps, one decimation step. So it is no surprise many people find the typical DAC's sound lacking, and prefer the sound of an NOS DAC, which as we've seen in the previous paragraph comes with its own sonic compromises.

 

- So what's a person to do? One excellent and relatively low cost alternative is to interpolate in the computer to 176.4 or 192kHz (or whatever the maximum input resolution of your DAC is) before feeding the signal to the DAC. An NOS DAC will then have much more headroom to work with when it employs its decimation filter, eliminating nearly all aliasing artifacts. A typical oversampling DAC will then do only one round of interpolation versus the usual three. Now there is nothing magic about doing interpolation in the computer; it has to be better interpolation than takes place in the DAC to result in better sound. One advantage of doing interpolation in the computer is that it can be done in one step rather than the typical three in a DAC chip. Another potential advantage is that the processing power of the CPU can be utilized to achieve more sophisticated filtering. If the interpolation in the computer is sufficiently good, there is an opportunity to achieve better sound than interpolating in the DAC, or not interpolating in an NOS DAC and suffering the sonic artifacts of steep filters. Some excellent interpolation filters are those employed by the XXHE and HQPlayer software for Windows; HQPlayer for Linux; and the iZotope software that comes bundled with Audirvana Plus for OS X.

 

Hope this has been helpful.

[bbreneman]

I use the BorderPatrol DAC SE (USB only version).   NOS DAC with R2R chip and no filtering.  I haven't heard every DAC by a long shot, but this is the most natural sounding one I've ever heard, and is amazing with Classical (85% of my listening) and Jazz (20%).   http://www.borderpatrol.net/DAC.htm.  Highly recommended.

[semente]

A few more educational posts by the Phasure NOS DAC designer:

 

On 04/02/2013 at 8:45 AM, PeterSt said:

NOS die-hards must know what they actually talk about;

 

Whether their Audio Notes can't have a digital filter because Audio Notes can't upsample because of the D/A chips used of whether it's a higher rate non delta-sigma (!) multi bit DAC which just does not provide upsampling to 16/44.1 Redbook material ... from the previous lesson it should be clear that the only NOS die-hards which are truly NOS lovers, just *can not* use any analog filtering at the same time. When they do, it is useless for the 30% harmonic distortion they *will* perceive (better : receive), but and however, very deep in the frequency band it may help amplifiers. So, when a 6dB filter is applied at 44100 corner frequency, at 88200 this is 12dB down, at 176400 it is 18dB down, at 352800 it is 24dB down, at 705600 it is 30dB down and somewhere there will be no output anymore.

In the mean time, there is (thus) -6dB at 44100 and -0dB at 22050.

 

Not that you can really understand from the above, but an analog filter as the only filter is useless.

... And this is how NOS die-hards should be talking always and ever about NOS/Filterless.

And they do !

 

While this is good to understand, it may also be good to understand that there is no use in a sort of shouting around that NOS is good because we can always apply an anlog filter. But might you ask your manufacturer to apply one with some usefulness, don't be surprised when he applies an e.g. 24dB filter at 44100 Hz, which is -18dB down at 22050, -12dB at 11025, -6dB at 5512.5 and loses its influence at 2756.25.

It would be the supiest thing. And therefore and again :

 

A NOS DAC is a NOS/Filterless DAC in reality. When not, it is inconcistent with its goal (which in the end is our own goal as a NOS die-hard).

 

 

What can I think more of ?

 

 

On 04/02/2013 at 9:28 AM, PeterSt said:

Ok, so a NOS(/Filterless !) DAC is more musical. How ?

 

Let's say we don't know, athough studies and experiments will exist to show what this can be about. Theoretically however, is should be about accuracy in the time domain vs. accuracy in the frequency domain. And FYI : a NOS DAC is highly INaccurate in the frequency domain but is perfect in the time domain. It does not ring - not post and not pre. No smear. But watch out, because what's no smear in the digital domain can result in high smearing in the analog domain (like speakers). Hint : it needs filtering.

 

NOS DACs sound far from muffled. But how come ? well, because the all too square waves they pass (hint : needs filtering) actually pass higher frequencies than originally there. So, when a 5000Hz frequency is output, you can bet it creates higher frequencies of that *IN* the audio band. Never mind the calculation, but this is the aliasing (someone called it "shadowing") which is in order. Look here for a 1000Hz signal and how it comprises of "stepping" or if you like a brickwall (hint ?) :

 

 

 

See ? these are actually squares. And hey, the better the DAC the more square that is. But you know what ? squares don't exist in electronics. They form from sines. Many many sines. The more real square, the more up to infinity the the number of sines to build the square. We call that "frequency" ...

Let's say that this 5000Hz stepped signal "needs" 1MHz of frequency to express those squares. Sadly though, because of how all works it is not allowed to have a higher frequency than half of the sample rate which is 44100 / 2 = 22050Hz. Because of this restriction all kind of anomaly products fold back in the audio band. Many frequencies will show while we only wanted our 5000Hz or whatever it was. This is called harmonic distortion.

In the mean time, these anomaly frequencies are correlated to the base frequency. They act as sheer overtones to the base frequency. Overtones without logic to the instrument, but they are there and you might perceive it as harshness. May, because it depends on the massiveness of the music playing and how large the mess gets. But two things :

1. It will sound more fresh because of the higher frequencies mixed with the actual signal;

2. It is perfectly accurate in the time domain.

 

We can add a #3 although highly suspect to strange thinking : Because of the accuracy in the time domain separate real squares in the original music, remain so. This is to be compared with proper filtering, which rings and smears and which does not allow the separation of the original pulse-like signals and where a violin becomes a (nice sine like) flute.

 

It should be key to the "musicality" that instruments remain to be recognized, while things like pace and proper timing could be a first for "musicality".

 

 

On 04/02/2013 at 10:06 AM, PeterSt said:

If you look again at the two pictures and where the right one is a zoomed version of the same left one, you can well look at it as if the left one is the upsampled version of the right one. This is really how it would look like. Now :

 

The steeper the squares or IOW the larger the voltage jumps needed to form them, the more frequency is needed to have them right. But this also counts the other way around : make the stepping smaller and less frequency is needed. So, upsample, and less frequency is needed. But merely :

 

The fewer frequency needed, the further away those backfolding anomalies stay from the audio band. Aha.

This is going to turn into a story against NOS die-hards ...

 

 

When I started out with my own NOS DAC and which was after owning Audio Notes (indeed) an other types of NOS DACs, I started that project for the sheer reason of being intrigued how in the world it could be possible that for most music a NOS DAC sounded way better than the oversampling types. It couldn't because they measure the most poor and I wanted to know why. To make a long story short (did I ?), I created a NOS DAC indeed, and no single analog filter is in there anywhere. Also no digital filter is in there or otherwise it would not be a NOS DAC.

I looked at the poor measurements of my own creation and thought how to improve on it while my NOS DAC still remained just that.

 

This is how the explicit (design) thought of having the filtering in-PC and which filtering is NEEDED.

Ehm, do you hear ?

It is needed for better figures while figures don't lie and it is needed for better sound quality because my ears don't lie just the same IF - and only IF you can compare apples with apples which I could with this DAC. Because remember, the DAC is and remains NOS/Filterless.

But there is one MAJOR difference :

 

I created that digital brickwall filter which does not ring. Not post and not pre. It needed 16x upsampling to do it right and might I think I need an analogue filter to protect my amplifiers I (or anyone) will stuff in that proper capacitor somewhere (see 2nd post or so).

 

And so, this DAC is perfect in the time domain, and is fairly accurate in the frequency domain.

Fairly ?

 

Yes, and this is the last subject of this sequence;

Eliminating the stepping distortion sufficiently is different from what we call a "reconstruction filter". Now what ?

 

When the (in-audio band) frequency gets too high, too few samples exist to form a normal sine. So, think of it : a 10000 Hz tone has 4 samples available to create its original sine. Uh-oh ... that is squares again !

Now, this is only partly solved by the upsampling filter. The result is that now images exist right beyond the audio band with half of the sample rate as the mirror. So, a 18Khz tone is also there at 22050 + (22050 - 18000) = 26100. It is 3dB or so down, but there. Same story and here too an analog filter can help.

Is this bad ?

 

Hey NOS lovers !?! do we think this is bad ??

Well, IMHO nobody should think this is bad, but NOS die-hards of course never will. They have all the in-band distortion out of the way (sorry guys, this NEEDS upsampling) and THD+N is now not 30% but 0.0018% (in my case) and ...

NO RINGING.

 

 

I hope to have shed some lights on this all. Not everything will have been understandable, especially not with my English, but you should understand one thing foremost : I am - no WAS that NOS lover. What came from it was for the better, while sustaining the key topic to what NOS is and feels ... the no ringing. But real NOS/Filterless for net result ? no please.

But NOS DACs, these days, are very very worth while, because they don't alter the sound other than their inherent good or less good electronics, while at the same time they 100% "depend" on what they are fed with. Like Jud suggested, take any of these in-PC filtering means and listen to how whatever means of filtering (including none) creates the sound. This is no joke. It is key. You are NOT stuck to the sound of the DAC when the filtering is not in there.

 

Regards,

Peter

[semente]

One more:

 

On 05/02/2013 at 10:15 AM, PeterSt said:

 

Yea, but now the why;

 

When you'd look at the output of an e.g. 18 KHz sine :

 

 

with the notice that this is the non-reconstructed form as NOS die-hards tend to listen to it :-)) ...

and when you take it from me that the -0dBFS point is at the +1 and -1 lines where some of the sample points rest ...

 

then it is easy to see how the *average* of that is less than the -0dBFS point. This is simply because most of the sample points incur for less output (voltage).

And notice that the explanation of it all is somewhat more complex and depends on how fast the DAC is on its slew rate etc. So, the slower it is, the more it will be a kind of more average mush because of overshoots on this near square wave (not that you'll see it like that, but it is) and knowing that any overshoot has its effect on the next sample.

 

 

 

Here you see a more "reconstructed" form of the same 18KHz. Notice the now higher (upsampled) sampling rate.

This rings a litte (see the three samples at the start) and already has a more equal top and bottom voltage. Notice though that this doesn't reach the +1 and -1 lines anymore, which is the effect of the so called "ripple" (which is a property to filtering results). So, while for this 18KHz frequency the tops are 50 mV under what they should be, for 17KHz and 19KHz it may be 10mV under or 10mV over.

 

Also notice the now clearly recurring longer term "wave" you see at the top and bottom. This is dangerous because it implies a frequency within itself (call it a resonance). If you look at the first picture from a distance you see similar, but since it is more spiky its resonance will be different, plus it is more subject to overshooting.

 

With a decent setup filter (which is not the last picture) top and bottom lines are equal for all frequencies (up to where the filter starts to roll off the high frequencies) which in the mean time implies nice sines. However, since the band where the roll of is allowed to be very small only (from hearing limit (is what ?) to 22049 Hz, the filter needs to be very steep and implies sheer infinite ringing when all further parameters are to be right. You can well look at this as "as infinitely" each sample interfering with the other (for its level (say volume) information) and *that* is muffled. It buzzes all the way, and you only know it can be different(ly sounding) when you first have it different.

And this is the fun with NOS DACs (which allow decent upsampling) because today you can just try it. And this for sure is NOT iZotope unless you have an analyser and are able to watch pictures like the above in combination with impulse response (which is what the NOS DAC is inherently good at and is the time domain I talked about). An oh, you can use iZotope all right, but all you would be doing is emulating that oversampling DAC.

 

If *that* is your intention, OK ... but it certainly isn't mine (and buzzes as hell again).

Or let's put that differently : I too don't use NOS/Filterless 100% legitimately, but at least I don't let it ring. To that there's really one solution at this moment that I know of, and this is the Arc Prediction filtering from XXHighEnd. Not that it is about that as a real subject, but any NOS die-hard should try its merits because it was 100% created for "you" and it is free to try. Works on the Mac just the same (Bootcamp).

 

Ok, done. ;-)

Peter

[Miska]

That beginning "rings", because it contains frequency components higher than 18 kHz and the filter cuts in. Any filter, analog or digital rings by definition. How much, depends mostly on order of the filter. All-pass filter doesn't, but it doesn't do (filter) anything either.

 

Those wave peaks are completely equal after proper reconstruction (to remove all frequency components above Nyquist frequency).

 

Here's Metrum Musette NOS DAC, 19 kHz sine wave at 44.1 kHz sampling rate (you can see the signal is totally unstable):

 

Here's the same file, upsampled to 384 kHz sampling rate using HQPlayer:

 

 

Another way to look at it, same DAC...

 

Spectrum analysis of 0 - 22.05 kHz sweep at 44.1 kHz sampling rate (you can see images around every multiple of 44.1k rate):

 

Same sweep, upsampled to 384 kHz sampling rate using HQPlayer (now there are images only around multiples of 384k rate - higher up and quite a bit lower level):

 

With better analog filter, those images around 384k could be even lower level (better reconstruction).

[davide256]

18 minutes ago, Miska said:

That beginning "rings", because it contains frequency components higher than 18 kHz and the filter cuts in. Any filter, analog or digital rings by definition. How much, depends mostly on order of the filter. All-pass filter doesn't, but it doesn't do (filter) anything either.

 

Those wave peaks are completely equal after proper reconstruction (to remove all frequency components above Nyquist frequency).

 

Here's Metrum Musette NOS DAC, 19 kHz sine wave at 44.1 kHz sampling rate (you can see the signal is totally unstable):

 

Here's the same file, upsampled to 384k sampling rate using HQPlayer:

 

 

So, if your 20 something maybe you care... why  would an audiophile population that's mostly older care, given our roll off in high frequency hearing? Most of us aren't listening for piccolo or pipe organ upper register harmonic structure and  orchestral instruments  fundamental tones end by 5 khz... your into 3rd order harmonics of their highest register.   Convince me that bad things happen at 12khz or below....

 

http://www.donbarbersound.com/media/libFrequencyRangeofInstrumentsAndVocals.pdf

 

[Miska]

13 minutes ago, davide256 said:

 

So, if your 20 something maybe you care... why  would an audiophile population that's mostly older care, given our roll off in high frequency hearing? Most of us aren't listening for piccolo or pipe organ upper register harmonic structure and  orchestral instruments  fundamental tones end by 5 khz... your into 3rd order harmonics of their highest register.   Convince me that bad things happen at 12khz or below....

 

http://www.donbarbersound.com/media/libFrequencyRangeofInstrumentsAndVocals.pdf

 

 

Because those high frequency high level directly correlated components easily generate intermodulation products in the audio band, down to 0 Hz...

 

Already, straight at the DAC output, the 19+20 kHz IMD test tone measurement doesn't look very pretty:

 

And this is already before you have any other electronics following it. And this is with only two tones, multi-tone signal with more tones makes it even worse.

[bbreneman]

I'm both an engineer and a trained and actively playing orchestral musician.  Music, both listening and playing, is my necessary escape from the former, so I tune out the moment the charts and graphs appear.  I apply a simple test to hardware when I'm evaluating it:  how much does it sound like sitting in an orchestra?  The more that I hear what I expect to hear, the more I value the solution, regardless of cost, technology, or whether it looks perfect on a scope.  

 

I will also add that I believe a well-played performance, properly recorded and mastered, makes more difference than any other single factor.  No amount of technology can rescue a lousy recording, and a great recording sounds great on pretty much any higher-end playback setup.  

 

As always, one man's opinion.

[Miska]

I want things to both sound and measure perfectly. I'm not giving up anything on either front.

 

[Miska]

Btw, if you people are worried about oversampling and digital filters, you should be much more worried about ADC and production side...

 

For ages, ADCs have been oversampling SDM type with digital decimation filters to convert into PCM. And nowadays typically masters are at least 96 kHz 24-bit. For RedBook output, that is then converted in software...

 

If you are worried about filter ringing, it is specifically the anti-alias decimation filter that generates ringing because it removes the high frequency components...

 

If you want to get rid of back-and-forth rate conversions, go for DSD content.

 

[ted_b]

1 hour ago, Miska said:

If you want to get rid of back-and-forth rate conversions, go for DSD content.

And if only there were a site for that. 

[PeterSt]

On 25-3-2018 at 11:12 PM, hvbias said:

Please feel free to share, positive, negative, whatever else! Thanks guys.

 

In the old days (say 10 years ago) when we used NOS "as is" hence without any filtering (digital nor analogue) NOS was famous for representing the individual instrument and the mess for when all comes together in masses like a full orchestra. With thanks to Ricardo (semente) for all the quotes of my texts which I apperantly ever back wrote (did I ?), back in 2009 or so I sat down to start out with that NOS for the individual instrument and build filters on top of it so the "masses" would sound OK too. So that is what I functionally did (and achieved with a by now always improving Phasure "NOS" DAC).

 

Soon after this was actually finished (say 2011) I already couldn't listen any more to large orchestras with Oversampling DACs. Why ? well, because the individual instruments could not be recognized. So somewhere in those quotes I mention flutes for violins, and to a large extent this really is so. It tortures your brain once you are used to your loudspeakers being able to render that all well. All is a matter of being used to and "reference", but Oversampling DACs just don't bring you there. Also not the similarly working filters.

It still is so that the ringing is the destructive party here.

 

Coincidentally yesterday I was listening to Jon Lord's (ex Deep Purple and ex everything by now) and I noticed how G-D beautiful the guy managed to put things together (by far superseding his similar attempt together with Deep Purple somewhere in the beginning of the 70's). What he does (and can be recognized from) is taking out whatever instrument from the orchestra and give it a solo (for instruments which are no solo instruments to begin with) and next you hear that same instrument among all of the others but still recognize the instrument. It feels like teaching (us). But is also testifies how oversampling wouldn't work out. You wouldn't be able to identify the instrument in solo very clearly, let alone you'll hear it back within the 110 other instruments.

 

Peter

 

[monteverdi]

Are there any conceptually differences between an oversampling DAC and a NOS DAC and feeding it oversampled data?  If the oversampling is happening in an external (computer) processor there is obviously more choice of software and processing power but then there also the problem of non optimal power supplies and cables etc.. or one can use some external oversampler e.g. from DCS.

 

What about digital/analog filters (so which NOS has none)?

[PeterSt]

14 hours ago, monteverdi said:

Are there any conceptually differences between an oversampling DAC and a NOS DAC and feeding it oversampled data?

 

Oversampling is not really the same as "upsampled data".

Very roughly, think that "oversampling" could be something as in "over" and that it's done more than actually required. These days I can explain it best my means of how an ADC could work : it may sample 10 times more than the output sampling rate requires. Now it takes the average of that and the result will be better (more accurate) than when one sample per output sample has been taken. It can see better into the noise.

 

Upsampling is the literal highering of the original sampling rate. Thus, how 44100 can turn into 176400 (or 706500 in the case of the Phasure NOS1 DAC).

 

The differentiation is elsewhere :

A genuine NOS DAC can only be a multi bit DAC and this in itself generally means R/2R. Inherently such a chip (or discretely set up resistor ladders (that's why the R/2R)) does not "oversample". It just converts. Such chips add nothing to the sound when executed well (which latter is far from easy because it is related to the always too low accuracy of the resistors and further tricks to get them there).

We could also say that the modern DAC chips are SDM based (Sigma Delta Modulation) and they inherently oversample (and meanwhile they also upsample).

 

The Upsampling step(s) is done to apply the digital reconstruction filter(ing). Say that this turns a 20000Hz digital registration (which can be done with 2 samples only when 44100 is the native sampling rate) into a nice sine again. Or something which approaches that. Or something less. Or not at all. And in this latter case we have genuine NOS again. Nothing is oversampled or upsampled.

The SDM chip always changes the data. But, this is for a good cause because it applies the "necessary" filtering meanwhile. "Necessary" between quotes because the NOS die-hard does not want that. Well, impossible with an SDM chip.

 

On a side note the story is more vague because chips exist that do not filter but are SDM anyway so they upsample anyway (say that they make DSD of it for easy thinking). The reason they don't filter is so the hardware designer is allowed to apply his own filtering means (can be another chip - can be other means like an arrangement of OpAmps doing the job, etc.).

 

What I started doing is explicitly use a NOS DAC setup (with multibit R/2R chips) and "slowly" turn that into a best filtering means for reconstruction (best = what my ears like best combined with better measurements) which - when done in software - is easily adjustable and apart from that can utilize all kinds of software doing it (as a plugin for your software player, for example). Back when I started doing this, such software existed but not for this reason. So, the DACs anticipating this did not exist and FYI AFAIK they still don't exist (OK, the Phasure). Thus, a Metrum (and quite some more) present themselves as "NOS" and that's it. They (not pointing at any brand in particular) claim - like in the old days - that 0.04% of THD+N is not so bad because it is about what you perceive from the music. Well, IMO that *is* bad, especially if we see that the 0.04% (@1KHz) can easily turn into 0.00063% like the Phasure NOS1 does it (this is 36dB better), and which is achievable with upsampling (16x) only and nice filtering in the mean time.

And still no oversampling, no SDM type and foremost : no ringing anywhere (this is dedicated to the Arc Prediction filtering).

 

Long story short : with an Oversampling DAC (chip) you always run into the unavoidable oversampling which R/2R does not apply inherently. The R/2R situation lends itself for being 100% guaranteed that nothing happens further down the line with your preciously applied upsampling/filtering in software. You can do what you want and you will know that your applied filtering "makes" the sound. And this latter really is so. Of course there's more to let a DAC sound good, but the filtering is crucial.

 

Such filtering can be let lose on an SDM chip just the same and it wil have effect. To this regard, think like whatever it is you apply to go from e.g. 44.1 to 176.4 will not be done by the chip any more. But it still will apply its own sauce (of further filtering).

 

15 hours ago, monteverdi said:

If the oversampling is happening in an external (computer) processor there is obviously more choice of software and processing power but then there also the problem of non optimal power supplies and cables etc..

 

You are right. OK, you are right for maybe everything outside of our (Phasure) own playback software because with that you won't even be able to see CPU cycles being applied to the playback process as a whole (for 44.1 to 705.6 upsampling/filtering). So there your mentioned issues just don't apply. But otherwise, yes. And true, if I was to upsample further than to 705.6 then I would have a bandwidth problem with the USB interface. Mind you, the 705.6 is 24 bit (32 bit actually) so there's where the bandwidth is eaten and there's where DSD (1 bit) can upsample way further. But now we're inherently in the SDM principle, and I took it that this is not what you want.

 

I hope this was clear a little !

Peter

[MrUnderhill]

Another vote for the Border Patrol SE, in my case sans USB - I use a Matrix SPDIF2 in front. Simply superb with a diet of 44.1/16, and has transformed my use of Qobuz.

 

M

[emcdade]

Just received my Borderpatrol SE today (USPS delivered it on Easter morning, WTF!).

 

I must say, very excited to put it in the chain tonight. 

[johndoe21ro]

On 3/27/2018 at 1:27 AM, skatbelt said:

 

What I quoted! But you already knew that. Speak for yourself, do not pretend it is an objective fact.

 

I had a Metrum Pavane and Adagio and an Aqua La Scala mkII in my system for several weeks. To my ears they sounded best when feeded native. And very, very good with plain redbook. Although not as good as my Chord DAVE...

Would you be kind to make a comparison between Chord DAVE and Metrum Adagio?

[skatbelt]

4 hours ago, johndoe21ro said:

Would you be kind to make a comparison between Chord DAVE and Metrum Adagio?

 

Both are very good DAC's. I loved the musicality of the Metrum Adagio. It excels in engagement and delivering emotion with a smooth, liquid (rounded?), romantic sound field. But it fails a little in layering and sound staging. DAVE is clearly in another league in this aspect. It shares the musicality with the Adagio but transparency, timbre, the reconstruction of the recording space and unraveling complex material are second to none with DAVE. Price-wise DAVE is also in another league. So in this case you get what you pay for.
 

The optimal digital connection on the Metrum was via AES/EBU fed by a Mutec MC-3+ USB. The USB implementation of the Metrum seemed of lower quality. Digital inputs on DAVE are all of the highest standard.

The only downside with DAVE for me is the build in headphone amp. It certainly isn't bad but it is not a competitor to my dedicated Bakoon HPA-21. I hoped it would be. I suspect the reason for this is that my Audeze LCD-3's are difficult to drive. The combination with the Bakoon on the other hand bettered the Metrum (but by a smaller margin than loudspeaker listening). Let me know if I touched all the things you expected.

[johndoe21ro]

Skatbelt, would you say that DAVE needs some body or punch? Did you find any fault with it?

[skatbelt]

46 minutes ago, johndoe21ro said:

Skatbelt, would you say that DAVE needs body or punch? Did you find any fault with it?

 

I don't think I understand the first part of your question. Do you mean up- or downstream? In most systems I expect other components to be the weakest part.

Apart from my remarks about the build in headphone amp no faults. I don't listen to DSD a lot but for people that do it can be a bit cumbersome that - for best SQ - you have to 'hard' switch between the PCM and DSD modes. Other observations: DAVE is pretty immune to power cord quality but benefits from good digital and analogue interconnects. Can things get better? Apparently yes when combined with the Chord Blu mk II. I did not hear that combination yet. I am waiting for a more cost friendly stand alone upscaler.

[gvl]

My opinion classical is probably one of the few materials that has a chance to sound reasonably good through a NOS DAC, and by NOS DAC I mean no digital filtering and no or very gentle analog filter. The reason is there is not that much high-frequency content in classical music, and as a result the level of ultrasonic images stays pretty low, which means low-levels of intermodulation distortion products in audio band, etc. Here is some classical peak spectrums for your viewing.

 

 

[gvl]

On 3/27/2018 at 3:42 PM, Miska said:

 

Because those high frequency high level directly correlated components easily generate intermodulation products in the audio band, down to 0 Hz...

 

Already, straight at the DAC output, the 19+20 kHz IMD test tone measurement doesn't look very pretty:

 

And this is already before you have any other electronics following it. And this is with only two tones, multi-tone signal with more tones makes it even worse.

 

One problem with this example is that -10dB is not a realistic level for 19/20kHz frequencies in actual music, you probably need to look hard to find material that has these frequencies above -40dB, it is 30dB lower than this synthetic example, which puts IMD products at what, -100db? That is lower than 16-bit SNR. Just saying.

[astrotoy]

What is NOS - New Old Stock?   My go to DAC is close to 20 years old. 

 

Larry

[esimms86]

Non oversampling, also called ladder DACs. Think Phasure NOS1a, Metrum Pavane, Holo Audio Spring, etc.

[gvl]

A ladder DAC can be oversampling, most are.